<p>A lightweight blade surface defect detection algorithm based on an improved YOLOv8n is proposed to enhance the accuracy of wind turbine blade surface defect detection and enable fast deployment on mobile devices. The backbone network structure of YOLOv8n is replaced with the MobileNetV3 model, and Depthwise Separable Convolution (DSC) is introduced to effectively reduce the model parameters. The Global Attention Mechanism (GAM) module is added to the neck network to improve the model’s global feature perception and enhance semantic and positional information in the features. The Weighted Bidirectional Feature Pyramid Network (BiFPN) module is introduced to optimize the fusion efficiency of features at different scales. Angle loss, distance loss, and shape loss are incorporated into the complete intersection over union loss function to enhance the detection accuracy of occluded defects. Experimental verification on a dataset of wind turbine blade surface defects collected in complex environmental backgrounds shows that, compared to the original YOLOv8n, the improved model achieves a 41.4% reduction in model parameters and a 42.9% decrease in the size of generated weight files. It also attains a 2.9% increase in precision, reaching 95.6%, and a 2.2% increase in mAP@0.5, reaching 97.5%. The image processing time is only 3.2 ms. These improvements outperform other mainstream object detection algorithms, providing solid theoretical support for the deployment and application of wind turbine blade surface defect detection on mobile devices.</p>