ED-YOLO: A Deep Learning and Edge Detection Based Turnout Switch Gap Monitor Method

A real-time monitoring method for turnout switch gaps based on deep learning and edge detection, named EDYOLO, is proposed to accurately monitor the fastening status of turnouts by identifying the scale readings of the turnout extension indicator rods. The difference between these readings and the baseline scale is calculated to obtain the distance of the switch indicator gaps. To transmit data from embedded clients deployed at the turnout edge to a remote monitoring center server, the system integrates two wireless communication technologies: 4G and LoRa. Additionally, Bluetooth communication technology is integrated to allow railway administrators to configure the embedded clients conveniently using handheld terminals. A prototype is implemented based on Raspberry Pi, and experimental results show that the maximum error of the proposed system in gap monitoring is less than 0.1 millimeters. Furthermore, the server can poll and remotely collect gap data from at least 60 clients per minute.