Research On Image Processing Technology Of Girth Weld DR Inspection Data

Nowadays,oil and gas pipeline industry developing fast,flaw detection can control over the quality of pipe,pipe nondestructive testing can not damage the pipe under the premise of evaluation,among them,the X-ray detection compared with other detection scheme has fast film imaging and evaluation of low cost,high accuracy,wide applicable scope,without being limited by the environment and other advantages,is one of the main detection methods of weld defect detection field.On the one hand,the image contrast is low,which is affected by noise,resulting in overlap between defects and noise,and small background gray difference.In order to improve the evaluation efficiency of inspectors,this paper focuses on the defect extraction based on image processing technology.On the other hand,in the current X-ray girth weld image detection by manual evaluation,it is easy to misjudge and miss judgment under fatigue state,and it is subjective.In this paper,the intelligent defect recognition based on deep learning is researched,and the evaluation system of man-machine combination is finally realized.In terms of image processing,contrast and noise reduction were carried out for the girth weld image problem,and algorithm was improved for the loss of detail information by mean filtering.The ability of noise suppression was proved through comparative experiments.The image defect area is disturbed by the welding process and the imaging process,and the edges of the image are blurred.Therefore,this paper designed an edge extraction strategy based on the combination of OTSU threshold segmentation method and Canny operator to obtain area.In terms of intelligent recognition,this paper realizes defect recognition based on deep belief network and defect recognition based on convolutional neural network.By comparison,it is proved that convolutional neural network has better effect on defect recognition of girth weld images.This paper improves the accuracy and efficiency of manual film evaluation by improving the contrast of X-ray images,image denoising,and extracting defect edges.By comparing the performance of different neural network algorithms,it is proved that the scheme based on convolutional neural network has a better effect on defect identification and improves the efficiency of film evaluation.