A survey on representation-based classification and detection in hyperspectral remote sensing imagery

This paper reviews the state-of-the-art representation-based classification and detection approaches for hyperspectral remote sensing imagery, including sparse representation-based classification (SRC), collaborative representation-based classification (CRC), and their extensions. In addition to the original SRC and CRC, the related techniques are categorized into the following subsections: (1) representation-based classification with dictionary partition using class-specific labeled samples; (2) representation-based classification with weighted regularization by measuring similarity between each atom and a testing sample; (3) representation-based classification with joint structured models to consider contextual information during recovery optimization; (4) representation using spatial features in a preprocessing or a postprocessing step; (5) representation-based classification in a high-dimensional kernel space through nonlinear mapping; and (6) target and anomaly detection with sparse and collaborative representations. Some open issues and ongoing investigations in this field are also discussed.