Design and testing of an underwater microscope and image processing system for the study of zooplankton distribution

A method that can monitor the density of zooplankton at an adequate spatio-temporal resolution is desired in oceanic ecosystem research. To address this need, we have developed a submersible microscope equipped with a noninterlace CCD camera. The target plankton for this microscope includes Copepoda, Ploima, and Ciliata, which are dominant species in the coastal waters around Japan. In addition, the requirements of systems for underwater imaging of zooplankton are discussed. The key issues investigated for their possible influence on system performance are lens selection, camera selection, and method of illumination. Higher order local autocorrelational (HLAC) masks are used to extract features from images. Combining these features with multivariate analysis, which is a two-step feature extraction method, results in a powerful tool for extracting general information from images. In our procedures, a set of these features provides a 33-dimensional vector. To identify and count zooplankton, canonical correlation analysis and discrimination analysis are performed. This allows zooplankton to be counted and classified into taxonomic units. Another canonical correlation analysis was made for the sizing of the plankton. Proof of the principle experiment is obtained with images of both preserved and living Copepoda.