Considering the issues of target occlusion and target scale changes encountered in UAV visual target tracking,and the complexity of the target tracking method that is strictly limited by the computing power of the UAV PTZ chip,a visual target tracking method based on UAV PTZ is proposed.Adaptive weight fusion based on multiple feature correlation filters is used to improve the accuracy of target position prediction.A scale change pool strategy with independent changes in length and width is introduced to solve the problem of target scale estimation in UAV target tracking.The occlusion status during the target tracking process is effectively detected and processed by setting up template detection and trajectory prediction modules.This method can a-chieve real-time performance in UAV PTZ chip,and achieved good performance in both public UAV target tracking datasets and self collected datasets.Compared with the baseline method,the method improves the suc-cess rate by 10.7％and improves the accuracy by 3％.
国家科技期刊平台
NSTL
万方数据
