基于高置信度综合判据的抗遮挡核相关滤波跟踪方法  

Anti-occlusion Kernel Correlation Filter Tracking Method Based on High Confidence Comprehensive Criterion

作　　者：袁振寰 娄静涛 李永乐 苏致远 徐友春 YUAN Zhenhuan;LOU Jingtao;LI Yongle;SU Zhiyuan;XU Youchun(Army Military Transportation University,Tianjin 300161,China)

机构地区：[1]陆军军事交通学院,天津300161

出　　处：《军事交通学报》2022年第7期80-88,共9页Journal of Military Transportation University

基　　金：军队重点学科专业建设项目

摘　　要：为提高复杂场景中目标跟踪的精度与鲁棒性,解决由于遮挡导致的目标跟踪失败的问题,提出一种基于高置信度综合判据的核相关滤波跟踪方法。首先,分析目标跟踪过程中的置信度,提出一种高置信度综合判据,实现对目标跟踪状态的判断;其次,在相关滤波位置估计模块引入三次指数平滑模型作为目标运动模型,在遮挡发生时,对相关滤波器给出的目标位置进行补偿,实现目标重定位,有效提高对目标的判别能力;最后,利用高置信度综合判据结果,实现自适应更新模板。实验结果表明,在OTB100数据集上,本算法能有效降低遮挡等因素对目标跟踪的影响,对测试视频序列取得较高的跟踪精度和成功率,整体性能优于对比算法。In order to improve the accuracy and robustness of target tracking in complex scenes and solve the problem of target tracking failure caused by occlusion,a kernel correlation filter tracking method based on high confidence synthesis criterion is proposed in this paper.Firstly,it analyzes the confidence in the process of target tracking,and proposes a high confidence comprehensive criterion to judge the target tracking state.Then,it introduces the cubic exponential smoothing model in the correlation filter position estimation module,in which the target position given by the correlation filter is compensated to realize the target relocation and effectively improve the ability to distinguish the target when occlusion occurs.Finally,it realizes adaptive update template by using the results of high confidence synthesis criterion.The Experimental results show that this algorithm can effectively reduce the influence of occlusion and other factors on target tracking on OTB100 data set and achieve high tracking accuracy and success rate for test video sequence,and the overall performance is better than the comparison algorithm.

关 键 词：三次指数平滑模型 抗遮挡 核相关滤波 目标跟踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN713[自动化与计算机技术—计算机科学与技术]