城市轨道交通司机驾驶视频分析系统识别功能评估

Urban Rail Transit Driver Driving Video Analysis System Identification Functionality Assessment

采用司机驾驶视频分析系统对司机的异常驾驶行为进行实时甄别,有助于提高司机的注意力,保障列车运行安全。但是,目前市场上各厂家的视频分析算法多由厂家自行测试,所提供的测试样本和测试方法不尽相同,不利于使用方的统一评判。因此需研究司机驾驶视频分析系统识别功能的评估方法和评估指标。提出了司机驾驶视频分析系统识别功能的评估指标和评估方法;通过视频样带测试和实车安装测试,对市场占有率较大的5家厂家的视频设备及分析算法进行了测试,分析、比较不同厂家司机驾驶视频分析系统可识别的事件类型和识别准确率;给出了司机驾驶视频分析系统识别功能技术指标的具体参数值。所提出的测试方法和技术指标可客观、公正地评估司机驾驶视频分析系统的识别功能。

The real-time identification of driver′s abnormal driving behavior through the implementation of a DVAS(driver driving video analysis system)contributes to enhancing driver attention and ensuring train operation safety.However,the manufacturers in the current market are mostly carrying out individual testing for their own video analysis algorithms,resulting in various test samples and methods,discouraging consistent assessment on user side.Therefore,it is necessary to explore evaluation methods and criteria for assessing the identification functionality of DVAS.Evaluation criteria and methods for assessing the identification functionality of DVAS are presented.Through testing by video sample analysis and on-site installations,the video devices and analysis algorithms offered by five major manufacturers with significant market share are tested,and comprehensive analysis and comparison are carried out to determine the types of identifiable events and the identification accuracy achieved by different DVAS.Additionally,specific parameter values for the technical indicators of the identification functionality of DVAS are provided.The proposed evaluation methods and technical indicators offer an objective and unbiased approach to assessing the identification functionality of DVAS.