关键词: 风电叶片, 缺陷检测, Cascade R-CNN, 目标检测, 深度学习

Abstract: As a key component of wind turbines, the detection of surface defects on wind turbine blades is crucial for
ensuring the stable operation and extending the service life of wind turbines. The traditional manual detection methods are
inefficient and susceptible to human interference. This article proposed a surface defect detection method for wind turbine
blades based on Cascade R-CNN. High-quality inspection data was obtained by capturing surface images of wind turbine
blades using high-resolution drones. The Cascade R-CNN model was adopted, which gradually improves the detection
accuracy through a cascade structure and could effectively identify defects of different scales, common damages such as
cracks, scratches, and corrosion. In order to improve the robustness and generalization ability of the model, the adopts
feature enhancement techniques and multi-scale feature fusion methods to enhance the sensitivity of the model to different
scales and complex features. This article also increased the diversity of training samples through data augmentation
techniques such as rotation, translation and scaling, thereby further enhancing the model's adaptability to different lighting,
backgrounds, and defect types. The experimental results show that this method achieves an average precision ( AP ) of
88. 2% and an average recall rate ( AR) of 75. 9%, significantly better than traditional detection methods, demonstrating
higher detection accuracy and efficiency. This method not only improves the accuracy and efficiency of defect detection for
wind turbine blades, but also provides strong technical support for intelligent monitoring of wind turbine blades.