Contour extraction and tracking of moving vehicles for traffic monitoring

Abstract

A robust algorithm is proposed in this paper for contour extraction and tracking of moving vehicles. A snake model is modified, in which directional information is utilized to guide the 'snaxel' behavior. Then an adaptive shape restriction is applied to govern the snake's motion, and a Kalman filter is employed to estimate the spatio-temporal relationship between successive frames. Moreover, several improvements are made to compensate for the snake's vulnerability to fake edges. All these steps contribute to a satisfying overall performance in contour extraction and tracking of moving vehicles for real traffic monitoring.