基于神经网络和Mean-shift的目标跟踪方法  

Target Tracking Approach Based on Neural Network and Mean-shift

作　　者：胡淑芳[1,2] 

机构地区：[1]重庆大学计算机学院,重庆400030 [2]重庆理工大学设备处,重庆400050

出　　处：《重庆工学院学报（自然科学版）》2009年第12期71-76,共6页Journal of Chongqing Institute of Technology

基　　金：国家"863"高技术研究发展计划资助项目(2007AA01Z423);重庆市自然科学基金资助项目(CSTC;2007BB2134)

摘　　要：提出了一种RBF神经网络和Mean-shift相结合的跟踪算法.此算法采用Bhattacharyya系数度量环境干扰程度,使用不同的比率因子将RBF神经网络对目标位置预测结果与Mean-shift算法跟踪结果加权,得到最终目标位置.实验结果表明,该方法在复杂环境下能对运动目标进行更加准确的跟踪.Traditional Mean-shift target tracking algorithm only uses the target＇s color feature. When there is disturbance in outside environment, tracking is easy to lose target. To solve this problem, a new tracking algorithm combining RBF Neural Network with Mean-shift is proposed in this paper. The Bhattacharyya coefficient is adopted to measure the disturbance circumstances. Then according to different disturbance circumstances, by weighting the estimating location of RBF Neural Network and the result of Mean-shift tracking algorithm by using different scale factors, the final position of the target is obtained. Experimental results show that the proposed method achieves more accurate target tracking in complex environments.

关 键 词：目标跟踪 MEAN-SHIFT 特征融合 RBF神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]