Yolo-inspection: defect detection method for power transmission lines based on enhanced YOLOv5s

Accurate identification of defective components in transmission lines and timely feedback to inspectors for timely maintenance can ensure the stable operation of the power system. A defect detection system based on “edge-cloud-end” collaboration is introduced to solve the problems of high bandwidth consumption and response delay in the cloud server-based approach. The system transfers the operation of image detection to the edge device, which reduces the data transmission and improves the response speed of the system. To balance the detection speed and accuracy of the algorithm, the YOLO-inspection algorithm applied on edge devices is proposed. The algorithm uses GhostNetV2 to reconstruct the C3 module in the YOLOv5 model, which reduces the computational complexity and captures the correlation between distant pixels so that it is more targeted to the critical region of the defective target. Meanwhile, based on the feature fusion network, a dynamic adaptive weight assignment module and cross-scale connectivity are designed to effectively reduce information loss and help the network learn fine-grained features. The improved algorithm is deployed on the NVIDIA Jetson Xavier NX platform, and the model is optimally accelerated using TensorRT. Experimental results show that the method proposed in this paper can accurately identify defective samples, and the YOLO-inspection algorithm has superior generalization ability under the harsh conditions of low light and snowfall weather conditions. On the edge computing platform, the mean average precision (mAP) can reach 94.3 % , and the inference speed can reach 63 frames per second (FPS). It can be proved that the method has good detection performance.