Application of Zernike moments for the quantitative analysis of polycyclic aromatic hydrocarbons based on fluorescence three-dimensional spectra

Abstract

A novel method combining three-dimensional fluorescence spectroscopy, Zernike moments, and generalized regression neural networks (GRNN) is proposed for the quantitative analysis of polycyclic aromatic hydrocarbons (PAHs)—such as acenaphthene and naphthalene—in mixture samples. The approach directly converts 3D fluorescence spectral data into grayscale images, from which Zernike moments are extracted as feature descriptors. These moments are used as inputs to a GRNN-based quantitative model, avoiding the need for complex pretreatment steps. By evaluating different orders of Zernike moments, an optimal set was identified, achieving average relative errors of 5.62% for acenaphthene and 5.70% for naphthalene across eight test samples. This strategy demonstrates the effective incorporation of image-based feature extraction into fluorescence analysis and offers a promising tool for the rapid quantification of components in complex environmental samples.