A Comprehensive Spectral Analysis of the Auto-fluorescence Characteristics of Three Algae Species at Twelve Discrete Excitation Wavelengths

Abstract

Harmful algae blooms are a growing concern around the globe, and therefore fast and reliable methods to detect and classify different types of algae in an automatic manner is highly desired. In this study, we explore the auto-fluorescence characteristics of different algae types to determine whether using their auto-fluorescence spectra could be used in automatic identification. Preliminary experimental results in this study for generating auto-fluorescence spectra of Anabaena flos-aquae (Cyanophyta phylum), Ankistrodesmus falcatus (Chlorophyta phylum), and Euglena gracilis (Euglenozoa phylum) demonstrate that this information could potentially be leveraged to discriminate between different types of algae. Future work must be done to explore the auto-fluorescence spectra of additional species within a given genus in order to determine the intra-class and inter-class variability of these auto-fluorescence spectra. Future work also includes using this technique to determine the concentration in a mixed sample, as well as determining the robustness in a sample with contaminants. Future work will also involve exploring the use of additional fluorescent images, absorption spectra, and morphological features to improve the performance of the classifiers. Finally, as data collection continues we will explore using data augmentation to deal with unbalanced class sizes.