Improvement of KCF Algorithm for Adaptive Model Updating

To address the shortcomings of the KCF algorithm in complex scenarios, we propose an improved target tracking algorithm that incorporates a multi-feature fusion approach, a scale-adaptive change mechanism, a multi-peak detection mechanism, and an adaptive model update strategy. The multi-peak re-detection mechanism is employed to mitigate drift caused by similar targets, thereby enhancing the accuracy of target localization. Additionally, a highly stable adaptive model update strategy is utilized to dynamically adjust the model update rate, effectively preventing erroneous updates. Experimental validation on the OTB50 dataset demonstrates that the improved algorithm achieves a tracking accuracy and success rate that are 11.3% and 17.1 % higher, respectively, than those of the traditional KCF algorithm, exhibiting better robustness and significant practical application value.