Particle Swarm Optimization Algorithm Establish the Model of Tobacco Ingredients in Near Infrared Spectroscopy Quantitative Analysis

57 tobacco samples were chosen, the near-infrared diffuse reflectance spectra at five different parts of tobacco leafs were averaged as the original spectra, the range of 4000-9000 wavenumber of spectral information was selected after wavelength selection, first-order differential was used to eliminate the impact of offset, PCA was used to reduce the dimensions and select the most representative of the principal components of the sample information. After these, use PSO and the data of chemical composition of nicotine and total nitrogen, to establish the models of near-infrared spectra of the main ingredient in tobacco quantitative analysis. In which, nicotine and total nitrogen quantitative analysis of modeling set correlation coefficients were 0.8886,0.9290; the prediction set correlation coefficients were 0.8786,0.8487. RMSEC values were 0.4737,0.2215; RMSEP values were 0.6417,0.2677. The results show that: PSO can be used to establish the model of nicotine and total nitrogen by near infrared spectroscopy quantitative analysis in tobacco. Keyword. Tobacco, Particle Swarm Optimization (PSO), Principal Component Analysis (PCA), Near infrared spectroscopy 1 Bibo Ma: Master. Research field: Spectral analysis. Haiyan Ji: Professor. Research field: Spectral analysis and sensing technique.