非重叠监控视频中基于颜色空间分布模型的行人匹配方法  

Human matching approach based on color spatial distribution model across non-overlapping camera views

作　　者：李博[1] 李国辉[1,2] 涂丹[1] 

机构地区：[1]国防科学技术大学信息系统与管理学院系统工程系,长沙410073 [2]国防科学技术大学信息系统工程重点实验室,长沙410073

出　　处：《计算机应用研究》2012年第2期797-800,共4页Application Research of Computers

摘　　要：针对非重叠视域监控系统中行人目标关联问题,提出了一种基于颜色空间分布模型的行人目标匹配方法。使用最邻近聚类方法对目标的像素进行聚类,得到目标的MC特征;建立目标的颜色空间分布模型,并进行特征转换;在建立的分布模型的基础上,计算目标相似度并进行匹配。实验结果表明,该算法能有效地去除目标检测与分割产生的边缘背景像素的影响,在对相同目标保持较高匹配率的情况下,能较好地处理由于目标颜色分布差异而造成的误匹配情况,同时对于摄像机视角的差异以及目标姿态的变化有较好的鲁棒性。For the human matching requirements,this paper presented a special approach of human matching based on color spatial distribution model across non-overlapping camera views.First,it obtained the MC feature by clustering the target pixels through the nearest cluster method.Secondly,built the color spatial distribution model and transformed the MC feature.Finally,calculated the target pair similarity based on the built color spatial distribution model and got the matching result.Experimental results show that the presented approach can effectively eliminate the influence of the edge pixels which come with the detection and the segmentation.While it maintained a high matching rate to the same target in different cameras,it is capable of coping with the wrong matching by the difference of the color distribution.The results also show that the proposed approach is robust to the difference of the visual angle and the change of the object gesture.

关 键 词：行人匹配 非重叠视域 颜色空间分布模型 主颜色谱直方图 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]