融合减法聚类与C-均值聚类的目标定位方法  

Joint Subtractive Clustering and C-Means Clustering Object Location Algorithm

作　　者：张君昌[1] 李明[1] 谷卫东[2] 

机构地区：[1]西北工业大学电子信息学院,陕西西安710072 [2]西安电子科技大学电子工程学院,陕西西安710071

出　　处：《计算机仿真》2012年第7期269-273,共5页Computer Simulation

摘　　要：视频运动目标的检测与定位是视频监控系统的主要技术之一。针对现有视频监控系统目标定位过程在目标被浅度遮挡或存在噪声时定位不准确的问题,提出了一种新的视频运动目标定位方法。采用减法聚类、聚类有效性函数与加权模糊C-均值聚类方法相结合。首先利用减法聚类,获得初始聚类中心,再通过加权模糊C-均值聚类算法对视频运动进行目标定位,避免了算法陷入局部最优而获取了全局最优。然后引入聚类有效性函数,获得视频序列中目标的最佳个数。仿真结果表明,改进方法对存在噪声或野点的情况具有较好的鲁棒性,并可以在不需要人为给定待检测图像目标个数的情况下,对存在浅度遮挡区域的目标进行准确定位。Moving object detection and location are core technologies in video surveillance system. In order to im- prove the present object location methods in video surveillance system, a novel moving object location method was proposed. The new algorithm combines the subtractive clustering, clustering validity function with the weighted fuzzy C-means clustering. The subtractive clustering was used to obtain initial clustering centers, and then the weighted fuzzy C-means clustering was used to locate the moving object, which makes the algorithm able to avoid local opti- mum and get the global optimum. Finally, the clustering validity function was introduced to obtain the optimal number of the clusters. Simulation results show that the proposed method has a good robustness even in a high noisy or outli- ers enviroment, and can accurately locate the moving objects which has overlap areas without the given number of the objects in the detected images.

关 键 词：减法聚类 模糊均值聚类 聚类有效性 目标定位 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]