Research On Preprocessing Technologics And Multicomponent Analysis Methods Of Raman Spectra

Raman spectroscopy is kind of spectroscopic analysis technology arising from Raman scattering effect, has non-destructive, rapid, high efficiency, real-time advantages, which is widely applied in material, petroleum, biology, geology and so on. Raman spectroscopy includes qualitative and quantitative analysis. Qualitative analysis has been mature, but until now quantitative analysis especially multicomponent samples has not obtained a general acceptance method. However mixture is universal in reality, so it is necessary to research on multicomponent Raman spectroscopy quantitative analysis.Raman spectroscopy was studied with Raman spectroscopy quantitative analysis method. The main content includes principle introduction, algorithm realization and experimental analysis of preprocessing technologies and multicomponent analysis methods of Raman spectra.To begin with, we introduce characteristics of Raman spectra and Raman spectra noise characteristic and de-noising methods; analyze detailedly Raman spectroscopy quantitative analysis steps Analyze and compare commonly used methods of Raman spectra preprocessing such as Smoothing, Standard Normal Variate, Orthogonal Signal Correction, Wavelet Transform. Analyze and compare commonly used algorithm such as Principle Component Regression, Partial Least Square, Artificial Neural Network, Genetic Algorithm.Next, we introduce theory and algorithms of Wavelet Transform used for de-noising Raman spectra. We introduce a new method which makes the improvement to the threshold function and threshold, then it was applied to pure Raman spectra and compare d with other preprocessing methods such as Smoothing, Wavelet Transform Modulus Maxima Method, Wavelet Transform Threshold Method. The experimental result shows that the new method is able to eliminate spectroscopic noises and interferences as well as reserve major information, superior to other methods.Finally, we use Immune Genetic Algorithm based on Vaccination program to analyze the Raman spectra data of ten species samples, design genetic operator and immune operator for decimal coding, discuss algorithm parameters such as population size, crossover probability, mutation probability, vaccination probability, termination conditions. The experimental result shows that the feasibility, accuracy and effectiveness of this algorithm.