Real-time TLD Target Tracking Algorithm Based on Improved Kernel Correlation Filtering for Maritime Targets

Aiming at the problem that the traditional KCF algorithm cannot adapt to the long-term target tracking in a complex offshore environment, TLD is a combination of track detection learning module tracking algorithms, which can carry out long-term target tracking, but its real-time performance needs to be improved. Based on the framework of the TLD target tracking algorithm, the tracking module adopts the improved multi-feature fusion and scale adaptive KCF algorithm, adaptive Kalman filter algorithm is used to predict the target position of the detection module, and the detection area is reduced. The detection speed of the algorithm is improved, and the detection results of occlusion detection and position estimation are used to modify the tracking results. In order to improve the accuracy of the tracker, an improved filtering template updating strategy is proposed. To improve the accuracy of the tracker, a new filtering template updating strategy is proposed. Then the accuracy and success rate of the proposed algorithm are verified on OTB50 dataset. Finally, a practical tracking experiment is carried out. By comparing the improved algorithm with KCF algorithm and DSST algorithm, the effectiveness of the proposed algorithm in complex Marine environment is verified.