Key points are not available for this paper at this time.
Owing to the frequent occurrence of extreme cold weather, new challenges are posed to the deicing of railway contact wire. In most cases now, it still relies on human standing on the train to break off the frozen ice. Using microwave heating is a kind of deicing work, but due to the Z-shaped wiring structure of the contact wire, the microwave generator cannot continuously take aim at the wire when the train is moving. There is little research on this deicing scene, so an automatic tracking system for railway contact wire is developed in this article. A scale down device is built indoors, and at a moving speed of 1 \ m/s, discriminative model prediction (DiMP) algorithm is used for tracking, and proportional-integral–differential (PID) is used to control the microwave generator to align with the contact wire. In addition, a deviation threshold is established to prevent motor misoperation. This deviation threshold is set to 10 \ px. In the tracking process, the DiMP algorithm updates its model every 20 frames, which ensures an accurate tracking by test. By comparing with Deeplabv3+, U-Net, Mask R-CNN, YOLOv8, YOLOv9, and YOLOv10, the intersection over union (IoU), robustness evaluation metrics (REM), and frames per second (FPS) of DiMP algorithm are relatively better. The confidence interval of PID control is 0. 78 \ cm.
Du et al. (Mon,) studied this question.