Plant species determination by coding leaf reflectance spectrum and its derivatives

Abstract

ABSTRACT The identification of plant species using spectrometry is utilized in various applications. Some researchers have reported the success of using the reflectance spectrum along with its derivatives and their coding to identify different Phenomena, including plants. In this study, based on these parameters, we developed a new method to improve the accuracy of identifying plant species. Two experimental datasets (LOPEX and ANGERS) containing the leaf reflectance spectrum of different plant species from different families were used. The proposed method includes two main stages. In the first stage, the unique regions of wavelengths were found for each plant family. By comparing an unknown species with each family only in the specified regions and coding results with the innovative coding method named Coding Based on Selected Threshold (CBSeT) and used similarity criteria, the family of unknown specie is determined. In the second step, similarly unique regions were found to separate the species within a family from each other using species standard deviation analysis, and the type of unknown species was similarly identified. The obtained results for the identification of plant species of unknown spectra show the accuracies of 91.66% and 90% for the proposed method with Similarity Coding Index (SCI) and Spectral Angle Mapper (SAM) criteria respectively. These numbers are more acceptable results compared to the other reflectance coding methods such as Spectral analysis manager (SPAM), Spectral feature-based binary coding (SFBC), Spectral derivative feature coding (SDFC), Spectral feature probabilistic coding (SFPC) with 38%, 37%, 29%, and 20% of accuracies improvement respectively.