Research On Fast Reconstruction Algorithm Of Multi-channel Raman Spectra And Its Software Implementation

As a powerful tool to characterize chemical and biological information,Raman spectral imaging technology has been widely used in biomedical and other fields.However,due to the weak Raman signals and long scanning time,which severely limits its development.As a new Raman imaging technique,multi-channel Raman imaging is one of the effective ways to solve this problem,which acquires the partial Raman signals of whole imaging area simultaneously through multiple channels.Reconstructing Raman spectra is one of the key techniques in multi-channel Raman imaging.In addition,due to the complexity of the system,nonlinear factors such as noise are often introduced in the experiments,which leads traditional linear algorithms cannot obtain good reconstruction results.In order to solve these problems,this paper explores the relevant Raman signal processing algorithms,and builds a multi-channel imaging system and writes supporting signal processing software.The main work of this article is as follows:A fluorescence background subtraction algorithm based on median filtering and un-uniform b-spline is proposed.Firstly,the spectral data is smooth by locally weighted linear regression algorithm.Then,the internal knots of the spectral data are adaptively selected by using difference calculation,setting thresholds and windows.Finally,the spectral baseline is fitted by median filtering and un-uniform B-spline algorithm.This method has good flexibility and universality,and the performance is better than the traditional algorithms.A Raman spectral reconstruction algorithm based on polynomial kernel partial least squares is proposed.Firstly,the calibration samples are adaptively optimized by calculating similarity factors.Then,the multi-channel response matrix is mapped by the kernel function,and the correspondence between the multi-channel response matrix and the full Raman spectra is established by partial least squares.Finally,the Raman spectra are reconstructed by the established coefficient matrixes.This method is suitable for Raman spectral reconstruction under nonlinear condition,and the reconstruction effect is better than the traditional algorithms.A Raman spectral reconstruction algorithm based on gaussian kernel principal component analysis is proposed.Firstly,the optimized calibration Raman spectra data set is mapped to a high-dimensional space by gaussian kernel function.Then,the basis functions are extracted by principal component analysis.Finally,the basis function coefficients corresponding to the multi-channel measurements are obtained by pseudoinverse.Compared with traditional algorithms,this method can effectively reduce the influence of nonlinear factors on the Raman spectral reconstruction,and the accuracy and robustness are greatly improved.A Four-channel Raman spectral imaging system is constructed and the supporting signal processing software is written.The software is written by C#and MATLAB.It implements all of the algorithms mentioned in this paper.The software achieves multichannel data reading,spectral reconstruction,Raman imaging and other functions.In addition,it can be effectively used in combination with the built system.