The Research of Mean-Shift UAV Tracking Algorithm by TWH and FB-Error Restriction

Abstract

In view of the phenomenon that the factors of occluded target and it's changed background lead to failure in the process of tracking, a MTF (Mean-shift by TWH and FBerror) tracking algorithm is proposed. Firstly, TargetWeighted Histogram (TWH) is introduced to describe target in Mean-shift tracking framework, i.e., using local-background of target to weaken inner-background features of all-region in order to make target features prominent in tracking process, secondly, FB-error restriction is introduced, the predicted results of target current position and the calculated results of mean-shift vector are combined together to estimate the final target location of time t by using weighted function about FBerror. The experimental results show that the proposed tracking algorithm has great breakthrough on tracking accuracy.