Standard Software for Automated Testing of Infrared Imagers, IRWindows TM in Practical Applications

Transcription

1 Standard Sotware or Automated Testing o Inrared Imagers, IWindows TM in Practical Applications Alan Irwin Santa Barbara Inrared, Inc. 32A North Nopal Street Santa Barbara, CA 9303 obert L. Nicklin Uncooled Inrared Imaging Products aytheon Systems Company 3532 North Central Expressway, MS 37 Dallas, Texas ABSTACT In the past, ad-hoc and manual testing o inrared imagers hasn t been a deterrent to the characterization o these systems due to the low volume o production and high ratio o skilled personnel to the quantity o units under test. However, with higher volume production, increasing numbers o development labs in emerging markets, and the push towards less expensive, aster development cycles, there is a strong need or standardized testing that is quickly conigurable by test engineers, which can be run by less experienced test technicians, and which produce repeatable, accurate results. The IWindows TM system addresses these needs using a standard computing platorm and existing automated I test equipment. This paper looks at the general capabilities o the IWindows TM system, and then examines the speciic results rom its application in the PalmI and Automotive I production environments. Keywords: inrared imagers, FLI, inrared imaging, automated testing, IWindows, PalmI, Automotive I, Automated MTD, AutoMTD.. INTODUCTION Automated FLI (Forward Looking Inrared) measurement equipment provides the ollowing beneits:. emoves human subectivity rom the measured data thus creating a reproducible measurement technique which can operate in a round-the-clock mode independent o operator skill or atigue level. 2. Drastically reduces measurement test time by a actor o 0+, which lowers product and/or maintenance cost. 3. Increases product quality by measuring all inrared parameters, and statistically monitoring key perormance parameters. The problem with automated FLI testing has traditionally been the ad-hoc preparation o the sotware systems, which requires specialized expertise and complex interacing between the control o hardware actuators and sources, the electronic interace to the data collection system, and the computationally intensive data analysis sotware. Also, much o the equipment required has been expensive and specialized. In an environment where system production is small and the ratio o trained personnel to production rates is high, then manual control, measurement, and data entry is oten a reasonable alternative. However, increased production volume and the need to use more production and ield personnel who are less well trained has increased the need or simple automation which can be quickly conigured or speciic testing requirements. aytheon TI Systems was aced with these concerns while developing the production line o their uncooled inrared cameras. Their solution was to use the IWindows TM FLI test system as developed by Santa Barbara Inrared, Inc.

2 . Santa Barbara Inrared, Inc. (SBI) SBI manuacturers inrared test equipment or both component and system level testing. SBI has developed the necessary sotware and hardware to accomplish automated FLI testing. This sotware product is called IWindows TM (IWindows). IWindows is a relatively new product introduced in 996. SBI and aytheon TI Systems have ointly participated in the product testing o IWindows. As a result o this eort, product enhancements have been incorporated, which improve the IWindows unctionality to a sophistication level required or perorming automated inrared system test and depot level testing..2 aytheon Systems Company (SC) SC produces ground and air based inrared detection systems. The ground based inrared products consist o high perormance cooled (detector) systems and medium perormance uncooled (detector) systems. SC has researched the use o automated inrared testing using IWindows on the uncooled product line since September 996. SC has implemented the IWindows measurement technology or use on the acceptance testing or PalmI and Automotive I (both commercial products). These two products system level testing were automated using IWindows during the second hal o 997. Future product implementation on the remainder o the commercial and military product line is being planned; including the Mobile (commercial surveillance I imaging system), the thermal weapon sight (W000 series) and the Drivers Vision Enhancer (DVE). The implementation o the IWindows measurement technology is also being developed or the Improved Bradley Acquisition System (IBAS), a high perormance military FLI, utilizing cooled, high resolution detector, as well as automated depot level testing. Military product testing is more stringent than commercial testing, thereore, a rigorous test plan must be completed to validate any new measurement approach rom the baseline measurements currently being utilized. esearch in automated uniormity measurements was also perormed or DVE during the second quarter o 997. Successul implementation o the IWindows measurement system is the result o lessons learned through research studies and measurement activities. This document describes the IWindows hardware, measurement deinitions and ormulas, the test plan approach, results o correlation eorts, and the implementation o the uncooled product line and test plan results. 2. IWindows HADWAE, MEASUEMENT DEFINITIONS, AND FOMULAS 2. Measurement Hardware The automated measurement system (see Figure ) consist o a unit under test (UUT), an inrared target proector, a data acquisition system, and a personnel computer (PC) containing the control, measurement, and analysis sotware IWindows. The inrared target proector consist o a relective (two mirror) collimator, a computer controlled blackbody controller and target wheel, a blackbody, and the various opaque targets. The data acquisition system consist o a commercial S-70 compatible digitizer. The computer measurement system is customized to measure key parameters o an inrared system, and along with pre-deined constants is capable o estimating the relative perormance normally described by a trained human observer. The IWindows sotware also allows an operator to conigure the parameters necessary to run a particular measurement, or load parameters rom a previous test. IWindows has a graphical type interace, and is easy to control and modiy the measurement parameters. Test results are displayed to the operator in a tabular or graphical display, or can either be exported to a text ile or dynamically linked to a master control program. And (optionally), entire test plans can be predeined as macros and run by clicking a button rom an operator s menu.

3 Figure Automated I Measurement System 2.2 Measurement Theory and Technology Various types o perormance parameters are used to measure the perormance o a FLI s ability to detect thermal radiation o a real world scene. Inrared electro-optical (EO) measurements include; the signal transer unction (SiTF), small area noise equivalent temperature dierence (NETD), small area uniormity (sigma divided by the mean), large area NETD, large area uniormity, modulation transer unction (MTF), minimum resolvable temperature dierence (MTD); and automated minimum resolvable temperature dierence (Auto MTD). Figure 2 (at the end o this section) is a graphic low diagram o how IWindows perorms Auto MTD. Each test produces a perormance parameter on its own, and the diagram shows how the elements are incorporated into the Auto MTD. Each o the tests is discussed in the ollowing sections (more rigorous deinitions can be ound in Holst, Dudzik 2 ) Signal Transer Function The signal transer unction (SiTF) determines the responsivity o an inrared imager. The SiTF provides inormation on gain, linearity, dynamic range and saturation. The SiTF is the slope o the output voltage versus the dierential temperature curve. The SiTF responsivity curve is typically shaped as a sideways S. The SiTF measures the variation in the signal dierence between the target and its background over a series o dierential temperatures (positive and negative). The SiTF is an automated measurement using the IWindows measurement system. The SiTF is measured by acquiring data rom an area on each side o a large area target. The video rame is captured using the rame grabber electronics. The inrared imagers output variation is sensed as a voltage dierence between the target image area and its background area. Thus, the target and background signal are acquired, and a voltage dierence corresponding to the temperature dierence is obtained. The dierential temperature o the blackbody controller is then commanded to the next temperature value, and the sotware obtains another data point on the SiTF curve. A minimum o two data points is required, however, at least ive data points is recommended. Ater completing all the pre-programmed dierential temperature measurement points, IWindows calculates the slope o the line, tangent to this curve to be the SiTF value. Frame averaging may be utilized to smooth out signal (temporal) variations during the SiTF measurements, however, this is dependent on the size o the measurement area utilized. I the measurement area is large enough, this tends to suiciently average out the temporal variations o the signal. Conversely, i the measurement area is small some amount o rame average may be necessary to eliminate temporal noise eects Noise Equivalent Temperature Dierence Measurement The noise equivalent temperature dierence (NETD) test measures the temperature dierence that produces a peak signal-tonoise ratio o (or unity) under equal (or lood) illumination. The NETD measurement is an excellent diagnostic indicator o system sensitivity because it veriies optimum system perormance.

4 NETD is determined by measuring the zero-signal noise, computing an MS and dividing by the SiTF. IWindows isolates the high requency component o the noise by removing ixed pattern (low requency) noise rom a captured image. The blackbody temperature or this measurement is usually set to 0 delta temperature. The NETD measurement can utilize other source temperatures (besides 0 delta T). This is usually or special uniormity tests, which can simulate environmental conditions or actual real world imagery. Other variations o noise measurements can be implemented by IWindows, which do not remove the ixed pattern low requency noise. This has advantages when measuring uncooled ocal plane arrays, which may have varying amounts o ixed pattern (describe as chicken wire ) noise. Since this noise intereres with the trained observer s ability to perorm manual system perormance measurements, it should be included in the unction, which predicts the Auto MTD measurement. This type o noise measurement is evaluated using a small area block, approximating the size o the bar target used in the manual perormance (MTD) measurement. The large area noise (uniormity) measurement analyzes a portion o the ield o view (FOV) or blemishes, blotches, and shading eects, which may be distracting to the observer. IWindows can be conigured to perorm this measurement. Uniormity is a measure o the luminance (or intensity) level at each point within the designated areas on the display as compared rom the total average luminance value. Frame averaging can be used to reduce temporal noise in order to match an observers real world scene interpretation (this is somewhere between 3-0 rames and is dependent on the person and the amount o scene movement). The mean and standard deviation o the pixels in the target area can then be calculated rom the selected area to be measured. Uniormity can be calculated by dividing the standard deviation o the pixels in the target area by the mean. Multiple small areas may also be measured within a single rame. However, large area uniormity (and noise) is obtained by including the entire FOV Modulation Transer Function The modulation transer unction (MTF) is a measure o how well the system will reproduce the scene. The highest spatial requency that can be reproduced is the system cuto or Nyquist requency. The MTF is calculated rom an edge input, which is dierentiated to obtain a line spread unction (LSF). A Fast Fourier Transorm (FFT) is perormed on the LSF to obtain the requency values. The requency output is normalized to get the 0% to 00% modulation values. IWindows perorms all these calculations on captured video data, usually averaging a large number o rames to reduce high requency noise eects on the calculations. Provisions or removing some well characterized noise sources (speciically the pedestal and component sample smoothing) are also included in the calculations Manual Minimum esolvable Temperature Dierence Observer detection o standard our bar targets is an industry-wide method o measuring inrared image quality and overall system perormance. The minimum resolvable temperature dierence (MTD) is a subective measurement, which depends upon the ability o the observer to resolve detail, and is inversely related to the MTF. The measurement sotware controls the target positioning, the starting temperature, and data recording/analysis. These are automated eatures which reduce the overall test time or this manual test Automated MTD The automated minimum resolvable temperature dierence (Auto MTD) is based on automated measurements and calibrated correction actors which incorporate an observer s eect on resolving each target. Ater calibration, Auto MTD does not require our bar targets or a trained observer. Vertical or horizontal measurements may be obtained providing the MTF has been perormed in that axis. The calibrated correction actors are called K, or K actors. Once calculated or a product line, then new MTD values can be calculated or an individual system rom its measured NETD and MTF. The Auto MTD ormula is: MTD K NETD MTF To get these results, an NETD test and an MTF test need to be perormed or each UUT. The K actors are derived rom the ormula:

5 K MTD MTF NETD where the MTD value reers to the results rom a manual MTD test perormed at the target spatial requency,. K actors or production testing are typically averaged using several units with several trained observers. K actors can be rechecked periodically to ensure reliable measurement predictions are maintained through the production lot. These K actors are stored in the IWindows coniguration ile (or a urther discussion o Automated MTD, see Holst ). SiTF dv dt ( dv )( dt ) 2 2 ( dt ) ( dt ) dv ( dt dt ) ( dt dt ) 2 Video Capture Extract Signal & Background Values Compute SiTF Value Video Capture Extract Area Values σ x, y B ( P B ) xy ( b ) Compute Noise 2 NETD NETD σ SiTF Trans Compute NETD K MTD MTF NETD Video Capture Extract Knie Edge Values Dierentiate Knie Edge or Line Spread Function Perorm FFT on LSF Normalize or MTF Measure MTD Manually (Still within IWindows) MTD dt dt 2 + Trans Statistically Average esults Over multiple Observers Measure SiTF (as above) Measure Noise (as above) Calculate NETD (as above) Measure MTF (as above) MTD K NETD MTF Automated MTD Figure 2 IWindows Calculation Dependencies Test Conigurations, Macros, and Data Files An IWindows coniguration ile is a customized data input ile, which describes parameters unique to an installation. In addition, each test procedure (SiTF, NETD, MTF, etc.) can have multiple conigurations, each o which can be executed as a speciic test. For instance, one SiTF coniguration could measure the signal 0 times at 0. o C intervals, and a dierent coniguration could perorm the SiTF taking 5 measurements at 0.5 o C intervals. The coniguration ile includes: FOV o the sensor Thermal source parameters Target wheel parameters Pass/ail criteria

6 K actors Test conigurations and macro programs Collimator transmission actor This test coniguration ile is controlled by the users resident system engineering sta and/or quality organization. The creation o this ile is based on UUT parameters, correlation studies, and production results. A test macro is a group o test conigurations which are executed in sequence. This is a way o encoding a complete test procedure which could involve many individual tests. These macros can include: UUT setup such as gain and ocus adustment UUT alignment utilizing the coordinate measurement sotware SiTF test Noise test Large area test MTF test Manual MTD test Auto MTD test Other tests as developed by SBI An event log displays testing and source control activity. Test results are displayed graphically, in tabular ormat, and can be exported or combined into a summary sheet. Pass/ail criteria can be compared to test results or decision making. Customized data sheets can be created by using the ile export commands in conunction with the Excel or any Windows TM application. Test results are collected into iles which can store all the results rom individual UUT s. These results iles (called UUT iles) can be loaded back into IWindows or urther data reduction at any time. 3. TEST PLAN AND COELATION MEASUEMENTS A test plan was created or the purpose o validating the perormance o IWindows. This test plan was conceived rom discussions held with personnel rom the US Army Night Vision and Electronic Sensors Directorate (commonly called NVL). NVL is a customer or the SC uncooled military product line. This test plan is shown in Figure 3. The test plan, simply stated, proves how well the measurement system predicts the MTD as compared to a trained human observer. Various test are used to degrade, improve, and alter the perormance o the inrared imager. Then the manual MTD is compared with Auto MTD. Additional correlation measurements are also perormed to determine how accurate the NETD and MTF measurements are as compared to the existing methods o measurement. Although it is important to always predict the MTD correctly, it is more important to never pass (predict a lower MTD value) on a system which is in question o ailing its speciication. However, i a system does not meet speciication it is less important to know exactly how bad it is, and general ranges may be acceptable.

7 Perorm manual measurements: (MTD, NETD, and MTF) Perorm automatic measurements: (MTD, NETD, and MTF). Compute MTD k actors. Degrade MTF o system by deocusing at each available bar target Do the automatic measurements correlate with manual meas.? NO Investigate Failure Do the automatic measurements correlate with manual meas.? YES etest again or repeatability. Time and position dependant. Do the automatic measurements correlate with manual meas.? YES Degrade the responsivity o system (I ilter and/or reduce aperture). NO Investigate Failure NO Investigate Failure YES Improve the system perormance using a 7x rame average real time signal processor. Do the automatic measurements correlate with manual meas.? YES Automatic MTD Prediction is accurate or all possible cases! NO Investigate Failure Do the automatic measurements correlate with manual meas.? YES NO Investigate Failure Figure 3 Auto MTD Measurement Test Plan 3. Correlation Measurements The process o correlating a new measurement tool involves perorming established measurements and comparing the results against the new technique. This is more important when implementing the new measurement tool on an existing product. 3.. Existing NETD Measurement Techniques SC has utilized several methods or the measurement o NETD. Some o these methods include: () the manual method, (2) the image evaluation lab (IEL) method, and (3) the IBAS method. The manual method utilizes the ollowing ormula: or, where, (delta temperature x collimator transmission actor) (signal variation / 6 sigma noise) ( T x K) (S/Np-p/6) T delta temperature o blackbody source. K collimator transmission actor S Signal variation between background and source (signal) in volts. The measurement is made using the peak to peak technique. Frame averaging is required to increase measurement accuracy; however, averaging tends to attenuate signals, so too much averaging invalidates the result. Np-p/6 6 sigma noise - the peak to peak noise is measured on a small area o the background then divided by 6 to compute the 6 sigma noise component. The target needs to be large enough such that MTF degradation is not a actor. The IEL method was developed by the SC image evaluation laboratory (IEL). This method closely matches the algorithms utilized by NVL. The IEL technique utilizes a computer (programmed with a customized Lab Windows

8 program) which interaces with a Tektronix sampling oscilloscope. The signal voltage, blackbody dierential temperature, and the collimator transmission actor are entered into the IEL computer program. The signal voltage is determined manually o the video output using the digital oscilloscope rom a single video line, averaged 256 times. Individual noise measurements are obtained rom sequential video lines. Each noise measurement is an MS noise measurement. The program then calculates an average NETD based on all the video lines measured. The IBAS NETD measurement method utilizes the ollowing equation and methodology: where: NETD ( T x τ coll x Noise MS ) (Signal AVG - NOISE AVG ) T Target to background temperature dierence (5 C) τ coll Transmission o collimator Noise MS MS o (0 rame average - single rame) Signal AVG Average o signal in single rame NOISE AVG Average o Noise in single rame Note: The single rame is not part o the 0 average The US Army Night Vision and Electronic Sensors Directorate method is described in the paper by Bell and Hoover 3. The uncooled detector actory s test station measures NETD using an automated process and is similar to the IBAS technique MTF Correlation Measurements The SC uncooled baseline manual MTF measurements is perormed using a sampling oscilloscope with the rame averaging unction set to around 00. Four bar MTD targets are presented to the sensor. Video scan lines are adusted to be in the center o the bar target. The peaks and valleys o the bar target are then measured. A single peak value is determined by averaging all the peaks and a single valley value is determined by averaging all the valley measurements. The ollowing ormula is then utilized: where: MTF (V-V2)x π (V+V2)x4 V average peak measurement V2 average valley measurement The π/4 is a correction actor to convert this contrast transer unction to a true MTF result. This actor is applied only or targets at /2 o. It should not be used or /4 o targets (lower requencies). Other automated MTF measurement techniques are perormed using a slit (smaller than the detector instantaneous FOV) illuminated using a high temperature source. The SC uncooled detectors are currently tested using this (slit target) approach as does the US Army Night Vision Lab system test acility. A video scan line is sampled across the slit target, averaged, and then an FFT is perormed on the measured signal. The disadvantage o this approach is the need or a precision slit (or each sensor s FOV) and the availability and increased time to settling o a high temperature blackbody source Manual MTD correlation measurements The Manual MTD measurements made by IWindows and NVL are virtually identical. Both assume a trained observing is determining the resolution threshold by making small changes to the dierential temperature. The only dierence is that IWindows measures the white hot (source plate warmer than target cutout) temperature beore the black hot (source plate cooler than target cutout) temperature, and some programs at SC (but not all, it varied) measured black hot irst. Comparison o the measured results were consistent. 3.2 esults o Correlation Measurements The results o the correlation measurement process was subdivided into initial, PalmI, and Automotive I phases.

9 3.2. Initial Measurements The initial measurements were perormed in the irst hal o 997 using a variety o uncooled inrared imagers over the course o the initial studies (9 FOV weapon sight, 2 FOV Nightdriver products, 5 FOV box camera, 2 FOV Mobile I system, and a 5 FOV TP prototype system). These early tests allowed the IWindows system to be conigured in a ashion which provided basic correlation with manual methods. The study utilized individuals who perormed the manual measurement, the IEL method, as well as the IWindows method. The results o this study showed that the IEL method produced higher than expected results as compared to the manual method, and that the SBI correlated with a 4 rame average. The IWindows MS (isolate high requency) noise measurement was later tested and was ound to correlate with the IEL results. These results also showed that the Auto MTD predicted optimistic results when the system was purposely degraded (by deocusing and aperture limiting). Further testing using the MTF algorithms, enhanced with oset and smoothing operators, improved the test plan perormance, but not to the expected level o providing oolproo test results. It would, however, provide an accurate indication o a degraded system Implementation o the Uncooled Product Line The automated inrared test measurement process was implemented in 997 or the PalmI and the Automotive I commercial products. Various NETD and MTF correlation studies were perormed during 997, which made the implementation o this technology possible. A uniormity study was perormed or the DVE program to investigate the easibility o implementing a non-subective technique or measuring the large area uniormity. The results o this study are discussed in a separate report. The Mobile (commercial) and the Drivers Vision Enhancer (military) applications are planned to be implemented during PalmI Implementation PalmI is the lowest cost commercial inrared imager sold by SC to date. PalmI is a hand held, battery powered system designed or industrial, security, and inspection applications. PalmI utilizes a commercial black and white, camcorder video display (viewinder). This viewinder is tested as a sub-assembly by the display s manuacturer. A decision to perorm automatic measurements on the video output was ustiied because most product ailures arise rom within the inrared imager, and not rom the display. In addition, the display is tested through a visual veriication beore inal shipment is made. PalmI validated 50 units using the manual and automated MTD measurements. The Auto MTD K actors were determined by averaging the results rom the initial 0 measurements, then were slightly modiied during the remaining units to achieve an acceptable error margin between the manual and Auto MTD measurements. A complete test plan analysis was not completed due to early implementation problems. However, an MTF and NETD absolute correlation eort was perormed. The IWindows Auto MTD was optimized or successul implementation by ensuring the lens was at optimum ocus beore perorming the automated testing. The use o the trained observer made this technique possible even though detailed manual testing is not required. This visual inspection would be necessary to ensure product quality. Accurate sensor positioning is required to ensure repeatable/reliable measurements. This is accomplished using the IWindows coordinate measuring system and an azimuth/elevation sensor positioning system. Table I is a summary o the Auto MTD accuracy. The PalmI absolute average percent deviation between the manual MTD and Auto MTD or the three target sizes are 0, 7, and 7% (% average). Table II is the list o the K actors used or the Auto MTD. Figure 4 shows the deviation (MTD - AutoMTD) measurements or each target s spatial requency, as computed rom 22 measurements. The use o the AutoMTD measurement technique has been successul and proven to be a valuable tool in achieving the high rate production o the PalmI product. Target (cy/mr) Spec. Value ( C) Table I PalmI Auto MTD Prediction Accuracy manual - manual - Auto MTD Auto MTD (maximum) (average) manual - Auto MTD (minimum) manual - Auto MTD % deviation (minimum) manual - Auto MTD % deviation (maximum) manual - Auto MTD Absolute deviation (average %) 0.9 (/4 o) % 2% 0% 0.38 (/2 o) % 26% 7.0% 0.76 (o) % 56.0% 7%

10 Table II PalmI K Factors Target K actors /4 o /2 o o % 60% Percent Deviation 40% 20% 0% -20% /4 o /2 o o -40% -60% Figure 4 Dierence between manual MTD and AutoMTD as a percentage o speciication Automotive I Implementation The Automotive I inrared imager is the next generation uncooled automotive product. It is designed to be installed under the hood o an automotive vehicle. This product is being designed to be the lowest cost inrared imager with manuacturing cost to be below the $,000 level. The display utilized or this product is a heads up display, which is presented to the driver in the lower portion o the windshield. The use o an automated measurement system will be required to meet the production rates, system cost, and product quality. Automotive I validated 8 units using the manual and Auto MTD measurement technique. The test was perormed similar to PalmI in that the video signal was analyzed (not including the display). New IWindows MTF sotware algorithms (developed as a result o the IBAS testing) was utilized on this system implementation. Several ailures were indicated using the Auto MTD, which suraced real problems (ocus & or noise). During the limited production run o 50 units, several alse ailures were manually re-tested (on the single target which ailed) to allow the system to pass the speciication. Accurate sensor positioning was also required to ensure repeatable/reliable measurements. This was accomplished using the IWindows coordinate measuring system and an azimuth/elevation sensor positioning system. Table III is a summary o the Auto MTD accuracy. The Automotive I absolute average percent deviation between the manual MTD and Auto MTD or the three target sizes is 6.7, 3, and 5.3% (or.7% average). Table IV is the list o the K actors used or the Auto MTD. Figure 5 shows the deviation (MTD - AutoMTD) measurements or each target s spatial requency, as computed rom the 8 measurements. Target cy/mr Spec. Value ( C) Table III Automotive I Auto MTD Prediction Accuracy manual - manual - manual - manual - Auto MTD Auto MTD Auto MTD Auto MTD manual - Auto MTD manual - Auto MTD

11 (minimum) (maximum) (average) % deviation (minimum) % deviation (maximum) Absolute deviation (average %) 0.9 (/4 o) % 8.7% 6.7% 0.38 (/2 o) % 26% 3.0% 0.76 (o) % 32.0% 5.3% Table IV Automotive I K Factors Target K actors /4 o 0.93 /2 o o % 20% Percent Deviation 0% 0% -0% -20% /4 o /2 o o -30% -40% Figure 5 Dierence between manual MTD and AutoMTD as a percentage o speciication 4. CONCLUSIONS The IWindows sotware, when properly implemented, provides an accurate automated inrared MTD measurement. This measurement has acceptable regions o uncertainty, which correlate equally with manual MTD measurements (especially when operator atigue or inexperienced operators are introduced). It is cost prohibitive to manually measure MTD on systems with a high rate o production. Although a signiicant amount o non-recurring engineering has been spent on perecting the measurement results obtained by IWindows, SC believes the pay o in increased product quality and lower production cost is worth the expenditure. Future improvements o IWindows may have adaptive mechanisms to shorten the implementation cycle time and make alignment less critical. And, as was mentioned previously, uture implementations on the remainder o the commercial and military product lines are being planned. 5. FUTHE IWindows DEVELOPMENT

12 In addition to the SC installations, IWindows is being installed and utilized in a variety o acilities or both production and development purposes. Further releases o this product will be introducing additional capabilities, including capture o digital camera data streams and additional test procedures such as 3D Noise and component (detector) level testing. 6. EFEENCES G. C. Holst, Testing and Evaluation o Inrared Imaging Systems, JCD Publishing Co., Maitland, FL, The Inrared and Electro-Optical Systems Handbook, Volume 4 Electro-Optical Systems Design, Analysis, and Testing, M. C. Dudzik, ed., Inrared Imaging Analysis Center, Ann Arbor, MI and SPIE Optical Engineering Press, Bellingham, WA 3 P. A. Bell and C. W. Hoover, Jr., Standard NETD Test Procedure or FLI Systems with Video Outputs, Inrared Imaging Systems: Design, Analysis, Modeling, and Testing IV, G. C. Holst, ed., SPIE Proceedings Vol. 969, pp , 993 Further IWindows inormation: SBI irwin@sbir.com; WWW: Telephone: (805) ; FAX (805) SC WWW: Telephone: (800)