A New Approach for Vision-based Rear Vehicle Tracking

A thorough research of dynamic target tracking problem in vision system on intelligent vehicle is conducted. A new approach to track of the rear part of a vehicle by a backward-looking video camera is put forward. Considered that the target´s relative speed is slow and the shape changes little, the vehicle fast tracking is obtained by locating the edge of target in tracking window. The tracking accuracy is improved by computing the target comparability feature to conduct self-evaluation of the location reliability, and the location deviation is emendated by template match algorithm. The stably tracking of vehicle target in complex background is achieved, and the tracking accuracy is higher than 90 percent. When the tracking window is shifting, the tracking can be resumed automatically by this algorithm.