Research For Target Tracking Based On Sparse Representation Of Local Sensing

The traditional tracking method based on sparse representation was led to poor real-time tracking when dealing with complex video sequences and may occur the phenomenon of tracking drifting in the tracking process and a novel approach that object tracking algorithm based on sparse representation of local sensing was proposed.The object template was divided into several sub-regions,these sub-regions will become trivial templates to matching,using BOMP algorithm to solve sparse sparse representation;the observation model was established by combined appearance model to predict the object motion state of next frame;introduced a new template correction mechanism,it has overcome the phenomenon of tracking drifting effectively.The proposed algorithm was compared with the target tracking algorithm in the recognized standard library based on different video sequences.The experimental results demonstrate that the proposed tracking algorithm has ideal effect of tracking,and it has well real-time tracking compared to traditional tracking method based on sparse representation when the multiple object tracking algorithm were tested under different video sequences.