Book Description

This is the first book to deal with automatic diatom identification. It provides the necessary background information concerning diatom research, useful for both diatomists and non-diatomists. It deals with the development of electronic databases, image preprocessing, automatic contour extraction, the application of existing contour and ornamentation features and the development of new ones, as well as the application of different classifiers (neural networks, decision trees, etc.). These are tested using two image sets: (i) a very difficult set of Sellaphora pupula with 6 demes and 120 images; (ii) 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. The results are compared with identification rates obtained by human experts. One chapter of the book deals with automatic image capture, i.e. microscope slide scanning at different resolutions using a motorized microscope stage, autofocusing, multifocus fusion, and particle screening to select only diatoms and to reject debris. This book is the final scientific report of the European ADIAC project (Automatic Diatom Identification and Classification), and it lists the web-sites with the created public databases and an identification demo.