Kernel Tracking Approach based on Fuzzy Kernel Histogram

摘要

In kernel tracking, background pixels in object model will induce localization error of object tracking, so we present a kernel tracking approach based on fuzzy kernel histogram in this paper. At first, a fuzzy factor is used in kernel histogram to form the fuzzy kernel histogram and a fuzzy membership functions are designed to acquire the fuzzy factor according to the center-surround approach and the background histogram. Secondly, details about kernel tracking algorithm based on fuzzy kernel histogram is provided. After that, the methods of performance evaluation and parameter selection are proposed. At last, a mass of experimental results show our approach has good localization precision of object tracking.