A robust algorithm of contour extraction for vehicle tracking

Contour extraction of moving vehicle is an important and challenging issue in traffic surveillance. In this paper, a robust algorithm is proposed for contour extraction and moving vehicle tracking. First, we establish a modified snake model and utilize the directional information of the edge map to guide the snaxels' behavior. Then an adaptive shape restriction is embedded into the algorithm to govern the scope of the snake's motion, and Kalman filter is employed to estimate spatio-temporal relationship between successive frames. In addition, multiple refinements are suggested to compensate for the snake's vulnerability to fake edges. All of them contribute to a robust overall performance in contour extraction and vehicle tracking. Experimental results in real traffic scene prove the effectiveness of our algorithm. The comparison with conventional snakes is also provided.