Real-time traffic vehicle tracking based on improved MoG background extraction and motion segmentation

A new method for real-time detection and tracking of multiple moving vehicles from traffic video is proposed. This method first uses MoG and texture based model to extract foreground from the scene, then detect moving targets using a modified version of timed motion history image (tMHI), and finally uses Kalman prediction filter to track these targets, which the full moving trajectories of the targets are obtained. Experiments on the real traffic scenes show that the method has good real-time performance and robustness against disturbance factors for outdoor traffic surveillance. Besides, it greatly improves the effects for detection in case of vehicle occlusion.