An Improved Mean Shift Algorithm for Target Tracking

摘要

Mean Shift algorithm is a non-parametric kernel density estimation based on the color histogram. It is a traditional algorithm of target tracking and is wildly used in the video monitoring for its simple calculation and lower time complexity. But it is inapplicable to the target that moves fast. In view of the fact, we propose a novel algorithm combining the bandwidth trial with the traditional algorithm in this paper. It is utilized to solve the problem of fast moving target tracking by comparing the Bhattacharyya coefficients to change the bandwidth of the kernel function automatically. Results of experiments indicate that the proposed algorithm not only can track the target more accurately, but also reduce the number of iteration.