Wind turbine blade damage detection based on YOLOv8

Aiming at the problem of low accuracy of wind turbine blade damage images due to multi-scale changes and difficulties in detecting small targets, a wind turbine blade fault identification and detection method based on Bidirectional Feature Pyramid Network is proposed on the basic framework of YOLOv8 (You Only Look Once 8). BiFPN (Bidirectional Feature Pyramid Network) module is used to replace the PAN (Path Aggregation Network) module in the Neck structure to extract multi-scale feature information, which effectively increases the network depth and extracts more useful feature information. The Squeeze-and-Excitation attention mechanism is used to focus the attention on the important features of the target and suppress the attention on unnecessary features. Finally, it is tested under the dataset of wind turbine blades and the experimental results are compared. The experimental results show that the improved algorithm detects an average accuracy mean value of $87.3 \%$, which is $3.7 \%$ higher compared to the original network, and is able to locate and identify the wind turbine blade damage images effectively.