Reconhecimento de objetos 3D através de atributos de superfícies de modelos poliedrais.

Abstract

This thesis presents a method of polyhedral 3D object recognition based on a multiview procedure using geometric attributes from surfaces. The recognition method consists on matching global, local, and relational attributes in order to make possible the identification of both the models present in the scene and the visible surfaces in the views. Experiments of real and simulated scenes were carryed on, where some scenes contain isolated objects and another ones contain more than one object. The reliability rate was very good, in some cases 100% of the objects and surfaces were correctly recognized. After the identification, the system stablishes the position of the model in the environment of the scene.