Hyperspectral Data Processing: Algorithm Design and Analysis

27.4 Experiments

To demonstrate the utility of VNVBS in hyperspectral signature characterization, two completely different data sets were used for experiments and two particular applications, signature discrimination and mixed signature classification/identification were of interest and further considered for comparative analysis with SAM and SID used as spectral similarity measures. Nevertheless, other applications can also be explored for VNVBS.

27.4.1 Hyperspectral Data

The hyperspectral data used for computer simulations presented in this section were the five AVIRIS reflectance spectral signature vectors, blackbrush, creosote leaves, dry grass, red soil, and sagebrush, shown in Figure 1.8. Each of these five spectral signature vectors has 158 bands after water bands were removed and can be considered as a 158-dimensional hyperspectral signature vector where each signature component is specified by a particular spectral wavelength. According to Chapter 2 in Chang (2003a), the spectral profiles of blackbrush, creosote leaves, and sagebrush were close to each other. In particular, the creosote leaves and sagebrush even have very close spectral values. A detailed quantitative analysis among these three signature vectors is provided in Chapter 2 in Chang (2003a). In this section, these five signature vectors constitute a spectral library or database to be used to evaluate the performance of VNVBS in three different applications, signature discrimination, classification, and identification. ...