Mikhail Kanevski

TL;DR: Two active learning algorithms for semiautomatic definition of training samples in remote sensing image classification, based on predefined heuristics, are proposed, which reach the same level of accuracy as larger data sets.

TL;DR: The main families of active learning algorithms are reviewed and tested: committee, large margin, and posterior probability-based, which aims at building efficient training sets by iteratively improving the model performance through sampling.

TL;DR: An effective solution to deal with supervised change detection in very high geometrical resolution (VHR) images is studied and the inclusion of spatial and contextual information issued from local textural statistics and mathematical morphology is proposed.

TL;DR: Experiments carried out in multi- and hyperspectral, contextual, and multisource remote sensing data classification confirm the capability of the method in ranking the relevant features and show the computational efficience of the proposed strategy.

TL;DR: The results of this study reveal the strengths of the classification algorithms, but evidence shows the need for relying on more than one method for the identification of relevant variables; the weakness of the adaptive scaling algorithm when used with landslide data; and the lack of additional features which characterize the spatial distribution of deep-seated landslides.