Data Exploration and Analysis of Hyperspectral images: Visualization and Symbolic Description
2.2 Spectral Space
The spectral response of a material can be defined as a spectral signature by the reflectance or radiance as a function of wavelength. Therefore, the spectral variations of each pixel can be drawn as a curve on the spectral space (see figure 1.b). Theoretically each class related to the composition of different material has its own shape and variance of the spectral curve. Some methods like “spectral matching” and “spectral angle mapper” use this property to distinguish the unknown spectral curve comparing with a series of pre-labeled spectral curve. Figure 3 shows the spectral signature of five different classes. Some basic statistics are calculated to depict the characteristics of the spectral variation. The mean curve represents the trend of the spectral variation. The standard deviations show the scattering to the mean. The maximum and the minimum values present the range of variation. One may find that the Grass-Trees and Hay-windrowed have very similar mean spectral curve, but present very different
stadard deviation, maximum and minimum values. Different spectral curves can be portrayed on one spectral space for comparison. Figure 4.a shows the overlaps of five different spectral signatures. The spectra can also be offset vertically to allow interpretation (see figure 4b).
Figure 3. The spectral signature of five different materials.
(a) Overlap spectral data
(b) Stack spectral data
Figure 4. Different spectral curves in the spectral space.
