Sparse and Low-Rank Representation With Key Connectivity for Hyperspectral Image Classification

Abstract

Combined techniques of sparse representation (SR) and low-rank representation (LRR) are commonly used for hyperspectral image (HSI) classification. Although they have the ability to capture the interclass representations of data for HSI classification, they ignore the adaptive key connectivity of the learned intraclass data representations in particular with the high-dimensional complex HSI data. It is well-known that the key connectivity of graph-based algorithms is crucial for subspace learning because of the guarantees of its good neighbors. For this purpose, a novel sparse and low-rank representation with key connectivity (SLRC) method is proposed for HSI classification. To be specific, the adaptive probability graph structure is developed to integrate the SR and LRR regularizations to formulate SLRC model, which flexibly perform discriminative latent subspace construction and preserve the key connectivity of intraclass representations. Then, extensive experiments are executed based on three popular HSI data sets, which demonstrates that the SLRC method outperforms the other popular methods. Index Terms-Sparse representation (SR), low-rank representation (LRR), key connectivity, hyperspectral image (HSI). I. INTRODUCTION YPERSPECTRAL images (HSIs) provide detailed structural and spectral information because they comprise hundreds of narrow spectral bands [1-4], which can effectively capture the subtle differences between different materials and facilitate better land-cover classification. The classification of HSIs, where each pixel is assigned one thematic class in a scene, has attracted much attention in many studies because it plays a vital role in various applications [5, 6]. In fact, effective feature expression is important in land-cover classification since they greatly affect the analysis of numerous HSI applications. Recently, graph-based learning algorithms have got widespread attention in representation learning because of their interpretability and effectiveness in practice [7, 8]. The critical step in graph-based learning aims at building good graph to

Graph-Based Semisupervised Learning With Weighted Features for Hyperspectral Remote Sensing Image Classification

“…In the past decade, HSI classification has been solved by some machine learning methods, including sparse representation [11], [12], support vector machine (SVM) [13], [14] and neural network [15]- [18]. But these machine learning methods rely on artificial feature extraction, and the extracted features are relatively single.…”