Título: Classificação de Imagens Codificadas por Cadeias Direcionais Utilizando Redes Neurais Artificiais
Autores: Oliveira, Kleyton C. de; Bauchspiess, Adolfo
Resumo: This paper presents a procedure for a robust classification of chain-encoded images by using artificial neural networks to recognize sets of features based on normalized Fourier series representation. A particular string-encoding technique and its normalization are described and its results are used in a procedure to determine the best topology of neural networks to be used. The results presented by Radial Basis Function Networks and Multilayer Perceptrons Networks are compared in order to determine the most appropriate architecture for such an application.
Palavras-chave: Pattern Classification; Chain-Enconded Images; Fourier Transform; Artificial Neural Networks
Código DOI: 10.21528/CBRN2001-017
Artigo em pdf: 5cbrn_017.pdf
Arquivo BibTex: 5cbrn_017.bib