Modern sensor technologies are capable of covering large surfaces of the Earth with exceptional spatial, spectral, and temporal resolutions. With the significant development of hyperspectral imaging technology, change detection, which discovers change knowledge about Earth surface, has emerged as a hot topic in remote sensing. In this paper, we propose a deep learning enhanced change detection methodology that leverages the power of traditional change vector analysis techniques by gaining in accuracy with autoencoding neural networks and clustering. Preliminary experiments performed evaluating the proposed methodology with bechmark data provide encouraging results, also when compared to recent state-of-the-art change vector analysis competitors.
Empowering change vector analysis with autoencoding in bi-temporal hyperspectral images
Appice A.
;Di Mauro N.;Malerba D.
2019-01-01
Abstract
Modern sensor technologies are capable of covering large surfaces of the Earth with exceptional spatial, spectral, and temporal resolutions. With the significant development of hyperspectral imaging technology, change detection, which discovers change knowledge about Earth surface, has emerged as a hot topic in remote sensing. In this paper, we propose a deep learning enhanced change detection methodology that leverages the power of traditional change vector analysis techniques by gaining in accuracy with autoencoding neural networks and clustering. Preliminary experiments performed evaluating the proposed methodology with bechmark data provide encouraging results, also when compared to recent state-of-the-art change vector analysis competitors.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.