Improved TLD Target Algorithm Based On Feature Fusion

A modified TLD tracking algorithm based on fusion of color model and ORB feature is proposed for the poor tracking performance of TLD algorithm for rotating target. The algorithm uses a feature fusion tracking algorithm instead of the optical flow method in the TLD tracking module, updating target tracking window and target template in the tracking module to improve the target tracking capability of rotating target, and compare the Bhattacharyya coefficient of the target model and the candidate model of with threshold to determine whether the tracking is failure or success, the detection module will reinitialize tracking if the tracking module failure. The experimental results show that, compared with the original TLD tracking algorithm, the improved algorithm could improve the ability of the rotating target tracking, and can meet the real-time requirements.