Research On X-ray Weld Defect Detection And Recognition Algorithm Based On Independent Analysis

With the development of pattern recognition technology and computer in the industrial field,computer automatic film evaluation has gradually replaced artificial identification and applied to the non-destructive testing field of important industries such as petroleum welded pipe with its unique advantages.However,in practical applications,the traditional machine identification requires relatively high samples and is easily affected by various factors,which may affect the detection rate and false detection rate of the detection and identification equipment.In this regard,for small sample objects,it is particularly important to study a method with high accuracy,good real-time and certain applicability for the identification of defects in petroleum welds.In this paper,X-ray inspection images of submerged arc welded pipe welds are taken as research objects.For the main defects such as pores and cracks in the weld area,three steps of image pretreatment,feature extraction and classification are used to realize the weld defects.Classification and recognition of images.Firstly,in order to accurately find the boundary information of the weld,a series of processing is performed by means of mean noise reduction,sin function enhancement,Ostu segmentation and edge detection,and the information of the boundary line is extracted by Hough transform.By comparing the threshold-based and density-based clustering-based segmentation methods,the defects in the weld zone are perfectly segmented.Secondly,the defects are described by using seven geometric features and shape features,and the corresponding feature vectors are established.The principal component analysis(PCA)and independent component analysis(ICA)are used to extract the feature features.Finally,the 7 types of shape feature parameters are used as the classification input,and the support vector machine is used to classify and identify the two types of defects.The results of 14 groups of experiments showed that the classification recognition rate of PCA was 85%.when extracting three principal elements,and the highest recognition rate was93%.when independent component analysis extracted 5 independent base images.Compared with PCA,with the increase of the number of components,the ICA algorithm will always be superior to the PCA algorithm,which can realize the identification of weld defects.