Surface Defect Detection of Cable Based on Threshold Image Difference

Abstract

When using computer vision technology to detect defects on cable surface of cable-stayed bridge, the image is easy to be affected by natural light and produce uneven illumination, which will lead to missing detection of defects in relevant areas during detection. Meanwhile, the non-linear edge shape of cable also brings inconvenience to the extraction of effective areas. Therefore, a method of surface defect detection for cable of cable-stayed bridge based on threshold image's difference is proposed. Firstly, the gray scale and gradient characteristics of cable images were analyzed, and based on the manual ROI extraction method, the automatic ROI extraction method of cable was designed based on Gaussian mixture model. Then, an appropriate image processing algorithm is selected to enhance the cable image. Finally, the threshold results of the image to be detected and the original image are differentiated, and the existence of defects is determined by calculating the maximum connected area. The experimental results on the bridge show that the effective area extracted by this algorithm is accurate, and it can effectively reduce the bright spots or shadows caused by uneven natural light. Meanwhile, it has a good detection result for the defects of cable surface, such as broken wire, deformation and wear, which has a high application value.