Remote Sensing Data Classification Using Combined Spectral and Spatial Local Linear Embedding (CSSLE)

Abstract

Feature learning and extraction have been used to reduce the complexity of the representation of remote sensing data. In this paper, a novel remote sensing data feature analysis method is proposed based on the locally linear embedding (LLE) techniques, which is an unsupervised manifold learning algorithm. If the remote sensing data variability is described by a small number of spatial features, we can treat the data as lying on a low dimensional manifold in the high dimensional space of remote sensing data. The proposed method, combined spectral and spatial local linear embedding (CSSLE) makes use the neighbors in both domains. Compared with LLE, CSSLE makes up the shortage that LLE ignores the relationship among the spatial neighboring pixels which is extremely important for remote sensing images. In this paper we provide experiment results from the analysis of remote sensing data using PCA, LLE and CSSLE. Classification results show that proposed method can give higher accuracies than the linear method of PCA and the nonlinear method of LLE.