Lake Biwa, Japan, is a significant freshwater source supplying industry and 14 million people. Over recent years, its water quality has suffered due to an increasing occurrence of algal blooms, comprising of three main cyanobacteria genera: Microcystis, Anabaena, and Planktothrix. Manually identifying algal species and quantifying their numbers is both labour intensive and costly. In this paper we present an image processing system for automatically detecting and classifying six cyanobacteria taxa, and report preliminary classification results. We viewed water samples at high magnification, and transferred images of cyanobacteria to the computer for processing. After image segmentation, we extracted properties such as shape, spectral, and textual features. Classification was implemented by a hierarchy of 2-class classifiers. Out of a total of 244 images examined, 7 were misclassified, indicating an error rate of approximately 3%. Although preliminary, the results suggest that accurate classification of algal specimens can be achieved via automated image processing.
Key words: Image processing, pattern recognition, cyanobacteria, Lake Biwa, algal bloom, texture, classification.