On development of a submersible microscopes and image processing system

The authors have developed a submersible microscope equipped with a non-interlace CCD camera. Ploima, cladocera, calanoida, cyclopoida, and ciliata are selected as target plankton to be counted automatically because those are the dominant species in the coastal waters of Japan. The size range of the target plankton is from a few tens of microns to several mm. The requirements of a submersible microscope for imaging zooplankton are also discussed. The selections of lenses, cameras and illumination is a key issue because these factors may influence system performance. A xenon flash lamp is employed as a light source. Three objective lenses with different magnifications (as x0.83, x3.3, x13.3) are used to take images of plankton. Acquired images are processed by an image processing system that extracts 33 dimensional second-order autocorrelational features from each image. To identify and count zooplankton, multivariate analysis is performed on a personal computer. Transformed features have properties similar to a vector. The direction of a vector represents a species and the its length indicates the number of the species. When more than two plankton are observed in one image, the transformed features are the linear combinations of each plankton´s vectors. Therefore, the authors can count and classify each zooplankton into its taxonomic unit. Through a multiple regression, a sizing of plankton can also be performed. As far as the authors tested, the accuracy of identification was more than 90% when discrimination analysis was performed between any two taxa of preserved plankton