Nonlinear Estimation of Hyperspectral Mixture Pixel Proportion Based on Kernel Orthogonal Subspace Projection

Abstract

A kernel orthogonal subspace projection (KOSP) algorithm has been developed for nonlinear approximating subpixel proportion in this paper. The algorithm applies linear regressive model to the feature space induced by a Mercer kernel, and can therefore be used to recursively construct the minimum mean squared-error regressor. The algorithm includes two steps: the first step is to select the feature vectors by defining a global criterion to characterize the image data structure in the feature space; and the second step is the projection onto the feature vectors and then apply the classical linear regressive algorithm. Experiments using synthetic data degraded by an AVIRIS image have been carried out, and the results demonstrate that the proposed method can provide excellent proportion estimation for hyperspectral images. Comparison with support vector regression (SVR) and radial basis function neutral network (RBF) had also been given, and the experiments show that the proposed algorithm slightly outperform than RBF and SVR.