DR Image Detection And Identification Of Defects In Metal Additive Manufacturing

Nowadays,metal additive manufacturing has made rapid achievements in actual production and processing,and safety problems in production have always been the top priority,but under actual conditions,after the workpiece of additive manufacturing is formed or in work,there may be defects or defects that expand,and the generation and expansion of some defects are unavoidable.At present,defect detection and identification is still manually tested by specialized inspectors based on personal knowledge and work experience,resulting in a lack of objectivity in the results.The rapid development of nondestructive testing technology and computer image processing technology has injected fresh vitality into defect detection and identification,X-ray-based DR imaging can clearly photograph defects that are not easy to observe inside the workpiece,and the application of image processing technology in defect detection can reduce the burden of manpower and reduce false positives,ensuring the classification recognition rate,It has important engineering practical significance for promoting the application of additive manufacturing technology and its industrial development.In this paper,in view of the low contrast and uneven illumination characteristics of X-ray-based metal additive manufacturing workpiece defect DR images,this paper proposes linear and nonlinear grayscale transformations to change their contrast and brightness,studies and analyzes the noise types in DR images,uses adaptive median filtering to filter and denois the impulse noise,and denoises the wavelet threshold denoising quantum noise,system noise,and fuzzy noise;proposes a method that combines fuzzy C mean clustering algorithm with mathematical morphology to segment DR defect images.This solution solves the problem of the target part sticking to the background,blurred edges,and excessive segmentation in the traditional segmentation algorithm to make the defect details lost;the 12 feature describers of compactness C,aspect ratio Z,gray mean,gray scale variance,gray scale entropy,and 7 invariant moments are used as feature parameters for defect feature extraction,making the data set more comprehensive and differentiated;The BP neural network is used as a classifier to identify and classify the DR defect images of metal additive manufacturing workpieces,and the recognition rate of various defects is above90%.It is proposed that the convolutional neural network based on deep learning has a transfer learning classification of the DR defect image of the metal additive manufacturing workpiece,which can directly train the image,extract the final output classification of feature extraction,and more automated and intelligent,and the final average accurate recognition rate through experimental verification reaches 81.48%,which proves that it is feasible to use the convolutional neural network to classify and identify the DR image of the metal additive manufacturing workpiece.