基于视频摘要的地铁工程施工Near-miss事件识别及自动响应研究

Research on Near-miss Incident Recognition and Automatic Response in Metro Construction Based on Video Summarization

随着我国城市地铁工程建设规模扩展,安全形势愈发严峻,加强Near-miss事件管理可以有效提高地铁工程施工安全管理水平,预防安全事故发生。远程视频监控是地铁工程施工现场安全管理的有效监管措施,传统的视频监控依赖人工监管,工作量大,效率低,不能及时发现Near-miss事件且其上报意愿不强。为提高地铁工程施工Near-miss事件监管效率,提出基于视频摘要技术的地铁工程施工Near-miss事件检测识别与自动响应方法。首先确定地铁工程施工Near-miss危险区域;然后利用运动目标检测和跟踪算法对进入危险区域的工人目标进行识别跟踪,根据Near-miss事件特征,抽取典型帧形成视频摘要,并生成结构化说明信息,再运用CBR推理算法查询地铁工程施工Near-miss知识库,获取纠正与预防措施;最后以基坑临边高处坠落Near-miss事件案例验证该方法的可行性。

The safety situation is getting more serious with the expansion of urban metro construction scale.Near-miss incident management can effectively prevent accidents and promote the safety supervision of metro construction.Remote video monitoring is an effective supervision measure for safety management of metro construction site.Traditional video monitoring relies on manual supervision which has a heavy workload and low efficiency and would fail to detect Near-miss events in time.In order to improve the supervision efficiency of Near-miss incident in metro construction,a method of Near-miss incident detection,identification and automatic response in metro construction based on video summarization technology is proposed.First step is determining the Near-miss dangerous area of metro construction,and then the moving target detection and tracking algorithm is used to identify and track the workers'targets entering the dangerous area.According to the characteristics of Near-miss events,typical frames are extracted to create video abstracts and structured description information is generated.Using CBR reasoning algorithm query metro engineering construction Near-miss repository for corrective and preventive measures.Finally,a case of Near-miss incident of falling at the edge of a foundation pit is used to verify the feasibility of this method.