Research On Construction Of Safety Features And Perception Technology Of Transportation Environment In Subway Station

In recent years,in our country,with the accelerating pace of urbanization,more people have chosen to enter the city,which has put tremendous pressure on urban rail transportation.In order to better serve society,ease traffic pressure,and improve the level of urban rail construction,it has become an important means of conforming to the urban development in the twenty-one century.The subway is an important part of urban rail transit,has many advantages: saving land,less noise,low energy consumption,low pollution,high punctuality,and high safety.It is precisely because of these advantages that the subway has become a means of travel for more and more people.In recent years,many provinces,cities and autonomous regions have continuously promoted the construction of subway,especially the construction of subways.With the continuous expansion of the subway network,new problems have also emerged,which have brought great challenges to the daily work of subway stations in my country.How to overcome difficulties has become the top priority of the subway development.At present,the management of urban rail transit in my country is mainly manual management.The low degree of systematization and insufficient degree of intelligence have caused many problems.However,foreign technical equipment has poor applicability and high price for China’s unique conditions.The existing technology and equipment are difficult to meet the actual operation needs of the subway network.Recently,computer based on deep learning and neural networks has developed rapidly.The existing security perception mainly relies on traditional image processing techniques and is basically based on manual selection for feature detection.The basic principle is to use a suitable classifier to classify feature vectors.In actual scenarios,a subway station often needs to install dozens or even hundreds of cameras,and traditional image processing methods are difficult to meet the requirements of real-time reliability.With the development of the explosive growth of the monitoring volume,deep learning and neural networks have gradually stepped onto the stage in recent years,which has an important impact on the development of video information.The research work of this article is mainly reflected in the following aspects:(1)Analyzed several factors that affect subway safety,and obtained the types of accidents affecting subway safety by analyzing and summarizing the statistical data of major foreign subway operations.(2)The law of subway safety is explored,focusing on the in-depth analysis of personnel factors,and in response to the increasing impact of terrorist attacks on subway safety,a detection model for abnormal pedestrian behavior is proposed.(3)In view of the many shortcomings of the commonly used pedestrian target detection algorithms,the KF-YOLOv3 algorithm is proposed.Used on the YOLOv3,this algorithm uses the Kalman filter algorithm and the Hungarian matching algorithm.The algorithm uses dimensional clustering to optimize the selection of preselection boxes.The algorithm uses non-maximum suppression methods to improve robustness.The algorithm improves the activation function and improves the ability to detect small targets.The experimental results show that this new algorithm overcomes the problems of crowded people and serious pedestrian occlusion caused by the sidewalk blockage of subway stations,and greatly improves the speed of video detection.Meanwhile,the algorithm has better detection performance and effect in the subway station.It has good processing ability for target occlusion.