Sistema neural para reconstruÃÃo de superfÃfices a partir de nuvem de pontos

AUTOR(ES)
DATA DE PUBLICAÇÃO

2006

RESUMO

Many applications need digital models of real objects, as machines, commercial products, sculptures, human organs, and others. The creation of these models is a process called Reconstruction. In this work, it is proposed a new method of surface reconstruction that combines two neural networks to produce a triangular mesh representing the shape of the object given a set of points on its surface. The system receives as input a set of cross sections of a real object, or a synthetic object. From these inputs, the points used by the neural network to reconstruct the shape of the object are acquired. First, the Reconstruction problem is presented with some of the existent solutions. Next, the Reconstruction system developed in the work is presented, especially the new method proposed for surface reconstruction. Finally, the results of the experiments carried out are presented and the achievements and limitations of the proposed algorithm for surface reconstruction are discussed, as well as the possible future improvements for the system

ASSUNTO(S)

surface reconstruction mapas auto-organizÃveis reconstruÃÃo de superfÃcies neural networks ciencia da computacao point clouds redes neurais self-organizing maps nuvens de pontos

ACESSO AO ARTIGO

Documentos Relacionados