A Study On Image Processing And Automatic Recognition Of X-Ray Weld Defects

With the continuous development of image processing technology and the moredetection efficiency, economy, and objectivity of the computer intelligent recognition, theartificial recognition of weld defects is gradually replaced by the computer intelligentrecognition. In order to replace artificial assessment piece, the computer intelligentrecognition accuracy must be increased. Therefore, the high recognition rate of defectsbecomes the aim of constant pursuit. The improvements of image processing and recognitionmethods are the research emphasis in the computer intelligent recognition field.This research focuses on the process of X-Ravs weld defect detection. First, thereal-time video stream is transferred to the receiving server via Ethernet. An acquisition cardis used to capture the fixed form images from the VGA video. Then, according to thecharacteristics of the defect images, a suitable image processing coniaining mean filter. Sinenhancement. Ostu segmentation. Sobel edge extraction and DBSCAN cluster eic.. is used tothe image with pertinence. The set of methods can effectively handle the weld image and ableto extract defects image and noise portion accurately. Third, the characteristic parameters arecalculated based on the external feature. To determine whether the types of the parameters aresuitable, the feasibility of parameters is veriifed through three-dimensional space experiments.Last, the pattern recognition is detected by the methods of CS and SVM based on theextracted images and characteristic parameters. The results showed that both detectionsensitivity and speciifcity are increased.