Título: Reconhecimento De Padrões Invariante A Rotação Utilizando Uma Rede Neural Morfológica Não Supervisionada
Autores: Rempel, Erico L.; Silva, Francisco A. T. F.
Resumo: This work presents the development of a rotation and translation invariant 2-dimensional image detection system using an unsupervised morphological neural network. Utilizing the translation invariance property of the operators of Mathematical Morphology, it is suggested a morphological pattern-matching mechanism based on the comparison between the input image and a set of rotated structuring elements, making possible rotation and translation invariant pattern detection by use of rotational weights applied to morphological neurons. A series of experiments were done using binary and gray scale images.
Código DOI: 10.21528/CBRN2001-024
Artigo em pdf: 5cbrn_024.pdf
Arquivo BibTex: 5cbrn_024.bib