Thermal imaging

Thermal imagers are imaging systems that generate images of the observed scenery using thermal radiation emitted by the scenery. These numerous imaging systems can be divided into several different groups.

First, according to a method of creation of two-dimensional image of the observed scenery, thermal imagers can be divided into two distinct groups: thermal cameras and airborne thermal scanners.

Second, according to application area, thermal imagers can be divided into two groups: surveillance thermal imagers and measurement thermal imagers.

Third, according to a spectral band, thermal imagers can be divided into: MW (mid-wave) thermal imagers and LW (long-wave) imagers. Sometimes SW (short wave) thermal imagers are added too.

Fourth, according to technology of IR detector (IR FPA), thermal imagers can be divided into at least three different generations.

1.1 Thermal cameras versus thermal scanners

Imaging thermal scanner is a thermal imaging system that provides creation of a two-dimensional thermal image of the observed scenery only when the system or the objects are moving.

Thermal cameras represent probably over 99% of all existing thermal imagers. Imaging thermal scanners are almost exclusively airborne systems used for reconnaissance applications because they offer very wide field of view (standard 120º) in contrast to the thermal cameras offering field of view not wider than about 30º.

Because of distinct differences in design of these two types of thermal imaging systems and narrow specialised market, the imaging thermal scanners are very expensive systems. Due to mass application of thermal cameras their prices are significantly lower. There exist numerous literature on both imaging thermal scanners and on thermal cameras. We can here only mention that detail presentation of a design of thermal cameras was presented in Refs. 1,2,3,4,5 and detail discussion on a design of imaging thermal scanners in Ref. 6.

As we mentioned earlier, thermal cameras are the most numerous group of thermal imagers. Practically, almost all thermal imagers are thermal cameras. Therefore both review of thermal imagers and later analysis of test methods in next chapters is mostly limited to thermal cameras. Next, the terms “thermal imager” and the term “thermal camera” will be used as equivalent terms.

1.2 Applications of thermal cameras

According to their applications, thermal imagers can be generally divided into two basic groups: surveillance thermal cameras and measurement thermal cameras. The surveillance thermal cameras are mostly used in military applications for observation of a battlefield in darkness or in difficult atmospheric conditions by creating the relative temperature distribution of the terrestrial scenery being observed. The measurement thermal cameras are used for civilian applications in industry and science; mostly for non-contact measurement of temperature distributions on the surface of the tested objects. Nowadays, the borderline between these two groups becomes more fluid as there are some cameras that can be used for both observation and measurement applications. However, this situation is still an exception from the rule as most surveillance cameras do not have capabilities to measure temperature of the observed objects and the image quality of the measurement systems is inferior to the image quality of the observation thermal cameras.

Image quality is the most important criterion for performance evaluation of surveillance (military) thermal cameras. In case of measurement (commercial) thermal cameras, the situation is more complicated.

Applications of measurement thermal cameras can be divided into two general groups: the applications that require only relative temperature measurement and the applications that require absolute temperature measurement. Although the same cameras can be typically used in both applications there are different criteria of assessment of camera suitability for these two groups of applications. If the camera is used in applications when only relative temperature measurement is needed, like in non-destructive thermal testing (NDTT), then the quality of the thermal image of the tested object is usually the most important criterion like in the case of the observation thermal cameras. If the measurement thermal camera is used in applications when an absolute temperature measurement is needed, then accuracy of measurement results is the most important criterion.

1.3 Spectral band

Objects of typical earth temperatures emits radiation mostly in the spectral region from about 3 m to about 15 m. Thermal radiation emitted by these objects dominate over the radiation reflected by them at this spectral range because the radiation emitted by sun, moon, stars and typical artificial sources is weak for wavelengths over 3 m. There are two „atmospheric windows” in the above mentioned range: the 3-5-m window and the 8-12-m window. Therefore there are two main types of thermal imaging systems: the middle-wave MW systems using the 3-5-m window and the long-wave LW systems using the 8-12-m window and rarely available commercially SW systems of spectral band located within 1-3-m range.

MWIR and LWIR mm spectral bands differ substantially with respect to background flux, scene characteristics, temperature contrast, and atmospheric transmission under diverse weather conditions. Factors which favour MWIR applications are: higher contrast, superior clear-weather performance (favourable weather conditions, e.g., in most countries of Asia and Africa), higher transmittivity in high humidity, and higher resolution due about 3 times smaller optical diffraction. Factors which favour LWIR applications are: better performance in fog and dust conditions, winter haze (typical weather conditions, e.g., in West Europe, North USA, Canada), higher immunity to atmospheric turbulence, and reduced sensitivity to solar glints and fire flares. The possibility of achieving higher signal-to-noise (S/N) ratio due to the greater radiance levels in LWIR spectral range is not persuasive because the background photon fluxes are higher to the same extent, and also because of readout limitations. Theoretically, in staring arrays charge can be integrated for full frame time, but because of restrictions in the charge-handling capacity of the readout cells, it is much less compared to the frame time, especially for LWIR detectors for which a background photon flux exceeds the useful signals by orders of magnitude.

To summarize, in general, the 8-14 mm band is preferred for high performance thermal imaging because of its higher sensitivity to ambient temperature objects and its better transmission through mist and smoke. However, the 3-5 mm band may be more appropriate for the hotter object, or if sensitivity is less important than contrast. Also additional differences occur; e.g. the advantage of MWIR band is smaller diameter of the optics required to obtain a certain resolution and that some detectors may be operated at higher temperatures (thermoelectric cooling) than it is usual in the LWIR band where cryogenic cooling is required (about 77 K). Therefore there is no definite, always valid answer which type of thermal imagers (MW thermal imagers or LW thermal imagers) should be preferred. Both types of thermal imagers have certain advantages and disadvantages.

1.4 Generations of thermal imagers

Thermal cameras are generally divided into three generations. Scanning cameras built using discrete detectors, simple non-multiplexing photoconductive linear arrays (typically PbSe, InSb or HgCdTe) of elements number not higher than about one hundred, or the SPRITE detectors are the first generation thermal cameras. They usually operate in 8-12-m spectral range, use the optics of F/2-F/4 number, and are characterised by temperature resolution NETD of about 0.2 K. Small quantities of first generation thermal cameras were introduced as military equipment in the 1970s, more in the 1980s. Thousands of these systems are still in military services, spare part will be available for many years. The US common module HgCdTe arrays that employ 60, 120 or 180 photoconductive elements are the prime example of Gen 1 thermal cameras.

Fig. 2. Principle of operation of scanning thermal cameras and array thermal cameras.

Scanning cameras built using linear or 2D focal plane arrays (FPA) of elements number higher than about 100 but lower than about 10000 are the Gen 2 thermal cameras. Temperature resolution NETD of these cameras is improved up to the level of about 0.1 K. They are also characterised by smaller weight and size and improved reliability. The 1980s is a period when most modern army forces started to use the second generation thermal cameras. The cameras of this generation are presently majority of all military thermal cameras. New version of these FPAs offered in a form of a single chip fully integrated with readout electronic are even now an attractive solution for many observation applications. Thermal cameras built using these improved linear FPAs are often termed Gen 2+. Temperature resolution NETD of Gen 2+ can be improved up to the level of about 0.05 K. Typical examples of these systems are HgCdTe multilinear 2884 arrays fabricated by Sofradir both for 3–5-m and 8–10.5-m bands with signal processing in the focal plane (photocurrent integration, skimming, partitioning, TDI function, output preamplification and some others).

Fig. 4. Exemplary Gen 2 thermal camera: Sophie (courtesy of Thales Optronique) .

Third generation cameras are non-scanning thermal cameras build using 2D array detectors (cooled FPA based on InSb, HgCdTe, QWIP technology or non-cooled FPAs based on microbolometer or pyroelectric/ferroelectric technology) that have at least 10⁶ element on the focal plane. These staring arrays are scanned electronically by circuits integrated with the arrays. These readout integrated circuits (ROICs) include, e.g., pixel deselecting, antiblooming on each pixel, subframe imaging, output preamplifiers, and some other functions. The opto-mechanical scanner is eliminated and the only task of the optics is to focus the IR image onto the matrix of sensitive elements.

Fig. 5. Exemplary cooled Gen 3 thermal camera: Catherine XP (courtesy of Thales Optronics).

Third generation thermal cameras have been offered since the beginning of the 90s to compete with their predecessor. First, they have been offered as cooled MWIR cameras (using InSb or HgCdTe technology) sensitive in 3-5-m atmospheric window in situation when for most geographic conditions LWIR thermal cameras are desirable. Cooled LW IR Gen 3 thermal cameras based on QWIP technology started to be commercially available at the end of the 1990s. Almost at the same time non-cooled thermal cameras based on microbolometer and pyroelectric/ferroelectric technologies became fully commercially available. Image quality of non-cooled thermal cameras is inferior to image quality offered by cooled cameras but is good enough to be used in many short and medium range applications. Due to a 2-4 times lower price than equivalent cooled systems, the number of non-cooled thermal cameras is growing rapidly in both military and commercial applications.

Parameters of thermal cameras from the same generation can vary significantly. Therefore it is not possible to form a single table enabling accurate comparison of parameters of thermal cameras from different generations. Table 1 was created on the basis of a review of the parameters of different observation thermal cameras offered during the last 30 years but should be treated as an estimation of the sophisticated situation on the market.

Fig. 6. Exemplary non-cooled Gen 3 thermal camera: ELVIR (courtesy of Thales Angenieux).

As we can see in Table 1, the Gen 2 thermal cameras are characterised by significantly better thermal and spatial resolution that the Gen 1 thermal cameras. This means that quality of the image and sensitivity offered by the latter cameras is significantly inferior. However, situation is not so clear if we compare Gen 2 and Gen 3 cooled thermal cameras. Thermal sensitivity of Gen 3 cooled thermal cameras is usually at least slightly better that of thermal resolution of Gen 2 cameras. However, image resolution of modern Gen 2 thermal cameras is superior to image resolution of typical Gen 3 cameras based on 320240 FPA, particularly in a horizontal direction. This inferiority of Gen 3 cameras can be eliminated by the use of microscanning technique, that can improve, up to two times, image resolution in both horizontal and vertical directions. However, the disadvantage of microscanning technique is the higher production costs and reduced reliability. The inferiority of image quality offered by typical Gen 3 thermal cameras in comparison to Gen 2 cameras can be fully eliminated if 640×512 or bigger FPAs are used.

A generation number is not connected strictly with image quality; it is more connected with mass, dimensions, manufacturing costs and reliability of the thermal camera. The generation number suggest rather potential of the detector module but does not describe quality of a thermal camera. Next, in order to evaluate properly thermal cameras, not only image quality (detection, recognition and identification ranges) but also other factors like mass, dimensions, resistance to harsh environmental conditions, ergonomics must be taken into account. Further on, there are, on the market, thermal cameras integrated with additional modules like GPS, laser range finder, goniometer, day light TV camera and laser pointer. These additional modules can significantly increase capabilities of a thermal camera. To summarize, evaluation and comparison of thermal cameras is a complicated and risky task that requires to take into account a set of factors that could vary, depending on the final user needs.

Fig. 7. Sophie MF – thermal camera integrated with laser range finder, goniometer, day light TV camera and laser pointer (courtesy of Thales Optronique) .

Detectors used in Gen 1, Gen 2 and partially Gen 3 of thermal cameras require cooling, typically to the temperature equal to 77 K. First thermal cameras were cooled using dewar coolers. The dewar cooler is essentially a “ vacuum bottle” filled with a coolant. Different liquid gases can be used as coolants. However, liquid nitrogen is used as a coolant in almost all dewars used in practice.

The cryogenic cooling is characterised by a few significant disadvantages like necessity to have a source of liquid nitrogen supply readily available, limited working time of the dewar after filling, and necessity to keep quasi-horizontal position of the thermal camera. Therefore later cooled thermal cameras employ Stirling coolers, or rather rarely Joule-Thomson coolers.

The Stirling cooler is fundamentally a closed-cycle compression-expansion refrigerator with no valves; instead, it incorporates a regenerator. The regenerator is a tube of porous material that has low thermal conductivity to maintain a temperature gradient and high heat capacity to act as an efficient heat exchanger. Typical Stirling coolers operate with a sealed charge of helium, which is mechanically compressed and then allowed to expand near the dewar cold finger. This expansion cools the detector , and the helium is then “recycled” through cooler’s compressor.

The Stirling coolers can cool the detector to the required temperature , usually after 3-5 minutes from the turn on. These coolers require recharging and service by the cooler manufacturer after a fixed period of time; typically after about 1000-10000 hours. Size and mass of these coolers depend on required cooling power. The power of about 0.2-0.6 W is enough to cool a small single detector but a few times higher is needed to cool an array FPA.

The Joule-Thomson cooler is an open cycle cooler that converts pressurized gas (typically nitrogen, argon, CO₂) to criogenic liquid gas. High pressure gas is cooled by expansion at the throttle valve, flows back through the counter-current heat exchanger and precools the incoming gas until the gas is liquefied as it leaves the throttle valve. Because Joule-Thomson coolers require the supply of pressurized gas they are rarely used in thermal cameras but they are typically used in IR guided seekers where the required working time is relatively short.

Both Stirling coolers and Joule-Thomson coolers are relatively expensive components that represent a significant portion of cost of a whole thermal camera. Therefore it was highly desirable to eliminate these components as it has been done recently by introduction of non-cooled FPA based on microbolometer and pyroelectric/ferroelectric technologies. However, please note that so-called non-cooled FPAs usually require temperature stabilisation and thermoelectric coolers are usually used in the non-cooled thermal cameras.

The thermoelectric coolers employ the effect of Peltier who found that (the temperature changes??) when current flows in a circuit consisting of two dissimilar conductors. In contrast to the cryogenic coolers, Stirling coolers and Joule-Thomson coolers, the thermoelectric coolers???? cannot be used to cool detectors down to very low temperatures; temperature difference of not more than about 50–70C to ambient temperature can be achieved. Next, significant difference is low cost of these coolers in sharp contrast to the Stirling coolers and the Joule-Thomson coolers.

Apart from the MWIR thermal cameras and the LWIR thermal cameras there are also SWIR cameras of a spectral band located within the spectral range 1-3 m. It is questionable whether the SWIR cameras are thermal cameras as in this spectral range the reflected radiation dominates over the emitted radiation for the objects of temperatures below about 100C. However, let us treat them as a group of thermal cameras because of very similar design to MWIR and LWIR thermal cameras.

The SWIR cameras are only a marginal group of thermal cameras. The SWIR cameras have been commercially available on the market for no more than a decade. This situation originated the fact that the SWIR range has not been an interesting range for both military and civilian applications. Due to dominance of the emitted thermal radiation and the atmospheric windows, military agencies were interested mostly in the MWIR and LWIR ranges. Because of sensitivity range of human sight and well developed silicon technology civilians were interested in the visible and NIR ranges.

This lack of significant interest created the situation when up to the middle of the 90s no well matured technology of detector arrays for SWIR range was available [7]. Currently, this vacuum is occupied by InGaAS arrays and the SWIR cameras found a few applications; mostly in telecommunication sector enabling accurate coupling of optical fibres working at 1.53 m and in museums for painting reflectography.

1.5 Technology trends

Thermal imaging is one of the technologies of paramount importance for military&security sector. Thermal imaging has found also numerous applications in a civilian sector. Therefore it is not strange that there is a lot of efforts to improve existing technologies of manufacturing thermal imagers and to develop new technologies.

Technology of non-cooled thermal imaging experienced very rapid growth during the last decade [8]. Parameters of non-cooled imagers improved so much that nowadays non-cooled imagers dominate on the market of short range surveillance thermal imagers in both military and civilian applications. The critical factor on this market is a price. Therefore now, the technology efforts concentrate on decrease in manufacturing costs but still keeping or even improving the image quality and reliability. Critical areas are two modules of non-cooled thermal imagers: infrared focal plane area and infrared optics.

The top end non -cooled thermal imagers offer 680x480 image resolution and are directed mostly towards more demanding military applications. Non -cooled thermal imagers of 320x240 image resolution are typically targeted to general surveillance and radiometric applications (security sector, automotive industry, non-contact temperature measurements, etc.). Imagers of 160x120 or lower resolution are targeted to mass applications in low-cost intruder detection systems or as non-contact imaging thermometers.

Technology of cooled thermal imagers is for the last decade under the pressure from non-cooled technology. Because of a need to use expensive cooler module, the cooled technology is inherently more expensive that non-cooled technology. Because of this situation, manufacturers of cooled thermal imagers concentrate on market of long/medium range surveillance or for applications that require dynamic surveillance of high speed scenarios. The efforts go into four directions. First, reducing costs of manufacturing of II and III Gen thermal imagers of medium resolution (up to 640x480). Second, development of high resolution thermal imagers of image quality comparable to quality of images offered by High Definition Television (minimal image resolution 1280x720 pixels is needed) [9,14]. Third, development of multi-band cooled imagers capable to employ spectral phenomenon as an effective tool in both surveillance and measurement applications [10,11]. Fourth, development of polarization-sensitive thermal imager to further improve the capabilities of cooled technology [12,13].

An increasingly noticeable trend appeared on the market to integrate thermal thermal imagers with other imaging and non-imaging sensors. Such integrated imaging systems (thermal imager, TV camera, laser range finder) have been used for quite a long time in airborne applications. Nowadays, however, modern airborne imaging systems consist of more sensors: high resolution four-FOV thermal imager, high resolution color TV camera, LLLTV camera, laser range finder, laser pointer [14]]. Next, ground portable thermal imagers are more frequently integrated with additional modules like GPS, laser range finder, goniometer, day light TV camera, and laser pointer. In some airborne, naval or ground applications, thermal imagers are integrated with classical radars or millimeter-wave radars. In all cases , such integration significantly increases capabilities of thermal imagers.

1.6 References

1. Campana Stehen B. ed., The Infrared & Electro-Optical Systems Handbook, Vol. 5: Passive Electro-Optical Systems , Chapt. 2 Forward Looking Infrared Systems, SPIE (1993).

2. Campana Stehen B. ed., The Infrared & Electro-Optical Systems Handbook, Vol. 5: Passive Electro-Optical Systems , Chapt. 3 Staring Sensor Systems, SPIE (1993).

3.Paul W. Kruse, Uncooled Thermal Imaging Arrays, Systems, and Applications, SPIE Vol. TT51 , 2001.

4.Herbert Kaplan, Practical Applications of Infrared Thermal Sensing and Imaging Equipment, SPIE Vol: TT75, 2007.

5.Gerald C. Holst, Common Sense Approach to Thermal Imaging, SPIE Vol: PM86, 2000.

6. Campana Stehen B. ed., The Infrared & Electro-Optical Systems Handbook, Vol. 5: Passive Electro-Optical Systems , Chapt. 1 Infrared Line Scanning Systems , SPIE (1993).

7. Baronti S. et al., Multispectral imaging system for the mapping of pigments in works of art by use of principal-component analysis, Applied Optics, 37, 1299-1309 (1998).

8.Paul A. Manning; John P. Gillham; N. J. Parkinson; Tej P. Kaushal, Silicon foundry microbolometers: the route to the mass-market thermal imager, Infrared Technology and Applications XXX Conference, SPIE Vol. 5406, 2006.

9. Anders G. M. Dahlberg, High-resolution QWIP thermal imager for AFV upgrade, Infrared Technology and Applications XXX Conference, SPIE Vol. 5406, 2004.

10. Rainer Breiter; Wolfgang A. Cabanski; Karl-Heinz Mauk; Werner Rode; Johann Ziegler, Portable sequential multicolor thermal imager based on a MCT 384x288 focal plane array, Infrared Technology and Applications XXVII Conference, SPIE Vol. 4369, 2001.

11. W. Randy Bell; Paul G. Weber, Multispectral Thermal Imager: overview, Algorithms for Multispectral, Hyperspectral, and Ultraspectral Imagery VII Conference, SPIE Vol. 4381, 2001.

12.Daniel W. Beekman; James B. Van Anda, Polarization-sensitive QWIP thermal imager, Infrared Detectors and Focal Plane Arrays VI Conference, SPIE Vol. 4028, 2000.

13.Olivier Cocle; Christophe Rannou; Bertrand Forestier; Paul Jougla; Philippe F. Bois, QWIP compact thermal imager: Catherine-XP and its evolution, Infrared Technology and Applications XXXIII Conference, SPIE Vol 6542, 2007.

14. Star Safire HD QWIP thermal imaging equipment - operator manual, FLIR Inc., P/N 4100070, 2006.

2 Parameters of thermal imagers

Thermal imagers generate images that can be seen by humans and it is possible to evaluate a thermal imager by looking on images generated by the imager. However it is surprisingly difficult even for an expert to precisely evaluate thermal imagers only by looking of images of typical scenery. Measurement of a number of parameters is needed in order to accurately evaluate tested thermal imagers.

Parameters are quantitative physical measures of thermal imagers. The measurement is typically done in laboratory conditions but generally parameters of thermal imager enable an expert to predict how this imager will perform under real observation conditions. Characterization of thermal imagers is relatively well standardized [1,2,15,17,19,26] and there exists also a numerous literature on a subject of testing thermal imagers [3,10,11,12,16,21,22,23,25,27,28]. On the basis of the mentioned above literature, parameters that describe performance of thermal cameras can be, in general, divided into eight groups:

Subjective image quality parameters give information about ability of the system: thermal camera – human observer to detect, recognize, and identify targets at different scenarios.

Response parameters give information about response of the thermal camera to variable size or variable temperature targets.

Noise parameters - about noise that limits camera sensitivity detecting low contrast targets.

Image resolution parameters carry out information about camera ability to perceive small details of high contrast images.

Geometric parameters give information about geometrical relations between the target and its image. Accuracy parameters give information about accuracy of temperature measurement using the thermal camera.

Operational parameters provide information about position of the observed target or position of human eye necessary for proper operation of the camera.

Finally, spectral parameters provide information about camera responsivity versus wavelength.

Subjective image quality parameters , response parameters , noise parameters , image resolution parameters , and geometric parameters give in general information about performance of thermal cameras in surveillance applications. Accuracy parameters are vital to evaluate thermal cameras for measurement applications. Operational parameters provide information about some practical aspects of work with the tested camera.

Attention: there is no correlations between parameters from different groups having the same number. Numbering was used only to make easier identification of different parameters.

Accuracy parameters are necessary to evaluate commercial thermal cameras to be used as non-contact thermometers. They are useless for imaging applications and will be excluded from further discussion.

Measurement of operational parameters of thermal cameras does not differ significantly from a measurement of the same parameters of other visual imaging systems. There is the same situation in case of spectral parameters . Therefore both operational parameters and spectral parameters will not be discussed here.

2.1 Subjective image quality parameters

Thermal cameras are imaging systems used to enhance human ability to see in darkness and poor visibility conditions. parameters of subjective image quality perceived by humans are considered as the most important parameters of military (observation) thermal cameras from the point of view of the user who wants to have the best ranges of detection, recognition and identification of enemy targets.

MRTD is a measure of ability to detect and recognize targets on non-uniform background and MDTD is a measure of ability to detect targets on uniform background. The targets of interest are usually located on non-uniform backgrounds and MRTD is considered as the most important measure of military (observation) thermal cameras.

2.1.1 MRTD

The MRTD is a subjective parameter that describes ability of the imager-human system for detection of low contrast details of the observed object. It is a function of a minimum temperature difference between the bars of the standard 4-bar target and the background required to resolve the thermal image of the bars by an observer versus spatial frequency of the target.

Generally, MRTD is measured by determining the minimum temperature difference between the bars of the standard 4-bar target and the background required to resolve the thermal image of the bars by an observer for 4-bar targets of different dimensions (spatial frequency). The measurement results of an exemplary long range military thermal camera are shown in Fig. 4.

Fig. 2. Image of five 4-bar targets of different spatial frequency at the same temperature difference.

As we can see in the above presented definition, MRTD is a subjective parameter that takes the observer into account. This subjectivity and time consuming measurements cannot be accepted in high volume production environment where, so-called, AutoMRTD is preferred.

AutoMRTD is a test methodology that proposes a quick objective method to measure MRTD of a thermal camera using an algorithm presented below.

Because to test new thermal cameras it is required to measure only NETD (noise equivalent temperature difference) and MTF (modulation transfer function), the measurement time is significantly shortened and additionally subjectivity of typical MRTD measurement procedure is eliminated. However, because determination of K() requires testing of many thermal cameras (at least 20–30) and AutoMRTD test method is profitable only for large production lines.

2.1.2 MDTD

The MDTD is a subjective parameter that describes ability of the imager-human system for detection of small size targets. It is a function of a minimum temperature difference between the circular (square) target and the background required to detect the target by an observer versus inverse spatial size of the target.

The MDTD is measured by determining the minimum temperature difference between the target and the background required to detect the thermal image of a target, for targets of different spatial dimensions.

2.1.3 Evaluation of ranges of effective observation

Detection, recognition and identification ranges of a target of interest is the prime criterion for evaluation of most surveillance thermal cameras.

It is possible to measure directly detection, recognition, and identification ranges of a target of interest and to evaluate the tested thermal camera on the basis of the test results. However, it is a risky solution. The ranges vary with observation conditions (atmosphere, background) and it is relatively easy to manipulate with the detection, recognition and identification ranges at real conditions if the observation conditions are not very precisely specified. Next, it is difficult to compare test results of different cameras tested at different time periods and different observation conditions.

It is possible to calculate theoretically the detection, recognition, and identification ranges of any target (man, tank, truck) at simulated conditions using a simulated thermal camera but due to different calculation methods existing in literature there can be later problems with precise interpretation of the calculation results. The safest way to evaluate surveillance thermal camera is to use the method proposed by NATO standard: STANAG 4347, Definition of nominal static range performance for thermal imaging systems, 1995. The standard defines precisely parameters of the standard target, standard atmospheric conditions and presents a way to calculate the detection, recognition and identification ranges of the standard target on the basis of the MRTD function of the tested thermal camera.

The detection and recognition ranges of the standard NATO target from an exemplary specifications are shown in Table 1.2.

Table 1.2. Exemplary specifications of a thermal camera using a concept of standard NATO target.

In order to calculate the detection and recognition ranges of the standard NATO target we must know MRTD function of the tested camera. Therefore a more common way to specify requirements on surveillance thermal cameras is to present requirements on MRTD characteristic in the specifications as shown in Table 1.3.

Table 1.3. Exemplary requirements for MRTD function of a long range surveillance thermal camera .

To summarize, we can say that MRTD is the most important characteristic of thermal cameras from the point of view of the user who wants to have the best ranges of detection, recognition, and identification of enemy targets. Having known MRTD functions of different thermal cameras we can calculate the ranges of detection, recognition of the standard NATO target and compare their performance (only cameras of almost the same field of view should be compared in this way). Therefore proper specifications of a thermal camera should specify precisely maximal values of MRTD function at a set of spatial frequencies. Then, the measured MRTD values must be lower than the values in the specifications if the camera is to pass the test.

Now, let us discuss the way to calculate the detection and recognition ranges on the basis of the measured and known MRTD function of the tested thermal camera using the recommendation from STANAG 4347 shown in Table 1.4.

Table. 1.4. Target parameters, atmosphere conditions and resolution criteria specified in the standard STANAG 4347.

The detection, recognition and identification ranges of the standard NATO targets can be calculated using the below presented algorithm.

In order to practise using this algorithm let us assume that we want to determine the detection, ranges of the standard NATO target using the thermal camera of MRTD function shown in Fig. 7. Calculation results are shown in Fig. 8 and we can conclude that the detection ranges are the following:

Calculation of the recognition and identification ranges can be done in the same manner.

Fig. 8. Converted MRTD function (for detection range) and the functions T_o[R] at different atmospheric conditions (rectangles – MRTD values, circle – T_o[R] at good transmission, triangles – T_o[R] at limited transmission) .

1.2 Response parameters

Response parameters give us information about system response to variable temperature targets or to the variable size targets.

Responsivity function is the system response to a large target of variable temperature. It provides information on gain, linearity, dynamic range and saturation level. The signal transfer function (SiTF) is the linear part of the responsivity function.

Aperiodic transfer function (ATF) is defined as a normalized dependence of system response to a variable size circular (square) target. It provides information on system ability to detect small targets.

Slit response function (SRF) is defined as a normalized dependence of system response to a variable size slit target. It provides information on system ability to detect long narrow targets.

1.2.1Responsivity function

Responsivity function is a function of an output signal (screen luminance, or electrical signal) versus target temperature (absolute or relative) in case of large, constant size target (Fig. 9, Fig. 10). It can be characterized by three digital parameters: SiTF, saturation level, and dynamic range that are determined on the basis of measurement results of the responsivity function.

Fig. 10. Responsivity function of a AC coupled thermal camera or DC coupled camera with AGC (automatic gain control).

The signal transfer function SiTF or the responsivity is the linear part of the responsivity function. It is calculated as tangent of the angle between linear part of the responsivity function and the temperature axis (the slope of the linear part).

Dynamic range is the ratio of the maximum measurable input signal and the minimum measurable input signal.

The situation is rather unclear what are really these two measurable input signals. There are at least three different definitions of a dynamic range used in specifications of thermal cameras.

First, the dynamic range is defined as a ratio of temperature difference generating the output signal equal to 95% (or 90%) of the saturation level value to the temperature resolution of the tested camera. It is typically assumed that the temperature resolution is equal to NETD of the tested camera.

Second, the dynamic range is defined as a ratio of the upper value to the lower value of the temperature difference range when the deviation between the response function RF(T) and its linear approximation is within specifications

Third, the dynamic range is defined as a ratio: the temperature difference T, when SiTF is crossing the saturation level, to the temperature resolution of the tested thermal camera.

2.2.2 Aperiodic Transfer Function

Thermal camera can detect targets of angular size smaller than its instantaneous-field-of-view (IFOV) calculated as a detector (single pixel) angular substance. Aperiodic transfer function (ATF) is the dependence of a normalized function of the output signal (voltage, current, digital) on a variable size circular (square) target.

The output signal generated by such a small size target depends on the target area. For an ideal thermal camera, the signal is proportional to the target area when the target area is smaller than IFOV; the signal does not depend on the target area when the target area is bigger than IFOV (Fig. 12).

The difference between the ideal and real ATF is caused by the image blur generated by optical and electronic systems. Therefore the target transfer function (TTF) calculated as the ratio of the real ATF to the ideal ATF provides useful information about this phenomenon.

The point visibility factor (PVF) is a point of TTF function determined for the conditions when the target area approaches zero. The PVF is also sometimes termed the ensquared energy (EE), or ensquared power(EP) or blur efficiency.

2.2.3 Slit Response Function

The SRF is defined as a function of the signal generated by a slit versus width of the slit normalized to the signal generated by a very wide slit. It can be treated as one dimensional ATF.

SRF generally provides directly information on the system ability to detect long narrow targets.

2.3 Noise parameters

Noise is a phenomenon that can significantly decrease image quality and limit system ability to detect low contrast targets. Noise parameters are very important performance measures of thermal cameras quality.

Noise present in thermal images can be in general divided into two groups: temporal noise and spatial noise. The temporal noise refers to temporal variations of the signals generated by detector pixels during observation of an uniform target: variations of the signal in a single line for scanning cameras or frame to frame variations of pixels signals for array cameras. The spatial noise refers to differences between the signals generated by different pixels during observation an uniform target that does not change from frame to frame. Both types of noise have its own noise power spectral density (NPSD).

Noise is a complex phenomenon, difficult for characterisation. In general, we can find in literature three different noise analysis approaches to characterise the noise phenomenon present in thermal images:

1.3.1.13D noise model

The 3D noise model is based on the concept of the D_i directional averaging operators that allow the mathematical derivation of eight noise components from the noise data set [8]. The operators average the data in the direction indicated by the subscripts.

Let us assume that a sequence of images generated by the tested imager was captured. Then, the captured data can be presented in form of 3D array N_TVH (Fig. 16). The T-dimension represents time or the framing sequence. The H-dimension and V-dimension give spatial information. In case of staring systems, m and n indicate the pixel location; in case of scanning systems, m refers to a pixel location but n refer to time or sample number in a digitized analog signal. Names of each components and information provided by each component are presented in Table 5.

Fig. 16. Three dimensional noise model coordinate system: m-row number, n-column number, N-frame number .

The noise components are calculated by converting the 3D array into a series of 2D or 1D arrays. The conversion formulas are presented in .

It is possible to get detail information about the nature of a noise phenomenon by the analysis of a sequence of the images generated by the tested thermal camera using the 3D noise model and this model is often used by manufactures of thermal cameras. On the other hand, the 3D noise model due to many parameters used for characterisation of a noise phenomenon is complicated and very rarely used by users of thermal cameras who prefer simple solutions. Because of these customer demands for simplicity, even manufacturers rarely publish data using 3D noise model.

1.3.1.2Single parameter approach

If we look into technical data offered by manufacturers of most thermal cameras we find there a parameter called “thermal sensitivity”, “thermal resolution”, “temperature resolution” or “ NETD” that provide information about the system noise.

The parameter has different names but usually means a noise equivalent temperature difference (NETD). The problem is that different definitions and measurement techniques of NETD are used in literature.

According to its classical definition, NETD is defined as the blackbody temperature difference between a target and its background required to produce a peak-signal-to-rms-noise ratio of unity at a suitable point in the output electrical channel.

This definition was developed at the time when all thermal cameras were the scanning thermal cameras. Although the definition does not state it clearly, NETD is a metric of only high frequency temporal noise along the video line (Fig. 17). Low frequency temporal changes are corrected (Fig. 18) before the NETD measurement. NETD gives no information about the spatial noise between different video lines. Actually we get different NETD values for different video lines.

Fig. 18. Signal profile of a video line after low frequency trends correction

where V_n is rms value of noise in the signal line, SiTF is the Signal Transfer Function of the tested camera. V_n can be presented in different signal units: digital levels, volts, etc. However, as long as SiTF is expressed as a ratio of the same units divided by K, then NETD is calculated in Kelvin degrees.

As it was presented above, the situation with NETD parameter as a metric of a noise phenomenon is relatively clear in case of scanning thermal cameras. It is a measure of high frequency temporal noise in a single line.

Situation is much more complicated in case of array thermal cameras. Different definitions of NETD are often used. NETD is still defined as a blackbody temperature difference between a target and its background required to produce a peak-signal-to-rms-noise ratio of unity. A sequence of images of uniform target is typically used as raw data for NETD calculations. However, the difference depends on the way how to calculate this rms noise. NETD can be calculated using different methods.

First, rms noise can be calculated as a standard deviation of temporal variation of a signal of a single pixel. NETD is now a measure of only temporal noise.

Second, rms noise can be calculated as a standard deviation from temporal and spatial variations of signals from a group of pixels. NETD is now a measure of total noise (both temporal and spatial noise components).

Both two methods can be used with or without low frequency trends correction. This means that depending if we use the correction or not, then NETD can be a measure of only high frequency noise or full bandwidth noise.

To summarize, NETD can be treated as a useful metric of a noise phenomenon but only if it is precisely known how it was measured; without this knowledge NETD data can be misleading. The difference in NETD values, get using different calculation methods, is very significant. NETD measured using the first method can be a few times lower than NETD measured using the second method. This situation is particularly clear in a field non-cooled thermal cameras when the total noise is typically a few times higher than its temporal component.

1.3.1.3Four parameters approach

Noise phenomenon is too sophisticated to be precisely characterised by a single parameter like NETD. Even if it is clearly defined how NETD was determined we get too little information about the noise. Two cameras can have the same NETD (defined using any of different methods presented earlier) but a human eye will immediately notice big differences in the noise in images generated by these two cameras. On the other hand, as it was stated earlier, 3D noise model, using eight components for precise characterisation of a noise phenomenon, is too sophisticated to be commonly accepted and used. It seems that in this situation a middle way, when the noise is characterised by a set of four parameters, is an optimal solution.

Four parameters approach is based on the assumption that the noise present in images generated by thermal cameras can be generally divided into two groups: temporal noise and spatial noise. Next, each group can be further divided into low frequency noise and high frequency component.

Temporal noise generates temporal variation of intensity of camera pixels even when target radiation does not change in time.

Spatial noise generates spatial variations of intensity of camera pixels even in case of uniform target filling the camera field of view.

Low frequency temporal noise generates slow temporal variations of intensity of camera pixels. This noise component creates an effect called 1/f noise. It is noticeable if we capture and compare the images generated by the camera, separated by a relatively long period of time (say at least a dozen or more minutes). We can see in Fig. 20 that some groups of pixels in the second or the third frame are clearly darker or brighter in comparison to the first frame.

High frequency temporal noise generates fast temporal variations of the intensity of camera pixels. If we refer to original interpretation of NETD in older scanning thermal cameras then NETD can be considered as a measure of this high frequency temporal component of the total noise. In case when high frequency temporal component is significant then it is clearly noticeable if we capture and compare a few neighbour images generated by the tested camera. We can see in Fig. 21 that of the pixels intensity depends on a frame number, in spite of the fact that they are neighbour frames and a time interval between them is very short (1/60 s in NTSC video system; or 1/50 s in PAL video systems).

Fig. 21. Three neighbour images of uniform target generated by a thermal camera of dominant high frequency temporal noise.

Low frequency spatial noise generates slow spatial variations of the intensity of camera pixels. This noise component creates an effect that is called non-uniformity. In case when low frequency spatial noise component is significant, then it is clearly noticeable if we capture and compare a few neighbour images generated by the tested camera. We could notice, present in every frame, low frequency spatial trend that does not depend on a frame number and the frames are almost identical (Fig. 22)

Fig. 22. Three neighbour images of uniform target generated by a thermal camera of dominant low frequency spatial noise

High frequency spatial noise generates fast spatial variations of the intensity of camera pixels. This noise component creates an effect that is called the Fixed Pattern Noise. In case when the high frequency spatial noise component is dominant then it is clearly noticeable if we capture and compare a few neighbour images generated by the tested camera. We could notice the high frequency spatial trend, present in every frame, that does not depend on the frame number and the frames are almost identical (Fig. 23).

Fig. 23. Three neighbour images of uniform target generated by a thermal camera of dominant high frequency spatial noise.

It is usually considered that the frequency of 150 kHz for NTSC video or 186 kHz for PAL video systems is the border between the high and low frequency components. It is possible to separate low frequency temporal and spatial components from the total noise using suitable low pass filter in a video channel; or to separate high frequency temporal and spatial components from the total noise using suitable high pass filter in a video channel.

Another more convenient method to separate different noise components is to capture a series of images generated by the tested thermal camera when its field of view is uniform. The captured data created 3D spatio-temporal array. It is possible then to calculate the measures of all four noise components by carrying out the data filtering using low or high pass filters.

Scanning or array thermal cameras can be treated as ergodic or non-ergodic sources of noise. If the thermal camera is ergodic then the detectors are considered as statistically dependent noise sources. The same average rms will be measured if the average is calculated from n different detectors, or the same rms noise of the same detector is the measured n times. Then, rms noise can be calculated as

where

is the variance of noise from i-detector or from the same detector but the measured i time; n is the number of detectors or indicator how many times were recorded the data from the same detector.

If the thermal camera is non-ergodic then the detectors are considered as independent noise sources. Each detector is statistically a different noise source. Then rms noise can be calculated as

Scanning thermal cameras can be considered to some degree as ergodic systems. Array thermal cameras are usually non-ergodic systems. In order to simplify the analysis let us assume that the tested imager is non-ergodic thermal camera as we can expect that array thermal cameras will dominate the market sooner or later. Anyway, the consequences of non-fulfilling this assumption are not really significant because the differences between rms noise values calculated using two presented earlier formulas are usually quite small (below 12%).

Now, we will present the methods that can be used to measure all four noise components.

1/f noise component is noticeable only if we analyse temporal trends in frames of a long duration video sequence; say at least a few minutes. The 1/f effect is not noticeable in short video sequences. Therefore 1/f noise is typically omitted in a noise analysis. It is acceptable because its influence on final image quality, perceived by a human observer, is smaller than from other noise components. However, we should remember 1/f noise if not taken into account then it changes the FPN effect. This means that FPN measured at different time points will be different.

High frequency temporal noise expressed by NETD is usually a dominant noise component in cooled thermal imagers. High frequency spatial noise expressed as FPN is typically a dominant noise component in non-cooled array thermal imagers. Low frequency spatial noise (non-uniformity) is typically much lower in cooled thermal imagers than in non-cooled thermal imagers. Examples of possible measurement results are shown in Table 7. Please note, however, that these are only the examples and within each technology test results can vary significantly. Next, measurements are usually done immediately after the imager internal calibration and the NU results are lower than during real work.

Table 7. Examples of measurement results of noise components of different thermal imagers.

2.4 Image resolution parameters

Image resolution parameters carry out the information about camera ability to perceive small details of high contrast images. There is a lot of confusions in literature in this area because many parameters are used to expressed this ability. Generally, parameters that represent thermal imagers ability to perceive small details (resolution) can be in four groups as shown in

2.4.1 Parameters based on specifications of the FPA

Infrared focal plane arrays are hearts of thermal imagers. Therefore it is not strange that some FPA parameters like number of detectors in FPA, number of lines in FPA, - IFOV, and DAS are often used to describe performance of thermal imagers.

The first parameter is the total number of detectors of the FPA and it is used to describe resolution of array thermal cameras.

The second parameter is the total number of vertical lines of the FPA. This parameter is used to describe resolution of scanning thermal cameras.

IFOV (instantaneous field of view) or DAS (detector angular substance), in spite of different names, are practically defined as angular dimensions of a single detector of the FPA used in the thermal camera.

where a is the detector dimension and f’ is the focal length of the optics. Please note, however, that a detector horizontal dimension can differ from a detector vertical dimension and therefore there can be a difference between a horizontal resolution and a vertical resolution of thermal cameras.

It is very easy to get these simple data about the FPA and the optics used in a thermal imager. Therefore not only manufacturers but also most community involved in IR technology like to express imager resolution using parameter based on FPA (Table 8). However, the presented earlier resolution parameters based on FPA specifications can be very misleading.

Table 8. Specifications of exemplary FPAs and calculated resolution of thermal cameras.

There are the cases when a thermal camera of smaller detector number can generate much sharper image than a thermal camera built using FPA of higher detector number.

When DAS is equal to x mrad, it does not mean that we will be able to resolve targets of x mrad angular dimension. In some cases we may be able to resolve two times bigger targets but there can be the cases when we will not be able to resolve even targets of 3x mrad or more angular dimension (three times bigger than DAS).

Therefore the resolution parameters based on FPA specifications should be treated as indicators of imager theoretical ability to resolve small details. More detectors in FPA means that thermal cameras should theoretically resolve smaller details. However, practically it is not always true.

The same is with DAS or IFOV. Lower values of DAS (IFOV) are welcome but they do not always indicate improvements in thermal camera ability to resolve small details.

2.4.2MTF and derivative parameters

MTF (modulation transfer function) is a function of the contrast of image of a sine pattern at a given spatial frequency generated by the tested camera relative to a contrast of an image of sine pattern at spatial frequency equal to zero. Spatial frequency is typically measured per cycles or line pairs per a unit angle or a unit length (in case of thermal cameras typically in cycles per milliradians)^¹ (see Fig. 24).

Images of sine patterns of different spatial frequency generated by a thermal camera of MTF presented in Fig. 25 are shown in Table 9.

Table 9. Images of sine patterns generated by a thermal camera of assumed MTF function.

I_max is the maximum intensity for a pattern of spatial frequency  ("white peak") and

I_min is the minimum intensity for a pattern of spatial frequency  ("black valley").

where C( = 0) is the contrast of image of the sine pattern at near zero frequency.

MTF can also be defined as the magnitude of a complex function Optical Transfer Function:

OTF is Optical Transfer Function, MTF is Modulation Transfer Function, and PTF is a function called Phase Transfer Function that represent the change in phase position as a function of spatial frequency^².

A perfect optical system would have MTF of unity at all spatial frequencies, and PTF equal to 0 at all spatial frequencies. In case of real imaging systems, MTF always decreases to zero at some spatial frequency. However, in most imaging systems PTF is not significant and therefore we can assume that OTF equals MTF. Therefore MTF, not OTF, is usually used as a measure of quality of imaging systems.

Optical Transfer Function is the Fourier transform of the point or line spread function - the response of an imaging system to an infinitesimal point or line:

where F is the Fourier transform operator, OTF is the function called Optical Transfer Function, PSF is the function called Point Spread Function, and LSF is the function called Line Spread Function.

Point Spread Function is two directional distribution of the flux in the image of an ideal point-like source. Line Spread Function is one directional distribution of the flux in the image of an ideal line-like source. Because imaging systems are usually symmetric and also due to simplicity OTF is usually calculated as a Fourier transform from LSF. Therefore narrow slit target is often used during MTF measurement.

As it was mentioned above, MTF (modulation transfer function) is a function that provides information about the contrast of image of the sine pattern at a given spatial frequency relative to contrast at zero frequency on pattern spatial frequency. Contrast transfer function (CTF) is defined in the same way with an exception that a square wave pattern is used instead of a sine wave pattern. The CTF is usually higher than the MTF. Because the difference is usually not great and it is easier to measure CTF than to measure MTF then sometimes CTF is used instead of MTF to evaluate an imaging systems.

MTF curve is excellent criterion of quality of the images generated by thermal cameras. However, interpretation of a curve is more complicated than interpretation of simple numerical parameters. Therefore five parameters based on MTF are also commonly used to characterize thermal cameras:

Calculation of the equivalent frequency is based on criterion Shade [24], who stated that perceived image quality can be described using the formula

where N_eis the equivalent frequency (called also equivalent line number or equivalent bandwidth). N_e is presented using spatial frequency units.

Effective resolution is presented using an angle or length units (in case of thermal cameras typically milliradians).

Experiments have also shown that perceived image sharpness is closely related to the spatial frequency where MTF is 0.5. This means that spatial frequency at which MTF drops to 0.5 can be a good indicator of imager quality. We can call this frequency as the half frequency HF. It is expressed in spatial frequency units.

Limiting resolution is defined as spatial frequency at which MTF equals from about 0.02 to 0.05. The definition is based on the fact that humans usually cannot distinguish high contrast sine pattern at frequencies where MTF drops below the level 0.020.05. Exact value of the limiting MTF level depends on the observer. We should also note that all evaluation based on MTF is valid only in case of high contrast targets based on an uniform background.

As we can see, in order to determine the first two parameters, we must carry out the calculations of integral from a square MTF function. Determination of next three parameters is more simple and they are more commonly used in spite of the fact that the first two parameters are more closely related to perceived image sharpness.

2.4.3Parameters based on imager response to point/slit sources

There are three parameters of image resolution that are based on imager response to point/slit sources:

Ensquared power (point visibility factor PVF) is defined as normalized centre pixel signal caused by point source. It is calculated as ratio of the center pixel signal to the sum of signals generated by the point source in both the centre pixel and the neighbour pixels:

Both the measurement spatial resolution MSR and the imaging spatial resolution ISR are Slit Response Function related parameters. The measurement spatial resolution MSR is defined as angular slit dimension for which the Slit Response Function (SRF) of the tested thermal camera is equal to 0.99. The imaging spatial resolution ISR is defined as angular slit dimension for which the Slit Response Function (SRF) of the tested thermal camera is equal to 0.5.

The imaging spatial resolution (ISR) is typically found among the parameters of older scanning thermal cameras. ISR is a good measure of camera ability to create a thermal image of small targets. However, this parameter does not give information whether the size of the tested object is high enough to assure negligible influence of this size on measurement results during non-contact temperature measurement. The information is provided by the measurement spatial resolution (MSR). When angular size of the tested object is higher than the measurement spatial resolution, then we can assume that influence of the size of this object on the temperature measurement results is negligible. We can say in other words that if the angular size of the tested objects varies, but is always higher than MSR, then the output temperature will be the same. However, the MSR is usually a few times higher than ISR and manufacturers prefer typically to present only the values of the second parameter.

2.4.4Subjective parameters based on resolution targets

There are myriads of resolution patterns developed during last century to measure resolution of optical instruments and later to measure resolution of image intensifier systems or television cameras. The USAF 1951 target, the EIA Resolution Chart 1956, the NBA 1963A Resolution Target can be considered as the most popular targets from this group. These resolution patterns can be potentially also used to characterize resolution of thermal cameras. However, there are some technical problems that make difficult direct use of commercially available resolution targets.

Typical glass does not transmit in spectral bands of thermal imagers. Next, emissivity of printed patterns can be similar to emissivity of the mylar/paper. These reasons create a situation when thermal cameras do not see typical resolution targets and they cannot be used for testing thermal cameras. However, it is possible to manufacture the resolution targets mentioned above on special substrates and create high contrast pattern/substrate in a far infrared region. Then, the resolution targets can be useful tools to test resolution of thermal cameras.

2.5 Accuracy parameters

So far we have discussed testing and evaluation of surveillance thermal cameras. In case of this type of thermal cameras, image quality is the most important figure of merit. However, in case of measurement (commercial) thermal cameras high quality thermal image is useful but high measurement accuracy is more needed.

There are two types of errors of a temperature measurement with thermal cameras: the external errors and the intrinsic errors [3]. Here, we will present parameters of measurement thermal cameras that describe camera performance when only the intrinsic errors are present. Such a situation occurs when emissivity of the tested object is close to unity and a distance camera-object is short. Then, the external errors due to unknown emissivity, reflected radiation and limited atmospheric transmittance can be treated as negligible.

Manufactures of measurement thermal cameras often state a parameter called “accuracy” that is measured as a range around the true object temperature T_ob in which the output temperature T_out is located when the external sources of errors are negligible. Typical values of this parameters are: 1% of the output temperature T_out but not less than 1 C for scanning thermal cameras or 2% of the output temperature T_out but not less than 2 C for matrix thermal cameras. The term “accuracy”, according to the international metrological organizations, is only a qualitative concept that should not be associated with numbers [14]. Therefore the name “accuracy” is improper for formal terminology. However, there are more serious limitations of usefulness of this parameter.

The “accuracy” parameter could potentially enable determination of the intrinsic uncertainty of a measurement thermal camera. Assuming a uniform distribution dispersion of a true temperature within the limits, determined by the “accuracy” parameter, we can write [9]

However, practically the “accuracy” parameter is not useful for estimation of the intrinsic uncertainty of the thermal camera because the conditions in which the “accuracy” is measured are not clearly defined by manufactures. The question is whether the “accuracy” is measured at optimal calibration conditions when the measurement errors are the smallest or it is measured at real measurement conditions when the errors can be many times higher.

Manufacturers state very rarely at what environment temperature the “accuracy” was measured. However, typical practice is the “accuracy” is measured at laboratory conditions when ambient temperature is equal to about 23C.

During real measurements, the environment temperature can vary significantly within wide limits determined by the manufacturer from about 10 C to about 40C. Changes of the environment temperature can have significant effect on the measurement results due to several reasons. First, radiation emitted by the optical elements of the camera depends directly on temperature of these elements and indirectly on temperature of the environment. Second, variation of the environment temperature can cause variation of the detector temperature and change detector’s sensitivity. Third, changes of the environment temperature cause direct changes of the temperature of the electronic blocks and indirect changes of the gain and the offset of these blocks.

Influence of the environment temperature on the measurement results can be corrected. Modern thermal cameras are equipped with software and hardware that should automatically correct this influence. However, only a partial correction of this harmful influence is possible. Therefore accuracy of measurements carried out in real measurement conditions can differ significantly from accuracy obtained in laboratory conditions.

There are also other parameters presented in the catalogues of measurement thermal cameras that give some indications about intrinsic errors of thermal cameras.

One of them is a parameter called “thermal sensitivity”, “thermal resolution”, “temperature resolution” or “ NETD” that provides information about the influence of noise in electrical channel on the measurement errors. It was shown in Ref. 5 that NETD equals the standard deviation of the output temperature dispersion caused by noise of the system. Therefore, the NETD can be treated as a good estimation of uncertainties due to system noise.

However, we must remember that the NETD depends on an object temperature. It is typically measured only for one fixed value of this temperature usually close to 30C and can be a few times higher for object temperatures at the lower limit of available temperature range close to 20C.

Manufactures of measurement thermal cameras present also, in specifications of their products, other parameters like MRTD, MDTD, IFOV and imaging spatial resolution ISR.

If the thermal camera is only to be used for non-contact temperature measurement on surfaces of the tested objects, then MRTD and MDTD parameters are practically useless because it is impossible to connect these parameters with the measurement errors of the thermal camera. However, MRTD and MDTD can be useful figure of merits if the cameras are to be used in non-destructive thermal testing (SAARS testing that can be treated as a part of NDTT technology) image quality is as important as the accuracy of temperature indications [6,4].

The instantaneous field of view IFOV is typically found among the parameters of modern measurement thermal cameras. It is related to the minimum angular size of the tested object for which influence of the size of the tested object on measurement results is still negligible. However, this minimum size depends also on a parameter of other blocks of the thermal camera like aberration of the optical block, diffraction effects, frequency bandwidth of the electrical channel and it is not possible to determine this minimum size on the basis of the IFOV only.

The imaging spatial resolution ISR (it was defined in 1.4) is a good measure of camera ability to create thermal image of the tested object. However, this parameter does not give the information whether the size of the tested object is high enough to assure negligible influence of this size on measurement results. This information is provided by the measurement spatial resolution MSR However, MSR is usually a few times worse than ISR and the manufacturers prefer typically to publish only ISR.

A set of four parameters (the minimal error ME, the noise generated error NGE, the temperature stability TS, and the measurement spatial resolution MSR) is sufficient for characterization of measurement thermal cameras [3].

The minimum error ME was defined as a range around the output temperature T_out in which the true temperature T_ob is located when the measurements are carried out in the conditions identical with the conditions during calibration of the thermal camera. The calibration conditions exist when the tested object is a sufficiently large blackbody, the distance between the tested object and the thermal camera is short in order to have negligible influence of limited transmittance of the atmosphere, environment temperature is of typical laboratory range 2030C, the object is located in the center of the system field of view, measurements are carried out for the shortest temperature span of the thermal camera, and averaging the effect of a dozen or more of measurement results is used. Practically, the minimal error ME is an equivalent of the “accuracy” parameters presented in data sheets of measurement thermal cameras.

Fig. 31. Exemplary measurement error ME parameters of several thermal cameras (squares – camera 1, triangles – camera 2, circles– camera 3, plus signs – limits according to “accuracy” parameter.

The noise generated error NGE is defined as the standard deviation of the output temperature dispersion caused by the system noise. As it was shown in Ref. 5 , NGE equals NETD in case of typical thermal cameras (systems of single spectral band).

NGE theoretically decreases with an object temperature. Practically, as we can see in Fig. 32 it does not always occur due to different reasons. One of them are neutral filters used to extend camera temperature measurement range that cause significant suppression of the signal coming to the detectors and increase in NGE (NETD) value.

The temperature stability TS was defined as a range in which the results of measurements, carried out at different environment temperatures, within the limits determined by the camera’s manufacturer, are located.

Fig. 33. Exemplary temperature stability TS parameters of several thermal cameras (the errors of temperature measurement T of a blackbody of the temperature T_bb = 90C with a few thermal cameras at different temperatures of the environment T_en ).

The measurement spatial resolution MSR is defined as the minimum angular dimension of the tested object when there is still no influence of limited size of this object on temperature measurement results. The measurement spatial resolution MSR can be measured as the angular slit dimension when the slit response function SRF equals 0.99. This is a classical measurement method of MSR based on SRF. However, the measurement spatial resolution MSR can be also measured using a concept of the slit temperature response function STRF proposed in Ref. 3. Then, the measurement spatial resolution MSR can be measured as the angular slit dimension when the slit temperature response function STRF equals 0.99.

The STRF is a modified version of the well-known slit response function SRF. The STRF is based on the normalized output temperature, when the concept of the SRF is based on the normalized output signal. Measurements of STRF of the tested thermal cameras are carried out by determination of a normalized difference between maximal temperature of the slit image and the uniform background temperature.

Practically the difference between SRF and STRF is small; almost negligible and both measurement methods of the MSR generate almost the same measurement result. Therefore it is matter of convenience which method is used to determine MSR of the tested thermal camera.

As we can see in Fig. 34, the MSR of the first camera equals 3.5 mrad and the MSR of the second camera equals 10 mrad. This means that the first thermal camera can be used for accurate temperature measurement of the argets as small as 3.5 mrad, the second- as small as 10 mrad.

2.6 REFERENCES

1. ASTM standard E 1213-2002 “Standard Test Method for Minimum Resolvable Temperature Difference for Thermal Imaging Systems”

2. ASTM standard E 1311-99 “Standard Test Method for Minimum Detectable Temperature Difference for Thermal Imaging Systems”

3. Chrzanowski K., Evaluation of commercial thermal cameras in quality systems, Optical Engineering, Vol. 41, No. 10 (2002)

4. Chrzanowski K., Park S.N., Evaluation of Thermal Imagers For Non-Destructive Thermal Testing Applications, Infrared Physics and Technology, 42 (2) 101-105 (2001).

5. Chrzanowski K., Szulim M., A measure of influence of detector noise on temperature measurement accuracy with IR systems, Applied Optics, 37, 5051-5057 (1998).

6. Chrzanowski, J. Fischer, W. Wrona, Testing of Thermal Imagers For Non-Destructive Thermal Testing Applications, ASTM Journal of Testing and Evaluation, 28, 395-402 (2000).

7 Curtis M. Webb, Gerald C. Holst, Observer variables in minimum-resolvable temperature difference, Infrared Imaging Systems: Design, Analysis, Modeling, and Testing III, SPIE Vol. 1689, 1992.

8. D'Agostiono, C. Webb, 3-D Analysis Framework and measurement methodology for infrared systems noise, Infrared Imaging Systems: Design, Analysis, Modeling and Testing, SPIE Vol. 1448, 110-121, 1991.

9. Guide to the expression of uncertainty in measurement, International Organisation for Standarisation-International Electrotechnical Commission-International Organisation of Legal Metrology-International Bureau of Weights and Measures, TAG 4/WG 3, 1993.

10. Holst G.C., Infrared Imaging System Testing, Vol.4, Chapt. 4 in The Infrared & Electro-Optical Systems Handbook, Michael C. Dudzik ed, SPIE 1993..

11. Holst G.C., Testing and evaluation of infrared imaging systems, JCD Publishing Company 1993

12. Holst G.C., The Infrared & Electro-Optical Systems Handbook, Vol.3: Electro-Optical System Design, Analysis, and Testing, Chapt. 4, pp. 206-207, SPIE (1993).

13. International Lighting Vocabulary, CIE Publ. No. 1 7.4, IEC Publ. No. 50(845) (1987).

14. International Vocabulary of Basic and General Terms in Metrology, International Organisation for Standarisation, 1993.

15. ISO 15529, Principles of measurement of modulation transfer function (MTF) of sampled imaging systems, 1999

16. Lloyd J. M., The Infrared & Electro-Optical Systems Handbook, Vol.3: Electro-Optical System Design, Analysis, and Testing, Chapt. 1,SPIE (1993).

17 MIL-I-24698(SH), Infrared thermal imaging systems, Department of Defense USA (1988).

18 MIL-STB- 2194, Infrared thermal Imaging Survey Procedure for Electrical Equipment, Department of Defense USA, (1992).

21 Miller Scott J., Backer Brian S., Kohin Margaret, Alonso Pascual, Whitwam Jason T. , Test methods and technology for uncooled imaging systems, Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XV,SPIE Vol. 5407, 30-37, 2004.

22. Ronald G. Driggers, Van A. Hodgkin, Richard H. Vollmerhausen, Patrick O'Shea , Minimum resolvable temperature difference measurements on undersampled imagers, Proc. SPIE Vol. 5076 Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XIV; 2003.

23 Ronald J. Pieper, Alfred W. Cooper, Mustafa Celik, Yucel Kenter, Objective MRTD experimental measurements compared with predictions based on the visibility model Publication, Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XIV, SPIE Vol. 5076, p. 196-207, 2003.

24. Schade OH (1948). Electro-optical parameters of television systems. 1. parameters of vision and visual systems. RCA Review, 9: 5-37.

25 Sousk Stephen, O'Shea Patrick, Van Hodgkin, Measurement of uncooled thermal imager noise, Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XVI, SPIE Vol. 5784, 301-308, 2005

26. STANAG 4349, Measurement of minimum resolvable thermal difference (MRTD) of thermal cameras, 1995

27 Stephen F. Sousk, Patrick D. O'Shea, Van A. Hodgkin, Uncertainties in the minimum resolvable temperature difference measurement, Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XV, SPIE Vol. 5407, p. 1-7, 2004.

28. Wood, Laboratory Bench Analysis of Thermal Imaging Systems, Opt. Eng., 15, G193-G197 (1976).

1 Attention: TV resolution is measured in line widths instead of pairs, where there are two line widths per pair, over the total height of the display

2 If case of a linear PTF, only simple lateral displacement of the image is observed. Non-linear PTF can adversely affect image quality. An extreme case is a phase shift of 180 degrees that produces a reversal of image contrast.

3 In the presented above exemplary case we get EP=0.4.

3 Test equipment

Some information about structure and requirements on test equipment for testing thermal imagers can be found in international standards that regulate testing thermal imagers [1,2,8] but value of this information is rather limited due to small size of these standards. More valuable information is presented in some literature sources like Ref. [6,7]. However, the most valuable information about test equipment can be get from analysis of real test systems offered by manufacturers of equipment for testing thermal imagers. Here in this chapter conclusions from such an analysis are presented. Due to limitations in information access most detail technical information refer directly to test equipment from a single manufacturer [41]. However, there are no basic differences in test systems from different manufacturers; only minor technical details. All properly working test systems independent on the manufacturer generate the same test results. Therefore the information about equipment for testing thermal imagers should be generally valid for all test systems available on international market.

3.1 Types of test systems

The task of the test system for testing thermal imagers is to generate images of some standard static targets of precisely known shape, dimensions and temperature. These images can be projected to the tested thermal camera by the test system or viewed directly by the tested camera. In both cases the tested camera generates distorted images of the original targets. Next, these images generated by the tested imagers are evaluated and important characteristics of the tested imagers are determined.

Fig. 1. Images of standard targets used during testing thermal imagers a)4-bar target, b)pinhole target, c)edge target d)slit target

Technically we can formulate four basic requirements on a test system for testing thermal imagers:

The requirements mentioned above can be fulfilled by several test systems using three different test principle:

The variable target measuring systems project images of targets fixed to a rotary wheel using an IR collimator as an image projector. The tested thermal camera is located at the output of IR collimator and the target at the collimator input (the focal plane). The distance between the target and tested camera is very short and usually not within focusing range of typical surveillance thermal cameras. However, due to use of IR collimator the camera “sees” the target as a very long distance object that is within its focusing range. A series of targets is fixed to rotary wheel. By rotating the wheel it is possible to exchange quickly targets. By changing target dimensions changes of distance are simulated.

The variable distance systems generate a thermal image at plane of the target fixed to a large area blackbody. The tested thermal imager sees this image directly and generate a distorted copy of the image generated by the test system. Because the variable distance systems do not use an IR collimator during tests then the distance between the target of the test system and the tested imager must be long enough to have the measuring system within the focusing range of the tested camera. This means that the testing system must be always located at distances longer than minimal focus distance of the tested imager. In case of surveillance thermal cameras the latter value vary from about 5m to about 50 m. This means that the tests cannot be carried in small rooms but should be carried out in the field or at indoor conditions using long corridors.

The angular size of the thermal images generated by the test system must be big enough to enables measurements of parameters of thermal imagers (in case of MRTD measurements should be done at both low and high spatial frequencies). Practically this means that due to of long distance between tested imager and the test system large area blackbodies and large IR targets are needed for variable distance test system. Therefore both blackbodies and IR targets used for variable distance test systems are several times bigger than needed for typical variable target test systems.

The variable target/distance test systems are practically the variable target test systems without an IR collimator to project a thermal image generated by a set: target/blackbody. Because there is no collimator and the generated thermal image must be seen directly by the tested thermal imager then the distance tested imager - test system must be bigger than minimal focus distance of the tested imager. At the same time typical commercially available rotary wheels are adapted to use target of small size (typical about 50 mm, max - about 100 mm). Due to requirement that the angular size of the target cannot be too small in order to enable extended testing of thermal imager the distance between tested imager and the test system cannot be too big. The value of the maximal distance depends on type of tested imager or parameters to be measured but usually when 50 mm diameter targets are used then the distance cannot be bigger than about 5 m. This creates situation when only thermal imagers of short minimal focus distance can be tested using the variable target/distance test systems. Commercial thermal imagers designed for observation of short distance target can always be tested using the discussed group of test systems. Most surveillance (military) imagers design for observation of long distance targets should not be tested using variable target/distance test systems.

All three types of earlier presented types of test systems posses some advantages and disadvantages.

The variable target measuring systems (or variable target projectors) of configuration shown in Fig. 2 can be considered as classical systems for testing thermal cameras. Due to high thermal inertia, use of baffles in IR collimator, motorized rotary wheels, and the fact that they are mostly used in laboratory conditions they are characterized by very good stability and measurement accuracy.

Any thermal imager is able to focus on optical infinity distance simulated by the IR collimator. Therefore all types of thermal cameras (surveillance or commercial, short distance of long distance) can be tested using the such test systems (on condition that the aperture of the imager optics is smaller then output aperture of the collimator). Big size of these systems (mostly due to the use a collimator) is their main drawback. Even in case of smaller collimators it is difficult to use such a test system outside laboratory due transport problems and necessity to align tested imager with the collimator output.

The variable distance measuring systems of configuration shown in Fig. 4 represent test systems of much more compact design than in case of variable target systems. Infrared collimators and rotary wheels are eliminated. Because of by small size and mass these systems are excellent measuring tools at field/depot applications when number of characteristics to be measured is limited. They can be packed in a large suitcase and easily transported to any location. Therefore, such test systems can enable comparison of quality of different thermal cameras offered by different manufacturers. Next, it is possible to test thermal cameras from some distance, and as a consequence, thermal camera can be tested without removing it from the carrier (tank or helicopter). Further on, several thermal imagers can be tested at the same time using a variable distance test system. Another advantage of variable distance systems is the fact that there is also no limitation on aperture of the tested cameras (in case of the previously discussed group (image projectors) the aperture of the tested thermal camera must be at least about 10% smaller than aperture of the IR collimator).

There are also several disadvantages of variable distance test systems. First, large blackbodies and larger IR targets and needed. This makes speed of regulation of blackbody temperature lower than in case of smaller blackbodies used by variable target test systems. There also technical problems to assure required high thermal uniformity and temporal stability of blackbodies. Second, due to possible use outside laboratory they are vulnerable to variation of environmental conditions. Accurate testing of modern thermal cameras often requires keeping stable temperature difference at level of several mK. It is extremely difficult to achieve such high temperature stability in case of these systems in non-stable environmental conditions. Therefore only blackbodies of extremely high thermal stability and thermal uniformity can be used in this type of test systems. Third, applications as a mobile measuring tool out side laboratory conditions can decrease reliability of variable distance test system.

The variable target/distance test systems represent a mixture of two earlier presented test concepts.

Their main advantage is lower costs and lower mass due to elimination of the collimator in comparison to variable target test systems. The disadvantage is the fact that the systems can be used only for testing thermal cameras of very short minimal focus range like typical commercial (measurement) thermal cameras.

Summary comparison of the presented three groups of test systems for testing thermal imagers is presented in Tab. 1.

First thermal imagers were designed for military application to enable observation of long distance targets. Minimal focus distance of these imagers was often more than 100 m. Therefore testing of such imagers was possible only using variable target test systems that used an IR collimator as an image projectors. Now, due to presence on the market of large number of thermal imagers designed also for short distance surveillance or non contact temperature measurement having much shorter minimal focus distance it is often possible to use other two types of test systems for testing modern thermal imagers although classical variable target test system are always acceptable choice. Anyway, if properly used all types of systems for testing thermal imagers should generate the same measurement results.

1.2 Blocks of test system

A set of the standard targets (metal sheets with holes) is fixed to the rotary wheel placed at the focal length of the IR collimator. One of the targets is within the field of view of the IR collimator (we can call it the active target). The differential blackbody is close behind this target. The temperature distribution on the target surface and the blackbody is projected by the IR collimator to the tested thermal camera and image at camera screen is created. Next, the image is evaluated by an observer or the image captured and analysed with help of specialized hardware/software module. In case of variable target/variable distance test systems the collimator is replaced by the shield box but the principle of test system work is practically the same.

Large area target is fixed at the bottom of the shield box. Large area blackbody is placed behind the target. The thermal camera looking on the target can see temperature distribution on the surface of the target and through target holes can see temperature distribution on the surface of the blackbody. The image generated by the tested thermal camera is evaluated by an observer or the image captured and analyzed with help of specialized hardware/software module.

The modules of variable target/distance test systems are practically the same as of modules of variable target test systems. Therefore we will not discuss here the variable target/distance test systems.

Now let us discuss principles of work and requirements on different blocks of test systems that should be fulfilled in order to get proper measurement results.

1.2.1 IR collimator

IR collimators are typical elements of laboratory set-ups used for testing thermal imaging systems. The function of the IR collimator is to generate a thermal image closely resembling the thermal scene at the target plate.

1.2.1.1 Collimator structure

Reflective two mirror collimators built using an off-axis parabolic collimating mirror and smaller directional flat mirror represent a typical design of the collimator to be used as a component of test systems for testing thermal imagers. Reflective collimators like the collimator of diagram shown in Fig. 9 dominate the market because of a few reasons:

Mirrors used in IR collimators are almost always front-surface mirrors of configuration shown in Fig. 10 and back-surface mirrors are excluded from discussion. As we can see this type of mirrors consists of three basic elements: substrate, reflective film and protective layer.

Mirrors surfaces can be distorted or damaged when subjected to wide temperature changes. Therefore low-thermal expansion substrate materials are critical to successful performance of imaging reflective optical systems. Surface accuracy and cosmetic quality of substrate material is important, too.

There are four materials that are most often used in mirrors fabrication: optical crown glass, low-expansion borosilicate glass (LEBG), synthetic fused silica and Zerodur as they fulfil relatively well earlier specified requirements [9].

Optical crown glass is an old and low cost material for mirrors. Crown glass has a relatively high coefficient of thermal expansion and is employed when thermal stability is not a critical factor.

Low-expansion borosilicate glass (LEBG) known by the Corning brand name - Pyrex - is well suited for high quality front-surface mirrors designed for low optical deformation under thermal shock.

Synthetic fused silica has a very low coefficient of thermal expansion. Fused silica mirrors can be polished to extreme accuracies, thereby minimizing wavefront distortion and scattering.

Zerodur is a unique glass-ceramic material whose thermal expansion is almost zero. This stability is essential in diffraction limited systems where the optical figure must not vary under thermal changes.

Parameters of the described earlier materials used for mirrors substrates are shown in Tab. 2. As we can see the best material for the substrate of mirrors seems to be Zerodur. However, it is also the most expensive material of the analyzed four materials. Therefore, Zerodur is used only sometimes when the collimator is be used at field conditions. In most cases mirrors in IR collimators are manufactured from Pyrex or fused silica; rather rarely from crown glass. However, we must state that if the collimator is to be used in laboratory conditions when temperature is stable and mirror is manufactured to get accuracy about /4 then all these materials are acceptable.

Reflective coatings used in manufacturing front-surface mirrors can be divided into two groups: metallic coatings and dielectric coatings. Metallic coatings are typically used as reflective coatings of front-surface mirrors. They are relatively insensitive to wavelength, angle, or polarization of incident light and mirrors using them represent a good mixture of performance and economy. There are six types of most often used types of metallic coatings: bare aluminum, protected aluminum, enhanced aluminum, UV-enhanced aluminum, silver and gold.

Protected aluminum coating is a bare aluminum coating with a dielectric overcoat that arrest the oxidation process. Silicon monoxide is commonly used for the dielectric overcoat. A layer with physical thickness of about 0.125m slightly increases reflectance and protects the surface against corrosion and abrasion. The overcoat usually allows the surface to be gently cleaned to remove dust or air-carried contaminant.

Enhanced aluminum coating is an improved version of the protected aluminum coating described above, offering higher reflectance in mid-visible regions. It is harder than the protected aluminum and offers good abrasion resistance.

UV-enhanced aluminum coating has a dielectric over-coating which prevents oxidation while preserving aluminum reflectance in the ultraviolet region. It is not as abrasion -resistant as the enhanced aluminum coatings. A film of magnesium fluoride is a common material for UV-enhanced aluminum coating overcoat.

Silver offers better reflectance in the near infrared than aluminum. and high reflectance across a broad spectrum. However, unprotected silver tarnishes rapidly. Therefore, overcoat of one or more dielectric layers is mandatory. Silver coatings reflectance is more subject to greater environmental degradation than aluminum and gold coatings are.

Gold offers consistently very high reflectance from about 0.8m to about 50 m. However, gold is so soft that protective layer is always needed.

1.2.1.2 Requirements on IR collimators

The task of the IR collimator is to project with negligible distortion an image of temperature distribution in the focal plane into direction of the tested thermal imager. The condition on negligible distortion can be fulfilled only when certain requirements on four parameters of IR collimators:

According to a popular myth the function of the collimator is to generate a parallel ray beam in direction of the tested imager. Practically, the collimator does not generate a single parallel ray beam; it generates an infinite number of parallel ray beams in different directions and the true function of the collimator is to generate a thermal image closely resembling the thermal scene at the test plate. In its ultimate form, an ideal IR collimator would be capable of generating a radiation pattern that exactly reproduces the real image. However, such quality is unattainable. Instead, a practical design condition should be adopted, based on the requirement that the collimator spatial resolution should match the spatial resolution capabilities of the tested thermal imager. We should remember that collimators of too low quality can become a source of significant measurement errors; collimators of too high quality can unnecessarily increase cost of the test system.

Manufacturers of IR collimators use different ways to characterize performance of these optical instruments and there is a confusion in area of evaluation methods.

Accuracy of manufacturing of the collimating mirror is typically presented as a collimator parameter [14]. However, manufacturing accuracy of the mirrors is not very useful if we really need to evaluate quality of the collimator we want to use in testing thermal imagers.

First, perfect mirrors do not necessary means that the collimator is perfect. Very precise alignment of these two collimator mirrors is required to obtain the maximal theoretically possible performance. Next, precision, zero thermal-expansion optical and mechanical elements must be used in collimator design. Practically this means that information about mirror accuracy gives precision information about mirrors performance but not about overall collimator performance. Practically, increasing accuracy of the collimating mirror do not always increases the collimator performance but always increases the costs of the collimator.

The most typical situation is that manufacturers claim that the collimator is diffraction limited [13,10,11]. However, let us look on the diffraction limited target frequency values for typical collimators presented by one of the manufacturers [12] that are shown in Tab. 4.

Tab. 4. Diffraction limited target frequency values (in cycles/mrad) for collimators of different optical apertures

As we see in Tab. 4 values of diffraction limited target frequencies are low; actually very low. The Tab. 4 suggests that resolution of typical collimators (aperture below 250 mm) during tests of LW thermal imagers is below 5 mrad^-1 due to diffraction limit of the collimator. This means that using typical off axis reflective collimators for projecting images of targets of frequencies over 5 mrad^-1 we should always get blurred images of these targets generated by the tested LW thermal imagers even if the thermal camera is perfect because the collimator is the limiting factor. It is not true as the author of this book tested some long wavelength thermal imagers that generated sharp images of targets of frequency over 10 mrad^-1; clearly over the suggested diffraction limit of the collimators.

The situation described above is possible because the formula that was used by the manufacturer [12] to calculate the values of limited target frequency _max is too pessimistic because is based on wrong assumption.

The manufacturer [12] used the following formula to calculate the values of limited target frequency _max:

where: D is collimator aperture, and  is wavelength. The formula is based on assumption that resolution of the collimator in real test system is limited by diffraction effect what is not true. Aperture of the tested imagers is always smaller than the aperture of the collimator. This means that quality of the image generated by the optics of the tested thermal camera is degraded by both aberration blur and the diffraction blur but quality of the image projected by the collimator is degraded only by its aberration blur. Therefore resolution of the collimators used in real test systems is limited only by aberration effects.

To summarize, diffraction blur is should not be used as criterion of collimators quality. It is a misleading parameter. Next, both mirror accuracy or aberration blur of the main off-axis mirror can be treated as indicators of possible collimator quality. However they are not the parameters that would give warranty about quality of collimator at user hands. The IR collimators should be characterized using a parameter called spatial resolution that depends on aberration blur and this parameter should be measured at final user facilities.

Such a precise condition on quality of IR collimator was proposed in Ref. [4]. The collimator spatial resolution _colwas defined as the frequency of the smallest bar pattern projected by the collimator that the observer is able to recognize. The resolution of the tested imager was defined as the Nyquist frequency _N that determines thermal imager theoretical limit.

It was shown in Ref. [4] that to have collimator influence on degradation of image generated by tested thermal imager negligible then the collimator resolution _col must be at least 5 times better than thermal imager resolution _N

The spatial resolution _N defined as Nyquist frequency of the thermal imager can be calculated using data provided by the manufacturers of thermal cameras as

where N is the number of pixels in horizontal (or vertical) direction of FPA used in imager design, FOV is imager field of view in horizontal (or vertical) direction.

The spatial resolution of the collimator _col cannot be calculated but it can be measured. It can be done using an measurement method that was proposed in Ref. [4]. The method proposes to carry out measurement of collimator spatial resolution in visible spectral range because aberration of typical reflective IR collimators do not change with spectral range.

Experiments carried out by the author of this book with infrared collimators used in different test systems¹ on the market showed that spatial resolution of these collimators can vary significantly from about 30 mrad^-1 to over 70 mrad^-1. The reasons for this significant dispersion of resolution of infrared collimators can be different: ageing processes, manufacturing errors of the mirrors, alignment errors, deterioration of coating properties, dust on mirror surfaces etc. Now, let us check if such collimators can be used as important blocks of measuring systems for testing thermal imagers. We can easily calculate the minimal acceptable collimator resolution using the earlier derived formulas and basic data of a few thermal imagers available on the web.

The calculation results are shown in Tab. 5. We can make two basic conclusions from the data presented in this table.

First, the requirements on spatial resolution of IR collimators significantly depends on field of view of the tested thermal imagers . The requirements are very low in case of imagers working in wide field of view mode but are many times higher in case of the same imagers working in narrow field of view mode.

Second, in case of testing short/medium range thermal imagers the collimators of resolution at 25 mrad^-1can be considered as acceptable. Collimators of resolution about 50 mrad^-1enable to carry out tests of all short/medium range thermal imagers and majority of long range thermal imagers. Please note that such collimators are acceptable even in case of LW IR imagers of very narrow FOV built using very large optics (240 mm) like Thermovision 2000 (FLIR Inc). IR collimators of spatial resolution higher than 70 mrad^-1 are recommended for testing long range imagers of very narrow field of view designed using very large optics and 640x480 resolution FPA. However, it should be noted that resolution over 70 mrad^-1can be practically expected only in case of some new collimators. After some time resolution typically deteriorate to level about 50mrad^-1 due to unavoidable dust on mirrors and small disalignent of the optical system of the collimator.

Tab. 5. Requirements on spatial resolution _colof IR collimators to be used in testing different thermal imagers

Collimator aperture is the diameter of the maximal ray beam that can be generated by the collimator when the point source is used. It is strictly needed for proper testing that collimator aperture must be bigger than diameter of the optics of the tested thermal camera. It is recommended that the collimator aperture should be at least 10% bigger than the diameter of the optics the tested thermal camera. If this this condition is not fulfilled then additional errors of measurement of parameters of thermal cameras are generated.

Diameters of optics used in modern thermal cameras varies greatly: from 10-30 mm diameter objectives typically used in commercial thermal cameras up to 200-240 mm diameter objectives in some long range surveillance thermal cameras. Therefore only big collimators of apertures over 250 mm can be used for testing all thermal cameras available on the market. However, bigger collimator means also more expensive, more bulky instrument. At the same time we should remember thermal cameras of optics diameter over 130 mm are very rarely met. We can expect also some problems testing thermal cameras of wide field of view with small (like commercial thermal cameras) using big collimator of long focal length because simply even the biggest targets will be quite small for the tested camera.

Therefore it seems that collimators of aperture about 150-200 mm represent an optimal choice. Such collimators enable testing almost all thermal cameras available on the market. In case when test area is limited to cameras of small aperture then smaller aperture collimators can be used.

F-number (ratio of focal length to aperture) of collimating mirrors used in IR collimators vary from about 5 (low F-number collimators) to about 10 (high F-number collimators). This means that focal length of IR collimators of the same aperture can vary significantly. For example in case of 150 mm aperture the focal length can vary from 750 mm to 1500mm.

There advantages and disadvantages of both types of IR collimators. Low F-number collimators are characterized by small size that enables to decrease dimensions of the complete test system. However, the cost of manufacturing low F-number mirrors is higher than in case of high F-number mirrors. Next, thermal stability of low F-number collimators is lower than in case of high F-number collimators. However, if large size high F-number collimators of long focal length are acceptable then from the point of view of the user the F-number or focal length of the collimator are not important. There is no relationship between image quality and focal length or F-number. Both low F-number collimators and high F-number collimators can project high quality images and the test results get using these collimators will be the same.

IR collimator must projects thermal radiation emitted by the blackbody and targets located in its focal plane at least within the spectral range of the tested thermal imager. This means that spectral range of IR collimator must cover two spectral bands used in thermal imaging: MW band (3-5µm) and LW band (8-14µm). Therefore we can conclude that for testing thermal imagers we need collimators of spectral range that cover at least the region from 3µm to 15µm. If the collimator is to be used also for testing visible imaging systems then the collimator is required to project radiation in the spectral range from 0.4µm to 15 µm.

Spectral range of the reflective collimators is determined by coatings of the mirrors. Metallic coatings are typically used as reflective coatings in IR mirrors. There are three types of most often used metallic coatings: aluminum, silver and gold. All three types offer high reflectivity over about 95% in the spectral range of interest: 3-15 m. As it was discussed earlier all mentioned above coatings needs some kind of dielectric overcoat that arrest the oxidation process or improve its mechanical properties.

Gold offers consistently very high reflectance (about 99%) from about 0.8m to about 50 m. . Silver offers slightly lower reflectance (about 96%) but broader spectrum from 0.3m to over 20 m. Aluminium coatings are characterized by lower average reflectivity (about 95%) and a certain reflectivity drop in near infrared. From the other point of view the aluminium coatings are characterized by the best durability and the lowest costs. Additionally reflectance of aluminium coatings increase with wavelength. Practically there is only a slight difference in 3-15 m spectral region between aluminium mirrors or gold mirrors but only in case of collimating mirrors where mirror surface is nearly perpendicular to the incoming beam. Silver/gold coated flat mirrors working at about 45 deg angle are characterized by much better reflectance than their aluminium equivalents.

To summarize, several guidelines on coating of mirrors for IR collimators can be formulated.

3.2.2 Shield box

The task of shield boxes used in variable distance test systems is to reduce influence of fluctuation of air temperature due to wind, ventilation, air turbulences etc. on temperature on the surface of the blackbody and on temperature on the target surface. Without a proper shield box real temperature on the surface of the blackbody radiator can be different than measured temperature inside the radiator. This can lead to significant measurement errors during testing thermal imagers.

Design of shield boxes is simple. The boxes are typically manufactured in form of an empty metal cubicoid with a few internal baffles. The surface of the walls inside the cubicoid are covered with high emissivity (low reflectivity) coatings to eliminate reflection of radiation coming into the shield box.

3.2.3 Targets

Targets for testing classical visible/near infrared imaging systems are manufactured by creating opaque or semi-transparent coatings on transparent glass substrate. When a diffuse light source is put behind the target then the tested visible range imaging system see a "target" formed by the coating on an uniform bright background (in case of positive contrast targets). The visible targets can be also manufactured by precise printing of images of different shape on high quality paper. These two techniques are rarely used to manufacture targets for testing thermal imagers because it is difficult to determine and to control temperature distribution on the surface of such targets. Targets for testing thermal imagers are usually manufactured by creating precision holes of different shapes in metal sheets. When a blackbody is put behind such a target, the tested thermal camera sees a "target" of shape determined by the holes on an uniform background. The apparent temperature of this "target" is equal to blackbody temperature; the apparent temperature of the background is equal to the temperature of the real target plate.

The targets manufactured using this technology are called “emissive targets”. There exist also so called “reflective targets” that are manufactured by creating a pattern that differ by its reflectivity from its background. However, the reflective targets are rarely used and will not be discussed here.

Fig. 12. Image of a 4-bar target (generated by a tested imager during MRTD measurement)

Fig. 13. Photo of a multiply 4-bar target used for MRTD measurement from two sides (courtesy of Inframet www.inframet.com)

In order to achieve high thermal uniformity on the surface of the target, the targets are manufactured by cutting the desired holes in a thick metal sheets of high thermal conductivity (typically copper alloys). Next, one side of the targets is coated using a special high emissivity black coating. This side of the targets should look into the direction of the IR collimator (or in the direction of the tested thermal camera). The second side of the target is coated using a high reflectivity coating in order to minimize influence of the blackbody thermal radiation on the target plate temperature. The targets are typically fixed to the rotary wheel or directly to the blackbody.

1.2.3.1 Target types

Different targets are manufactured to enable measurement of different parameters of thermal imagers. We can distinguish at least fourteen types of IR targets:

1.2.3.2 Requirements on targets

A process of cutting holes in metal sheets necessary to manufacture IR targets looks simple. However, practically it is quite difficult to manufacture proper IR targets that fulfil presented below requirements.

It is typically required that thermal target uniformity cannot be worse than 0.01C when the blackbody of temperature difference equal to 5C is located close to the target. The uniformity of the temperature distribution on the target plates should be a few times better than MRTD or MDTD values obtained with these targets. In order to fulfil requirements no 1 the targets should be manufactured using material of high thermal conductivity. Copper or copper alloys are acceptable but steel sheets should be avoided. Next, the metal sheets cannot be to thin as thin sheets are characterised by low thermal conductance even if manufactured from proper material. It seems that 0.3-0.5 mm can be considered as minimal thickness of metal sheets for target manufacturing even in case of copper sheets. Next, the area where the target is think? should be kept as low as possible.

As general guidelines should be considered following tolerances: 2% for patterns of minimal dimension over 1 mm and 4 % for patterns below 1 mm but over 0.3 mm, and 8% for pattern below 0.3 mm.

It should be required that target emissivity should be at least 0.97 to resemble ideal blackbody surface.

It should be required that the target should be polished or coated to get of reflectance at least 0.9 in order to eliminate influence of blackbody temperature on target temperature.

3.2.4 Rotary wheel

The task of rotary wheels is to enable speedy exchange of the target to be projected by the collimator or to be observed directly by the tested thermal imager. This task can be also done by horizontal or vertical sliders with different targets but due to bigger simplicity rotary wheels are used typically to exchange the targets in the test systems.

Two basic types of rotary wheels are available on the market: manual rotary wheels or motorized rotary wheels. A touch of human hand can change slightly temperature of the wheel. In situation when modern thermal imagers are becoming extremely sensitive any such temperature variation can cause some measurement errors. Therefore motorized rotary wheels are recommended when testing high sensitivity thermal imagers.

Next, air flow can cause some variations of target temperature. This effect can be reduced if the wheel with targets is put inside a closed enclosure that eliminates exchange of air inside/outside the wheel block.

Number of hole for target can vary. However, there are usually from 6 to 12 holes for targets in the wheel. Wheels are usually covered with a black, high emissivity coating. The requirements on coating are similar like in case of targets.

Fig. 15. Photo of a motorized rotary wheel and differential blackbody (courtesy of Inframet www.inframet.com)

During tests of thermal imagers we need to know temperature of the blackbody radiator and temperature of the target. Temperature of the blackbody radiator is typically measured using a small temperature sensor inserted to a hole in the radiator. This direct contact measurement method cannot be used to measure temperature of the targets because they are rotating. One solution to solve the problem is to use a sliding temperature sensor that slightly touches the active target and measure its temperature. However the problem is that direct contact of the sensor with the target when the latter is moving generates some heat and create some measurement errors. Another solution is measure target temperature indirectly. The sensor is attached to the rotary wheel and measures temperature of the rotary wheel. If the wheel is designed properly and there is sufficient thermal contact between the targets and the rotary wheel case then difference between target temperature and wheel temperature is negligible and the indirect measurement method can be used.

The target plates and the wheel should be put inside an enclosure that would reduce changes in air flow close to the target plates and prevent unnecessary radiation from reaching the wheel. The enclosure should have good thermal contact with the IR collimator or with the base of the blackbody.

3.2.5 Blackbody

Blackbody is an ideal body that completely absorbs all radiant energy striking it and, therefore, appears perfectly black at all wavelengths. The radiation emitted by such a body when heated is referred to as blackbody radiation.

A perfect blackbody has an emissivity equal to unity. Emissivity of real technical blackbodies is close to unity. There are generally two methods to design technical blackbodies of emissivity nearly . equal to unity.

First method is to manufacture a cavity in a block of material of high thermal conductivity (mostly metal alloys) and regulate temperature of this block using a heating element. Emissivity of cavity blackbodies is typically over 0.995 even when emissivity of cavity walls is much lower (about 0.5-0.8). Most medium, or high temperature blackbodies are designed using cavity concept. Typical commercially available cavity blackbodies are characterised by relatively small emitting aperture about 1 inch aperture and quite high thermal inertia.

Second method is to put layer of high emissivity material over a flat element of regulated temperature. Some special black (or gray) paints of emissivity eaqual to 0.97 are put over emitters manufactured from high conductivity metals (usually copper). It is possible using the second method to achieve bigger emitting apertures then using the first method. The balckbodies designed using the second methods are called area blackbodies. The area blackbodies are also characterised by lower thermal inertia than the cavity blackbodies. However, due to lack of high emissivity coatings resistible to high temperatures the area blackbodies can be used only at low temperatures (typically the maximal temperature is no more than 400°C).

The emitting area of the blackbodies used in systems for testing thermal imagers must be bigger than the target plates. This creates requirement that emitting area must be typically at least equal to about 40-50 mm. Next, blackbody for the test systems should simulate objects of rather low temperature to simulate typical obsevation condition. Further on, both positive and negative temperature differences (between the target and the blackbody) are needed during testing thermal imagers. Because of these requirements thermoelectrical area blackbodies dominate among blackbodies used in systems for testing thermal imagers.

Themoelectrical area blackbody is an blackbody that uses Peltier element for temperature control of the emitting element. Peltier element (Peltier module) is a semiconductor-based electronic component that functions as a small heat pump. By applying a low voltage DC power source to a Peltier element (called also often TEC(thermoelectrical cooler because of its ability to cool) heat will be moved through the element from one side to the other. One face of the element will be cooled while the opposite face simultaneously is heated. Consequently, a thermoelectric element may be used for both heating and cooling by reversing the polarity (changing the direction of the applied current). This ability makes TECs highly suitable for precise temperature control applications as well as where space limitations and reliability are paramount. Due to ability to heat or cool relative to ambient temperature the thermoelectric blackbodies are usually called differential blackbodies.

There are a few important advantages of differential blackbodies that made this type of blackbodies an ideal choice for testing modern thermal imagers.

First, standard temperature range of differential blackbodies from about -25C to about +75C fits perfectly to temperature range needed during tests of thermal imagers. Second, it is possible by careful design to develop differential blackbodies of excellent temperature resolution, stability and uniformity. Third, these blackbodies are characterized by low mass and low inertia in contrast to cavity blackbodies (liquid based blackbodies or pipeline blackbodies) and enable speedy measurements of different characteristics of thermal imagers. All these features created situation when differential thermoelectric blackbodies are practically the only type of blackbodies used by manufacturers of professional test systems for testing surveillance thermal imagers. Please note however, that the cavity blackbodies are frequently used in testing accuracy of commercial thermal imagers to be used for non contact temperature measurements.

1.2.5.1 Design of differential blackbodies

Differential blackbodies to be used as components of test systems are offered by different manufacturers [38,39,41,42,43]. The blackbodies are typically built from two basic blocks: the radiator block and the controller block Now we will discuss in details design of a differential blackbody offered by one of manufacturers [41] but the conclusions will be generally valid for all high performance differential blackbodies present on the market.

The TCB Series blackbodies offered by Inframet are built from two blocks: the RTCB radiator block and the CTCB controler block.

The diagram of RTCB radiator block is shown in Fig. 16. As it can be seen this block is built from: the radiation emitter, the Peltier element, two temperature sensors (the sensor of temperature of the radiation emitter and the sensor of temperature of the front wall), the cooling radiator integrated with the fans, the electronic module, and the block case.

The radiation emitter is manufactured as a several plates fixed together. The plates are manufactured from material of high thermal conductivity in order to achieve very good temperature uniformity on the surface of the emitter. The front side of the emitter plate is painted using a special high emissivity coating. This technique enables to achieve emissivity of the emitter plate equal to 0.97±0.01. Emissivity of the emitter plate can be improved up to 0.985 using additionally so called the microcavities technique when an array of small microcavities is created in the surface of the emitter.

Two high quality platinium resinstance thermometers (PRT resistors) are placed in a hole drilled in the emitter and in the hole drilled in the front wall of the of the RTCB radiator. The task of the second sensor is to measure the ambient temperature.

The PRT resistors are characterized by high linearity, temporal stability and high temperature coefficient . Due to good thermal contact of the PRT resistor with the emitter plate or with the front wall it is possible to measure accurately temperature of the emitter plate and temperature of the front wall (or further of the MRW rotary wheel when the wheel is connected to the TCB blackbody). Because of time inertia the measurement is typically done with frequency not higher than 1 Hz. Next, due to fixing of the PRT resistors in high thermal conductivity material the influence of the effect of resistor self heating on its resistance is minimized. Further on, the emitter was placed inside a cavity made from low thermal conductivity material in order to minimize influence of air random vortexes on temperature distribution on the surface of the emitter

The back side of the emitter plate is closely attached to the Peltier thermo-element of temperature dependent on the applied voltage. It is possible to heat or cool the emitter applying proper voltage. The other side of the Peltier thermoelement is closely attached to the cooling radiator equipped with fans. The task of the cooling radiator is to dissipative quickly the heat from the Peltier element.

Measurement of temperature of the emitter plate and temperature of the front wall is done by measurement of temperature dependent resistance of the PRT resistors. The resistance is converted to voltage using resistance bridge made from resistors of high linearity, temporal stability and temperature stability. Next, the output signal is amplified using a low noise preamplifier with corrected temperature drift. Finally the analogue voltage is converted to digital signal using 24-bit A/D converter and that digital signal is sent to the CTCB controller.

The CTCB controller consists of 3 basic modules: microcontroller 1 module, microcontroller 2 module, and power supply module.

There are 3 basic functions of the microcontroller 1. First, it enables communication with the RTCB radiator block, with PC computer through RS 232 port, and with the microcontroller 2. Second, the microcontroller 1 converts values of the output voltage into values of temperature. Third, the microcontroller 1 controls, through the D/A converter, the voltage applied to the Peltier thermoelement in the RTCB radiator.

The microcontroller 2 controls the keyboard and enables the user to set the required absolute or relative temperature. Next, the mikrocontroler 2 controls the screen and enable visualization of current and required values of temperature of the emitter plate, or the temperature difference between the emitter temperature and the front wall temperature.

Task of the power supply module is to power the microcontroller 1 module, the microcontroller 2 module and the Peltier termoelement and analog electronic module of the RTCB radiator block.

Fig. 18. Photo of the TCB-2D blackbody: RTCB radiator block and CTCB controller block

The main task of the CTCB controller is to control and stabilize absolute or differential temperature of the emitter in the RCTB radiator. Stabilisation is a temperature control process of the emitter with aim of achieving its stable temperature equal to the value set by the user. Temperature control of the emitter plate is done via precision control of the voltage applied to the Peltier element.

The easiest way to shorten time necessary to achieve the stable temperature of the emitter plate is to decrease thermal capacity of the emitter plate. However, lower thermal capacity means also lower thermal uniformity of the emitter and therefore the only reasonable way to shorten stabilisation time is to optimise control algorithm of the voltage applied to Peltier element. Classical PID(proportional-integral-derivative) algorithm can be used for this voltage control. However due to its low speed a combination of DMC (Dynamic Matrix Control) and PID algorithm is used in CTCB controller. All data required by these algorithms is determined on the basis of statistical analysis of measurement results obtained during experimental trials.

Before TCB blackbody can work propely it must be calibrated. Calibration of TCB blackbody is a process when the relationships between temperature of the sensors and output voltages generated by the temperature sensors are determined. The calibration process is carried out using an external temperature probe inserted to a hole in the emitter plate close to the temperature sensors. High quality certified external temperature meter of temperature resolution 1 mK and stability 2 mK is used during calibration.

The relationship between temperature and voltage is determined at 10 temperature points. Later the measurement data is interpolated using a tenth degree polynomial and the calibration function is generated in a form of a table. Values of the calibration table are saved in EPROM memory what enables easy editing during cyclic recalibrations.

The presented above description and photos refer precisly to old model of blackbodies offered by Inframet. New models look a bit different but basically the work concept is the same as presented earlier.

The user can control the TCB blackbody not only using CTCB controller but also from a PC using ThermoDriver computer program. This way of communication between the user and the TCB blackbody is more convenient as the user can use large keyboard, mouse and monitor of the PC instead of a small keyboard and small screen of the CTCB controller. Next, the ThermoDriver software enables the user visualization and recording temperatures versus time and recalibration of the blackbody. It also support MRTD and MDTD measurements. This means these are still manual subjective measurements and the final decisions but be taken by the user but the software helps him to arrange, analyse, visualize and record measurement results.

3.2.5.2 Requirements

Temperature resolution NETD of modern cooled thermal cameras can be as low as about 10 mK. Temperature resolution NETD of uncooled thermal cameras is worse (typically about 100 mK) but is improving rapidly. If we want to test accurately thermal imagers then we need blackbodies of temperature resolution a dozen or more times better than imager temperature resolution (NETD). Temperature resolution of 10 mK is acceptable in case of non-cooled thermal cameras. However, if cooled thermal imagers are to be tested then blackbodies of 1 mK resolution are needed. In some extreme cases when imager MRTD at low frequency range is below 10 mK then even 0.1 mK resolution is usefull².

During measurement of such characteristics like MRTD, MDTD, and noise parameters differential temperature range of no more than 5C is needed. During measurement of SiTF, SRF, MTF or during calibration process much wider temperature range is needed. It can be estimated that absolute temperature range from 0C to 75C fulfils typical requirements but sometimes extended temperature range from from -15C to 100C is useful.

Theoretically it is possible to use grey bodies during tests of thermal cameras and later correct influence of difference between emissivity of the real blackbody and emissivity of an ideal blackbody on condition that the emissivity of the real blackbody is not lower than 0.9 due to problems with reflected radiation. However, a better option is to use real blackbodies of emissivity close to one when almost no correction is needed, particularly during differential measurements. Therefore emissivity not least than 0.97 can be considered nowadays as a standard requirement.

Measurement of some characteristics is carried out for both positive and negative contrast. Later the results are averaged. Theoretically this approach enables full correction of systematic errors of measurement process (mostly caused by offset phenomenon). However, practically this correction works properly only when blackbody does not changed its temperature with time. Practically this means that blackbodies of high emporal stability are recommended. Blackbodies of temporal stability not worse than 3mK are usually acceptable but there are cases when temporal stability at level 1mK is needed.

Non-uniformity of temperature distribution on surface of the blackbody emitter should not influence the measurement results. The requirements on non-uniformity are highest during measurements of subjective characteristic like MRTD. The observer should not notice any non-uniformity on blackbody surface during MRTD measurement. Temperature difference during MRTD measurements is usually not higher than 5C and therefore it is usually cosidered that non-uniformity is acceptable if it is below 10mK at 5C.

Tests of thermal cameras can be quite time consuming; particularly MRTD measurements. Therefore blackbody speed (stabilization time) is an important parameter because long settling time means long tests. There is no strict limits on settling time but it is convenient if blackbody temperature stabilize at time no longer than about 1 minute in case of small 2 inch blackbodies or 90 sec in case of blackbodies of size more than 4 inches. .

Due to differential nature of measurement of most characteristics of surveillance thermal imagers accuracy of absolute temperature of the blackbody emitter is not typically a crucial parameter. However, accuracy is important during calibration process of measurement (commercial) thermal cameras. Absolute accuracy of modern commercial thermal cameras is not better that 2C (rarely 1C). Therefore it seems that temperature uncertainty of blackbodies at level 0.1C should be acceptable but 0.01 C should be considered as the recommended value.

3.2.6 Image acquisition/analysis module

Image acquisition/analysis module is built from the following blocks:PC, frame grabber (video card),

and test software. It is practically a specialized PC to carry out tasks needed in testing thermal imagers. The module should enable acquisition of the output signal from the tested thermal imager, analysis of the captured images and semi-automatic determination of important characteristics of thermal imagers.

3.2.7 PC

In general PC should enable processing of the input data from the frame grabber, and calculation and visualization of characteristics of thermal imagers. Practically all modern PC can handle such tasks.

3.2.8 Frame grabber

The task of frame grabber (video card) is to capture and record on PC hard disk sequences of images generated by tested thermal imager. The frame grabber should accept input data in typical electronic standards: PAL, NTSC, Fire Wire, USB 2.0, Camera Link. Next, it is critical that there should be no noticeable degradation of image quality caused by the frame grabber. Please note that commercially available frame grabbers are designed with aim to be used in applications where quality of the captured images is not critical and most of them cannot be used to capture video sequences generated by thermal imagers.

Each frame grabber device should provide a proper driver, which should contain routines compatible with one of existing, commonly used APIs (Application Programming Interfaces). Image acquisition applications (software modules) can base on such standards as e.g. TWAIN, or more native to Microsoft’s operating systems interfaces as DirectShow or WIA (Windows Image Acquisition). Main functionality of image acquisition software module is to get image from video capture device in form of separate frames or – in most cases – in form of video sequence. Such collection of image data can be passed on for further processing and analysis

3.2.9 Test software

There are three basic tasks of test software used in test systems for testing thermal cameras enable the following functions:

.The tasks mentioned above can be handled in different way by different computer programs. Here we will formulate basic requirements and recommendations for test software.

3.2.10 Optional blocks

It is impossible or at least very difficult to carry out testing of thermal imagers without such blocks like: collimator, blackbody, rotary wheel, targets, PC, frame grabber, and test software. There are also some blocks that are not strictly needed but still can be useful.

3.2.10.1 Temperature chamber

Thermal imagers are usually tested at laboratory conditions in situation when they are expected to work properly at extreme ambient temperatures. It is a commonly forgotten truth that parameters of thermal imagers can vary significantly with ambient temperature. This dependence or in other words imager temperature stability can be measured by placing the tested imager inside a temperature chamber of variable ambient temperature and then by measuring imager characteristics. There are two methods to measure temperature stability: the first by putting to the chamber both the imager and the blackbody, the second – by putting to the chamber the imager and using a chamber with a transparent window. The latter method is more convenient for the test crew.

There are many commercially available temperature chambers on the market. However, they are typically not optimized for testing thermal imagers: they are too big, have too high thermal inertia, and the chambers often are not equipped with infrared transparent windows. Therefore It is recommended to use for testing thermal imagers small temperature chambers of low thermal inertia equipped with an infrared transparent window. Temperature range of the temperature chamber should fit to the temperature range of the environment where the tested imager is to be used.

3.2.10.2 Optical table

Special expensive antivibration optical tables are not generally needed as a place where the test equipment is to be located to enable precision alignment. Simply the required accuracy of alignment of the test system is much lower than of some laser laser or holographic equipment. Test equipment and tested thermal camera can be well align on any large, heavy and stable wooden (stone, metal) table in most cases. However, special care should be taken during testing long range thermal imagers of very narrow field of view. Vibration of the table can influence image quality of the image from the tested imager and distort measurement results. In this case typical tables should be equipped with vibration damping parts or special antivibration optical tables used.

3.3 REFERENCES

1. ASTM standard E 1213-2002 “Standard Test Method for Minimum Resolvable Temperature Difference for Thermal Imaging Systems”

2. ASTM standard E 1311-99 “Standard Test Method for Minimum Detectable Temperature Difference for Thermal Imaging Systems”

3. Chrzanowski, J. Fischer, W. Wrona, Testing of Thermal Imagers For Non-Destructive Thermal Testing Applications, ASTM Journal of Testing and Evaluation, 28, 395-402 (2000).

4. Chrzanowski K., Evaluation of IR collimators for testing of thermal imaging systems, Optoelectronics Review, 1/2007.

5. Chrzanowski K., Lee H.C., Wrona W., A condition on spatial resolution of IR collimators for testing of thermal imaging systems, Optical Engineering., 39 (5), 14137-1417 (2000).

6. Holst G.C., Testing and evaluation of infrared imaging systems, JCD Publishing Company (2008).

7. Holst G.C., The Infrared & Electro-Optical Systems Handbook, Vol.3: Electro-Optical System Design, Analysis, and Testing, Chapt. 4, pp. 206-207, SPIE (1993).

8. STANAG 4349, Measurement of minimum resolvable thermal difference (MRTD) of thermal cameras, 1995

15. Chrzanowski K., Szulim M., A measure of influence of detector noise on temperature measurement accuracy with IR systems, Applied Optics, 37, 5051-5057 (1998).

16. Chrzanowski, Evaluation of commercial thermal cameras in quality systems, Optical Engineering, Vol. 41, No. 10 (2002)

17. Digital Signal Processing Application with the TMS320 Family, Theory, algorithms and implementations, Vol.2, Texas Instruments, 1990, 570-573.

18. Guide to the expression of uncertainty in measurement, International Organisation for Standarisation-International Electrotechnical Commission-International Organisation of Legal Metrology-International Bureau of Weights and Measures, TAG 4/WG 3, 1993.

19. Holst G.C., Infrared Imaging System Testing, Vol.4, Chapt. 4 in The Infrared & Electro-Optical Systems Handbook, Michael C. Dudzik ed, SPIE 1993..

23. International Lighting Vocabulary, CIE Publ. No. 1 7.4, IEC Publ. No. 50(845) (1987).

24. International Vocabulary of Basic and General Terms in Metrology, International Organisation for Standarisation, 1993.

25. ISO 15529, Principles of measurement of modulation transfer function (MTF) of sampled imaging systems, 1999

26. K. Chrzanowski, Evaluation of commercial thermal cameras in quality systems, Optical Engineering, Vol. 41, No. 10 (2002)

27. K. Chrzanowski, Non-contact thermometry-measurement errors, SPIE PL, Research and Development Treaties, Vol. 7, Warsaw, 2000.

28. Li .X. An investigation into the stability of industrial platinum resistance thermometer, Hart Scientific

29. Lloyd J. M., The Infrared & Electro-Optical Systems Handbook, Vol.3: Electro-Optical System Design, Analysis, and Testing, Chapt. 1,SPIE (1993).

31. MIL STD 1858, Image intensifier assemblies, performance parameters of,1983

32. Ronald G. Driggers, Van A. Hodgkin, Richard H. Vollmerhausen, Patrick O'Shea , Minimum resolvable temperature difference measurements on undersampled imagers, Proc. SPIE Vol. 5076 Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XIV; 2003.

33. Schade OH (1948). Electro-optical characteristics of television systems. 1. Characteristics of vision and visual systems. RCA Review, 9: 5-37

34. STANAG No. 4351, Measurement of the minimum resolvable contrast (MRC) of image intensifiers,1987

35. TAVENER J.P., Common errors in industrial temperature measurement., Isotech Journal of Thermometry Vol.6 No.2 1995, 20–27.

36. WALKER R., Achieving 0.25mK uncertainty with an integrated-circuit resistance thermometer readout., Hart Scientific.

37. Wood, Laboratory Bench Analysis of Thermal Imaging Systems, Opt. Eng., 15,
G193-G197 (1976).

1Attention: The warranty period of most of the tested collimators expired but the collimators are used usually many time over the warranty period.

2Such tests should be carried out at stable environment.

1. Review of thermal imagers