Reconhecimento de edifícios utilizando a transformada de Gabor, a transformada wavelet e a rede neural perceptron de múltiplas camadas

Abstract

Recognition of buildings is a very difficult task, because the images used have different angles and lighting conditions, partial obstructions of trees, moving vehicles or other buildings. Due to all these difficulties, this paper proposes a recognition system which uses Gabor wavelet representation for the extraction of features, the wavelet transform for dimensionality reduction and the multi-layer perceptron neural network for recognition. In order to check the efficiency of this system, we used a database of images of new and old buildings. The performance of the proposed method was satisfactory compared to other methods found in the literature.