AUTOMATIC PLANKTON IMAGE RECOGNITION

Plankton form the base of the food chain in the ocean and are fundamen tal to marine ecosystem dynamics. The rapid mapping of plankton abunda nce together with taxonomic and size composition is very important for ocean environmental research, but difficult or impossible to accompli sh using traditional techniques. In this paper, we present a new patte rn recognition system to classify large numbers of plankton images det ected in real time by the Video Plankton Recorder (VPR), a towed under water video microscope system. The difficulty of such classification i s compounded because: 1) underwater images are typically very noisy, 2 ) many plankton objects are in partial occlusion, 3) the objects are d eformable and 4) images are projection variant, i.e., the images are v ideo records of three-dimensional objects in arbitrary positions and o rientations. Our approach combines traditional invariant moment featur es and Fourier boundary descriptors with gray-scale morphological gran ulometries to form a feature vector capturing both shape and texture i nformation of plankton images. With an improved learning vector quanti zation network classifier, we achieve 95% classification accuracy on s ix plankton taxa taken from nearly 2,000 images. This result is compar able with what a trained biologist can achieve by using conventional m anual techniques, making possible for the first time a fully automated , at sea-approach to real-time mapping of plankton populations.