Target Recognition and Evaluation of Typical Transmission Line Equipment Based on Deep Learning

The traditional method of conducting regular inspections for high-voltage transmission towers is mainly based on manual inspection. Workers need to board the tower for visual inspection during operation, which is not only unsafe, but also inefficient. With the gradual popularization of drones in the use of power industry, automated inspections based on artificial intelligence and image processing technology have become possible. In the complex background of aerial images, doing defect detection for key equipment components of high-voltage transmission lines is a challenging problem, and achieving the goal of target recognition of transmission line equipment is the basis of defect detection. Based on the deep learning technology, this paper researches the target recognition of transmission line equipment by using aerial image, focusing on insulator and anti-vibration hammer. The process is as follows: Firstly, the image data should be preprocessed. Secondly, the data should be marked and the data set should be divided. Then the two networks of Faster R-CNN and YOLOv3 are used for training. Finally, the trained model is evaluated.