Mean shift based orientation and location tracking of targets

Abstract

The mean shift based tracking algorithm has achieved considerable success in tracking targets' location due to its simplicity and robustness. It finds local minima of a similarity measure between the color histograms or kernel density estimates of the model and target image. However, it can't track the targets' orientation. This paper proposed a novel mean shift based tracking method that can track both location and orientation of targets. It was realized by proposing an orientation tracking method that utilized the probability density distribution of the target gradient angle as the feature and constructed a similarity function that can be optimized by mean shift method, thus orientation tracking was transformed into an optimization problem. Thanks to the fast convergence of mean shift, this method can be run in real-time. A complete tracking method was constructed by using alternate iteration of the orientation tracking and Meer's location tracking algorithm.