Improved Algorithm for Foreign Object Recognition and Localization of High Voltage Transmission Lines Based on YOLOv5 and ORB-SLAM2

The use of drones to clean foreign objects on high-voltage transmission lines is currently the main development direction for high-voltage line maintenance. In order to achieve the recognition and localization of foreign objects, in the recognition part, this paper uses the YOLOv5 training dataset, and the training results are [email protected]. It is 0.602. In the positioning section, this paper proposes an improved binocular stereo vision algorithm based on ORB-SLAM2. By combining the detection boxes output by YOLOv5 to crop binocular images, the computational complexity of binocular stereo vision algorithms has been reduced. The experiment shows that the algorithm proposed in this paper reduces the computation time by 156.10 ms, effectively improving the real-time positioning performance.