Hyperspectral Data Processing: Algorithm Design and Analysis

13.7 Synthetic Image Experiments

This section conducts two sets of experiments, synthetic image and real image experiments, to demonstrate the utility of FLSMA in mixed pixel classification and quantification. For AC-FLSMA, only AFCLS-FLSMA that is abundance fully constrained FLSMA is conducted to compare its counterpart of LSMA, AC-FCLS which is actually the FCLS developed by Heinz and Chang (2001).

In order to substantiate FLSMA, a synthetic image similar to the real scene in Figure 1.15(a) was simulated. It has the size of pixel vectors and 20 panels with various sizes arranged in a matrix and located at the center of the scene shown in Figure 13.1(a). The five panel signatures in Figure 1.16 were used to simulate these 20 panels.

Figure 13.1 A synthetic image, (a) 20 simulated panels; (b) background simulated by a grass signature corrupted by an additive Gaussian noise with SNR 20:1, (c) a synthetic image with the 20 simulated panels in (a) implanted in the background simulated in (b).

For row i, the panel signature p_i was used to simulate four panels in each of columns where the panels are a -pixel panel, {,,,} in the first column, a -pixel panel, {,} in the second column, a ...