Research On Defect Detection And Identification Based On X-ray Weld Image

In 2020-2022,China’s oil and gas pipeline industry will continue to grow despite the impact of the new crown pneumonia epidemic at home and abroad,and according to statistics,by the end of 2022,the cumulative mileage of oil and gas long-distance pipelines built in mainland China reached 185,000 km.By 2025,the total mileage of long-distance pipelines in China is expected to exceed 240,000 kilometers.The quality of pipeline welds will directly affect the transportation safety of oil and gas pipelines.The traditional manual evaluation method is very subjective and prone to more misjudgments and omissions,which is no longer suitable for digital image evaluation.Computer-aided evaluation method has gradually become a major research direction in the field of non-destructive testing because of its advantages of objectivity and stability.Based on the analysis of the feature of X ray weld,this paper realizes the detection and identification of weld defect by means of image processing and pattern recognition.The main research is as shown below:(1)For the traditional defect detection algorithm is vulnerable to the influence of the weld background,an improved background estimation weld defect detection algorithm is proposed.Firstly,the low contrast of the image and the removal of noise interference are improved by preprocessing,then the maximum threshold segmentation and affine transformation are used to obtain the corrected weld area,then modified median filter background estimation algorithm and morphological filter significance detection are used to improve the defect and background contrast,then the multi-directional multi-level gradient is used to further remove the background edge residue,and finally the defects are obtained using adaptive threshold segmentation.The algorithm tests the advantages of the proposed algorithm through comparative experiments.(2)Aiming at the problem of multi-classification identification of weld defects,a classification algorithm based on PCA and binary tree SVM is proposed.Firstly,the algorithm extracts the geometric features and grayscale features of the weld defect,constructs the corresponding feature vector,and then uses the principal component analysis method to analyze the principal element,deletes the redundant data,reduces the dimension,and finally the optimised feature vector is used as the feature parameter of the weld image binomial tree support vector machine for recognition,and the classifier is used to identify the weld defect.Experiments show that the proposed algorithm can improve the defect recognition rate effectively and has a good effect.(3)Complete the design and integration of X-ray weld image defect intelligent identification platform.Taking a weld image public dataset as the research object,the overall framework of the software platform is first analyzed and designed,and then the defect detection and classification algorithm is embedded in the platform,and the specific functions of different modules are individually designed and implemented,and finally test cases are written for the functions contained in the software platform for complete testing.