上海交通大学学报(自然版) ›› 2015, Vol. 49 ›› Issue (08): 1213-1219.

• 自动化技术、计算机技术 • 上一篇    下一篇

人机交互中的人体目标跟踪算法

张铁,马琼雄   

  1. (华南理工大学 机械与汽车工程学院, 广州 510641)
  • 收稿日期:2014-06-30 出版日期:2015-08-31 发布日期:2015-08-31
  • 基金资助:

    广东省产学研项目(2012B090600028),广东省科技计划项目(2011A091101001,2012B010900076),中山市科技计划项目(201207A002)资助

Human Object Tracking Algorithm for Human-Robot Interaction

ZHANG Tie,MA Qiongxiong   

  1. (School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou 510641, China)
  • Received:2014-06-30 Online:2015-08-31 Published:2015-08-31

摘要:

摘要:  针对机器人在与人交互过程中对指定人体目标的跟踪容易受到周围其他人体干扰的问题,提出了一种人机交互中的人体目标跟踪算法.将所有干扰区域看作候选目标,通过建立基于重叠率的粒子分布模型,确保粒子集可以通过均值偏移收敛到所有的候选目标,并减少粒子数量.以权重距离总误差和目标尺寸作为聚类条件,将粒子划分到相应的候选目标粒子集中,最后选择最优的候选目标作为跟踪结果.实验结果表明:该算法能够避免周围相似物体的干扰并准确跟踪目标,具有较好的鲁棒性和实时性.

关键词: 粒子滤波, 均值偏移, 粒子聚类, 目标跟踪, 人机交互

Abstract:

Abstract: Proposed in this paper is a human object tracking algorithm for HRI (Humanrobot interaction) as a solution to the interference by similar human objects when a robot is tracking a specified human object. In the proposed algorithm, all interference regions are regarded as candidate targets, and a particle distribution model based on overlapping rate is established to make particles converge to all candidate targets via a mean shift with a reduced number. Then, by taking the total error of weight distance and the target size as the clustering conditions, particles are divided into corresponding candidate target particles, and the best candidate target is selected as the tracking result. The experimental results show that the proposed algorithm can track targets accurately in the presence of surrounding similar objects, and  it possesses strong robustness and good realtime performance.

Key words: particle filter, mean shift, particle clustering, target tracking, humanrobot interaction

中图分类号: