An improved image stitching method for blades of wind turbine based on online repair technology

When machine vision technology is used for online defect detection in wind-turbine blades, existing image-stitching methods have difficulty detecting image features, correctly matching rates, and accurately registering images. Therefore, an image-stitching method suitable for the online repair robot platform of wind-turbine blades is proposed based on the improved accelerated-KAZE (AKAZE) method. The feature points of the wind-turbine blade crack image are detected using the AKAZE algorithm and described using a binary robust invariant scalable keypoint descriptor. A grid-based motion statistics algorithm is used for feature prematching, and the random sample consensus algorithm is used to optimize the feature matching results and calculate the image transformation model. A weighted blending algorithm is used to blend the overlapping areas of the images to obtain a high-resolution and complete image of the wind-turbine blade cracks. The stitching effect of the proposed method was verified on cracked wind-turbine blade images, comparing the method with other algorithms in terms of feature-point detection, correct matching rates, stitching quality, and efficiency. Experimental results show that the proposed method effectively implemented high-resolution wind-turbine blade crack-image stitching. Therefore, the improved AKAZE image-stitching method can support the overhaul task of a wind-turbine blade repair robot based on online repair technology.