计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (30): 183-187.

• 图形、图像、模式识别 • 上一篇    下一篇

基于动态贝叶斯网络的多特征目标跟踪

吴孟俊1,付 钿1,刘建平2,牛玉刚1   

  1. 1.华东理工大学 信息科学与工程学院,上海 200237
    2.河北石家庄职工大学,石家庄 050049
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-10-21 发布日期:2011-10-21

Multi-feature target tracking based on dynamic Bayesian network

WU Mengjun1,FU Tian1,LIU Jianping2,NIU Yugang1   

  1. 1.School of Information Science and Technology,East China University of Science and Technology,Shanghai 200237,China
    2.Shijiazhuang Staff and Workers University,Shijiazhuang 050049,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-10-21 Published:2011-10-21

摘要: 通过将动态贝叶斯网络模型应用到人体目标跟踪中,提出了一种多特征融合跟踪算法。该方法基于动态贝叶斯网络建立状态模型,分别针对形变、遮挡、有干扰三种情况提取运动中人体的颜色和梯度特征,利用粒子滤波方法对颜色特征和梯度特征进行融合。实验表明,提出的多特征跟踪算法能较好地解决复杂环境下的目标跟踪问题,相比传统的利用单一目标特征的跟踪算法具有更好的鲁棒性和准确性。

关键词: 动态贝叶斯网络, 颜色直方图, 梯度直方图, 粒子滤波

Abstract: In this paper,the dynamic Bayesian network model is applied for the tracking of human,and a multi-feature fusion tracking algorithm is proposed.This algorithm firstly establishes the state model based on the Dynamic Bayesian Network,and then extracts the features of body color and gradient under the three different cases,i.e.,deformation,occlusion,and interference,respectively.Both the color feature and gradient feature are integrated by using particle filter method.The experiment results show that the proposed multi-feature algorithm is of better robust and accurate in the problem of human tracking,compared with the traditional single feature in complex environment.

Key words: dynamic Bayesian network, color histogram, gradient histogram, particle filter