Target Tracking Based on Improved Kernel Correlation Filter

Abstract

In response to the shortcomings of using a single feature to describe targets and the limitations of target scale invariance in traditional kernel correlation filter tracking algorithms, this paper proposes a scale adaptive kernel correlation filter algorithm that integrates color attribute features. Using principal component analysis to obtain color attribute features with strong discrimination, in order to reduce computational complexity and achieve color adaptation, and designing scale adaptive filters to dynamically adjust the target scale. Compared with the original kernel correlation filter and its improved algorithm, this method has good adaptability in situations such as occlusion, deformation, rotation, and background cluster interferences.