An adaptive mean shift particle filter for moving objects tracking

Abstract

In this paper, we integrate mean shift algorithm into the particle filter as the proposal distribution, and adaptively modify the tracking step according to the process of mean shift algorithm. It can be used through the searching for the most matching region with the feature, such as color, texture, etc, which should first be drawn from the targets. Experiments show that it is an efficient and powerful method for tracking fast moving objects in clutter even with occlusions.