基于智能视频分析的轨道交通异物入侵检测方法研究  

Research on foreign body intrusion detection method in rail transit based on intelligent video Analysis

作　　者：杨岗 王富斌 刘孜学[1] 虞凯[1] YANG Gang;WANG Fubin;LIU Zixue;YU Kai(China Railway Eryuan Engineering Group CO.LTD,Chengdu,610031,China)

机构地区：[1]中铁二院工程集团有限责任公司,四川成都610031

出　　处：《长江信息通信》2021年第12期8-11,共4页Changjiang Information & Communications

摘　　要：目前在轨道交通异物入侵检测领域中,采用基于深度学习的视频目标检测方法要达到可以应用的实时检测速度,对设备资源的要求较高且成本不低。文章通过计算视频图像敏感网格区域,采用数字图像处理方法实时地对移动物体进行初检测,然后将检测到的异常视频图像发送至异物入侵识别模型进行更加精准的智能识别分析和研判报警,节约计算时间和设备资源。文章基于相同的构思提出一种异物入侵检测装置,通过建立信号传输机制高效地控制多台设备并行处理视频目标检测任务。At present, in the field of foreign body intrusion detection in rail transit, the video object detection method based on deep learning needs to achieve the applicable real-time detection speed, which requires high equipment resources and the cost is not low. In this paper, the moving object is initially detected by calculating the sensitive grid area of video images and combining with the digital image processing method.Then the detected abnormal video images are sent to the foreign body intrusion recognition model for more accurate intelligent recognition analysis and alarm, saving computing time and equipment resources.Based on the same idea, this paper proposes a foreign body intrusion detection device which can efficiently control multiple devices to process the target detection task in parallel by establishing a signal transmission mechanism.

关 键 词：轨道交通 异物入侵 智能识别 检测装置 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]