基于支持向量机的针对ATM机的异常行为识别  

The abnormal behavior recognition of ATM based on support vector machine

作　　者：陈敏智[1] 汤一平[2] 

机构地区：[1]浙江工业大学计算机科学与技术学院,浙江杭州310032 [2]浙江工业大学信息工程学院,浙江杭州310032

出　　处：《浙江工业大学学报》2010年第5期546-551,共6页Journal of Zhejiang University of Technology

基　　金：国家自然科学基金资助项目(60673177);浙江省科技厅重大科技项目(2006C11202)

摘　　要：针对广泛存在的ATM机中安装假门禁、闲逛等行为,提出一种基于支持向量机(Supportvector machine,SVM)的针对ATM机的异常行为检测方法.首先对所获取全景视频图像进行透视展开,用基于高低更新率的自适应混合高斯算法提取出前景人体对象;其次对前景人体对象用块匹配跟踪算法进行跟踪并提取出运动轨迹,接着按照一定的语义规则对运动轨迹进行预处理得到有效跟踪轨迹;最后用SVM算法对有效跟踪轨迹信息进行异常行为识别.实验表明:该方法具有较好的鲁棒性,能有效的识别出多种针对ATM机的金融犯罪行为.In allusion to the widespread behaviors such as sham guarding,loitering and so on,a novel algorithm is proposed in this paper to detect ATM abnormal behaviors based on Support Vector Machine.Firstly,the omni-directional video images are perspectively unwrapped,the foreground human objects are extracted based on adaptive Gaussian mixture algorithm of high-low update rate.Secondly,blocking-matching algorithm is used to track the foreground object and extract the trajectory.Then,the effective trajectory is acquired by the pre-processing of trajectory based on semantic rules.Finally,the SVM algorithm is used to identify the abnormal behavior in the effective tracking trajectory.The Experiments show that the method has a good robustness and can effectively identify a variety of financial crimes for ATM machines.

关 键 词：支持向量机 目标检测 目标跟踪 运动轨迹 行为识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]