<title>Mixed pixels classification</title>

Abstract

There are two major approaches in spectral unmixing: linear and non-linear ones. They are appropriate for different types of mixture, namely checkerboard mixtures and intimate mixtures. The two approaches are briefly reviewed. Then in a carefully controlled laboratory experiment, the limitations and applicability of two of the methods (a linear and a non-linear one) are compared, in the context of unmixing an intimate mixture.

Does Independent Component Analysis Play a~Role in Unmixing Hyperspectral Data?

“…This can be understood by computing the correlation factor between sources and . Assuming that is independent of and that is independent of , we have (9)- (11), shown at the bottom of page, where was invoked to obtain the right hand side of (10) and (11). We conclude then that signature variability does not increase source correlation.…”

“…If the multiple scattering among distinct endmembers is negligible and the surface is partitioned according to the fractional abundances, then the spectral radiance upon the sensor location is well approximated by a linear mixture of endmember radiances weighted by the correspondent fractional abundances [4], [10], [11], [33], [48].…”

“…Depending on the mixing scales at each pixel, the observed mixture is either linear or nonlinear [4], [11]. The linear mixing model holds when the mixing scale is macroscopic [12].…”

Context dependent spectral unmixing.

“…The resulting endmembers are shown in figure 6. 31. As it can be seen, the identified endmembers are unaffected by the noise points, and are similar to the results of RCDSU in figure 6.27(b).…”

“…Mixing models can be characterized as either linear or nonlinear [1,31]. The linear mixing model (also known as the convex geometry model ) holds when the mixing scale is macroscopic [32] and the incident light interacts with just one material, as is the case in checkerboard type scenes [33,34].…”