High Confidence Tracking with Offline Historical Learning and Online Correlation Filter Updating
Target tracking is one of the most challenging tasks in computer vision.In this paper,the high confidence tracking(HCT)algorithm is proposed by combining the offline historical learning network with online correlation filter updating model.First,the weighted historical targets are introduced into the offline learning network,which solves the problem of target loss caused by inaccurate tracking of the previous frame.Second,the targets confidence detection mechanism is proposed,and added to the correlation filter tracking algorithm,so that the model drift is avoided.Finally,we form a new high confidence tracking algorithm with offline learning.Compared with the state-of-the-art tracking algorithm,our algorithm performs outstandingly on benchmark OTB13 and OTB15,while ensuring real-time performance.
High confidence Target tracking Offline learning Online updating
Shou-dong HAN Hong-wei WANG Xin-xin XIA
Key Laboratory of Image Processing and Intelligent Control,Ministry of Education,School of Artificia Key Laboratory of Image Processing and Intelligent Control,Ministry of Education,School of Artificia
国际会议
武汉
英文
6-11
2020-01-12（万方平台首次上网日期，不代表论文的发表时间）