管控区域人员异常行为检测系统与方法研究  

DETECTION SYSTEM AND METHOD FOR ABNORMAL BEHAVIOR OF PERSONNEL IN CONTROL AREAS

作　　者：周晓芳[1] 吴松洋[1] 韩玮[1] 丁昊[1] 孙伟华[1] 邱瑾[1] Zhou Xiaofang;Wu Songyang;Han Wei;Ding Hao;Sun Weihua;Qiu Jin(The Third Research Institute of Ministry of Public Security,Shanghai 201204,China)

机构地区：[1]公安部第三研究所,上海201204

出　　处：《计算机应用与软件》2022年第5期92-97,共6页Computer Applications and Software

基　　金：十三五国家重点研发计划“公共安全风险防控与应急技术装备”重点专项“重要场所安全保卫关键技术研究”的“暴狱、脱逃行为检测技术与应用示范”课题(2016YFC0800405)。

摘　　要：针对重点人员动作类型复杂、干扰多、识别难度大等问题,提出一套异常行为检测系统与方法。搭建集行为数据采集、上传、处理和识别等功能于一体的硬件和软件拓扑系统,并利用卡尔曼滤波器和牛顿迭代算法等对数据进行处理和优化;在获取人员行为轨迹数据的基础上,把时间、空间、上下文信息与个体行为相关联,实现个体异常行为和多人异常行为的识别;基于传感器加速度和角速度特征值,选取SVM分类算法,实现暴力动作和非暴力动作的分类。实验结果表明,所提系统与方法能够实时准确地检测出人员闯入、脱管、滞留、异常接触和暴力动作等异常行为。Aiming at the problems of complex action types,disturbances and recognition difficulty of key personnel,we propose a set of abnormal behavior detection system and method.We built a hardware and software topological system integrating with behavioral data collection,upload,processing and recognition.Kalman filter and Newton iteration algorithm were used to process and optimize the data.On the basis of access to trajectory data of personnel behavior,the individual behavior was associated with time,space and context information,so as to realize the detection of abnormal behavior of individual and group.Based on the characteristic values of acceleration and angular velocity of sensors,we adopted the SVM classification algorithm to realize the classification of violent action and non violent action.Experimental results show that the proposed method can accurately detect the abnormal behaviors in real time,including breaking in,escape,detention,abnormal contact and violence action.

关 键 词：管控区域 异常行为 位置 传感器 检测算法 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]