TY - GEN
T1 - Mahalanobis kernel for the classification of hyperspectral images
AU - Fauvel, M.
AU - Villa, A.
AU - Chanussot, J.
AU - Benediktsson, J. A.
PY - 2010
Y1 - 2010
N2 - The definition of the Mahalanobis kernel for the classification of hyperspectral remote sensing images is addressed. Class specific covariance matrices are regularized by a probabilistic model which is based on the data living in a subspace spanned by the p first principal components. The inverse of the covariance matrix is computed in a closed form and is used in the kernel to compute the distance between two spectra. Each principal direction is normalized by a hyperparameter tuned, according to an upper error bound, during the training of an SVM classifier. Results on real data sets empirically demonstrate that the proposed kernel leads to an increase of the classification accuracy by comparison to standard kernels.
AB - The definition of the Mahalanobis kernel for the classification of hyperspectral remote sensing images is addressed. Class specific covariance matrices are regularized by a probabilistic model which is based on the data living in a subspace spanned by the p first principal components. The inverse of the covariance matrix is computed in a closed form and is used in the kernel to compute the distance between two spectra. Each principal direction is normalized by a hyperparameter tuned, according to an upper error bound, during the training of an SVM classifier. Results on real data sets empirically demonstrate that the proposed kernel leads to an increase of the classification accuracy by comparison to standard kernels.
KW - Classification
KW - Hyper-spectral images
KW - Mahalanobis kernel
KW - Probabilistic principal component analysis
KW - Support vector machine
UR - https://www.scopus.com/pages/publications/78650897460
U2 - 10.1109/IGARSS.2010.5651956
DO - 10.1109/IGARSS.2010.5651956
M3 - Conference contribution
SN - 9781424495658
SN - 9781424495665
T3 - International Geoscience and Remote Sensing Symposium (IGARSS)
SP - 3724
EP - 3727
BT - 2010 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2010
T2 - 2010 30th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2010
Y2 - 25 July 2010 through 30 July 2010
ER -