Research On Pulsed Eddy Current Thermal Imaging And Image Analysis Of Defects In Ferrous Components

In the process of casting and using some metal parts in electric power,military industry,rail transit,pipeline transportation and other equipment,all-round non-destructive testing should be used to ensure the safety.In the process of using pipeline transportation equipment,metal parts may have hidden defects or fine crack defects due to long-term use,which may lead to potential safety hazards during operation.Therefore,it has practical significance for the research of metal parts nondestructive testing.In the process of nondestructive testing of metal parts by pulsed eddy current thermal imaging,the thermal image of metal parts has some problems,such as blurring of defect edge,serious non-uniform background interference due to the uneven thermal emissivity of the tested materials and external interference,which also increases the difficulty of infrared thermal image defect target detection.Therefore,this paper uses pulse eddy current thermal imaging technology to detect defects in iron parts,and conducts in-depth research around infrared thermal image preprocessing and target defect detection.The main research content has the following aspects:(1)This paper analyzes the basic principles of non-destructive testing and infrared non-destructive testing of pulsed eddy current thermal imaging,and analyzes the characteristics of infrared thermal images.In order to improve the detection effect and reduce unnecessary interference,the Block-Matching and 3-D Filtering(BM3D)and improved Top-hat algorithm were studied and the implementation steps were introduced in detail to achieve denoising and suppression of non-uniform background.(2)The infrared thermal image preprocessing based on improved RPCA and Weighted Nuclear Norm Minimization(WNNM)is studied.The disadvantages of traditional robust principal component analysis(RPCA)model to treat each singular value equally are redefined The weight distribution method adopts the allocation strategy of large singular values with small weights and small singular values with large weights,and uses an inexact solution algorithm.Secondly,a Weighted NuclearNorm Minimization denoising model is constructed,and finally the enhancement of the defect target is achieved.(3)Focusing on the analysis of Pulse Coupled Neural Networks(PCNN)defect target detection algorithms,the initial threshold adaptive selection method of the maximum inter-class variance algorithm is used to select the initial threshold of the pulse-coupled neural network algorithm.For the pre-processed image,the maximum inter-class variance algorithm is used to traverse the image pixel values,the maximum value of the variance in the image is solved,and the corresponding gray value is used as the optimal threshold.Use this threshold as the initial threshold for the first iteration of the PCNN algorithm,segment the image,and finally get the target information.