Investigation into the Application of Classification Algorithms to Field Collected Plankton Data

This paper describes an investigation into the automatic identification of plankton with a view to duplicating some of the results obtainedfi-om research carried out at the University ofMassachusetts (UMass) from images collected in situ. A range of classification algorithms were applied to a variety of features that had been extracted from the collected images. Some of the classifiers employed were the same as those used by UMass and some new ones were introduced in an attempt to increase classification accuracy and thus improve upon the previously reported results. Plankton species are a good indicator of large scale marine environment conditions and so an accurate identification of their types and distributions is important. Identification by human means is slow and often inaccurate leading to a need for automatic techniques including those described herein. It was found that that there was a high correlation between the results obtained in this research and those from UMass. However, no significant improvements on the UMass results could be obtained either from the classifiers used in that research or from those introduced for this paper.School of Computing, Communications and Electronic