Improved target tracking method based on ECO-HC

Target tracking methods based on correlation filtering often face challenges such as tracking accuracy degradation or even failure due to interference factors like fast movement, occlusion, and scale changes during the tracking process. To address these issues, this paper proposes an improved target tracking method based on Efficient Convolution Operators with hand-crafted features (ECO-HC). Firstly, a multi-scale fusion strategy is introduced into the target scale estimation part of the ECO-HC algorithm. The proposed multi-scale fusion strategy considers target information from multiple scales simultaneously, achieving more accurate target scale estimation by fusing feature response maps from different scales. Additionally, an adaptive model updating strategy is proposed, which dynamically adjusts the frequency and method of model updates based on the actual situation during the target tracking process, thereby adapting to changes in the target's appearance and shape. Experiments conducted on the OTB100 dataset show that the proposed method significantly improves tracking accuracy, success rate, and robustness compared to the original ECO-HC algorithm.