A backpropagation ANN algorithm based on RGB images for the identification of granite-forming minerals

摘要

Granite is a rock widely used in the built Cultural Heritage in the NW of Iberian Peninsula. Nowadays, one of the most studied cleaning procedure of the built Cultural Heritage is the laser application because it is gradual and selective. Considering the laser cleaning of granite, it is of great interest to perform the identification of the forming-minerals in the stone surface in order to avoid damage due to the overexposure, improving the treatment results. The aim of this work is the optimization of a back propagation artificial neural network in order to obtain rapid and reliable identification of forming-minerals in granitic stones by means of RGB images. Our goal is, eventually, in-situ monitor the laser cleaning of granite stoneworks. The artificial neural network results obtained were compared with the results of the modal analysis and it was detected a high degree of correct identification of the minerals.