Sampling Training Data for Accurate Hyperspectral Image Classification via Tree-Based Spatial Clustering
The classification of hyperspectral images is a challenging task due to the high dimensionality of the task (i.e. large amount of pixels described over a high number of spectral channels) coupled with the small number of labeled examples typically available for learning. In the last decades, Support Vector Machines (SVMs) have gained in popularity in the field of the hyperspectral image classification as they address large attribute spaces and produce solutions from sparsely labeled data. However, they require “representative” training samples of the unknown class distribution to be accurate. In general, these samples are manually selected by expert visual inspection or field survey. This paper describes a learning schema, where the most suitable pixels to train the classifier are automatically selected via a spectral-spatial clustering phase. This reduces the expert effort required for sampling training pixels. Experimental results highlight that the proposed solution allows us to achieve a classification accuracy that outperforms the accuracy of both random and baseline sampling schemes.
Sonja Pravilovic’s research was supported by the Ministry of Science of Montenegro, Higher Education and Research for Innovation and Competitiveness (INVO/HERIC). She received the national scholarship for excellence (1/10/2016-1/10/2017) funded by the proceeds of a loan from the International Bank for Reconstruction and Development. Authors thank Francesco Dammacco for his support in developing the algorithm presented and anonymous reviewers for useful suggestions provided to improve this paper. This work is carried out in partial fulfillment of the research objectives of the European project “MAESTRA - Learning from Massive, Incompletely annotated, and Structured Data (Grant number ICT-2013-612944)” funded by the European Commission, as well as the ATENEO 2012 project “Mining Complex Patterns” and the ATENEO 2014 project “Mining of network data” funded by University of Bari Aldo Moro.
- 1.Appice, A., Guccione, P.: Exploiting spatial correlation of spectral signature for training data selection in hyperspectral image classification. In: Calders, T., Ceci, M., Malerba, D. (eds.) DS 2016. LNCS, vol. 9956, pp. 295–309. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-46307-0_19 CrossRefGoogle Scholar
- 3.Appice, A., Džeroski, S.: Stepwise induction of multi-target model trees. In: Kok, J.N., Koronacki, J., Mantaras, R.L., Matwin, S., Mladenič, D., Skowron, A. (eds.) ECML 2007. LNCS, vol. 4701, pp. 502–509. Springer, Heidelberg (2007). https://doi.org/10.1007/978-3-540-74958-5_46 CrossRefGoogle Scholar
- 8.Dong, P., Liu, J.: Hyperspectral image classification using support vector machines with an efficient principal component analysis scheme. In: Wang, Y., Li, T. (eds.) ISKE 2011. AISC, vol. 122, pp. 131–140. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-25664-6_17 CrossRefGoogle Scholar
- 15.Plaza, A., Benediktsson, J.A., Boardman, J.W., Brazile, J., Bruzzone, L., Camps-Valls, G., Chanussot, J., Fauvel, M., Gamba, P., Gualtieri, A., Marconcini, M., Tilton, J.C., Trianni, G.: Recent advances in techniques for hyperspectral image processing. Remote Sens. Environ. 113(Suppl. 1), S110–S122 (2009)CrossRefGoogle Scholar
- 16.Rajadell, O., Garcia-Sevilla, P., Dinh, V.C., Duin, R.P.W.: Semi-supervised hyperspectral pixel classification using interactive labeling. In: 2011 3rd Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS), pp. 1–4 (2011)Google Scholar
- 17.Stearns, S.D., Wilson, B.E., Peterson, J.R.: Dimensionality reduction by optimal band selection for pixel classification of hyperspectral imagery. vol. 2028, pp. 118–127 (1993)Google Scholar