TLD Tracking Algorithm Based on Feature Fusion

Various places in life are related to target tracking. In order to ensure that the tracked objects can be correctly identified, the accuracy of the algorithm needs to be continuously improved. This article focuses on the problem that the recognition rate of TLD (Tracking-Learning-Detection) algorithm decreases and even tracking drift occurs when the target undergoes rapid movement, rotation, and deformation. These situations are not expected to happen. Therefore, this paper proposes to add HSV color space and local binary pattern (LBP) to the traditional TLD algorithm. Through this method, the recognition accuracy of the tracked target is improved when the target is moving, rotating, and deforming rapidly, thereby improving the robustness of the algorithm. After a large number of experiments and analysis in this paper, this method can achieve correct detection and accurate tracking of target objects in complex scenarios.