摘要

Rayleigh and Raman scatterings inevitably appear in fluorescence measurements, which make the quantitative analysis more difficult, especially in the overlap of target signals and scattering signals. Based on the grayscale images of three-dimensional fluorescence spectra, the linear model with two selected Zernike moments was established for the determination of humic acid, and applied to the quantitative analysis of the real sample taken from the Yellow River. The correlation coefficient (R-2) and leave-one-out cross validation correlation coefficient (R-cv(2)) were up to 0.9994 and 0.9987, respectively. The average recoveries were reached 96.28%. Compared with N-way partial least square and alternating trilinear decomposition methods, our approach was immune from the scattering and noise signals owing to its powerful multi-resolution characteristic and the obtained results were more reliable and accurate, which could be applied in food analyses.