Robust Tracking of Dynamic Visual Object with Complex Environmental Factors
Jiahe Zhang, Bingxiang Wang, Jinyuan Niu · 2024
Aiming at a series of problems caused by complex environment changes, such as target tracking instability, target retracking deviation after occluding or disappearing and high similarity interference, a target tracking algorithm based on TLD framework is proposed in this paper. By using CLAHE(Contrast Limited Adaptive Histogram Equalization) image enhancement technology to preprocess the image, the color contrast of different pixels can be enlarged to increase the image recognition ability. To ensure the accuracy and real-time performance of the model, Kalman filter is used to predict the possible position of the target when it is blocked or disappeared, and detect and track the target immediately after it reappears. Then, the Median-Flow optical flow tracker with CSRT assisted TLD algorithm makes use of its dual reliability mechanism of channel and space. As the result, the target tracking task can be fulfilled effectively and stably. Experimental results show that the proposed algorithm can achieve high tracking accuracy and robustness under complex environment.