A self-adaptive edge matching method based on mean shift and its application in video tracking

Abstract

A self-adaptive edge matching method based on mean shift adjustment is proposed in this paper. Such method uses the local mode seeking character of mean shift to adjust the edge information of each model to a stable state before matching, which can effectively avoid the deviation problem of traditional method and raise the successful matching rate. Furthermore, the interfering vector with a self-adaptive coefficient is proposed to optimise the matching performance in complex background. Compared with a pre-set constant coefficient, the self-adapted coefficient has a better perception of background edge complexity so as to control the initial adjusting position more rationally, and thus increases the robustness and accuracy of matching. This matching method is applied in an improved particle filtering tracking framework, and experimental results prove the validity and rationality of the theoretical analysis, and show that the proposed matching method performs a robust and efficient tracking.