With the continuous improvement of the complexity of modern equipment, the importance of its maintenance support has become increasingly prominent. The traditional maintenance method mainly relies on manual inspection, which is not only inefficient, but also easy to miss inspection due to human factors. With the development of computer vision technology and augmented reality technology, object detection technology is gradually applied to the field of equipment maintenance deployed in AR glasses. In this paper, a maintenance target detection algorithm based on improved YOLO v10 is proposed. Receptive Field Attention Convolution (RFAConv) was introduced to fuse with the C2fCIB module in the backbone network to improve the accuracy of target feature extraction. The Spatially Enhanced Attention Module (SEAM) is added to the neck network to improve the detection ability of small targets that are greatly affected by the background and occluded. The loss function based on Normalized Wasserstein Distance (NWD) is used to reduce the influence of bounding box position deviation on the detection results of small targets and improve the detection accuracy of the model. Experimental results show that the improved detection model is 1.9%, 1.8% and 1.6% higher than the original YOLO v10 model on the self-made datasets with mAP 0.5, P and R, respectively.