Plankton image classification using novel parallel-training learning vector quantization network

At the base of the food chain in the ocean, plankton have a large impact on marine ecosystem dynamics. Rapid mapping of plankton abundance together with taxonomic and size composition is very important for ocean environmental research but difficult or impossible to accomplish using traditional techniques. In this paper, the authors develop a new pattern recognition system to classify large numbers of plankton images detected in real time by a towed underwater video system. The difficulty of such classification is compounded because the data sets are not only noisier but the plankton are non-rigid, projection-variant, and often in partial occlusion. The approach described combines traditional invariant moments features and Fourier boundary descriptors with gray-scale morphological granulometries to form a feature vector capturing both shape and texture information of plankton images. With a novel parallel-training learning vector quantization network classifier, the authors achieve 95% classification accuracy on six plankton taxa taken from more than 2,000 images, making possible for the first time a fully automated, at-sea approach to real-time mapping of plankton.