A robust multiple classifier system for a partially unsupervised updating of land-cover maps

We propose a system for a regular updating of land-cover maps based on the use of temporal series of remote sensing images. Such a system is composed of an ensemble of partially unsupervised classifiers integrated in a multiple classifier architecture. The updating problem is formulated under the complex constraint that for some images of the considered multitemporal series no ground-truth information is available. With respect to the authors’ previous works on this topic [1-3], the novel contribution of this paper consists in: i) developing partially unsupervised classification algorithms defined in the framework of a cascade-classifier approach; ii) defining a specific strategy for the generation of an ensemble of classifiers, which exploits the peculiarities of the cascade-classifier approach. These novel aspects result in the definition of more robust and accurate classification systems.