Multi-wavelength laser-induced fluorescence spectroscopy with KLDA and Fisher score feature selection for pollen classification

This is the authors' abstract. We don't add key points for this paper.

This study proposes a method for pollen classification using three-wavelength laser-induced fluorescence (LIF) spectroscopy combined with kernelized linear discriminant analysis (KLDA). Fluorescence spectra were acquired from eight pollen species using LIF with excitation wavelengths of 266, 360, and 405 nm. Fisher score was applied to remove redundant features and identify representative spectral information. The selected features were then subjected to KLDA for dimensionality reduction, and classification was performed using a Random Forest algorithm. The proposed method achieved a classification accuracy of 100% across all pollen species tested. Among the three excitation wavelengths, the 405 nm wavelength provided the most distinct spectral features. Integrating spectral data from all three wavelengths further enhanced classification performance. The results demonstrate that combining multi-wavelength LIF spectroscopy with KLDA effectively differentiates pollen types based on fluorescence spectral characteristics.