Classification of hyperspectral remote sensing images is affected by two main problems: high dimensionality of the acquired signatures and scarce availability of labeled samples. Learning a low dimensional manifold and active learning represent two approaches that have been investigated in the literature to mitigate these effects. However they are usually applied independently from each other. In this paper we propose a method in which feature extraction and active learning are combined. In particular, a new reduced feature space is learned by large margin nearest neighbor (LMNN), a metric learning strategy that takes advantage of labeled information. The method is applied in conjunction with k-nearest neighbor (k-NN) classification, for which a new sample selection strategy is proposed. Experiments on a real hyperspectral dataset confirm the effectiveness of the proposed method.

Combining active and metric learning for hyperspectral image classification / Pasolli, E.; Yang, H. L.; Crawford, M. M.. - 2014-:(2014), pp. 1-4. (Intervento presentato al convegno 6th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing, WHISPERS 2014 tenutosi a che nel 2014) [10.1109/WHISPERS.2014.8077529].

Combining active and metric learning for hyperspectral image classification

Pasolli E.;
2014

Abstract

Classification of hyperspectral remote sensing images is affected by two main problems: high dimensionality of the acquired signatures and scarce availability of labeled samples. Learning a low dimensional manifold and active learning represent two approaches that have been investigated in the literature to mitigate these effects. However they are usually applied independently from each other. In this paper we propose a method in which feature extraction and active learning are combined. In particular, a new reduced feature space is learned by large margin nearest neighbor (LMNN), a metric learning strategy that takes advantage of labeled information. The method is applied in conjunction with k-nearest neighbor (k-NN) classification, for which a new sample selection strategy is proposed. Experiments on a real hyperspectral dataset confirm the effectiveness of the proposed method.
2014
978-1-4673-9012-5
Combining active and metric learning for hyperspectral image classification / Pasolli, E.; Yang, H. L.; Crawford, M. M.. - 2014-:(2014), pp. 1-4. (Intervento presentato al convegno 6th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing, WHISPERS 2014 tenutosi a che nel 2014) [10.1109/WHISPERS.2014.8077529].
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11588/837363
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? 0
social impact