Utilizing Push-Broom Fluorescence Hyperspectral Imaging for Classification of Underwater Flora and Fauna

This study explores the application of fluorescence hyperspectral imaging (FHSI) for marine ecosystem monitoring, integrating laser-induced fluorescence (LIF) with hyperspectral imaging (HSI), an approach rarely used for ecological analysis. FHSI enables precise identification of photopigmented marine flora based on their unique fluorescence signatures. Key pigments, including chlorophyll-a, chlorophyll-b, carotenoids, and accessory pigments, emit distinct fluorescence signals when excited by specific wavelengths. This capability addresses challenges in marine ecology, such as overlapping optical properties and complex underwater conditions. Using a 450 nm blue laser and a push-broom HSI system, along with a one-dimensional convolutional neural network, we classified macroalgae, eelgrass, and blue mussels. The method demonstrates high efficiency and precision for detecting photopigment-bearing organisms while revealing limitations in low-fluorescence organisms, such as mussels. To evaluate the ecological applicability of the setup, we further conducted a beta field test in Roskilde Fjord. The laser provided sufficient excitation to resolve clear chlorophyll-a fluorescence despite sun-induced backscatter; however, assembling push-broom scans into stable hyperspectral cubes was challenged by object motion and potential instability of the imaging system. These findings underline both the promise of FHSI for benthic habitat analysis and the need to address motion related constraints before robust deployment in automated marine ecosystem monitoring.