The power generation efficiency and service life of wind turbines are related to their surface integrity. This study aims to address the issue of inaccurate detection results and long detection time in traditional surface defect detection methods for wind turbines. A surface defect detection model for wind turbines is designed. Firstly, lightweight convolution technology is integrated into the model, effectively enhancing the ability of information exchange between channels and improving the detection effect of small-sized defects through richer feature information. Secondly, the visual attention network module has been introduced into the backbone network, enriching the contextual information and improving the feature extraction ability of convolutional neural networks. Then, a coordinated attention mechanism is introduced into the neck network to capture the location information of defects through spatial orientation. Finally, modify the loss function of bounding box regression to WIoU to develop an appropriate gradient gain allocation strategy. The experimental results show that the improved detection model improves the detection accuracy by 4.14% compared to the original model, significantly enhancing the detection ability of small defects. At the same time, the parameter count of the improved model was reduced by 2.29 M, while the parameter count was reduced by 6.2 G. The detection speed was significantly improved, meeting the real-time detection requirements of the model.