real world are represented and where the information from the real world might be degraded by the systems components. Examples of such components are the systems optics, detector elements, system electronics, storage media, digitizing procedures, computer based algorithms, display units, printers and the abilities of the systems operators. These factors all need consideration during any forensic examination. Imaging technologies are designed to achieve an optimal image quality, but features like noise patterns and deviations in signal-levels are often the key issues in forensic investigations. The technical approach for this investigation was to first analyze the FLIR sensing system and, in particular, its SPRITE/TED detector. Some typical fixed pattern noise induced by the scanning mechanisms, among other factors, was immediately detected and recognized on the FLIR imagery. The results from this examination proved valuable in the digitizing process, noise reduction and for estimating the tracking error in the 3D reconstruction. The findings of the systems examination were also quite important to explain why small objects cannot be trusted to be correctly represented in the thermal images from the FLIR system. The next step was to analyze the FLIR videotapes in depth. Because analyzing evidential images often means extracting information that is not directly obvious to our eyes and brain, image enhancement methods were used to provide images that were easier to interpret. Various analytical methods were used to provide complementary numerical values, such as the position of a flash or the FLIR sensor position. The main focus of this part of the investigation was to reconstruct the flash geometry. This work has been performed with awareness to the generic problems of detecting small objects, false detection, and the specific imperfections of the FBI Nightstalkers long-range thermal imaging system used on the April 19, 1993. The electrical signals stored on the FLIR videotapes were successfully used to authenticate the April 19, 1993, FLIR recordings. Video machines leave individual and detectable traces on the videotape during recording. To determine if the recordings were interrupted, tampered with, erased, altered or recorded over, the following electrical signatures were analyzed; RF-envelope for video signal, RF-signal carrier frequency for video signal, RF-envelope for FM-audio signal, RF-signal carrier frequency for FMaudio signal, dihedral error measurements and control pulses (CTL- pulses).
4. IMAGING The system used to digitize and analyze the recorded images on the April 19, 1993, FLIR videotapes utilized dpsReality/Velocity software and a frame grabber processing board. This was hosted in an Intel Pentium PC with a Windows NT 4.0 operating system. Data was stored uncompressed on an IBM SCSI raid disc with a capacity of 226 Gb. Seven additional IDE discs, of 20 to 26 Gb storage capacities were used for uncompressed storage of the remaining digitized video. In order to digitize and store such a large amount of data, while retaining as much as the image quality as possible, analogue measurements and system tests were performed prior to digitizing. This was an important phase in the video authentication investigation, as deviations in measured values are also useful to trace technical tampering. Significant effort was put into creating sequences to be used for image enhancements. As a result, the FLIR videotapes were digitized at three different levels of the available video signal. First, the dpsReality/Velocity systems default settings were used to maintain the FLIR videotapes dynamic range. System defaults in this case resulted in some bit reduction. As the graphical presentations in the FLIR image peripherals were of slightly higher intensity than the saturation level by the FLIR sensor, this bit reduction was not critical. When digitizing the video signal, the white-level of the horizontal part in the AZ presentation, figure 3.1, was used as reference as this was the highest video signal level. This was set to correspond to pixel value 255 in an 8-bit grey scale image, which also corresponds to the cut of level of the system used for this investigation. Loss of single frames was only detected on three occasions, a rate that was considered small enough to be neglected considering the large amount of frames. Secondly, interesting sequences from FLIR videotape Q-4 were digitized, with the dynamic range in low contrast areas stretched before digitizing. This saturated the upper and lower parts of the dynamic range. In this way, a significant enhancement of the contrast for the middle grey levels was achieved. These sequences were specifically designed to search for persons or other objects of very low contrast relative to the background and proximate to the flashes. Third, several sequences from FLIR videotape Q-4 were digitized multiple times in the same manner to be used for noise reduction. All of FLIR videotape Q-4 was digitized. Selected sequences from FLIR videotapes Q-1, Q-2 and Q-5 were also digitized. FLIR videotape Q-3 was omitted as the content and quality were similar to FLIR videotape Q-1. FLIR videotapes Q-6 and Q-7 were not used since they were first generation copies. In total, approximately 3 hours of video from the FLIR videotapes were digitized, which still left space on the system for temporary storage of processed sequences.

Figure 5.1. Regular appearance of flashes from the same part of the single story roof at 11:43:36, 11:45:20, at 11:47:07 the roof was partially clouded and a flash only faintly visible, at 11:49:02,
11:51:00, 11:55:46, 11:57:26 and 11:59:03. At 11:53 no reflection was detected. At 12:01 and 12:03 the single story roof was only partially seen in the FLIR images. At 12:05:15 a flash appeared again from this same part of the single story roof. Two-minute turns are commonly used in aviation. From the FLIR imagery it was clearly seen that one turn of the FBI Nightstalker takes about two minutes to complete. See for example figure 5.1. On April 19, 1993, the FBI Nightstalkers pilot probably flew so the operator could keep the complex in the FLIR sensors field of view. In all probability this caused the pilot to fly under visual flight rules (VFR). Variations in the FBI Nightstalkers flight path were believed to have been caused by the strong wind, which made the aircraft drift relative to the complex. As the flight path played an important role during recording of the FLIR videotapes, the first task was to investigate the various positions of the FBI Nightstalker and thereby the FLIR sensor, section 5.4. It was discovered that the flashes seemed fixed in position relative to the complex. Based on the predictable nature of the flashes, it was decided to reconstruct the flash geometry in order to confirm the hypothesis that the flashes were reflections from glass or other kind of debris, section 5.5. Reconstructing the geometry for the flashes in table 2 became the main focus of this investigation due to the characteristics of the flashes, the results from the image enhancement and the lack of persons proximate to any of the analyzed events. A three-dimensional (3D) reflection model based on the physical laws of reflection formed the basis for the reconstruction of the flash geometry. Essential parameters included the position of the complex, the position of the sun and the FBI Nightstalkers flight path from which the FLIR sensor position was determined. The reconstruction model was based on the geometrical relation between the two-dimensional (2D) information in the FLIR thermal images and the true size and location of objects existing in the real world. The time was set within the model based on the time information in the images and the reliable frame rate of the video standard. In all probability, the FLIR operator set the time and date information when he initialized the FLIR system. One of the problems encountered when reconstructing 3D information from FLIR videotape Q-4 was the lack of detailed platform data. Fortunately, the Nightstalkers FLIR sensor is well described in research literature. Additional information requested and provided by the Office of Special Counsel, were reliable measurements of the complex and weather conditions for April 19, 1993. As the solar specular reflection hypothesis did not fully explain the appearance of some of the flashes, the characteristics of the FLIR sensor and the video format were also studied. The Fort Hood flight trials were also analyzed, section 5.5. This analysis was motivated from the results in section 5.1, where it was concluded that the FLIR system could distort the image representation of small objects such as the flashes.

sequences play back at half the normal speed as each processed field was repeated twice to form one new frame. The upper left quadrant of the screen contains a cropped version of the above sequence A. The upper right quadrant depicts the corresponding sequence with increased resolution, sequence F super-resolution using frame fusing techniques. The bottom left and right quadrant of the screen depicted from left to right shows:
sequence C, with stretched dynamics to enhance the contrast sequence D, with motion deblur noise reduction technique applied sequence E with total variation noise reduction technique applied
Sequences D) -- G) utilized the sequences A) -- C) as input. The split sequences were created for comparison of processed sequences to the same unprocessed sequences. By analyzing the split sequences it was verified that no significant artifacts were induced.
Figure 5.10. A view from the split-screen sequence from event 53 at 12:09:00. The results from reviewing sequences A) C) was that several moving objects were detected, however nothing was detected in close proximity to any of the listed flashes. Several events were also analyzed in depth utilizing the procedures D) G). For example, event 5 during 11:18:20 24, event 50
during 12:08:30 34, and event 53 during 12:08:59 12:09:03. Nothing in close proximity to any of these three flashes was detected within any of these sequences. When analyzing event 50, moving objects were detected and the sequence 12:08:34 12:08:39 were enhanced by the same techniques A) G) and further investigated in section 5.3. 5.3 Motion Detection, Tracking and Analysis An individuals radiated temperature difference relative to the background will vary as he moves. This can cause a person to appear alternately as colder or warmer than the background due to the wider temperature interval of the complex background. For this reason moving persons, in most cases, are easier to detect than persons standing still are. Occasionally a persons radiated temperature difference can be the same as the background. Consequently, a person can appear thermally invisible. This same person, however, will most probably be traceable before and after entering the area where his temperature is about the same as the background. This is illustrated by an example from the Fort Hood Lynx copy videotape (app. 2 item 17), see figure 5.11. FLIR videotape Q-5 also contains several examples of this phenomenon where persons attending the fire momentarily disappeared when walking across the water-cooled roof of the tornado shelter.

Figure 5.19. A simplistic model of the solar reflection from a planar surface, diffuse reflection component (upper left) and specular reflection component (upper right), from [Vince] p. 82-83. The behavior of light reflection can simply be expressed as diffuse or specular reflections. Lights reflected by a diffuse surface will be radiated equally in all directions and are independent of the position of an observer. Rough surfaces, like carpets and textiles, exhibit mainly diffuse reflection properties. Specular reflections create a highlight that can be seen only within a limited distance from the reflection ray. Polished surfaces, like glass and metals, can cause specular reflections for which the visibility strongly depends on the position of the observer relative to the specular reflection ray. This reflection model holds for visual as well as infrared wavelength bands. In the simplistic reflection model of one light reflection, the intensity I, of a pixel at column x, row y and time t summarizes the reflection components of a light source; ambient; diffuse; and specular reflection components. I(x,y,t)= Iambient(x,y,t) + [Idiffuse(x,y,t)+ Ispecular(x,y,t)] I(x,y,t)= Ia(x,y,t) Ka + [Ii(x,y,t)* Kd (L n)+ Ii(x,y,t) Ks cosg] (1) (2)
Ii is the intensity of the light source. Ka and Kd are surface reflection components. Ks is a colorindependent specular coefficient. The angle represents the observer V offset from the reflection ray R. L is the direction to the light source, n the surface normal and L n denotes the scalar product. g is the materials reflection component and the factor cosg and describes how the observation angle affects the intensity of the specular reflection component. The angle represents the angle of the reflection, which depends on the substance of the surface among other factors. This reflection model defines the intensity of image pixel values for an illuminated object as a function of the position of the camera and the intensity contributions from ambient, diffuse and specular reflections.
Figure 5.20. The expansion of the specular reflection model into 3D for reconstructing the geometry of the flashes on FLIR videotape Q-4 with the distribution cone of the solar specular reflection (red) in where the specular reflection can be observed. On April 19, 1993, the sun was the light source and any solar specular reflections would have resulted in reflection cones in the air above the Branch Davidian complex. As the FBI Nightstalker aircraft approached one of these narrow cones where the flashes would be observable, the angle between the reflection ray R and the observing FLIR sensor V, approached zero and the specular component became dominant. This agrees with the way a solar specular reflection is modulated by the cosg function. In fact, equations (1) and (2) can in fact be further simplified. I(x,y,t)= c1 + c2 cosg (3)

f denotes the focal length of the camera. The corresponding camera model is illustrated in figure 5.22. Even if the ideal camera model is mainly a theoretical model, it could be used to reconstruct the FLIR sensor position. The distance between the FLIR sensor and the complex was estimated directly from the images utilizing the FLIR system field of view and some reliable sizes of objects in the depicted scenery of the Branch Davidian complex. Thereby the geometrical relationship between the sun, the complex and the FLIR sensor could be numerically calculated for the flashes in table 2 utilizing specially developed software scripts.
Figure 5.22. Perspective projection (left), initially from [Vince] p. 48, and the corresponding camera model (right). The FLIR sensor position was estimated without considering the rotational movements of the image plane, which were believed to increase the error in the aircraft azimuth angle relative to the complex. To deal with the rotational movement of the FLIR sensor, MatchMover by Realviz was used to more accurately estimate the FLIR sensor positions. This software inserts virtual objects into a real-world video recording by reconstructing the camera positions. This technique is widely used in the movie industry. In this case only the 3D position of the FLIR sensor camera and the complex coordinates were of interest. The tracked FLIR sensor position from MatchMover and the 3D model was oriented towards North. Finally, the bearing to the sun was added to complete the reconstructed flash geometry. The calculated distance between the complex and the FLIR sensor position varied. However, the reconstructed azimuth and elevation angles were accurate and reliable. By numerically calculating a set of reference distances from the image of the tower roof in figure 5.18, all tracked distances from MatchMover were calibrated for accurate distances. The error in the reconstructed flash geometry was estimated to be somewhat larger than it would have been had a staring camera with known focal lengths been used instead of the FLIR system. The errors induced from the tracking procedure were mainly caused by misplacements of the 2D tracking points, see figure 5.23, or from image displacements caused by the FLIR system, illustrated in figure 5.9. By manually calculating the 3D points from the 2D FLIR images and comparing these data to the results, from both methods used for 3D reconstruction, the reconstructed flash geometry proved to be accurate and reliable.

Figure 5.26. The window from the demolished catwalk.
Figure 5.27. Event 50 at 12:08:31 tracked (red and red arrows) and compared FLIR sensor fields of view at 12:06:33 (green) and 12:10:35 (blue).
Figure 5.28. Debris causing reflections during the Fort Hood flight trial on the Lynx FLIR imagery. A flash that was small, elongated and short in duration (left) and a flash that was larger and longer in duration (right). 5.5.7 Appearance of Flashes This section analyses the remaining flashes from the previous sections 5.5.3 and 5.5.6 in depth, by also taking into account the characteristics of the FLIR system and by comparing the flashes on the FLIR videotapes to the flashes seen on the imagery from the Fort Hood Flight trial, March 19, 2000. The remaining events were; event 8 at 11:24:3035; event 12 at 11:26:27; event 51 at 12:08:48; and event 52 at 12:08:51. The aim was also to investigate any similarities between the analyzed flashes on FLIR videotape Q-4 and the muzzle blasts flashes seen on the videotapes from the Fort Hood flight trials March 19, 2000. The following data was reviewed: the trial imagery, protocol and report, Nightstalker FLIR videotape (app. 2, item 16), Lynx FLIR videotapes (app. 2, items 17-18), photos (app. 2, items 37-41), flight trial protocol (app. 2, item 68) and Imagery Analysis report FLIR TRIAL Fort Hood, Texas 19 March 2000 in (app. 2 item 65). Noted was that (app. 2, items 17-18) were converted from PAL to NTSC, which have been considered in the investigation. The results from an initial visual review was that the muzzle blast flashes that could be detected on the Fort Hood FLIR videotapes were different in appearance than the solar reflections in the debris area from the same videotape. The detected muzzle blast flashes on the Fort Hood trial videos were also different in appearance compared to the flashes seen on FLIR videotape Q-4, one example is shown in figure 5.29. Moreover, the flashes on FLIR videotape Q-4 were very similar in appearance to the solar reflections in the debris area from both the Fort Hood Lynx and Nightstalker FLIR videotapes, illustrated in figure 5.28. Noted was the similarity in the flashes horizontally elongated shape. Pulsating solar reflections were detected on the Fort Hood Nightstalker videotape in the debris area. These pulsating solar reflections were similar to the pulsating flashes on FLIR videotape Q-4. Event 8 at 11:24:30 35, event 12 at 11:26:27 and the flashes from 12:08:48 and onwards, including event 57, on FLIR videotape Q-4 were compared to the pulsating solar specular reflection flashes recorded by the Fort Hood FBI Nightstalker. The visual review confirmed that events 8, 12, 51 and 52 on the FLIR videotape Q-4, were similar in appearance to the pulsating solar reflections in the debris area on the Fort

Figure 5.33. To the left is a ThermaCAM image taken at 12:14 p.m., where the specular solar reflection was captured. In the middle image, taken less than a minute later from another position at the about the same time, no visually detectable reflection was detected. To the right is a plot of the geometrical relation between the positions of the mirror, ThermaCAM camera and sun.
The solar specular reflection was located visually and captured by the ThermaCAM camera. Solar specular reflections were located and captured at 12:14 p.m. and could not be detected at any nearby position outside the distribution cone. The geometric relationship between the positions of the mirror, bearing to the sun, and the camera position were reconstructed, see figure 5.33. The positions of the sun and the camera position were reconstructed in the same manner as for the images from FLIR videotape Q-4. The results from the tests proved that the solar reflection could be detected in the far IR band. It further confirmed that solar reflections could be detected in the far IR band only within a limited distance from the reflection angle. Solar specular reflections were further located and captured at 12:17 p.m. and later at 1:23 p.m. The geometrical relationship between the position of the mirror, bearing to the sun and camera position were reconstructed, figure 5.34. This result verified that the reflection geometry varies with the position relative to the sun in the same way as for the analyzed flashes on FLIR videotape Q-4.
Figure 5.34. To the left is a ThermaCAM image taken at 12:17 p.m., in a position where the specular solar reflection was captured. To the right is a plot of the geometrical relation between the positions of the mirror, ThermaCAM camera and sun. 5.5.9 Results Events 6 and 42 were only normal thermal radiation from debris/objects set in motion by the CEVs and were omitted from the remaining of the analysis of the flashes. Based on the results from the visual reviews, the correlation of reconstructed data to the physical laws of solar specular reflection geometry, the fact that several long duration flashes fluctuate in intensity near their peak values, the lack of persons on the roofs and in other areas proximate to the flashes, and the results from the experiment, it is concluded that all of the analyzed flashes were caused by solar or heat reflections.

Figure 6.2. Helical scanning, from [Panasonic, Basic video guide] p. 7. During recording, the rotation of the drum is locked in speed and phase to the actual frame-rate of the supplied video signal. In case of NTSC, for which the frame rate is about 30 Hz, the drum will make 30 revolutions per second. Thus, one revolution will correspond to two fields. The drum is also angled so that half way around the drum corresponds to approximately the width of the videotape, as showed in figure 6.3. This angle between the drum and the videotape is called helix angle.
Figure 6.3. Helix angle, from [Panasonic, Basic video guide] p. 7. The resulting recording of the video signal on the tape will be diagonal tracks as shown in figure 6.4. Each track starts with a new field only if the drum is synchronized at a correct speed and phase during recording. Since the videotape wraps a little bit more than halfway around the rotating drum and the video heads are connected in parallel during recording, each track will end with the same information as the next track starts with. The recorded signal on these tracks is a Radio Frequency (RF) carrier that is frequency modulated (FM) by the luminance part of a video signal. Under normal circumstances there is also a chrominance part to record, and this chrominance signal is distributed as a carrier in the composite video signal. This chrominance carrier will be down converted by electronics inside the video machine from 3.58 MHz (NTSC) to 627 kHz. The converted signal will be used by recording circuits to amplitude modulate the luminance RF-signal.
Figure 6.4. Tapepath, from [Panasonic, Basic video guide] p. 9. In this case, when the video signal was generated by the FBI Nightstalker FLIR system, the FLIR sensor information was recorded as luminance signal. The frequency of the RF-carrier varies between 3.8 MHz and 4.8 MHz, where 3.8 MHz represent the lowest level (0 volt) and 4.8 MHz the highest level (1 volt) of the video signal that modulates the carrier. Black in an FLIR image will correspond to approximately 4.0 MHz. During recording, a Control signal (CTL) is created from the video signals vertical synchronization of the first field in a frame, and formed into a square wave signal, that is recorded into a longitudinal CTL-track into the lower edge of videotape, figure 6.4. During playback the CTL-pulse serves as a reference for the drum servo circuit, which will cause the video heads to match the recorded track on the videotape. The recording format of a CTL-track is similar to a normal audio track, but the CTL-track is recorded without a bias current. Thus, the recording of a square wave will end up as spikes. A typical playback CTL-pulse is illustrated in fig 6.5.

6.4.2 RF-envelope and Carrier Frequency for Recorded Video Signal The playback RF-envelope on the entire recorded part, for each of the FLIR videotapes, was observed using an oscilloscope. This aim was to determine if there were any changes or level fluctuations that could indicate an attempt to tamper with the recorded material. The lowest frequency of the video signals RF-carrier has been monitored with an oscilloscope to detect if there were abnormal changes in carrier frequency. Such a change would indicate the use of other recording devices. The lowest stable frequency available for examination occurs during vertical synchronization. By using an oscilloscope to capture several cycles of the RF-carrier it was possible to calculate the frequency, and also to exclude changes that would indicate a different video machine. The results from this inspection were:
FLIR videotape Q-1. Only normal variations have been detected. Carrier frequency is about 3.44 MHz. FLIR videotape Q-2. Only normal variations have been detected. Carrier frequency is about 3.43 MHz. FLIR videotape Q-3. Only normal variations have been detected. Carrier frequency is about 3.43 MHz. FLIR videotape Q-4. A fast disturbance in RF-envelope occurs at 10:47:16, according to the time information displayed in picture. Both the RF-envelope and carrier frequency remained similar, before and after the disturbance. The disturbance was identified as a stop during recording. Referring to the time information displayed in the picture there is an interrupt in recording that last between 10:47:16 and 10:51:57. Carrier frequency is about 3.44 MHz. FLIR videotape Q-5. Only normal variations have been detected. Carrier frequency is about 3.42 MHz. FLIR videotape Q-6. Only normal variations have been detected. Carrier frequency is about 3.31 MHz. FLIR videotape Q-7. Only normal variations have been detected. Carrier frequency is about 3.31 MHz.
The results from analyzing the RF-envelope and carrier frequency is that only normal variations in RF-envelope and carrier frequency recorded on FLIR videotapes Q-1 Q-7 were observed through the entire recorded portions of the FLIR videotapes. The visible break in recording that occurs on FLIR videotape Q-4 from 10:47:16 to 10:51:57 is with a disturbance in the RF-signal. Both the RF-envelope and carrier frequency signals remained similar before and after the disturbance. This indicates that Q-4 was made utilizing the same video machine. The break in recording that occurs on FLIR videotape Q-7 at 10:47:15 and lasts until 10:51:57 does not have a disturbance in the RF-signal. This indicates that Q-7 is a copy and that the recording was continuous during the copying process. 6.4.3 RF-envelope and Carrier Frequency for FM-audio Since the recorded sound on videotapes Q-1 Q-5 are FM-type and deeply modulated into the surface of the videotape, it is not possible to edit or erase this track without also tampering with the video

signal. Therefore, this part of the analysis was performed in conjunction with the analysis of the RFenvelope for the video signal. The carriers average center frequency of both channels has also been measured during playback. The results from this inspection were;
FLIR videotape Q-1. Only normal variations of the playback envelope have been detected. Carrier frequency: left channel = 1306.5 kHz, right channel = 1705.3 kHz. FLIR videotape Q-2. Only normal variations of the playback envelope have been detected. Carrier frequency: left channel =1305.3 kHz, right channel = 1704.8 kHz. FLIR videotape Q-3. Only normal variations of the playback envelope have been detected. Carrier frequency: left channel =1303.5 kHz, right channel = 1703.8 kHz. FLIR videotape Q-4. A disturbance in RF-envelopes occurs at 10:47:16. Both the RFenvelopes and carrier frequencies remained similar, before and after the disturbance. The disturbance was identified as a stop during recording. According to the time information displayed in picture the stop lasts between 10:47:16 and 10:51:57. Carrier frequency: left channel =1307.0 kHz, right channel = 1705.5 kHz. FLIR videotape Q-5. Only normal variations of the playback envelope have been detected. Carrier frequency: left channel =1306.0 kHz, right channel = 1705.0 kHz. FLIR videotape Q-6. No carriers were recorded. FLIR videotape Q-7. No carriers were recorded.
The conclusion after analyzing the FLIR videotapes Q 1 Q-5 is that the recorded audio FM carrier on FLIR videotapes Q-1 Q-5 are original recordings without any erased or edited parts. The visible break in the recording that occurs on FLIR videotape Q-4 at 10:47:16 and lasts until 10:51:57 has a simultaneous disturbance in the RF-signals. Both RF-envelopes and carrier frequencies signals remained similar before and after the disturbance. This indicates that a single video recorder made Q-4. Videotapes Q-6 and Q-7 were not recorded with deep modulated FM-carriers for hi-fi sound. 6.4.4 Dihedral Error Measurement It is possible to estimate the dihedral error from the time shift in the video signal that occurs at the head switch position. By measuring the time shift related to the specific recording and the playback video machine a time shift signature is produced. Variations in this signature could indicate if more than one

Waco Investigation: Image Analysis and Video Authentication 7.3 Conclusions
Finally, it is concluded with a confident level of certainty that all of the analyzed flashes seen on FLIR videotapes from the April 19, 1993, between 10:41 a.m. - 12:16 p.m. are caused by solar or heat reflections from single or multiple objects. The characteristics of the SPRITE/TED detectors and scanning mechanisms and the interlaced video format are factors that have contributed to distort the appearance of the flashes on the April 19, 1993, FLIR videotapes. Moreover, no humans were detected on the FLIR videotapes in any area in the vicinity of any of the flashes. Only moving debris were detected. The results from this investigation have shown, with a confident level of certainty, that the flashes on the FLIR videotapes from the April 19, 1993, between 10:41 a.m. - 12:16 p.m. cannot form evidence of gunfire. It is also concluded, with a confident level of certainty, that FLIR videotapes Q-1 Q-5, are original recordings. There is no sign that the recorded portions on FLIR videotapes Q-1 Q-5 have been edited or erased after they were produced. It is concluded, with a confident level of certainty that the interrupt on FLIR videotape Q-4, between 10:47:16 and 10:51:57, was not created afterwards but during the time of recording. It is also concluded, with a confident level of certainty, that FLIR videotapes Q-6 Q-7 are copies.

October 4, 2000

REFERENCES [AA] Astronomical Applications Dept. U.S. Naval Observatory, Washington, DC Internet: http://aa.usno.navy.mil/AA/data/docs/AltAz.html. [Baker] Baker I. M., Hastings M. P., Hipwood L. G., Jones C. L. and Knowles P., Infra-red Detectors for the year 2000, GEC REVIEW, Vol. 10, No. 31, p. 148-160, 1995. [Bear] Bear M. F., Connors B. W. and Paradiso M. A., Neuroscience, exploring the brain, Williams & Wilkins, 1996. [Cuthberthson] Cuthberthson G. M. Thermal Imaging, GEC REVIEW, Vol. 2, No. 1, p. 21-31, 1986. [Elliott] Elliott C. T., The SPRITE Detector. In: Institution of Electrical Engineering, Conference Publication Advanced Infrared Detectors and Systems, 29-30 October 1981, p 1-6, 1981. [Fraden] Fraden J., AIP Handbook of Modern Sensors. Physics, Design and Applications, AIP Press, New York, 1993. [Guichard] Guichard F. and Rudin L., Velocity Estimation from Image Sequence and Application to Super Resolution. In: IEEE Proceeding on International Conference on Image Processing, Vol. 3, p. 527-531, 1999. [HolstCCD] Holst C., CCD Arrays, Cameras and Displays, SPIE Press, 1998. [HolstIR] Holst C., Testing and Evaluating Infrared Imaging Systems, SPIE Press, 1998. [Klasn] Klasn L., Image Analysis. In: Proceedings of the 12th INTERPOL Forensic Science Symposium, October 20-23 1998, Lyon, France, p. 261-301, 1999. [Panasonic] Basic Video Guide, Panasonic Matsushita Electric Industrial Co., Ltd., unknown year for publication. [Philips] Camera tubes and accessories, Image Intensifiers, IR detectors, Data Handbook, Electronic Components and Materials, T10-05-83, Part 10, 1983. [Rudin] Rudin L. and Osher S., Total Variation Based Image Restoration with Free Local Constraints In: IEEE Proceeding on International Conference on Image Processing, Vol. 1, p 31-31, 1994.

Waco Investigation: Image Analysis and Video Authentication 50. 51. 52. Stake positions of Mount Carmel Compound generated by the FBI 6 days after the siege ended and a drawing of first floor plane revised 4/22/93. A 3D model of the complex, <Mount Carmel.DFX>. Six drawings of the complex external and interior views. Reports 53. 54. 55. 56. 57. 58. Allard E. F., Analysis of the April 19, 1993 Waco FLIR Videotapes, March 1, 2000. Cox M., Sun Reflection Geometry, technical report, revision 1, 22 March 1999. Johnson F., System Engineering and Laboratories Corporation, Mount Carmel Complex Structural Damage Evaluation. Prepared for Caddell & Chapman, Houston, Texas. Ziegel F., Radian Inc., Report of Ferdinand Ziegel, 29 February 2000. Zimmelman J. B., Report of Jack B. Zimmelman. Ertem M. C., Pierson R., Burchick D., Maryland Advanced Development Laboratory, Analysis of Flashes by an Infrared Sensor, report to US Department of Justice, February 28, 2000. Ertem M. C., Pierson R., Burchick D., Maryland Advanced Development Laboratory, Long Wave Infrared Muzzle Flash Duration Experiment, report to US Department of Justice, February 28, 2000. Ertem M. and Burchick D., Maryland Advanced Development Laboratory, Muzzle flash detection by Infra-red Cameras, report to US Department of Justice, February 28, 2000. Pierson R., Maryland Advanced Development Laboratory, notes prepared for Office of Special Counsel from IR Video review of Fire Development Analysis report, report to US Department of Justice, February 28, 2000. Burchick D., Maryland Advanced Development Laboratory, Potential for Solar Reflections, report to US Department of Justice, February 28, 2000. Ginsberg I. W., Technical Review of Allard Report, report to US Department of Justice, February 29, 2000. Corley G. W., Kosel H. C., Stejskal B. G., Construction Technology Laboratory Inc., Analysis of Structural and Egress Issues Relating to Events of April 19, 1993 at the Branch Davidian Compound, Waco, Texas, report to US Department of Justice, February 29, 2000. Oxley D. D., Evans N. M., Ayres P., Vector Data Systems (U.K.) Ltd., Imagery Analysis Report The Events at Waco Texas 19 April 1993, prepared for Office of Special Counsel.

60. 61.

62. 63. 64.
Waco Investigation: Image Analysis and Video Authentication Related Literature 66. 67. 68. A summary of NIIRS scale. Sea Owl Passive Identification Device, product specification GEC V5040, 1 page. Flight trial protocol. Protocol for a Forward Looking Infra-red Imagery Trial, prepared for Office of Special Counsel and the United States District Court for the Western District of Texas. Aircraft handbook. Duplicate of the Nightstalkers Airplane Flight Manual, requested from Office of Special Counsel. Audio transcripts from videotape Q-1, Q-2, Q-4 and Q-5. Network International Forensic Science Division, FLIR (Nightstalker) Tape Transcripts, prepared for The Office of Special Counsel. A summary of public statements made by experts regarding flashes at FLIR videotape Q-4. Miscellaneous 72. Climatic data for April 1993. U.S. Department of Commerce, Asheville, N. C., prepared for The Office of Special Counsel. Surface weather observations for March 18-19, 2000, Fort Hood, TX. Activity list from Vector Data Systems (U.K.) Ltd. An early and abbreviated version of the events by Vector Data Systems (U.K.) Ltd.