Plants are one of the most important parts of the ecological system and demand a reliable method for accurate classification. In this study, the first-derivative fluorescence spectral curves (FDFSCs) based on laser-induced fluorescence technology were proposed for the characterization of plant species. The measurement system is mainly composed of a spectrometer, an excitation light source (the two excitation wavelengths are 460 and 556 nm, respectively), and an intensified charge-coupled device camera. FDFSCs were calculated from the deviation between the fluorescence values at each wavelength, plus and minus one band, divided by the wavelength range. Principal component analysis was utilized to analyze the FDFSCs by extracting the main attributes and reducing the dimensionality of variables. A support vector machine was used to evaluate FDFSC performance for the identification of plant species. Plant species that are difficult to distinguished by the naked eye, can be identified effectively using the proposed FDFSCs. For the 556 nm and 460 nm excitation wavelengths, the overall identification rates of the six plant species evaluated were 93.3% and 91.7%, respectively. Experimental results demonstrated that the combination of the FDFSCs with multivariate analysis could provide a simple and reliable method for the characterization of plant species.
Keywords: classification; first-derivative; laser-induced fluorescence; multivariate analysis.
Copyright © 2016 John Wiley & Sons, Ltd.