Publication White Paper:
Extracting Physical Evidence from Digital Photographs for use in Forensic Accident Reconstruction
David Danaher, P.E., Jeff Ball, Ph.D., P.E., and Mark Kittel, P.E., D.F.E.
Typically accident scene investigators or law enforcement officers will document evidence which they consider to be “significant events”, such as the beginning of a tire mark, the first gouge mark or the rest position of the vehicles. Although this information gives the forensic engineer a snap shot of the events during the accident, it may not give the detail required to do a thorough analysis. The evidence and information located between the major events can provide further details as to the speed and dynamic motion of the accident vehicles that may be necessary for a complete analysis.
To determine the position of the vehicle between the points documented by law enforcement, a forensic engineer can often use the photographs taken at an accident scene which show the physical evidence such as the tire marks or roadway gouges. To determine the location of the physical evidence depicted in the accident scene photographs several methods can be used such as camera matching, photogrammetry, and photo rectification. In order to accurately place the evidence, basic dimensions of the roadway and surroundings should be determined from measurements taken by the officers at the scene or by subsequent inspection. To determine the exact roadway or median geometry, a laser survey can be utilized by either a forensic engineer familiar with the equipment or an independent surveying company.
Once the available data is collected, camera matching, photogrammetry, or rectification can be performed. The following demonstrates each of those three processes.
Camera Matching
Camera matching involves the use of the accident scene photographs depicting various points of evidence at the scene of the accident. Camera matching begins with a close review of the available photographs to determine the extents of the accident site which must be surveyed. The next step is to perform an accurate 3-dimensional survey of the accident site to document reference features which are depicted in the photographs. The survey data is then used to create a three dimensional model of the roadway surface using CAD software. The survey data is then combined with the accident scene photographs, both of which are imported into a three dimensional software package, such as 3D Studio Max. Using camera matching techniques, a virtual camera can be positioned relative to the 3D roadway surface with the same specifications and orientation as the camera used by the investigator. Once the camera is properly positioned, the physical evidence can then be mapped from the photograph onto the 3D roadway. Data obtained in this manner can then be used in the creation of a two or three dimensional accident scene drawing.
Photogrammetry
Another method of documenting physical evidence is through the use of photogrammetry. Photogrammetry is a technique that determines the three-dimensional geometry of object on the accident scene from two dimensional photographs. The three dimensional coordinates of the objects in the photographs are determined after the virtual camera is positioned in the virtual space. With the virtual camera properly positioned, the specifications of the camera and the vanishing point in the photographs can be determined.
In some cases, the only evidence available may be photographs. Even though the vehicles or tire marks may no longer be available, the photographs can be used to extract “lost” evidence. For example, a vehicle involved in a rollover may have been crushed or salvaged years before the engineer’s involvement, but the deformation to the vehicle is still crucial to the investigation. To evaluate the damage, photogrammetry can be used on the available photographs to quantify the extent of the damage. With sufficient photographs taken from several positions around the vehicle and knowing certain dimensions of the vehicle, such as the wheel base, a scaled three dimensional model of the crushed vehicle can be produced.
Photographs from several viewpoints can be imported into software such as PhotoModeler and then the forensic engineer can select points common in each photograph. After the common points are selected on the photographs and the camera specifications are entered into PhotoModeler, then the software, which is based on known principles of optics and photogrammetry, can calculate the location of each selected point in a three dimensional coordinate system. The data can then be exported to a 3D software program to create an accurate scaled model of the damaged vehicle. The forensic engineer can then use the model of the crushed vehicle to aid in determining vehicle positions, speeds, and intrusion into the occupant compartment, as a result of the roll sequence.
Photographic Rectification
Photographic rectification is another tool the forensic engineers have at their disposal to analyze physical evidence that may not have been measured by the investigating officer. Two dimensional (2D) rectification is the process of transforming a single photograph which is oblique to a planar surface into an orthographic image or a top-down view. This simplified form of photogrammetry is applicable only to photographs in which evidence is located on a relatively flat, planar surface such as a roadway (a typical occurrence in vehicle accident investigation).
A computer program such as PC-Rect can be used to import and rectify a digital photograph or digital scan of a photograph. The process involves the forensic engineer defining and locating known roadway dimensions in the photograph, such as lane line spacing, lane widths, etc. The program uses these dimensions to calculate the position, orientation, and specifications of the camera. If some or all of the information about the position of the camera or specifications are known to the engineer, the data can be entered into the software for increased accuracy.
The rectification process itself can be visualized as the reverse of the photographic process. When the photograph is taken, photons of light are projected from the road surface, through the camera lens and onto the image plane (the film). For the 2D rectification, it is assumed that the point on the road, the focal point of the camera and the point on the image plane are collinear. To rectify the photograph, the “light” is projected from the camera position (the focal point), through the image plane (the photograph, located a distance of the lens focal length times the image magnification factor from the focal point) and onto a planar surface. The resulting bitmap image has the appearance of taking the photograph and stretching it onto the roadway. With the proper definition of the area to be rectified and good reference dimensions in the photograph, high accuracy can be achieved, resulting in scale images in which measurements of evidence important to the accident investigation can be taken.
Conclusion
Overlooked or undocumented evidence can be retrieved and quantified as long as photographs of such evidence are available. Using photographs of the accident scene or of the vehicle, “lost” evidence can be accurately determined using several scientifically based methods, such as camera matching, photogrammetry, and rectification.
The accuracy of the photogrammetry methods are a function of the quality of the photographs, the available dimensional data, and skills of the forensic engineer. The more information available, documented, and collected by the forensic engineer, the greater the potential accuracy of the analysis.
The available techniques to retrieve the evidence such as camera matching, photogrammetry, and rectification are well-accepted and published scientific methods and have been accepted by State and Federal courts and have successfully passed Daubert challenges.
References
Massa, David J. “Using Computer Reverse Projection Photogrammetry to Analyze an Animation.” Society of Automotive Engineers (SAE) paper 1999-01-0093 (1999).
Cliff, William E., Duane D. MacInnis, and David A. Switzer. “An Evaluation of Rectified Bitmap 2D Photogrammetry with PC-Rect.” Society of Automotive Engineers (SAE) paper 970952 (1997).
Neale, William, T.C., Steve Fenton, Scott McFadden and Nathan A. Rose. “A Video Tracking Technique to Survey Roadways for Accident Reconstruction.” Society of Automotive Engineers (SAE) paper 2004-01-1221 (2004).
Fenton, Stephen, and Richard Kerr. “Accident Scene Diagramming Using New Photogrammetric Technique.” Society of Automotive Engineers (SAE) paper 970944 (1997).
Grimes, Wesley D. “Computer Animation Techniques for Use in Collision Reconstruction.” Society of Automotive Engineers (SAE) paper 920755 (1992).
Campbell, A. T. III and Richard L. Friedrich. “Adapting Three-Dimensional Animation Software for Photogrammetry Calculations.” Society of Automotive Engineers (SAE) paper 930904 (1993).
Grimes, Wesley D. “Classifying the Elements in a Scientific Animation.” Society of Automotive Engineers (SAE) paper 940919 (1994).
Grimes, Wesley D., Charles P Dickerson and Corbett D. Smith. “Documenting Scientific Visualizations and Computer Animations Used in Collision Reconstruction Presentations.” Society of Automotive Engineers (SAE) paper 980018 (1998).
Bohan, Thomas L. and April A. Yergin. ‘Computer-Generated Trial Exhibits: A Post-Daubert Update. Society of Automotive Engineers (SAE) paper 1999-01-0101. (1999).
Pepe, Michael D., James S. Sobek, D. Allen Zimmerman. “Accuracy of Three-Dimensional Photogrammetry as Established by Controlled Field Tests.” Society of Automotive Engineers (SAE) paper 930662. (1993).
Tumbas, Nicholas S., J. Rolly Kinney, and Gregory C. Smith. “Photogrammetry and Accident Reconstruction: Experimental Results.” Society of Automotive Engineers (SAE) paper 940925 (1994).
Rentschler, Walter and Volker Uffenkamp. “Digital Photogrammetry in Analysis of Crash Tests.” Society of Automotive Engineers (SAE) paper 1999-01-0081 (1999).
Smith, Gregory C., and Douglas L. Allsop. “A Case Comparison of Single-Image Photogrammetry Methods.” Society of Automotive Engineers (SAE) paper 890737 (1989).
Switzer, David A. and Trevor M. Candrlic. “Factors Affecting the Accuracy of Nonmetric Analytical 3-D Photogrammetry Using Photomodeler.” Society of Automotive Engineers (SAE) paper 1999-01-0451 (1999).
Pepe, Michael D., James S. SObek and Gary J. Huett. “Three-dimensional Computerized Photogrammetry and its Application to Accident Reconstruction.” Society of Automotive Engineers (SAE) paper 890739 (1989).
Pepe, Michael D., Eric Grayson and Andrew McClary. “Digital Rectification of Reconstruction Photographs.” Society of Automotive Engineers (SAE) paper 961049 (1996).
Kullgren, A., A Lie, and C. Tingvall. “The Use of Photogrammetry and Video Films in the Evaluation of Passenger Compartment Measurement and Occupant-Vehicle Contacts.” Society of Automotive Engineers (SAE) paper 950239 (1995).
Main, Bruce W., Eric A. Knopf. “A New Application of Camera Reverse Projection in Reconstructing Old Accidents.” Society of Automotive Engineers (SAE) paper 950357 (1995).
Wester-Ebbinghaus, Wilifried and Ulrich E. Wezel. “Photogrammetric Deformation Measurement of Crash Vehicles.” Society of Automotive Engineers (SAE) paper 860207 (1986).