Abstract:
With the development of target decomposition theorems for polarimetric synthetic aperture radar (PolSAR) data, various informative polarimetric descriptors could be obtai...View moreMetadata
Abstract:
With the development of target decomposition theorems for polarimetric synthetic aperture radar (PolSAR) data, various informative polarimetric descriptors could be obtained. The redundancy among these descriptors poses a hindrance to accurate classification. In this paper, we propose a tensor-based dimension reduction technique, which aims to obtain a lower-dimensional intrinsic feature set from the high-dimensional polarimetric manifold. We combine 48 polarimetric features together and formulate them as a third-mode tensor. The spatial information is taken into consideration for feature reduction. Experimental results in comparison with principal component analysis (PCA), independent component analysis (ICA) and Laplacian Eigenmaps (LE) proves its effectiveness.
Published in: 2014 IEEE Geoscience and Remote Sensing Symposium
Date of Conference: 13-18 July 2014
Date Added to IEEE Xplore: 06 November 2014
Electronic ISBN:978-1-4799-5775-0