吴松伟 刘军 范长军 陈久红 程球 柳明



摘要:在刑侦过程中通过基于视觉的行为分析与理解技术,及时地发现和识别潜在作案者的异常行为,对犯罪预警具有极为重要的现实意义。文章分析了视频监控系统在异常行为检测方面存在的问题,对异常行为检测系统的实现形式、系统架构以及异常行为检测方法进行研究。设计了视频监控异常行为检测系统的软硬件系统框架;研究了基于视频的行人检测方法与异常行为检测识别方法;完成了相关算法在上述系统上的实现与开发.系统测试表明,该系统可以准确、高效地检测异常行为,能够满足当前刑侦中视频监控产品的需要。
关键词:视频监控;异常行为检测;行人检测;刑侦
中图分类号:TP391.4
文献标识码:A
文章编号:1006-8228(2020)09-67-05
Design and implementation of video-based abnormal behavior detection system for criminal investigation
Wu Songwei', Liu Jun2. Fan Changjun3, Chen Jiuhong3, Cheng Qiu3, Liu Ming3
(I. Investigation Technology Division of Air Force Political Worle Department, Beijing 100843, China: 2. Chine.se People's Liberation Army /Vo.63650; 3. No. 52 Re.search Institute of China Electronic Technology Corporation)
Abstract: In the process of criminal investigation, it has great practical significance to detect and identify abnormal behaviors ofpotential perpetrators by visual-based behavior analysis and recognition technology. In this paper, the problems of abnormal behaviordetection in current video surveillance systems are analyzed. and implementation forms, system architectures and related methods ofabnormal behavior detection system are studied. After designing the overall hardware and software framework of the system.studying the methods of pedestrian detection and abnormal behavior detection, and realizing the related algorithms for the system, avideo-based abnormal behavior detection system is implemented. The experimental results show that the system can detect abnormalbehavior accurately and efficiently, and can meet the needs of current video monitoring products in criminal investigation.
Key words: video monitoring; abnormal behavior detection; pedestrian detection; criminal investigation
0引言
安防監控产品日益普及,覆盖银行、地铁、停车场、超市、学校、主干道等公共场所,在刑侦、安保等领域发挥了重要作用。现有的视频监控数据量庞大,依赖人力已无法及时有效地处理,而大多数监测系统只是对监控范围内发生的情况进行简单记录和存储,并不进行检测处理。即使有人值守,由于各种复杂场景等原因,监控的实时性也不高[1]。为了解决上述问题,出现了无人值守的智能视频监控系统。
目前,一些最新的智能视频监控产品开始具备视频分析的功能,能对简单的人体行为进行检测并报警,可以辅助刑侦人员提高工作效率。国内的部分智能监控产品实现了对预先定义的简单异常事件进行检测,包括入侵检测,移动物体检测,移走物体检测等[2]。……