测绘通报 ›› 2022, Vol. 0 ›› Issue (9): 23-28.doi: 10.13474/j.cnki.11-2246.2022.0258

• 交通建设工程测绘技术应用研究 • 上一篇    下一篇

基于深度学习的视频识别及动态监测技术应用

吴真真1, 唐超2, 杨晓飞2   

  1. 1. 中国安全生产科学研究院, 北京 100012;
    2. 北京城建勘测设计研究院有限责任公司, 北京 100101
  • 收稿日期:2022-07-06 发布日期:2022-09-30
  • 作者简介:吴真真(1988—),女,工程师,研究方向为城市安全与应急管理。E-mail:594002507@qq.com
  • 基金资助:
    北京市科技专项(Z201100008520038);北京市科协金桥工程种子资金 A 类项目(ZZ20002)

Application of video recognition and dynamic monitoring technology based on deep learning: taking the rail transit construction project as an example

WU Zhenzhen1, TANG Chao2, YANG Xiaofei2   

  1. 1. China Academy of Safety Science and Technology, Beijing 100012, China;
    2. Beijing Urban Construction Exploration & Surveying Design Research Institute Co., Ltd., Beijing 100101, China
  • Received:2022-07-06 Published:2022-09-30

摘要: 基于国内外城市轨道交通建设工程安全管理及信息化建设现状,本文分析了当前轨道交通建设过程中安全管理面临的难点问题,论述基于Faster R-CNN的施工现场视频识别方法,并引入图像空间特征,提高了施工现场人员不安全行为的识别率;采用基于Apriori关联分析算法构建施工环境、设备动态监测与隐患数据间的潜在强关联关系的分析处理流程,提高了施工现场隐患的发现与处置效率。在此基础上提出轨道交通建设安全管理平台总体架构,重点介绍了安全风险隐患管理、视频监测目标识别等在内的安全管理核心模块的应用实践。应用成果表明,本文算法与平台可有效提高施工现场安全信息处理效率,降低安全事故总体发生率,为城市轨道交通建设安全管理提供有效技术支撑。

关键词: 安全管理, 深度学习, 关联分析, 轨道交通建设, 信息化平台

Abstract: Based on the current situation of safety management and informatization construction of urban rail transit construction projects at home and abroad,this paper analyzes the difficulties faced by safety management in the current rail transit construction process and discusses the video recognition method based on Faster R-CNN,which introduces image space features to improve the recognition rate of unsafe behavior. Also,this paper builds the potential strong correlation relationship between the construction environment,equipment dynamic monitoring and hazard data based on the Apriori,so as to improve the efficiency of hazard discovery and disposal.On this basis,the author proposes the architecture of the safety management platform for rail transit construction,and introduces the application practice of the core modules of safety management,including safety risk management and video monitoring.The results show that the algorithm and platform described in this paper can effectively improve the safety information processing efficiency,reduce the overall incidence rate of safety accidents,and provide effective technical support for the safety management of urban rail transit construction.

Key words: safety management, deep learning, correlation analysis, rail transit construction, information platform

中图分类号: