An effective object tracking based on spatio-temporal context learning and Hog

This paper proposes an improved object tracking approach based on spatial-temporal context and hog descriptor of image to improve the accuracy and real-time of object tracking. Hog is an effective feature to represent the image in visual tracking. We extract the hog feature instead of raw pixel. In order to fully use the information of background, the object tracking can be regarded as the spatio-temporal model. A confidence map is found by computing the spatio-temporal model. The target location is decided by likelihood function. Experimental results show that the proposed method outperforms favorably against others tracking approach based on kernel method in many complex conditions.