Infrared Detection and Identification of Wind Turbine Blade Defects Based on Bimensional Filtering Empirical Mode Decomposition and Threshold Segmentation

The study focuses on the infrared nondestructive detection of inclusion-type internal defects in glass-fiber-reinforced plastic (GFRP) wind turbine blade specimens, which were designed to simulate the laminated material structure and typical hidden defects of in-service blades. To address the difficulty of detecting internal defects in in-service wind turbine blades, this paper establishes an active thermal imaging defect detection and recognition system using a halogen lamp as the infrared thermal excitation source and a high-resolution thermal imaging camera as the detection component. To improve the recognition of defect contour information in infrared images, a method combining bidimensional filtering empirical mode decomposition (BFEMD), Gaussian filtering, and Otsu threshold segmentation is proposed. The BFEMD procedure decomposes the infrared image into bidimensional intrinsic mode function components and residual components, Gaussian filtering suppresses noise in the selected components, and Otsu threshold segmentation extracts the defect contours. Experimental results show that the combined algorithm can enhance defect targets in infrared images, improve visibility and contour integrity, and provide a higher detection rate for wind turbine blade defects under different defect depths and materials.