Pulse coupled neural networks for automatic urban change detection at very high spatial resolution

Abstract

In this paper, a novel unsupervised approach based on Pulse- Coupled Neural Networks (PCNNs) for image change detection is discussed. PCNNs are based on the implementation of the mechanisms underlying the visual cortex of small mammals and with respect to more traditional neural networks architectures own interesting advantages. In particular, they are unsupervised and context sensitive. The performance of the algorithm has been evaluated on very high spatial resolution Quick- Bird and WorldView-1 images. Qualitative and more quantitative results are discussed. © 2009 Springer-Verlag Berlin Heidelberg.