Research On Data Processing Method In Eddy Current Thermography

As a new type of active thermal imaging nondestructive testing technology,eddy current thermal imaging has been widely used in the detection of cracks,corrosion,residual stress and other defects of metal materials.The defect feature enhancement extraction algorithm has attracted the continuous attention of researchers because it can enhance the contrast of defect images,reduce the amount of data of detection results and improve the detection efficiency.At present,eddy current thermal imaging processing methods mainly include Thermographic Signal Reconstruction(TSR),Fourier Transform(FFT),Principal Component Analysis(PCA),Independent Component Analysis(ICA),Sparse Decomposition(Sparse Decomposition),etc.However,there is no unified thermal image processing effect evaluation system,which hinders the implementation of the algorithm in engineering.Based on the physical model of infrared thermal imaging,this paper analyzes the infrared thermal response characteristics of surface and subsurface defects,and the theoretical basis of five common eddy current thermal imaging data processing methods and their physical meanings in eddy current thermal imaging.Then,according to the characteristics of eddy current thermal imaging sequence in time domain,frequency domain and space domain,the data processing results of different eddy current thermal images were compared and analyzed by using the objective evaluation indexes with reference and without reference.Theoretical and experimental results show that in the absence of reference,for surface defects,TSR has the best thermal image processing results under average gradient,information entropy and standard deviation;sparse decomposition has the best thermal image processing results under gray mean and spatial frequency;for subsurface defects,sparse decomposition has the best thermal image processing results under average gradient,gray mean and spatial frequency,while PCA performs best under information entropy and standard deviation.Under the frame of reference,for surface defects,FFT has the best thermal image processing results under SNR,MSE and cross entropy,TSR has the best thermal image processing results under SNR,MSE,joint entropy and cross entropy;for subsurface defects,TSR has the best thermal image processing results under SNR,MSE and cross entropy,and PCA has the best thermal image processing results in joint entropy.In this paper,different thermal image processing methods are compared under the same evaluation index,and the quantitative evaluation results of each method in different scenes(such as surface and subsurface defects)are given.The research can not only provide theoretical basis for the research of new processing algorithms,but also provide guidance for the selection of data processing methods in engineering application,which has important theoretical and engineering values.