Polarization fingerprint for microalgae classification

Abstract

Microalgae classification is important for aquatic environment monitoring, but it is difficult in some in-situ cases. The Mueller matrix is known to encode the physical properties including the sophisticated microstructure of microalgae cells. In this paper, the concept of polarization fingerprint based on the Mueller matrix is presented to extract the parameters to classify microalgae. An experimental setup based on the particulate Mueller matrix polarimetry (PMP) is used to measure 25 kinds of microalgae covering seven phyla to form a Mueller matrix dataset. The proposed sixteen polarization parameters with explicit physical meanings and associations with the structural properties are selected to compose the polarization fingerprint. As a result, different microalgae can be effectively classified by the polarization fingerprint and machine learning algorithms. The contribution of the parameters in the polarization fingerprint is analyzed to provide the specific features for microalgae classification, which gives insights into microalgae's exclusive physical properties. The extensibility and flexibility of the polarization fingerprint are discussed. This work shows the polarization fingerprint's power to classify microalgae, which can help retrieve in-situ information on the community structures of microalgae, and further monitor the aquatic environment.