Image Mosaic Using SIFT for Bridge Cable Surface Flaw Detection

Abstract

A machine vision system is developed to detect the cable surface damage of cable-stayed bridge. In the system, four cameras are employed to acquire images around the cable surface. So the same one defect may be split into several images. Image mosaic had to be done to obtain a complete defect image for further process. The feature of cable surface image is simple and its illumination is heterogeneous. So the Scale Invariant Feature Transform (SIFT) feature matching algorithm is suitable for the image mosaic. Firstly, cable surface images should be preprocessed. Secondly, the SIFT algorithm achieves the detection, extraction, description and matching of feature points for defect images. Finally, image fusion is implemented to acquire the integrated image and a complete defect will be showed in the image. Experimental results show that using SIFT for the cable defect image mosaic has good effect on improving the detection accuracy and integrity for cable surface defects.