Study on Technologies of Overhead Line Recognition and Obstacle Distance Measurement by Patrol Robots Based on Deep Learning

At present, the power inspection work is mainly carried out manually, and there are problems such as low work efficiency, heavy workload, and insecurity. Based on this, people have studied a large number of inspection robot methods based on computer vision to achieve inspection work. The more important part is line identification and obstacle distance measurement. The current research on wire recognition problems mainly have problems such as insufficient research, high application conditions, and low robustness. Therefore, this paper designs a deep learning-based route, obstacle detection, and ranging model, which can be well applied to a variety of Complex environment; by constructing the overhead line data set, performing pixel-level edge labeling, processing the labeling area through Gaussian heat mapping, and sending it to the neural network for pixel classification, you can get a good line recognition effect, and then you can identify obstacles. The position and obstacle type are convenient for the robot to cross obstacles. Experiments show that the method proposed in this paper is highly robust and has extremely high value in the application scenarios of inspection robots.