Target Detection and Tracking Algorithm based on the Improved Multi-feature Camshift

Abstract

In this paper, a vision-based target detection and tracking approach using improved multi-feature camshift is presented. Because the traditional camshift based on single feature is not robust when the background color, illuminate and target deformation change, it may lead to losing target or failure of tracking. In order to improve this problem, we presented target detection and tracking algorithm based on improved multi-feature camshift. The foreground image was obtained by Gaussian Mixture Model (GMM), which was used to modify the kernel function of target model. Due to the advantage of colour and texture feature in describing the target appearance, these features were used as recognition features. Additionally, EKF was integrated in tracking system to improve the accuracy of target tracking and predict the object position when the target was occluded. This paper details the architecture of the presented method and gives some experimental results to verify the effectiveness of the proposed method.