Abstract: Under the background that computer vision is widely-used in the field of defect detection, for the new application scenario of defect detection of wind turbine blades, a data enhancement method based on improved Cycle-GAN is proposed to generate high-quality images of small cracks on the surface of wind turbine blades so as to overcome the application bottleneck caused by the data-driven limitations of the detection network. This method constructed a basic training defect image set by transferring references and image fusion. On the basis of Cycle-GAN, NAM attention mechanism and multi-scale feature fusion module were introduced to improve the quality of generated images. The experimental results show that the method can generate fan surface crack images with high similarity to the real defects, and the average accuracy (mAP@0.5) of the YOLO network trained by the generated images reaches 83.7%, which has good engineering application value.