Design and testing of an underwater microscopy for the study of zooplankton distribution

A method that can monitor zooplankton density at adequate spatio/temporal resolution is desired in the field of oceanic ecosystem studies. We have developed a submergible microscopes equipped with a non-interlace CCD camera. As target plankton, ploima, cladocera, calanoida, cyclopoida, ciliata are selected those are dominant species in coastal sea in Japan. Requirements of an system for an underwater imaging of zooplankton are discussed. Selections of lens, camera and illumination are key issues, as they may influence on performance of the system. Second order auto-correlational masks are used to extract features from images. The combination of local auto-correlational masks and multivariate analysis, which is a two-step feature extraction, is a powerful tool as a general information extracting method from images. In our procedures, a set of these masks is 33 dimensional vector. To identify and count zooplankton, canonical correlation analysis is performed. Therefore, we can count and classify zooplankton into its taxonomic unit. Through another multiple regression, sizing of plankton can also be performed. Proof of principle experiment was also performed with images of living copepoda.