Automatic Diatom Identification 2002. (Series in Machine Perception and Artificial Intelligence, 51). illus. 328 p. Hardcover.

Probably the first book to deal with automatic diatom identification.Background information concerning diatom research, development o felectronic databases, image preprocessing, automatic contour extraction, application of existing contour and ornamentation features and the development of new ones, application of different classifiers (neural networks, decision trees, etc.). They are tested using two image sets: 1)a very difficult set of Sellaphora pupula with 6 dems and 120 images, 2) a mixed genera set with 37 taxa and approximately 800 images. The results are excellent, and recognition rates well above 90% have been achieved on both sets. Much additional basic information.