Abstract
Obstacle detection plays an important role in train automatic operation. To overcome the low accuracy and poor real-time performance of traditional detection methods, and better detect medium and long distances obstacles, the Improved-YOLOv4 network based on deep learning was proposed. The D-CSPDarknet was designed as feature extraction network. A combination of path aggregation and feature pyramid networks were used in feature fusion network, and a spatial pyramid pooling network was set up at each fusion layer. A method of dividing the ROI using a mask was proposed to improve the accuracy of the model while the processing speed can reach 0.004 s. Data augmentation, transfer learning and phased training strategies were used to improve model performance. Based on the data collected in the real operating environment of the train, Improved-YOLOv4 obtained the mAP of 93% on NVIDIA Jetson AGX, which is more suitable for the obstacle detection of rail transit.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
