基于改进KCF的移动机器人视觉行人跟踪系统  

Visual Pedestrian-tracking System Based on Improved KCF with a Mobile Robot

作　　者：宋仁和 林名强 戴厚德 姚瀚晨 富巍 SONG Renhe;LIN Mingqiang;DAI Houde;YAO Hanchen;FU Wei(School of Electrical Engineering and Automation,Xiamen University of Technology,Xiamen 361024,China;Quanzhou Institute of Equipment Manufacturing Haixi Institutes,Chinese Academy of Sciences,Quanzhou 362216,China;School of Artificial Intelligence,University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]厦门理工学院电气工程与自动化学院,福建厦门361024 [2]中国科学院海西研究院泉州装备制造研究中心,福建泉州362216 [3]中国科学院大学人工智能学院,北京100049

出　　处：《控制工程》2024年第4期627-635,共9页Control Engineering of China

基　　金：富巍横向科研发展基金资助项目(200811061302);中国科学院国际伙伴计划对外合作重点项目(121835KYSB20190069)。

摘　　要：针对视觉跟踪机器人在跟踪目标被大面积遮挡和短暂丢失的场景下容易跟踪失败的问题,提出了一种基于改进核相关滤波(kernelized correlation filter,KCF)的移动机器人视觉行人跟踪系统。在传统KCF算法的基础上引入平均峰值相关能量(average peak correla-tion energy,APCE),结合APCE响应值和最大响应值判断目标丢失情况,并加入了两级重检测机制。在OTB2013数据集的目标丢失(out of view,OV)场景中,改进算法的成功率和精度分别达到0.626和0.592像素,比传统KCF算法分别提升了8.7%和10.9%,有效提高了算法在目标遮挡和短暂丢失场景下的跟踪鲁棒性。在有行人干扰的场景中进行跟踪实验,实验结果表明所提视觉行人跟踪系统能够稳定跟踪目标行人。For the problem that the tracking robot is prone to tracking failure when the tracking target is blocked by a large area and temporarily lost,a visual pedestrian-tracking system based on improved kernelized correlation filter(KCF)with a mobile robot is proposed.The average peak correlation energy(APCE)is introduced into the conventional KCF algorithm,the APCE response value and the maximum response value are combined to judge the target loss,and a two-stage redetection mechanism is added.In the out of view(OV)scenarios of OTB2013 dataset,the success rate and accuracy of the improved algorithm reach 0.626 and 0.592 pixels respectively,which are 8.7%and 10.9%higher than that of the conventional KCF algorithm,the tracking robustness of the algorithm in the scenarios of target occlusion and transient loss is effectively improved.The tracking experiment is carried out in the scenarios with pedestrian interference,and the experimental results show that the proposed visual pedestrian-tracking system can track the target pedestrian stably.

关 键 词：行人跟踪机器人 改进KCF 虚拟弹簧模型 平均峰值相关能量 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]