多路监控的目标识别和行为分析系统的研究与设计

发布时间：2018-02-08 21:37

本文关键词： 智能监控目标识别行为分析智能检索　出处：《北方工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着校园不断的扩建,社会人员不断涌入校园,使得校园内的安全隐患逐渐暴露了出来,国家教育部和各地的高校对于"平安校园"安防项目越来越重视,但是普通高校可用于视频监控系统的投资金额有限,且安保人员有限,导致校园诸多安防死角,本文的研究内容就是在这种供求不平衡的背景下应运而生的。传统的视频监控系统多数只停留在录像阶段,靠人工监控。当高危区域入侵时,没有任何防护作用。耗费人力事后调取录像信息判断视频内容,常错失最佳追踪和处置时机。本论文针对以上问题及对系统功能需求分析的基础上,提出了一种可用于目标识别和行为分析的多路监控系统的设计方案并将该方案采用通用化和智能化的方式加以实现。该智能视频监控系统采用两个型号为DS-2CD3310D-1的海康威视红外阵列半球网络摄像头作为视频采集前端,通过以太网的连接,在PC终端对视频信息进行目标识别、行为分析、检索和抓拍等智能化"主动"处理。本文设计的智能视频监控系统在传统Vibe算法的基础上做出了对于空洞现象的改善,改进了目标检测算法,同时采用卡尔曼滤波算法实现了目标跟踪。在目标识别的基础上,通过入侵检测算法、周界警戒算法实现了对于运动目标的智能行为分析。在无人值守的状态,当运动目标触发异常报警后,系统对于含有异常事件的视频帧图像可以实现自主抓拍,并将图像存储到本地。在事后调查的过程中,可以根据检索条件调取所有符合条件的监控视频帧图像。该系统设计并实现模块化功能后,在实验室进行了系统联调,通过大量实验证明可以实现对于运动目标的目标识别和行为分析,灵敏度较高。
[Abstract]:With the continuous expansion of the campus, social personnel are constantly pouring into the campus, which gradually exposes the hidden dangers of security in the campus. The Ministry of Education and colleges and universities around the country pay more and more attention to the "safe campus" security project. However, the amount of investment that ordinary colleges and universities can use in video surveillance systems is limited, and the number of security personnel is limited, resulting in a lot of security dead spots on campus. The research content of this paper is under the background of this kind of imbalance of supply and demand. Most of the traditional video surveillance systems only stay in the video stage and rely on manual surveillance. When the high-risk area invades, There is no protective effect. It takes manpower to get video information to judge the video content, and it often misses the best time to track and dispose the video content. Based on the analysis of the above problems and the functional requirements of the system, this paper aims at the above problems, and analyzes the functional requirements of the system. A design scheme of multi-channel monitoring system for target recognition and behavior analysis is proposed, and the scheme is implemented in a general and intelligent way. The intelligent video surveillance system uses two types of DS-2CD3310D-1 sea. Conway infrared array hemispherical webcam is used as the front end of video acquisition. Through the connection of Ethernet, the target recognition and behavior analysis of video information are carried out at the PC terminal. The intelligent video surveillance system designed in this paper improves the void phenomenon based on the traditional Vibe algorithm and improves the target detection algorithm. At the same time, the Kalman filter algorithm is used to realize target tracking. On the basis of target recognition, the intelligent behavior analysis of moving target is realized by intrusion detection algorithm and perimeter alert algorithm. When the moving object triggers the abnormal alarm, the system can capture the video frame image with abnormal events and store the image locally. According to the retrieval condition, all the qualified video frame images can be obtained. After the system is designed and modularized, the system is combined in the laboratory. Through a large number of experiments, it is proved that the target recognition and behavior analysis of moving targets can be realized, and the sensitivity is high.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN948.6;TP391.41

【参考文献】