Face Based Browsing for Surveillance Applications
Surveillance applications often collect a large amount of video data. Currently the surveillance applications do not allow the user to quickly search the collected data for an occurrence of a particular individual. This project gives the user this capability by displaying a pallet filled with the faces of all the individuals observed by a particular camera during a particular time. The user can then navigate to the video segments that contain a particular person. It is further possible to arrange the people in order of similarity with respect to a particular individual. This is an application of the Video Warehouse project.
Background & Objective: Surveillance applications are meant to help users understand "the Four W's" (Who, What, Where and When) of a situation. Current surveillance applications only allow search on the last two of these, the where and when. This makes browsing surveillance data a very tedious and time consuming process. By adding the "Who" component to the searching capability of a surveillance application, we hope to relive the tedium of the process and make it faster.
Technical Discussion: The browsing application leverages face detection, object tracking and face recognition technologies in order to provide the user with a most concise list of faces possible. Efficient database techniques have been utilized to make searching through image metadata as fast as possible.
Technology Area: Computer Vision
Modification Date: January 23, 2007