In view of the low monitoring efficiency of compliance use of key safety equipment such as safety helmet and safety belt for working at heights in manual inspection and maintenance of electric power, a detection model based on YOLOv8s is proposed to reduce labor costs and improve detection efficiency. By designing the C2f _ PTB feature extraction module, combining the global information capture of Transformer and the local feature extraction ability of convolutional neural network, the detection efficiency of the model for small size and scattered targets is improved; The normalized Gaussian Wasserstein distance (NGWD) loss function is introduced to enhance the stability and accuracy of the model for the detection of small safety equipment; Lightweight backbone network C2f _ star module based on StarNet is designed to reduce network parameters. Experimental results show that the mAP of the improved model reaches 93.7% on the power safety equipment data set, the detection accuracy is improved by 5.6% and the detection speed is improved by 10 frames per second compared with the benchmark model, which proves that the proposed method can effectively improve the detection effect.