High-Voltage Tower Nut Detection and Positioning System Based on Binocular Vision

Ultra High Voltage (UHV) transmission is the most advanced transmission technology in the world. However, it is difficult for the daily maintenance of high voltage power towers. Based on the development of robots and in-depth learning, this paper proposes a visual-based pylon climbing robot to detect high-voltage tower nuts. An improved yolov5 is developed by adding coordinate attention (CA) module to the backbone, and assigning different weights to different levels of features, replacing the Concat of neck species with Full-Concat. Experiment results showed that our proposed scheme can detect and locate nuts very well, and our trained model can also be well applied in our devices.