系统仿真学报 ›› 2016, Vol. 28 ›› Issue (6): 1281-1288.

• 仿真建模理论与方法 • 上一篇    下一篇

基于粒子滤波节点选择的协同跟踪

陈莹, 程雪雪   

  1. 江南大学物联网工程学院,江苏 无锡 214000
  • 收稿日期:2015-01-05 修回日期:2015-04-02 出版日期:2016-06-08 发布日期:2020-06-08
  • 作者简介:陈莹(1976-), 女, 浙江丽水, 博士, 副教授,研究方向为多传感融合、模式识别;程雪雪(1990-), 女, 安徽滁州, 硕士生, 研究方向为视频传网。
  • 基金资助:
    国家自然科学基金(61104213),江苏省自然科学基金(BK2011146)

Node Selection of Collaborative Tracking Based on Particle Filter

Chen Ying, Cheng Xuexue   

  1. School of IoT Engineering, Jiangnan University, Wuxi 214000, China
  • Received:2015-01-05 Revised:2015-04-02 Online:2016-06-08 Published:2020-06-08

摘要: 针对视频传感网中的目标协同跟踪问题,基于粒子滤波提出一种新的节点选择协同目标跟踪方法。采用粒子滤波解决目标跟踪问题获到目标的后验概率密度分布,并得到节点的信息熵估计,以此评价节点对目标状态估计的不确定性;通过背景建模和相除法完成目标斑点的有效提取,计算斑点所占像素数用来目衡量节点捕捉目标信息的能力;基于以上这两个因素确定节点的置信度,比较置信度大小来实现优化节点选择并进行粒子滤波目标协同跟踪。实验结果表明,与同类方法相比,该算法能够有效提高目标跟踪精度,实现相机节点的选择分配,完成节点间协同跟踪。

关键词: 节点选择, 协同跟踪, 斑点提取, 粒子滤波, 置信度, 信息熵

Abstract: Due to the problem of target tracking in video sensor networks, a new algorithm was proposed for cooperative monitoring and tracking of node selection based on particle filter. The method obtained posterior distribution through particle filtering in tracking problem and got the information entropy to evaluate the estimate uncertainty. Background modeling and the phase division were applied to extract target blob, and pixels number of the target blob was calculated to measure detection information. Confidence measure of nodes was determined based on the two factors to realize the optimization of the selected node and then track by particle filter. Experiment results show that the proposed method can effectively improve the tracking accuracy in comparison with the similar method, which selects camera nodes to achieve.

Key words: node selection, cooperative tracking, blob extraction, particle filter, confidence measure, information entropy

中图分类号: