Hyperspectral Image Classification Based on Non-Uniform Spatial-Spectral Kernels

In this paper, several criteria to measure the nonuniform distribution of the spatial-spectral information were investigated in hyperspectral remotely sensed images, then anovel weighted spatial- spectral kernel is introduced. They are proportional to combined (linear / constrained linear andnonlinear) distance measures. Then, the extracted weight vectoris applied to both spatial and spectral features. Several spectralspatial distance criterion including Bhattacharyya distance, theMutual Information (MI), Spectral Angle Mapper (SAM) and the spatial Markov Random Fields (MRF) energy function arecalculated and used in the SVM kernel design. Their combination with three aspects (Linear, constrained linear and nonlinear)regard to convex convergence conditions are examined. The proposed method was implemented for spatial-spectral kernel design. For performance evaluation we used, the data set ofIndiana Pines with 190 spectral bands. Comparison of average accuracy, overall accuracy and Kappa coefficient of classificationare presented along with class-maps. Experimental results achieved better accuracy and reliability, particularly in terms oflimited training samples, in comparison to some classification methods (ECHO and EMP).

Keywords:

Hyperspectral image classification , spatial energy function , kernel , non-uniform distribution information , Bhattacharyya distance , the Mutual Information , Spectral Angle Mapper , SVM , spatial Markov Random Fields energy function