Application of principal component analysis to selection of characteristic wavelengths with total light scattering

作者：Tang Hong^*; Zheng Wen Bin; Li Xian Xia

来源：Optics and Precision Engineering, 2010, 18(8): 1691-1698.

DOI：10.3788/OPE.20101808.1691

• Contribution rate
• Differential extinction
• Experimental data
• Extinction spectra
• Extinction values
• First-order
• Inversion errors
• Monomodal
• Numerical simulation
• Particle size measurement
• Particle systems
• Principal Component Analysis(PCA)
• Principal Components
• Reference material
• Second orders
• Selection methods
• Selection of characteristic wavelength
• Spectral region
• Total light scattering
• Visible-infrared region
• Wavelength selection

To select the characteristic wavelengths which contain important information about particle sizes and refractive indexes in a total light scattering particle sizing method, a wavelength selection method was proposed by performing the Principal Component Analysis(PCA) for the extinction spectra, the first-order differential extinction spectra and the second-order differential extinction spectra of monomodal R-R particle system in visible and visible-infrared regions. This method firstly used the principal component transform to dispose the first-order differential extinction spectra, and then regarded the contribution rate of the principal component to the first-order differential extinction spectra at each wavelength as the mainly criterion to select the characteristic wavelengths. Meanwhile, the boundary wavelengths in the spectral region were also selected as the characteristic wavelengths. The numerical simulations on the monomodal and biomodal R-R distributions were performed, and the experimental data of reference materials were also validated.Experimental results show that the inversion errors of reference materials based on the PCA selection method are less than 3%. It is concluded that the selection method has advantages of convenient computation and easy realization, and can ensure to the extinction values of selected characteristic wavelengths to show the most valuable information about the particle size distribution.

• 出版日期2010
• 单位中国计量大学