China's wind power generation is rich in resources and mature technology, but has the problems of harsh power generation environment, high operation and maintenance costs due to complex operating conditions, and serious consequences of failures. For this reason, this paper proposes a more efficient defect identification method for wind turbine blades that have the longest downtime due to faults. Firstly, starting from the characteristics that the blade defects are darker than the surrounding and distributed in block or point shape, the blade images taken by UAV cruise are processed by grey scaling, filtering, histogram equalization and Grab-cut foreground segmentation. Secondly, a wind turbine blade defect recognition algorithm based on the adaptive parameter region growth algorithm is proposed, where the number of seed selection points and location information are planned through the results of image preprocessing and the conventional defect features of wind turbine blades; and a threshold adapted to a variety of defects is determined through the results of filtering and equalization. Finally, the image of defect recognition is demonstrated through morphological algorithm and framing optimization, the Mean Intersection over Union (MIoU) performance evaluation index is analyzed, and the effectiveness of the algorithm is verified through experimental data comparison.
Keywords: Adaptive parameters; Defect identification; Fan blade; Region growing algorithm; Wind power generation.
© 2024. The Author(s).