In-service wind turbine blade damage location method based on UAV inspection image stitching

This study addresses the challenges of accurately capturing and localizing surface damage on large wind turbine blades using UAV-based inspection systems. We propose an integrated framework consisting of:(1) An advanced image feature extraction module that combines U-Net architecture with the Sobel operator for precise edge detection;(2) A robust sequence image stitching algorithm based on equal-width feature matching criteria. The methodology integrates a damage localization system via a Cartesian coordinate framework, achieved by fusing a damage recognition model with geometric transformation equations. These equations establish quantitative links between physical measurements and pixel dimensions, enabling a coordinate transformation pipeline to map defect locations from image pixels to actual blade dimensions. Experimental validation shows the proposed algorithm achieves localization accuracy within a 5% error margin across diverse simulated damage scenarios. This performance enables reliable quantitative analysis for engineering applications, representing a significant advancement in wind energy infrastructure inspection technology.