Hyperspectral Data Processing: Algorithm Design and Analysis

19.6 Exploration-Based Applications

The exploitation-based applications are a key component in hyperspectral information compression proposed in Figures 19.2(a), (b), and 19.3 because the other two components, DR/BS and 3D compression, are developed to support and enhance its functionalities. It is this component to make a hyperpscetral information system versatile and adaptive. Its success is determined by the exploitation criterion used for hyperspectral information compression that represents the information extracted from data for future information retrieval and data processing. While considering all possible exploitation applications is impossible, in this section we describe only four exploitation applications of particular interest in hyperspectral image analysis, anomaly detection, subpixel target detection, spectral unmixing, and endmember extraction, each of which requires a different level of target information to be compressed. For example, anomaly detection and endmember extraction requires no target information at all compared to the spectral unmixing that needs complete target knowledge of image endmembers in the data. The subpixel target detection is somewhere in between and needs only the target information of interest while discarding all other target knowledge.

19.6.1 Linear Spectral Mixture Analysis

Linear spectral mixture analysis (LSMA) has been widely used to perform spectral unmixing. For LSMA to work well, we need to find an appropriate set of image endmembers ...