Real-time Front-end Detection Technology of UAV for Transmission Line Defects Based on YOLOv5

With the development of smart grid technology, safety monitoring of transmission lines has become particularly important. A real-time UAV front-end detection technique based on the YOLOv5 algorithm is proposed for identifying and localizing defects in transmission lines. YOLOv5 is widely used in real-time target detection tasks because of its fast speed and high accuracy. In this paper, the basic principle of target detection model YOLOv5 and the inference acceleration method of edge-oriented end neural network model are firstly elaborated in detail, and then an improved transmission line defect recognition model is constructed based on the YOLOv5 model. By collecting a large number of UAV inspection images under different light and weather conditions, four types of inspection object images are obtained, such as transmission line gold tools, insulators, wires and pins., from which the images are classified, and then these images are labeled with data and trained with data. The test results show that the model proposed in this paper has a significant reduction in the number of parameters, the amount of computation and the model size, which are reduced by 15.36%, 49.5% and 15.8%, respectively. Although the mAP is slightly lower than that of YOLOv5-benchmark, the efficiency improvement of the model in this paper is significant, indicating that the model achieves lighter weight and more efficient performance while maintaining higher detection accuracy.