YOLO-WTDL: A Lightweight Wind Turbine Blades Defect Detection Model Based on YOLOv8

Wind turbine blades are important part of wind power equipment, whose health is very vital for the safe and stable operation of wind power station. However, at present, the defect detection of wind turbine blades requires a lot of computing resources and difficult to ensure the detection accuracy. Therefore, this paper proposed a lightweight wind turbine blades defect detection model based on YOLOv8, named YOLO-WTDL. First, the RFCBAMConv block is introduced to replace the original convolutional blocks in YOLOv8, which enhances the feature extraction capability of YOLO-WTDL. Secondly, the model integrates context anchor attention module with high-level screening feature fusion pyramid module, referred to as CAA-HSFPN, to replace the original feature fusion component in the neck. While reducing the model's complexity, it effectively achieves multi-scale feature fusion and enhances the model's feature expression capability. Finally, the task align dynamic detection head is used to improve the detection accuracy while further reducing the number of parameters, making the model easier to deploy on devices with limited computing resources. The experimental results show that the YOLO-WTDL model is more efficient, accurate and lightweight than the original YOLOv8 model, and has great application value in the field of wind turbine blades defect detection.