Focus Area: Digital and Multimedia Forensics
Description: The vehicle image search tool performs searches, matches, and exclusions of vehicle images in video and image files. The application provides the capability for post-event forensic analysis of video and digital image depictions with an enhanced feature to automatically extract and classify vehicles from video streams. The tool identifies the make and model of the vehicle in a still image or a video clip to narrow the field of candidates that must be searched in an investigation. It can associate the same vehicle across two different multimedia files. The application performs 1:N searches of surveillance quality images of a vehicle against an enrolled set of images. The user has the capability to highlight or delineate (“sketch”) key features of the vehicle and the angle of view for boosted searching (e.g., rear window shape, vehicle outline and approximate angle of view will allow the system to consider off-angle views). For high quality video streams, the algorithms and the application are capable of automatically sorting and clustering all vehicles presented to the system.
Project Cost: $969,265
Project Duration: 18 months
Operational Impact: Much larger amount of evidence and information about vehicles in video and image files is available.
Deliverable: A software application/tool that searches video and image files to locate and identify vehicles and then determine the make and model and make individual matches when possible.
Performer: SRI International
End Users: Federal Bureau of Investigation, Intelligence Community
Transition: Disseminated to two federal agencies and commercially available from the SRI International.