Three-dimensional fluorescence spectrum feature extraction method based on two-dimensional Gabor wavelet

Abstract

The invention discloses a three-dimensional fluorescence spectrum feature extraction method based on two-dimensional Gabor wavelet. Comprises the following steps: S1, acquiring original three-dimensional fluorescence spectrum data; S2. Removing the Rayleigh scattering in the original spectrum by Delaunay triangle interpolation; S3. Removing the Raman scattering in the original spectrum by subtracting the blank background solvent; S4. Using multidimensional S-G smoothing for spectral denoising; S5. Convolutional filtering is performed on Two-dimensional Gabor wavelet functions generated by different scales of expansion and contraction and rotation and three-dimensional fluorescence spectra after pretreatment of S1-S4 obtain different Gabor coefficient characteristic matrices. S6. Reducing the characteristic dimension of the Gabor coefficient characteristic matrix obtained from S5 by using the method of block statistics; S7. The different characteristic matrices obtained by the method instep S6 are recombined to obtain the final characteristic matrices of the three-dimensional fluorescence spectrum. The method improves the efficiency of extracting texture information from 3D fluorescence spectra and lays a foundation for water quality monitoring based on 3D fluorescence spectra.

Description

technical field [0001] The invention relates to a feature extraction method of three-dimensional fluorescence spectrum, in particular to a three-dimensional fluorescence spectrum feature extraction method based on two-dimensional Gabor wavelet. Background technique [0002] Three-dimensional fluorescence spectroscopy is a kind of fluorescence spectrum obtained by fluorescence analysis method. The three-dimensional fluorescence spectrum is a three-dimensional spectrogram obtained with the excitation wavelength and emission wavelength as variables, and can obtain the fluorescence intensity when the excitation wavelength and emission wavelength change simultaneously. According to the principle of fluorescence generation and emission, material molecules have different absorption efficiencies for different wavelengths of excitation light, and different excitation efficiencies. Therefore, in the detection of organic substances based on three-dimensional fluorescence spectroscopy, ...

AI Technical Summary

Problems solved by technology

These methods can be verified for applicability in specific research fields and application scenarios, but there are certain limitations in the field of identification of organic pollutants in pipe network water

Among them, the parallel factor method needs to specify the number of factors at the beginning of the algorithm, which may not be suitable for unknown pollutant scenarios; the principal component analysis method for feature extraction of the three-dimensional fluorescence spectrum is to pull the three-dimensional fluorescence spectrum in the direction of the row vector or the direction of the column vector Straight into a two-dimensional matrix, a large amount of spectral image texture information is lost, and it cannot be well adapted to the identification of organic species when the spectral characteristics of pollutants are close

Images