Research And Application Of Spectral Reflectance Reconstruction Method Based On Multi-kernel Support Vector Regression

The spectral reconstruction method acquires the spectral images in different channels by multi-spectral imaging technology,and reconstructs the surface reflectivity of the object by using a large amount of spectral information and spatial information of the image,which helps to obtain the color information of the object.The imaging device used in the method has low cost,and there is no strictly limits on the area and flatness of the object.The reflectance can be obtained quickly,efficiently,and accurately with fewer channels,and has certain real-time and applicability.In this paper,the spectral reflectance reconstruction algorithm and color restoration are studied,and the main contents include:(1)Research on spectral reflectance reconstruction based on multi-kernel support vector regression.When the common spectral reflectance reconstruction method is used to process high-dimensional and nonlinear spectral data,there are problems of low reconstruction accuracy,poor reconstruction ability and generalization ability of the reconstruction model.Therefore,the spectral reflectance reconstruction method of multi-kernel support vector regression is studied in this paper.This method uses the product of the Cauchy kernel function and the polynomial kernel function as the multi-kernel kernel function of the support vector regression machine.The tensor product is used to characterize the spectral data with higher dimensional features,which is beneficial to the enhancement of model performance and the improvement of reconstruction accuracy.The kernel parameters were selected by trial and error method to avoid the influence of parameters on the performance of the model,and the paper's experiments were carried out with ral color card and simulated murals as samples.The results show that compared with the pseudo-reverse reconstruction method and the single-kernel support vector regression reconstruction method,the reconstruction algorithm based on multi-kernel support vector regression improves the reconstruction accuracy and reduces the chromatic aberration.(2)Mural color restoration based on spectral chromatic aberration model.A spectral chromatic aberration recovery model was designed in this paper for the problem that the fading of mural pigments was difficult to recover due to factors such as dust.The model simulates the pigment fading process and combines the pigment reflectivity obtained by multi-kernel support vector regression spectroscopy to establish a sample library of pigment fading.Moreover,The spectral space is mapped to the chromaticity space.By analyzing the pigment fading law,the spectral chromatic aberration recovery function of the linear relationship between the pigment color difference and the dust amount is constructed.The experimental results show that the color recovery of mural fading pigments can be achieved by using each pigment color difference recovery function combined with the pigment fading sample library,which provides an effective method for the study of mural fading pigments.