Fatigue crack image stitching method for steel bridge decks based on improved SURF algorithm

ObjectiveComputer vision is an efficient approach to detect fatigue cracks on orthotropic steel bridge decks, where crack image stitching is involved. Traditional image stitching method, based on scale invariant feature transform algorithm, has the limitations of unstable feature point pairing, severe image distortion, indelible stitching traces and long stitching time. Therefore, this study proposes an innovative fatigue crack image stitching method for steel bridge decks based on speeded up robust features algorithm.MethodThe proposed method improved the feature point pairing accuracy by improving the matching algorithm, and enhanced the calculation efficiency of transformation matrix by optimizing the random sample consensus algorithm. Moreover, the shape correction function was introduced to control the images deformation, and the image fusion automatic capture algorithm was used to improve the stitching effect. Furthermore, fatigue crack image detection tests were conducted on steel bridge decks to validate the method. The stitched images were quantitatively evaluated using several indexes, e.g., RMSE, spatial frequency, entropy and stitching time.ResultThe proposed innovative method can effectively eliminate the wrong matching of feature points, generating stable and uniformly distributed matching pairs with parallel alignment. It significantly mitigates image deformation after geometric transformation and yields seamlessly fused images with no visible stitching seams or color discrepancies. Consequently, it achieves the high-precision stitching and panorama generation of crack images, and solves the pre-existing problems of stitching error even stitching failure.ConclusionThe proposed innovative method's stitching accuracy is higher, the stitching effect is better, and the stitching time is shortened by 19% on average, which has evident advantages compared with the traditional stitching method. It would be further applied in the practical engineering.