A wind turbine damage detection algorithm designed based on YOLOv8

Due to operational conditions, wind turbines may suffer from various types of damage, including cracks and wear. Traditional methods of wind turbine damage detection face challenges such as low detection accuracy and high computational resource consumption. This study proposes a wind turbine damage detection algorithm designed based on the YOLOv8 to address these issues. Firstly, the C2f-FocalNextBlock module is added to the algorithm's backbone network, enhancing the feature extraction capability of the main network. Then, the ResNet-EMA module is incorporated into the algorithm's neck network. This module effectively captures cross-dimensional interactions and establishes dependencies between dimensions, thereby enhancing the algorithm's feature extraction capability. Finally, a slim-neck structure is introduced into the neck network of the algorithm to better integrate multi-scale features of targets and background information, thus improving the algorithm's performance. Experimental results demonstrate that the wind turbine damage detection algorithm designed based on YOLOv8 achieves an mean average precision mean (mAP) of 79.9%, accurately detecting wind turbine damage.