Pattern recognition system using artificial neural networks and wavelets for taxonomic identifications

This research focuses on the development of a pattern classification system that is capable of identifying spiders to genus and species. This system attempts to emulate the eyes and expertise of arachnologists. It takes as input images of epygina, representative of a genus or species of spider, encodes it by means of a Daubechies 4 wavelet transform, and uses a fast and efficient cascade correlation artificial neural network to identify specimens of that genus or species. If the samples are chosen well then the system is able to perform accurate identification of members of the specified genus or species. In tests involving samples from the family Lycosidae, the system created by this research achieved an accuracy of 100% in the identification of genera and up to 88% accuracy in the identification of species.