基于改进的均值漂移算法的运动汽车跟踪  

Moving Vehicle Tracking Based on Improved Mean Shift

作　　者：雷飞[1] 孟晓琼 吕露[1] 黄涛[1] 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100124

出　　处：《计算机技术与发展》2017年第2期106-109,共4页Computer Technology and Development

基　　金：北京市教育科技计划面上项目(KM201210005003)

摘　　要：交通领域的智能视频监控系统有效解决了车辆的实时跟踪问题。针对运动车辆的特点,提出一种均值漂移(Mean Shift)和粒子滤波相融合的跟踪算法。该算法以HSV颜色直方图为核心建立运动汽车目标模型,利用Bhattacharyya距离度量粒子区域和目标模型的相似性,并根据相似性来更新粒子权值。使用Mean Shift聚类偏移粒子,通过观测模型和再估计过程使得这些粒子的候选区域能更加接近真实的目标位置。实验结果表明,该算法具有较强的实时性和鲁棒性,能实现对感兴趣运动汽车的稳定跟踪。The intelligent video surveillance system effectively solves the problem of real-time tracking of vehicles in transportation field.According to vehicle characteristics,a newalgorithm combined of Mean Shift and particle filter is proposed to track the target.The algorithm takes the HSV color histogram as the core to establish the target model of moving vehicle,using the Bhattacharyya distance to measure the similarity between particle region and the target model and updating the particle weights according to the similarity. After that,Mean Shift is used to cluster offset particles whose candidate region is closer to real target location through the observation model and re-estimation.Experimental results showthat the algorithm has strong real- time performance and robustness,and can achieve the stable tracking of interest moving vehicles.

关 键 词：均值漂移 粒子滤波 采样 目标跟踪 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]