Abstract
b Dipartimento di Ingegneria Industriale (DIIn), Università degli Studi di Salerno, via Giovanni Paolo II 132, 84084 Fisciano (SA), Italy, michele.miccio@unisa.itc Università degli Studi di Palermo, Italyignazio.sammarco@you.unipa.itMicroalgae play a crucial role in various sectors, such as biofuel production or environmental monitoring. The ability to accurately classify and analyze microalgae species from optical images is vital for advancing research and applications. Generally, the data regarding the population of microalgae constitute a valuable input for machine-learning algorithms whose aim is to classify real data derived from optical images of microalgae cells. However, obtaining a diverse dataset of microalgae populations to train machine-learning models can be challenging and resource-intensive. This paper presents a machine-learning algorithm based on a dataset of algal records generated by a simulation model. The simulation model uses a combination of mathematical models, probabilistic distributions, and biological knowledge to create realistic data on microalgae populations. The dataset generated from the developed model serves as a resource for training and validation of the subsequent machine-learning model proposed for the microalgae classification task. The machine learning model, trained on this synthetic data, can subsequently be applied to efficiently classify and analyze optical images of real microalgae populations, leading to improved precision and reliability in the identification of species useful to produce biofuels.
