Automatic recognition of coccoliths by dynamical neural networks

Abstract Systeme de Reconnaissance Automatique de Coccolithes (SYRACO) is a tool for automatic recognition of coccoliths by neural networks. Previous versions of this tool were able to identify individual species of coccoliths with a high reliability, but failed in that many coccoliths were overlooked. SYRACO was able to identify only about half the coccoliths present in a field of view. We have now developed a new type of Neural Network, which includes a dynamic view of the object analysed. We have added parallel neural networks, which perform five types of simple image transformation to the general back-propagation neural network. This new dynamic version of SYRACO is able to identify individual coccoliths even more reliably at the same time as it is able to recognize almost all coccoliths present in a field of view. The only remaining problem concerns the inclusion of objects that are not coccoliths. This problem can be solved partly by performing a secondary SYRACO analysis of output images. The performance of the current system is demonstrated using the study of 21 EUMELI sediment trap samples.

[1]  D. Dollfus,et al.  Reconnaissance de formes naturelles par des réseaux de neurones artificiels : application au nannoplancton calcaire , 1997 .

[2]  A. Mcintyre.  Coccoliths as Paleoclimatic Indicators of Pleistocene Glaciation , 1967, Science.

[3]  P. Culverhouse,et al.  Automatic classification of field-collected dinoflagellates by artificial neural network , 1996 .

[4]  Jörg Bollmann,et al.  Morphology and biogeography of Gephyrocapsa coccoliths in Holocene sediments , 1997 .

[5]  Lawrence D. Jackel,et al.  Backpropagation Applied to Handwritten Zip Code Recognition , 1989, Neural Computation.

[6]  Michael Knappertsbusch,et al.  Morphologic variability of the coccolithophorid Calcidiscus leptoporus in the plankton, surface sediments and from the Early Pleistocene , 1997 .

[7]  M. Berthod,et al.  Feasibility study of automatic identification of planktic foraminifera by computer vision , 1996 .

[8]  K. Geitzenauer,et al.  Coccoliths as Late Quaternary Palaeoclimatic Indicators in the Subantarctic Pacific Ocean , 1969, Nature.

[9]  Françoise Fogelman-Soulié,et al.  Multi-Modular Neural Network Architectures: Applications in Optical Character and Human Face Recognition , 1993, Int. J. Pattern Recognit. Artif. Intell..

[10]  S. Heussner,et al.  The PPS 3 time-series sediment trap and the trap sample processing techniques used during the ECOMARGE experiment , 1990 .

[11]  A. D. Poularikas,et al.  Automated sizing, counting and identification of zooplankton by pattern recognition , 1984 .

[12]  J. Young.  Size variation of Neogene Reticulofenestra coccoliths from Indian Ocean DSDP Cores , 1990, Journal of Micropalaeontology.

[13]  A. Mix,et al.  ENSO-like forcing on oceanic primary production during the Late Pleistocene. , 2001, Science.

[14]  S. Zeki.  Functional specialisation in the visual cortex of the rhesus monkey , 1978, Nature.

[15]  D. Dollfus,et al.  Fat neural network for recognition of position-normalised objects , 1999, Neural Networks.

[16]  David H. Foster,et al.  Internal representations and operations in the visual comparison of transformed patterns: Effects of pattern point-inversion, positional symmetry, and separation , 2004, Biological Cybernetics.

[17]  S. P. Almeida,et al.  Automated Identification of Diatoms , 1982 .

[18]  R. Shepard,et al.  Mental Rotation of Three-Dimensional Objects , 1971, Science.

[19]  A. Bory,et al.  Downward particle fluxes within different productivity regimes off the Mauritanian upwelling zone (EUMELI program) , 2001 .