A defect detection algorithm of wind turbine blades based on an improved YOLOv8

Aiming at the problem of decreased detection accuracy and low reliability caused by the large variation of wind turbine blade defect scale, this paper proposes a wind turbine blade defect detection method based on improved YOLOv8. First, the CBAM attention mechanism is used to optimize the YOLOv8 network, and the C2f module is replaced by the C2f-CBAM module in the Neck part, which enhances the network's ability to extract and dynamically adjust the important features, so that it is able to detect the information of defects at different scales more effectively. Second, the GIoU loss function is used to replace the IoU loss function, which provides more stable and accurate gradient information and improves the positioning accuracy of the bounding box by calculating the generalized IoU of the predicted box and the real box. The experimental results show that for wind turbine blade defects at different scales, the CBAM attention mechanism has better applicability than the SE attention mechanism and the LSK attention mechanism, and the YOLOv8-CBAM-GIoU network proposed in this paper has significantly improved detection accuracy and robustness than the YOLOv8 network.