基于视频监控的地铁施工不安全行为检测预警

Research on detection and early warning of unsafe behavior in metro construction based on video monitoring

为了解决地铁施工人员不安全行为的检测问题,从地铁施工人员的不安全行为入手,提出一种较为全面的检测框架.将人的不安全行为检测归纳为人身检测、侵入性检测和多人协同作业检测三个方面.人身检测采用支持向量机(SVM)与方向梯度直方图(HOG)特征组合的方法进行判断;侵入性检测采用基于视频监控、传感器以及现场人员反馈信息的多源数据融合方法进行判断;多人协同作业采用多监控协同检测方法,对图像中人员数量与间隔实时判断.以苏州地铁3号线施工为例,对地铁隧道场所光线偏暗、光照不均的问题进行分析.

To solve the detection problem of unsafe behavior in metro construction,a comprehensive detection framework was proposed from the perspective of the subway constructors’unsafe behavior.The detection of unsafe human behavior could be divided into three aspects,which were personal detection,intrusive detection and multi-person cooperative work detection.By using SVM(support vector machine)and HOG(histogram of oriented gradient)feature combination method,personal detection could be judged.By using multi-source data fusion method based on video monitoring,sensors and feedback information from field personnel,intrusive detection could be judged.By using multi-monitoring cooperative detection method,the number and interval of personnel in the image could be judged.The construction of Suzhou Metro Line 3 was taken as an example to analyze the problem of dim light and uneven light in the metro tunnel site.