Geoscience Reference
In-Depth Information
Table 11.2 Accuracy Statistics for Classification
Based on MAP Selection from Conditional
Probabilities Computed Using Different
Methods: Gaussian Maximum Likelihood
(GML), Ordinary Indicator Kriging (OIK),
and Bayesian Integration of GML and OIK
Probabilities (GML/OIK)
GML
OIK
GML/OIK
Overall accuracy
0.73
0.73
0.80
Kappa
0.44
0.44
0.59
Producer's accuracy
Forest
0.92
0.88
0.91
Shrub
0.44
0.52
0.63
Rangeland
0.30
0.39
0.51
User's accuracy
Forest
0.82
0.78
0.86
Shrub
0.48
0.61
0.64
Rangeland
0.55
0.63
0.68
Table 11.3 Class Proportions from Reference and
Classified Maps Based on MAP Selection from
Conditional Probabilities Computed Using
Different Methods: Gaussian Maximum
Likelihood (GML), Ordinary Indicator Kriging
(OIK), and Bayesian Integration of GML and OIK
Probabilities (GML/OIK)
Reference
GML
OIK
GML/OIK
Forest
0.65
0.73
0.73
0.69
Shrub
0.21
0.19
0.18
0.21
Rangeland
0.14
0.08
0.09
0.10
(a)
(b)
Figure 11.5
Pixel-specific accuracy values for GML-derived classes (a) and for GML/OIK-derived classes (b).
