Improved Small-scale Defect Detection Algorithm for Wind Turbine Blades Based on YOLOv8

Aiming at the problems of difficult small-scale defect detection, high false detection rate and low detection accuracy in the inspection task of wind turbine blades under complex environmental background, an improved small-scale defect detection algorithm for wind turbine blades based on YOLOv8 is proposed. Firstly, the orthogonal channel spatial attention(OCSA) module is proposed to suppress confusing irrelevant information interference and enhance the model's ability to locate and perceive defect features. Secondly, the fast spatial pyramid module is optimized. By introducing dilated convolution and global average pooling, the model can obtain rich context information. Finally, the GMS-C2f module is designed, that is, the grouped multi-scale convolution(GMSConv) is applied to the C2f module to capture multi-level semantic features and reduce the computational complexity of the module. The experimental results show that the mAP50 and mAP50: 95 values of the proposed algorithm on the validation dataset are 89.3% and 63.1%, respectively, which are 3.9% and 7.4% higher than those of the classical algorithm. The detection performance of small-scale defects is improved, and the problem of false detection is effectively alleviated, which proves the effectiveness of the proposed algorithm.