Classification of Pansharpened Urban Satellite Images

Publication: Research - peer-reviewJournal article – Annual report year: 2012

Standard

Classification of Pansharpened Urban Satellite Images. / Palsson, Frosti; Sveinsson, Johannes R.; Benediktsson, Jon Atli; Aanaes, Henrik.

In: I E E E Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 5, No. 1, 2012, p. 281-297.

Publication: Research - peer-reviewJournal article – Annual report year: 2012

Harvard

APA

CBE

MLA

Vancouver

Author

Palsson, Frosti; Sveinsson, Johannes R.; Benediktsson, Jon Atli; Aanaes, Henrik / Classification of Pansharpened Urban Satellite Images.

In: I E E E Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 5, No. 1, 2012, p. 281-297.

Publication: Research - peer-reviewJournal article – Annual report year: 2012

Bibtex

@article{704ed5cb7933454ea0d23fc5f408d175,
title = "Classification of Pansharpened Urban Satellite Images",
keywords = "Classification, Mathematical morphology, Morphological profile, Pansharpening, Spatial consistency, Spectral consistency",
publisher = "I E E E",
author = "Frosti Palsson and Sveinsson, {Johannes R.} and Benediktsson, {Jon Atli} and Henrik Aanaes",
year = "2012",
doi = "10.1109/JSTARS.2011.2176467",
volume = "5",
number = "1",
pages = "281--297",
journal = "I E E E Journal of Selected Topics in Applied Earth Observations and Remote Sensing",
issn = "1939-1404",

}

RIS

TY - JOUR

T1 - Classification of Pansharpened Urban Satellite Images

A1 - Palsson,Frosti

A1 - Sveinsson,Johannes R.

A1 - Benediktsson,Jon Atli

A1 - Aanaes,Henrik

AU - Palsson,Frosti

AU - Sveinsson,Johannes R.

AU - Benediktsson,Jon Atli

AU - Aanaes,Henrik

PB - I E E E

PY - 2012

Y1 - 2012

N2 - The classification of high resolution urban remote sensing imagery is addressed with the focus on classification of imagery that has been pansharpened by a number of different pansharpening methods. The pansharpening process introduces some spectral and spatial distortions in the resulting fused multispectral image, the amount of which highly varies depending on which pansharpening technique is used. In the majority of the pansharpening techniques that have been proposed, there is a compromise between the spatial enhancement and the spectral consistency. Here we study the effects of the spectral and spatial distortions on the accuracy in classification of pansharpened imagery. We also study the performance in terms of accuracy of the various pansharpening techniques during classification with spatial information, obtained using mathematical morphology (MM). MM is used to derive local spatial information from the panchromatic data. Random Forests (RF) and Support Vector Machines (SVM) will be used as classifiers. Experiments are done for three different datasets that have been obtained by two different imaging sensors, IKONOS and QuickBird. These sensors deliver multispectral images that have four bands, R, G, B and near infrared (NIR). To further study the contribution of the NIR band, experiments are done using both the RGB bands and all four bands, respectively.

AB - The classification of high resolution urban remote sensing imagery is addressed with the focus on classification of imagery that has been pansharpened by a number of different pansharpening methods. The pansharpening process introduces some spectral and spatial distortions in the resulting fused multispectral image, the amount of which highly varies depending on which pansharpening technique is used. In the majority of the pansharpening techniques that have been proposed, there is a compromise between the spatial enhancement and the spectral consistency. Here we study the effects of the spectral and spatial distortions on the accuracy in classification of pansharpened imagery. We also study the performance in terms of accuracy of the various pansharpening techniques during classification with spatial information, obtained using mathematical morphology (MM). MM is used to derive local spatial information from the panchromatic data. Random Forests (RF) and Support Vector Machines (SVM) will be used as classifiers. Experiments are done for three different datasets that have been obtained by two different imaging sensors, IKONOS and QuickBird. These sensors deliver multispectral images that have four bands, R, G, B and near infrared (NIR). To further study the contribution of the NIR band, experiments are done using both the RGB bands and all four bands, respectively.

KW - Classification

KW - Mathematical morphology

KW - Morphological profile

KW - Pansharpening

KW - Spatial consistency

KW - Spectral consistency

U2 - 10.1109/JSTARS.2011.2176467

DO - 10.1109/JSTARS.2011.2176467

JO - I E E E Journal of Selected Topics in Applied Earth Observations and Remote Sensing

JF - I E E E Journal of Selected Topics in Applied Earth Observations and Remote Sensing

SN - 1939-1404

IS - 1

VL - 5

SP - 281

EP - 297

ER -