Image Tracking Algorithm Improvement Based on TLD Frame

Image target tracking is an important part in computer vision field, and has a broad application prospect in modern society. Lack of online learning function, the traditional target tracking algorithm was based on the single track as long as possible. The target deformation and occlusion will lead to target tracking lost. TLD (tracking-learningdetection) is the long-term effective algorithm in tracking single target proposed. The algorithm requires less prior knowledge and can achieve long-term tracking, but when target encounter obstacles, there will be drift phenomenon in the short term. In order to overcome this shortcoming, the paper improve the tracking module of TLD. A method is proposed by using Harris feature point instead of the original Grid sampling method. The experimental results show that this method can effectively suppress the tracking drift, and improve tracking precision.