摘要
针对小型移动机器人跟踪人体运动目标时,因光照变化或人体快速运动导致跟踪准确率降低甚至跟踪失败的问题,提出人体足部运动模型引导下的核相关滤波跟踪算法(kernel correlation filter guided by motion model,MMKCF).该算法通过建立足部运动模型预估视频跟踪中双脚位置,获得核相关滤波跟踪算法的目标检测区域,有效解决了跟踪框的漂移.基于外场实际拍摄的4组人体行走视频进行了目标跟踪实验,对比了MMKCF,KCF等5种算法的跟踪性能.实验结果表明,在光照变化以及目标快速运动时,MMKCF算法的平均跟踪准确率约为0.77,远高于其他4种跟踪算法.最后在机器人操作系统(robot operating system,ROS)下将MMKCF算法应用于TurtleBot机器人目标跟踪,成功地完成了人体快速移动时的足部跟踪,验证了所提算法具有较强的鲁棒性和实时性.
Aiming at the problem that the tracking precision of small mobile robot decreases or even fails due to the change of illumination or the fast motion of human body,a kernel correlation filter guided by motion model(MMKCF)is proposed.By building a feet motion model to predict the position of the feet in video tracking,the algorithm obtains the target detection area of the kernel correlation filter tracking algorithm,which effectively solves the drift of the tracking box.Based on four groups of human walking video,target tracking experiments are carried out,and tracking performance of five algorithms such as MMKCF and KCF are compared.The experimental results show that the average tracking precision of MMKCF algorithm is about 0.77 when the illumination changes and the target moves rapidly,which is much higher than the other four tracking algorithms.Finally,the MMKCF algorithm is applied to the target tracking of TurtleBot robot under the robot operating system(ROS),and the foot tracking of human body during fast motion is successfully completed,which proves that the proposed algorithm has strong robustness and real-time performance.
作者
陈丹
姚伯羽
吴欣
Chen Dan;Yao Boyu;Wu Xin(School of Automation and Information Engineering,Xi’an University of Technology,Xi’an 710048)
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2020年第12期1967-1975,共9页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(61671375)
榆林市科技计划(2019-146)
西安理工大学研究生竞赛培育项目(252051834).
关键词
移动机器人
目标跟踪
运动模型
核相关滤波
机器人操作系统
mobile robot
target tracking
motion model
kernel correlation filter
robot operating system(ROS)
