 |
 |
   |
|
|
|
Fusion
of Spectral Bands MSS4 to MSS7
and Out-Image Data
using Bayes' Rule
Bayes' rule benefits from the additional information to make a decision
on the classification of a pixel. However, it has a very severe behaviour
as shown by the large number of not classified pixels (93.03%) (table "Rate
of classification").
In a severe behaviour, unanimity is sought in order to make a decision. If only one source does not agree with the others, decision becomes impossible, and classification fails. That explains the very high rate of not classified pixels.
On the other hand, when unanimity is reached, it is almost sure that the good class has been found for the pixel. That is shown on the image by the localization of the classified pixels which are exclusively around the samples areas. The unanimity reinforces the reliability of the decision.
Unclassified pixels: 93.03%
| Rate of classification | |||
| Class | Number of pixels
correctly classified (A) |
Number of pixels
in samples (B) |
Rate of pixels
correctly classified (A/B) |
| 1 | 63 | 459 | 13.73% |
| 2 | 88 | 459 | 19.17% |
| 3 | 80 | 306 | 26.14% |
| 4 | 44 | 391 | 11.25% |
| 5 | 100 | 459 | 21.79% |
| 6 | 185 | 459 | 40.31% |
| 7 | 62 | 459 | 13.51% |
| 8 | 37 | 459 | 8.06% |
| 9 | 237 | 459 | 51.63% |
| TOTAL | 896 | 3910 | 22.92% |
|---|---|---|---|
