Research On Active Anti-collision Detection System Of Metro Based On Vision And Lidar

With the continuous improvement of China’s economic and the demand of residents’ travel continues to grow,the transportation industry has been rapidly developing.The level of automation and intelligence of metro is constantly improving,which puts forward higher requirements for the safety of metro.The obstacles in the running area greatly affect the safety of metro.At present,the detection system has the defects of low accuracy,low efficiency and is easy to be affected by the environment.Research on active anti-collision technology of urban rail transit train can improve the safety of metro train operation,realize intelligent automatic identification of obstacles,and improve detection efficiency,thus it has important engineering application value.The specific work of this paper is shown as follows:Firstly,the deep learning model is applied to detect obstacle in the metro environment,and the intelligent and real-time obstacle detection is realized based on YOLO V3 model.Because the defect of recognition accuracy of YOLO V3 model in detecting obstacle is not high.Dense connection network and spatial pyramid network are used to improve the feature extraction ability of the recognition model.Data augmentation and transfer learning are used to speed up the training of the model.The effectiveness of the improved model and its recognition accuracy is proved.The experimental results show that the model has satisfactory recognition performance and can meets the requirements of obstacle visual recognition in metro environment.Then,a method of metro obstacle detection model based on vision and lidar information fusion is proposed.The vision sensor has the defects of insufficient perception dimension,easily affected by lighting conditions and environment,which will affect the image quality.Lidar can actively perceive the environmental information through the return signal of the laser emitted by itself,which makes up for the deficiency of visual information detection.For the 3D point cloud of lidar in the metro environment,the off-ground point cloud segmentation is completed based on the ground plane fitting algorithm,and the non-ground point cloud is clustered by Euclidean clustering method.The decision level fusion is used to complete the metro obstacle detection by the fusion of vision and lidar information,and relevant experimental verification is carried out,which effectively improves the detection performance.Finally,an active anti-collision system for metro is designed.The system can intelligently identify the obstacles,which provides theoretical basis and technical support for intelligent detection of metro obstacles.