Abstract
Conventional remote sensing classification techniques that model the data in each class with a multivariate Gaussian distribution are inefficient, as this assumption is generally not valid in practice. We present a novel, independent component analysis (ICA) based approach for unsupervised classification of hyperspectral imagery. ICA, employed for a mixture model, estimates the data density in each class and models class distributions with nonGaussian structure, formulating the ICA mixture model (ICAMM). We apply the ICAMM for unsupervised classification of a test image from the AVIRIS sensor. Four feature extraction techniques namely principal component analysis, segmented principal component analysis, orthogonal subspace projection and projection pursuit have been considered as preprocessing steps for reducing the data dimensionality. The results demonstrate that the ICAMM significantly outperforms the K-means algorithm for land cover classification of hyperspectral imagery implemented on reduced data sets. Moreover, datasets extracted using segmented principal component analysis produce the highest classification accuracy.
| Original language | English |
|---|---|
| Article number | 1182251 |
| Pages (from-to) | 29-35 |
| Number of pages | 7 |
| Journal | Proceedings - Applied Imagery Pattern Recognition Workshop |
| Volume | 2002-January |
| DOIs | |
| State | Published - 1 Jan 2002 |
Fingerprint
Keywords
- Feature extraction
- Gaussian distribution
- Hyperspectral imaging
- Hyperspectral sensors
- Image segmentation
- Image sensors
- Independent component analysis
- Principal component analysis
- Remote sensing
- Testing
Cite this
}
ICA mixture model based unsupervised classification of hyperspectral imagery. / Shah, Chintan A.; Arora, Manoj K.; Robila, Stefan; Varshney, Pramod K.
In: Proceedings - Applied Imagery Pattern Recognition Workshop, Vol. 2002-January, 1182251, 01.01.2002, p. 29-35.Research output: Contribution to journal › Article › Research › peer-review
TY - JOUR
T1 - ICA mixture model based unsupervised classification of hyperspectral imagery
AU - Shah, Chintan A.
AU - Arora, Manoj K.
AU - Robila, Stefan
AU - Varshney, Pramod K.
PY - 2002/1/1
Y1 - 2002/1/1
N2 - Conventional remote sensing classification techniques that model the data in each class with a multivariate Gaussian distribution are inefficient, as this assumption is generally not valid in practice. We present a novel, independent component analysis (ICA) based approach for unsupervised classification of hyperspectral imagery. ICA, employed for a mixture model, estimates the data density in each class and models class distributions with nonGaussian structure, formulating the ICA mixture model (ICAMM). We apply the ICAMM for unsupervised classification of a test image from the AVIRIS sensor. Four feature extraction techniques namely principal component analysis, segmented principal component analysis, orthogonal subspace projection and projection pursuit have been considered as preprocessing steps for reducing the data dimensionality. The results demonstrate that the ICAMM significantly outperforms the K-means algorithm for land cover classification of hyperspectral imagery implemented on reduced data sets. Moreover, datasets extracted using segmented principal component analysis produce the highest classification accuracy.
AB - Conventional remote sensing classification techniques that model the data in each class with a multivariate Gaussian distribution are inefficient, as this assumption is generally not valid in practice. We present a novel, independent component analysis (ICA) based approach for unsupervised classification of hyperspectral imagery. ICA, employed for a mixture model, estimates the data density in each class and models class distributions with nonGaussian structure, formulating the ICA mixture model (ICAMM). We apply the ICAMM for unsupervised classification of a test image from the AVIRIS sensor. Four feature extraction techniques namely principal component analysis, segmented principal component analysis, orthogonal subspace projection and projection pursuit have been considered as preprocessing steps for reducing the data dimensionality. The results demonstrate that the ICAMM significantly outperforms the K-means algorithm for land cover classification of hyperspectral imagery implemented on reduced data sets. Moreover, datasets extracted using segmented principal component analysis produce the highest classification accuracy.
KW - Feature extraction
KW - Gaussian distribution
KW - Hyperspectral imaging
KW - Hyperspectral sensors
KW - Image segmentation
KW - Image sensors
KW - Independent component analysis
KW - Principal component analysis
KW - Remote sensing
KW - Testing
UR - http://www.scopus.com/inward/record.url?scp=84948992696&partnerID=8YFLogxK
U2 - 10.1109/AIPR.2002.1182251
DO - 10.1109/AIPR.2002.1182251
M3 - Article
VL - 2002-January
SP - 29
EP - 35
JO - Proceedings - Applied Imagery Pattern Recognition Workshop
JF - Proceedings - Applied Imagery Pattern Recognition Workshop
SN - 2164-2516
M1 - 1182251
ER -