Multiple Vehicle Visual Tracking from a Moving Vehicle

In order to realize collision warning and avoidance, a robust and real-time dynamic vehicle detection and tracking system regardless of occlusion is required for driver assistant system (DAS). In this paper, a novel classify vehicle tracking method is presented for the DAS, which can be applied to general road condition and the tracking stability under swerve and occlusion conditions is well resolved. With the thought of the intelligence tracking device, the targets are classified into three classes. Edge feature localization combined with template match is used for normal state target tracking, the tracking window drifting problem is solved and tracking precision is improved by the template update mechanism realized with target feature comparability. Multiple vehicle targets tracking under swerve and occlusion conditions in complex traffic circumstance is realized by the reformed template match tracking algorithm with cutting pieces template and reverse template