摘要
针对特定人跟踪过程中的遮挡、形变以及长时跟踪易导致跟踪丢失等问题,提出一种基于特定人步态信息的移动机器人跟踪方法。采用轻量级网络模型通过对步态特征识别进行特定人检测,并结合核相关滤波(KCF)跟踪框架增强跟踪目标;改进遮挡判断策略,自适应协调KCF与卡尔曼滤波算法(KF)对特定人位置的跟踪。选取RGB-D数据集作为实验数据集,将算法和其他4种主流算法通过仿真实验进行性能比较,证明方法的可行性,将方法移植到Turtlebot2移动平台进行实际环境特定人跟踪实验。通过实验对比,算法在光照变化、运动模糊和目标遮挡等情况下的表现具有较高的稳定性和准确性;在成功率和跟踪精确度上都优于其他4种算法。
In this study,aiming at the problems of occlusion,deformation and tracking loss caused by long-time tracking in the process of specific person tracking,a tracking method for mobile robot based on specific person gait detection is proposed.In this paper,a lightweight network model is used to detect the specific person through gait feature recognition,and kernel correlation filtering(KCF)tracking framework is used to enhance the tracking target;The occlusion judgment strategy is improved to adaptively coordinate KCF and Kalman filter(KF)to track a specific person′s position.Select RGB-D data set as the experimental data set,compare the performance of this algorithm with the four mainstream algorithms through simulation experiments,prove the feasibility of this method,transplant the method to the turnlebot2 mobile platform for real environment specific person tracking experiment.Experimental results show that the performance of the proposed algorithm is stable and accurate in the case of illumination change,motion blur and target occlusion;It is superior to the other four algorithms in success rate and tracking accuracy.
作者
张皓诚
王晓华
王文杰
Zhang Haocheng;Wang Xiaohua;Wang Wenjie(School of Electronics and Information,Xi'an Polytechnic University,Xi'an 710048,China)
出处
《国外电子测量技术》
北大核心
2022年第4期8-14,共7页
Foreign Electronic Measurement Technology
基金
国家自然科学基金(51905405)
陕西重点研发计划(2019ZDLGY01-08)项目资助
