An improved mean-shift moving object detection and tracking algorithm based on segmentation and fusion mechanism

Abstract

The mean-shift moving object detection and tracking algorithm is an important technique for analyzing human motion. It is widely used in military defense, video surveillance, human-computer interaction, medical diagnostics as well as in commercial fields such as video games. However,the general mean-shift model does not perform well when dealing with serious occlusions. In this paper, an improved mean-shift moving object detection and tracking algorithm based on segmentation and fusion mechanism is proposed in order to address the occlusion problem. Firstly, the detection algorithm detects and extracts the target by processing a rectangular target input. Secondly, the mean-shift method of segmentation solves the sheltering problem. Finally, the fusion of weights of various segmentations is used to improve the tracking speed. Through fusion, several segment's information are integrated, which provides more space information. The experiments we carried out demonstrated that, the proposed algorithm not only improved the performance in sheltered or occluded cases, while not significantly increased the computation cost.