Reconhecimento de Peças para Robô de Montagem utilizando Redes Neurais Artificiais: abordagem baseada em Invariância Geométrica

Abstract

Automated Assembly Systems, based on simple and repetitive tasks of manipulation, with high level of tolerance, in controlled environment, uses artificial vision to manage tolerances and imperfect parts, and errors of positioning. This paper describes a Neural Network based Pattern Recognition System, that can recognize patterns even they are deformed by transformation of rotation, scaling and translation. Geometric Invariance is the main characteristic presented by the proposed system. Another important characteristic is the system's ability to work with noisy images. The final structure was designed to enable a parallel implementation, specially designed to run on the CPAD [3] Architecture.