Adaptive segmentation of weld defects based on flooding

Abstract

RADIOGRAPHYTo improve the speed and accuracy of defect segmentation for automated radiographic NDT, this study proposes a segmentation method based on flooding (SMBF), in which an original line-flooding algorithm and a novel adaptive thresholding method are suggested. The adaptive thresholding method is based on the growth of the flooded area. SMBF firstly detects a defect by analysing the grey level intensity profiles of a radiographic image and labels the defect found with a seed point. Then, the flooding is carried out using the line-flooding algorithm, in which water starts from a seed point and flows to the neighbour region in column-by-column order. After flooding, the threshold value is determined by detecting an abnormal growth of flooded area. In addition, comparisons of SMBF with a segmentation method based on watershed and a segmentation method using background substation are provided. The experiments demonstrate that SMBF is able to significantly reduce the segmentation time and obtain a more accurate result.

Defect Inspection Based on Segmentation and Defective Tracking in Radiographic Image

“…In the defect detection in NDT&E the two major problems that possed a challenging design is the low quality of the RT image and algorithms with improper segmentation [10,11]. The RT image must have enough details and enhanced contrast to obtain a proper result.…”

Multi‐step radiographic image enhancement conforming to weld defect segmentation

“…Two problems restrict the accuracy improvement of defect segmentation in industrial NDT&E. One is an improper segmentation algorithm, and the other is the low quality of RT images [4,5]. To solve the former, we proposed an adaptive defect segmentation algorithm based on flooding in our previous papers [5,6]. To solve the latter, we plan to improve the quality of low-quality RT images in a follow-up study.…”