Kernel Multivariate Analysis in Remote Sensing Feature Extraction
Jerónimo Arenas-Garcia and Kaare B. Petersen
Kernel Methods for Remote Sensing Data Analysis
Feature extraction has become an important topic in remote sensing due to the usually very high dimensionality of data, as well as the high redundancy among spectral bands. This chapter reviews different Multivariate Analysis techniques for feature extraction, providing a uniform treatment of Principal Component Analysis (PCA), Partial Least Squares (PLS), Canonical Correlation Analysis (CCA) and Orthonormalized PLS (OPLS). Non-linear extensions are derived by means of kernel methods. Special attention is paid to a sparse version of Kernel OPLS, which can significantly reduce the computational burden both in training and operational phases. The discriminative power of these MVA methods has been studied on a classification scenario, showing the superiority of kernel over linear versions, and the convenience of using the target data for training the feature extractors.