Objective
This project is concerned with the assessment of utility of two artificial neural networks (back propagation of error and radial basis function) for the automatic categorization of marine dinoflagellates. Over 2.000 photomicrographs of dinoflagellates are being collected, validated by panels of experts and digitized into a computer. This validated data set will be used to make a comparative assessment of neural network algorithms with classical image processing techniques such as principle component analysis.
Work is also being undertaken on improving the robustness of artificial neural network algorithms to noise, clutter, and multiple specimens within images using connectionist techniques being developed by the cognitive psychology community and others. Additionally, a prototype computer-based training tool using the validated data set is being developed. This tool will be disseminated on a trial basis by satellite link to a number of sites in Europe and will be evaluated as a training aid to ecologists working in marine and coastal management.
Fields of science
Topic(s)
Call for proposal
Data not availableFunding Scheme
CSC - Cost-sharing contractsCoordinator
PLYMOUTH
United Kingdom