Research on Power Equipment Detection Method Based on Improved YOLOv5s

The traditional inspection method of electric power equipment mainly relies on manpower, which needs to consume a lot of manpower and material resources. With the rapid development of deep learning, the use of target detection algorithm to detect power towers, poles, nails, power engineering cars and insulator equipment has become the main way of inspection. Aiming at the problems of low accuracy and poor real-time performance of traditional target detection algorithms for small targets, an improved algorithm model based on YOLOv5s is proposed in this paper. First, the backbone layer of YOLOv5s is improved, and the original Spatial Pyramid pool (SPP) module of YOLOv5s is adjusted to Spatial Pyramid Pool-Plus (SPPP) module, which has better feature extraction while reducing parameters. Then, the original C3 module of YOLOv5s backbone layer is adjusted to C2f module to obtain more abundant gradient flow information. Finally, a small target detection layer of 160×160 is added to improve the detection performance of the model. The experimental results show that the average accuracy of validation set and test set is increased by 2.25% and 3.1% respectively compared with the original algorithm, and the detection frame rate reaches 120.94 fps.