Tracking Algorithm Based on Attention Mechanism and Template Update

We propose a tracking algorithm based on attention mechanisms and template updating to solve problems like target deformation, occlusion, and poor tracking performance in complex background conditions caused by the Fully Convolutional Siamese Network's (SiamFC) simple structure. Introducing channel attention mechanisms directs the neural network's focus towards crucial channel information, thereby enhancing the network's discriminative ability. Utilizing a template updating strategy dynamically refreshes the template, improving the model's resilience against drift and refining the network's precision in localization and adaptability. Experimental results on the OTB100 target tracking dataset validate the efficacy of the proposed algorithm. Compared to the baseline SiamFC algorithm, the improved algorithm demonstrates a 6.1% increase in success rate and a 9.2% increase in precision rate.