Research On Image Model Identification Of Pipeline Weld Defects

The quality of pipeline weld affects the safety of oil and gas pipeline operation.With the development of big data,X-ray imaging and deep mining technology of data,the automatic identification and analysis techniques of pipeline weld images have become an important research field in the oil and gas industry.One of the significant methods of non-destructive detection is X-ray imaging.Currently,the identification of weld negatives is only determined by manual.The criterion of subject is subjectively influenced by the individual,and it is susceptible to the work intensity and environmental conditions during the discriminating process,which further affects the efficiency and accuracy of the test.In addition,the preservation of the ray film is time-limited,which brings great inconvenience to subsequent inspection and verification.Therefore,it is necessary to make it intelligent through computer-aided evaluation technology,improve the identification accuracy of weld,and ensure the quality and safety of pipe weld.Based on the digital images of X-ray films,this paper detected the possible defects in the weld images of oil and gas pipelines and identified weld defects through image preprocessing,image transformation and enhancement as well as feature calculation and edge detection of image defects.The results achieved are as follows:(1)The types of noises produced in digital images of weld films were analyzed,and the images of X-ray films were preprocessed by the median filter and mean filter.Fourier transformation were performed on the digital images of X-ray followed by image contrast enhancement processing,A method for maximal inter-class variance threshold segmentation and principal component analysis dimension reduction processing for weld image defect regions is established;(2)Through the transformation of the pixel size of the weld image,the edge of the weld image is processed by morphological operation,Considering the size and direction of structural elements affecting the results of morphological edge detection,a multi-operator fusion processing technique is proposed;(3)The defect features of the weld images were described and their images were analyzed.and geometric and texture parameters were selected as the characteristic parameters of image identification.An improved full local ternary mode(CLTP)algorithm is proposed to extract image texture features,the feature values of images were extracted by contour tracking;(4)The global optimization classifier method(M-SVM)was applied to study the weld defects,a support vector machine(SVM)model of classifying weld defect characteristics based on the binary tree was established to identify and classify the weld defect images of pipeline achieved;(5)Based on the above work,an automatic identification and evaluation system of pipeline weld films was developed in this study in order to identify the weld defects of oil and gas pipelines accurately.