Improved YOLOv5x for Offshore Wind Turbine Blade Defect Detection

With the rapid development of offshore wind power, operation and maintenance (O&M) has become an increasingly important issue. As a key component of offshore wind turbines, wind turbine blades are essential for the safe and stable operation of the system. This paper addresses the challenges of high risk, low efficiency, and low accuracy in the O&M of offshore wind turbine blades. We propose an improved YOLOv5x-based offshore wind turbine blade defect detection model. The proposed model introduces the ECA attention mechanism to enhance the network's perception of input features and uses the NWD loss function to improve the detection of small targets. The model is trained on a dataset of offshore wind turbine blade defect images. Experimental results show that the improved YOLOv5x achieves an mAP of 64.01%, precision of 70.59%, and recall of 61.32%. Compared with the original YOLOv5x, the mAP is improved by 2.56%, precision is improved by 5.85%, and recall is improved by 4.23%. The improved model is a promising algorithm for offshore wind turbine blade defect detection.