Wind turbine rotor blade encoding marker recognition method based on improved YOLOv8 model

With the increasing demand for renewable energy, the efficiency and stability of wind power generation have become a research focus. The wind turbine rotor blade, as a core component of wind power generation systems, its state monitoring is crucial for ensuring system performance. This study proposes a recognition method based on an improved YOLOv8 model for the automatic identification of encoding markers on wind turbine rotor blades. By introducing spatial and channel reconstruction convolution (ScConv) and Wise-IoU (WIoU) loss functions, we developed the YOLOv8-ScConv-WIoU model and trained and evaluated it on a self-built dataset of encoding markers with different rotation speeds and pasting positions. Experimental results show that the improved model achieved precision (P), recall (R), and mAP@[0.5–0.95] of 99.3%, 99.1%, and 74.3%, respectively, which are improvements of 2.0%, 2.58%, and 3.5%, respectively, compared to the original YOLOv8 model, with faster convergence speed. This achievement not only improves the accuracy and robustness of encoding marker recognition but also provides more effective technical support for the monitoring and maintenance of wind power generation systems. Furthermore, the methods and findings of this study also provide new perspectives and references for the application of other object detection tasks in the industrial field.