Real-Time Measurement of Catenary Steady Arm Orientation Angle Based on YOLO

Abstract

In the catenary system, the steady arm is an important component used to support and guide the catenary wire, and its orientation angle is crucial to determine the height of the catenary wire. This paper mainly detects and calculates the orientation angle of the steady arm. Two methods are proposed. One is a two-step method that uses improved YOLOv5E-based YOLOv5 to detect the steady arm and plate targets, followed by calculation of the orientation using a line segment detector (LSD). Another method uses an end-to-end YOLOv8-pose model to directly locate terminals of the steady arm for angle measurement. We trained the model using a self-made steady arm dataset and compared the detection results of different versions of YOLO. The results show that the YOLOv5E combined LSD algorithm is the best, with an accuracy of almost 95% and a detection speed of 97 FPS. The findings determine that the method using YOLOv5E combined with LSD linear detection is more robust and more suitable for practical applications.