Research On Multi-channel Raman Imaging And Its Key Techniques

Raman imaging is a non-destructive,label-free spectral imaging technology.As a powerful tool to characterize chemical and biological information,it has been widely used in many fields such as biomedicine and material detection.However,the spontaneous Raman signal is relatively weak.Therefore,it takes a long time-integratio to obtain a Raman image of a high signal-to-noise ratio,which directly affects the quality of the Raman dynamic imaging.In order to solve this problem,a fast,high-performance,integrated multi-channel Raman imaging system is developed,in this paper.Then a series of special Raman spectrum calibration-reconstruction algorithms are proposed,as well as the processing methods of Raman spectrum and image.Finally,the high-quality Raman spectra and images can be quickly reconstructed,which will promote the application of Raman imaging technology in dynamic fields.The main research content of the paper is as follows.A set of four-channel Raman imaging system has been built and the corresponding visual processing software also has been written in the paper.The software is written by MATLAB,and its functions mainly include real-time adjustment and control of the imaging system’s CCD camera,completion of multi-channel imaging data reading,rapid reconstruction of Raman spectra and Raman images of the band of characteristic peak,and post-processing of the reconstructed Raman spectrum and image.Aiming at the problem of Raman spectrum reconstruction,three special Raman spectrum calibration-reconstruction algorithms based on sample optimization and nonlinear variable correction have been proposed.That is.Raman spectral reconstruction algorithm based on global weighted linear regression,Raman spectral reconstruction algorithm based on sample optimization and PCA,Raman spectral reconstruction algorithm based on KECA.Meanwhile,two optimization algorithms are proposed to optimize the sample,that is,an optimization algorithm based on sample similarity and an optimization algorithm based on sample response values.In order to reduce the interference from nonlinear factors,polynomial regression,PCA,and KECA are introduced,simultaneously.The experimental results show that the three reconstruction algorithms have completed the reconstruction of the Raman spectrum better,and the reconstruction accuracy has been improved by nearly 30%～40%compared with the traditional algorithms.Among those algorithms,KECA has 8.17%and 5.45%improvement compared with global weighted linear regression and sample optimization and PCA,respectively.More importantly,the time to reconstruct the Raman image based on our algorithms is much lower than that of traditional imaging mode,which provides theoretical support for the further application of Raman imaging technology in fast dynamic systems.In order to remove the noise and fluorescence background in the reconstructed spectrum,two algorithms are proposed,that is the optimized EMD algorithm based on LMS adaptive filtering and the baseline correction based on adaptive window spline fitting.The experimental results show that the two algorithms can effectively complete post-processing of the reconstructed spectrum.At the same time,they have a fewer parameters and simple structure,which are suitable for processing of large data.Aiming at the problem of noise and fluorescence background in the reconstructed Raman image,two processing methods are proposed,which are processed based on the numerical principle and physical principle.The experiment results show that both methods can complete the post-processing of the reconstructed image.However,the first method ignores the physical meaning of the image itself,which causes the loss of useful data from the Raman image.The second method removes the noise and baseline of the reconstructed spectrum,firstly.Thereafter the reconstructed Raman image is obtained,therefore,the original useful information of the image is greatly preserved.