Measurement ( IF 3.364 ) Pub Date : 2020-09-15 , DOI: 10.1016/j.measurement.2020.108436 Yasheng Chang; Jianmin Gao
In this study, we present a novel method that can measure the similarity of two digital radiography (DR) images from pressure pipeline welds. Firstly, in order to obtain a uniform weld image, a new tilt angle detection algorithm is proposed based on the fitting line of edge midpoint, which can quickly and accurately detect the tilt angle of a pipeline in a weld image. Secondly, a new algorithm based on the convex points of the pipeline edge is proposed to extract and segment the weld area. Thirdly, local entropy of the image is introduced to enhance the texture of the weld area. Finally, the Kullback–Leibler divergence is used to calculate the similarity between two images. The proposed method provides an accurate calculation of the similarity between pressure pipeline weld DR images. The method is evaluated using a set of 100 pipeline girth weld DR images. It shows significant efficacy for tamper detection in DR weld images with an accuracy of 98%. This work contributes towards the improvement of the management of field operators, and achieves the purpose of improving welding quality.