|
|
|
Image Processing
|
Self-Organizing Feature Map for Multi-Spectral Spot Land Cover Classification
Table 2 MLC classification matrix using spatial features.
| Class |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
Producer's purity (%) |
| 1 |
2322 |
0 |
0 |
0 |
178 |
0 |
0 |
92.9 |
| 2 |
0 |
2347 |
0 |
0 |
138 |
15 |
0 |
93.9 |
| 3 |
6 |
0 |
1760 |
0 |
662 |
73 |
5 |
74.4 |
| 4 |
64 |
0 |
3 |
2366 |
41 |
0 |
26 |
94.6 |
| 5 |
63 |
2 |
176 |
0 |
2254 |
5 |
0 |
90.1 |
| 6 |
0 |
68 |
37 |
0 |
132 |
2263 |
0 |
90.5 |
| 7 |
49 |
0 |
24 |
0 |
75 |
0 |
2352 |
94.1 |
| User's purity (%) |
93 |
97.1 |
88.3 |
100 |
64.8 |
96.1 |
98.7 |
|
Overall purity=90%, coefficient=0.88
Table 3 SOM classification matrix using spatial features.
| Class |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
Producer's purity (%) |
| 1 |
2348 |
0 |
0 |
0 |
152 |
0 |
0 |
93.9 |
| 2 |
8 |
2408 |
25 |
0 |
59 |
0 |
0 |
96.3 |
| 3 |
18 |
0 |
2361 |
0 |
121 |
0 |
0 |
94.4 |
| 4 |
75 |
0 |
21 |
2378 |
3 |
0 |
23 |
95.1 |
| 5 |
75 |
3 |
215 |
0 |
2207 |
0 |
0 |
88.3 |
| 6 |
3 |
47 |
110 |
0 |
45 |
2295 |
0 |
91.8 |
| 7 |
148 |
0 |
34 |
0 |
0 |
0 |
2310 |
92.4 |
| User's purity (%) |
87.8 |
98 |
85.4 |
99.7 |
85.3 |
100 |
99.0 |
|
Overall purity=93.2%, coefficient=0.921
Conclusions
A large variety of back-propagation methods are used to train the networks. The convergence of a learning process is sensitive to the selection of a training data set, and the learning method often requires a large number of iterations and much computational time. The method is a black box approach that is difficult to give physical meaning to weights connecting the neurons. The utility of the SOFM lies mainly in its fairly rapid convergence, it can capture the probability distribution of the inputs, and it is easy to interpret. From the experiments, it gets pretty good results compared to other methods.
Reference
-
Andrews, H. C., 1972. Introduction to Mathematical Techniques in Pattern Recognition. Wiley, N. Y.
- Kohonen, T., 1982. Self-organized formation of topologically correct feature maps. Biol. Cybern., vol. 43, pp. 59-69.
- Kohonen, Teuvo, 1984. Self-Organization and Associative Memory. volume 8 of Springer Series in Information Sciences, Springer, New York, pp. 184-189.
- Zheng,Yi and Greenleaf, James F, 1996. The effect of concave and convex weight adjustments on self-organizing Maps. IEEE Trans. Neural Networks, vol. 7, no. 1, pp. 87-96.
|
|
|
|
|
|
|
