Welding Line X-ray Image Processing Based On Compressive Sensing

Theer could be some defects exist in the welding line x-ray image in the welded pipe ofoil tube making industry. It is especially important to judge whether there are defects or not onthe pipe X-ray welding line image duirng the production process, the judgment can ensuresecure operation of oil and gas pipelines and prolong their servicc life. ITiis paper focuses onthe theory of compressive sensing in the petroleum pipe X-ray welding line image rebuildingand dcfcct detection.This paper applies compressive sensing (CS) theory directly to the welding line imageswhich has captured by X-ray real-time imaging system, A redundant dictionary of defects andnon-defects models is constructed based on the CS theory, the sparse representation of dcfcctsimages and non-defects images are calculated with the basis of the redundant dictionary. Theimage is rebuilt by OMP algoirthm and defects detection experiment is judged by imagecoefficients weight sum standard at last. According to the detection expeirment, the paper hasgot the sensitivity and specificity curve of expeirment results.It is not easy to analyze the welding line x-ary image and identify its defects because ofits characteristic of lower contract, noise, higher fluctuate and change of background, andfuzzy welding edge of the image. According to these characteristics, some preprocessingoperation on the image should be made including noise iflleirng, image enhancement, edgedetection, Hough transformation，segmentation and normalization. The preprocessing hashighlighted the defects feature, and got the segmented original image.The compressive sensing theory is applied to the subtracted preprocessed X-ray weldingline images. The subtracted sample images arc dcfccts and noises images. A redundantdictionary of defects and noises is constructed based on compressive sensing theory. On thebasis of the redundant dictionary, the sparse representation of detecting defects and detectingnon-defects are calculated, the dcfccts detection experiment is judged by two standards: imagecoefifcients weight sum standard and the nonc-zcro coefficients standard. The paper gets thesensitivity and specificity3D diagram of expeirment results. Based on compressive sensing,the paper analyzes non-preproccsscd and preprocessed X-ray welding line image experimentresults and selects one sample to judge the X-ray welding line image accurately whether there are defects or not.The CS theory has proved a good application effect on the X-ray welding line imagedefects detection according to the research.