Accurate Classification of Time-Varying Microalgae by Stokes Imaging With Multiple Polarization Illumination

Microalgae are crucial to the global aquatic ecosystem, and their in situ accurate recognition is essential for early warning of harmful algae blooms. Based on the existing Stokes and Mueller matrix imaging methods of polarization microscope, this article introduces a method for classifying microalgae using Stokes imaging with multiple polarization illumination (SiMPI), effectively addressing the challenges of similar species (which troubles Stokes imaging) and time-varying states (which troubles Mueller matrix imaging), further enhancing the feasibility of polarization microscopy for in situ detection. The SiMPI method achieves 91.6% accuracy in classifying 29 representative harmful microalgae species. Specifically, increasing the polarization illumination number improves the classification accuracy of the eight lowest performing species to 86.3%. Field data from coastal waters show recognition accuracy exceeding 95%. The study highlights that high imaging resolution, optimal polarization illumination combinations, and balanced databases enhance the SiMPI method's performance. After quantifying the time variation of microalgal cells, it is convincingly demonstrated that the SiMPI method outperforms Mueller matrix imaging. These findings underscore the SiMPI method's potential as a feasible and robust tool for future aquatic environment monitoring.