Research Of Hyperspectral CT Nondestructive Testing Based On Principal Component Analysis

X-ray Computed Tomography(X-CT)is able to obtain agreeable structural imaging without causing physical damage to the object.Therefore,in the field of non-destructive testing and medical imaging,X-CT has become a mainstream detection method.However,it has been found that,practically,X-rays are not single-energy electromagnetic waves.The interaction between X-rays and substances exist a series of reactions dominated by the photoelectric effect and Compton scattering,which means the characteristics of the attenuation varies in different substances.X-ray spectroscopy provides a theoretical basis for the analysis of materials within the object.Spectral CT is a kind of the X-CT system combined with the photon counting detector,in which X-ray spectroscopy can be applied to CT scanning,so that traditional X-CT is no longer limited to imaging the internal structure of the object,which means it has the potential to analyze internal materials.In this paper,a multi-color CT system based on principal component analysis was proposed,which was connected with the application of Hyperspectral CT expending the dimension of X-ray absorption spectrum(XAS)to obtain better material recognition results and multi-color CT images.On one hand,this paper introduced the CT reconstruction algorithms in detail.In order to lessened the effects that Filtered Back-Projection(FBP)is extremely sensitive to data noise and beam hardening effects,the statistical iteration algorithm significantly reduced the noise and artifacts of the images.On the other hand,the combination of principal component analysis and Artificial Neural Network was applied to Hyperspectral CT technique.After the process of reconstructing 7 kinds of 3D printing supplies and the other interference polymer separately,the XAS of test set specimens were reconstructed in the region of interest for classification.The principal component analysis was used again to extract the features of the attenuation from the CT images of different energy ranges,then the images was used to produce a multicolor CT image.The experimental results show that even if the training set samples and the test set samples are reconstructed in stages,as long as the consistency of the experimental conditions and agreeable precision of reconstruction is guaranteed,Hyperspectral CT still has the ability to build the library of reconstructed XAS for identifying.The corresponding multi-color CT images will have more intuitive material resolution.