ECO-HC Based Tracking for Ground Moving Target Using Single UAV

摘要

To overcome the problems of target missing and serious target occlusion in ground target tracking with UAV, this paper introduces the tracking confidence and detect module on the basis of ECO-HC. We use the Peak to Sidelobe Ratio(PSR) to express the track confidence. The YOLOv3 is taken as the detect module, the position and feature of detecting results are used to match with the target. The algorithm descreases the frequence of calling YOLOv3 by PSR, and reduces the influence of YOLOv3 to real-time through the way of offline training to enhance the speed. At the same time the accuracy of detecting is guranteed with the target relevance matching. The results show the algorithm can find the target-lost back and has obvious improvement in tracking accuracy.