Weld Slim Line Defects Extraction Based on Adaptive Local Threshold and Modified Hough Transform

Abstract

Computer aided evaluation is objective, scientific and standardized, and automatic defects detection technology based on X-ray digitized images is the core of computer aided evaluation. However, for the slim line defect whose average width is not more than three pixels, such as some weak incomplete penetrations, cracks and so on, its contrast to background is extremely low, and normal X-ray defects detection algorithm can not detect it effectively. This paper proposes a special defect segmentation algorithm to detect this kind of defects. Firstly adaptive local threshold is performed column by column to extract slim line defects preliminarily, and then the improved Hough transform is utilized to eliminate the noise, and segment the line defects. Corresponding experimental results show that the proposed algorithm is effective to detect low-contrast slim line defects in X-ray digitized images of different grey ranges Keywords-weld defects detection; X-ray images ; image processing ; Hough transformI.

Neural network segmentation methods for fatigue crack images obtained with X-ray tomography

“…Thresholding the gray level histogram is the simplest and most widely used method for image segmentation, but it is highly sensitive to noise and artifacts and thus not very efficient in the vicinity of the crack tip where the crack opening is low, even if the sample is under load. Feature extraction methods, including Hough transform [18], finite plane integral transform [19] or filtering based on Hessian matrix [20], assume that the crack has a prescribed shape in the image such as a line (in 2D) or a plane or a portion of plane (in 3D). Such methods have been used for example in the case of bones [12] or concrete [21] for which crack opening can be very small (lower than 1 µm).…”