An efficient YOLO v3-based method for the detection of transmission line defects

Abstract

The UAV inspection method is gradually becoming popular in transmission line inspection, but it is inefficient only through real-time manual observation. Algorithms are available to achieve automatic image identification, but the detection speed is slow, and video image processing is not possible. In this paper, we propose a fast detection method for transmission line defects based on YOLO v3. The method first establishes a YOLO v3 target detection model and obtains the a priori size of the target candidate region by clustering analysis of the training sample library. The training process of the model is accelerated by adjusting the loss function to adjust the learning direction of the model. Finally, transmission line defect detection was achieved by building a transmission line defect sample library and conducting training. The test results show that compared with other deep learning models, such as Faster R-CNN and SSD, the improved model based on YOLO v3 has a huge speed advantage and the detection accuracy is not greatly affected, which can meet the demand for automatic defect recognition of transmission line inspection videos.