Abstract:Aiming at the problem that the traditional object detection algorithm has low recognition accuracy and inaccurate effect of construction safety equipment in power distribution construction operation scenarios, this paper proposes a YOLO-Rotating algorithm for the identification of construction safety equipment in distribution network. Based on YOLOv8, the algorithm uses deep separable convolution instead of part of Conv to design C2f-R modules, which reduces the number of model parameters and improves the receptive field. Use the GAM attention module to enhance feature extraction capabilities, improve semantic information and reduce interference; Finally, the rotating target detection module is added to make the detection frame fit the target contour better and improve the detection accuracy. The experimental results show that the average accuracy mean (mAP) of the YOLO-Rotating algorithm reaches 84.6% on the distribution safety equipment dataset, which is 2.5% higher than the original algorithm and the accuracy is improved by 2.07%. The algorithm has higher detection accuracy and better practical application value, meets the requirements of edge computing equipment, and is suitable for the construction scenario of power distribution live network operation.