TABLE 3: 
Classification accuracies of different band subsets for the 
eight band-moment data set as well as for the GEOBOTANY band 
set (Table 2). The different band subsets were determined with 
a feature selection approach using a branch and bound algorithm 
(average pairwise transformed divergence). 
Weighted Mean 
Classification Accuracy 
Data Set 
Band Subset 
(± standard error) 
GEOBOTANY 
4 
57.03 
+ 
10.59 
4,7 
82.24 
+ 
5.33 
3,5,7 
92.21 
± 
2.09 
1,4,5,6 
94.54 
± 
1.50 
1,2,4,5,6 
96.46 
± 
1.04 
1,2,4,5,6,8 
97.73 
± 
0.72 
1,2,3,4,5,6,8 
98.19 
± 
0.61 
1,2,3,4,5,6,7,8 
98.59 
± 
0.45 
BAND-MOMENTS 6 
61.38 
± 
12.49 
2,5 
85.34 
± 
4.04 
1,7,8 
94.86 
± 
1.74 
1,6,7,8 
96.65 
± 
1.21 
2,4,5,6,7 
97.91 
± 
0.86 
1,3,4,6,7,8 
99.24 
± 
0.43 
1,2,4,5,6,7,8 
99.33 
± 
0.32 
1,2,3,4,5,6,7,8 
99.41 
± 
0.27 
n 
If calibration 
data applied 
Figure 1: Radiometric preprocessing steps implemented in the Imaging Spectrometer Data Analyzer 
(ISDA) for PMI data sets. 
557