International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
    
  
Ke Predicted Class ae 7 ey 
WT UI IV BR BQ SB NG JP 
WT 223 0 1 0 2 0 0 0 98.67 
Ul 0 852 1 55 0 43 3 5 88.84 
IV 0 ] 522 2 1 0 0 0 99.24 
BR 0 12 0 1402 0 3 3 21 97.20 a] 
BQ 0 0 0 0 81 0 0 0 100 
SB 0 19 0 2 0 327 25 5 86.51 ? De 
NG 0 4 0 36 0 99 57 5 28.36 
JP 0 8 0 45 0 47 12 63 36.00 
Average accuracy (%) = 79.36 Overall Accuracy (%) = 88.46 
Table 2. Classification matrix for the study area by using BPNN 
KEY V 
At Predicted Class ga ey 
ABSTE 
WT UI IV BR BQ SB NG JP 
WT 222 0 1 0 3 0 0 0 98.23 Polarim 
Ul 0 9 ] 4 0 29 0 8 0 8 95.3 ] proport 
IV 0 6 520 0 0 0 0 0 98.86 synthes 
BR 0 39 0 1367 0 4 0 31 94.86 derived 
BQ 0 0 0 0 81 0 0 0 100 image « 
SB 0 63 0 4 0 270 6 35 71.43 Rajski 
NG 0 8 0 28 0 86 63 16 31.34 for text 
JP 0 11 0 43 0 21 3 97 55.43 differer 
Average accuracy (%) = 80.68 Overall Accuracy (%) = 88.64 probabi 
inspect 
Table 3. Classification matrix for the study area by using RBFC Eid : 
slightly improve the classification accuracy compared with 5. ACKOWELDGEMENT accurac 
BPNN. It can provide fuzzy classification results, more expand 
appropriate in the case of mixed, intermediate, or complex The authors wish to thank the Earth Data Analysis Center 
cover pattern pixels. The structure of this algorithm is ~~ (EDAC) at UNM for the Landsat data set and grateful help. We 
composed of a limited number of fuzzy rules, which are also thank the support and discussion of students and staff of 
interpretable and can be modified by human knowledge. the Autonomous Control Engineering (ACE) Center at UNM. The a 
of the | 
6. REFERENCES validity 
classifi 
Berenji, H. R. and Khedkar, P. S., 1993, "Clustering in Product networ 
Space for Fuzzy Inference”, IEEE Fuzzy, pp. 1402-1407. scatteri 
corresp 
Duda, R. O., Hart, P. E. and Stork, D. G., 2000, Pattern inform: 
Classification, Jone Wiley & Sons (Asia) Pte. Ltd., New York, scatteri 
2001. al., 19€ 
S"R^a 
Foody, G.M. 1992, “A fuzzy sets approach to the inform: 
representation of vegetation continua from remotely sensed that or 
data: An example from lowland heath”, Photogramm. Eng. technic 
Remote Sens., 58(2), pp. 443-451. accura 
inform 
Heermann, P.D. and Khazenic, Nahid, 1992, “ Classification of image 
S nk Multispectral Remote Sensing Data using a Back-Propagation arisen 
vds NX Neural Network", [EEE Trans. Geoscience and Remote used f 
se Urban Impervi Sensing, 30(1), pp. 81-88. such € 
[7] trrigated Vegetation (IV) LL] Barren (BR) ent 
Bosque (BQ) 2 Shrubland (SB) Vassilas, N. and Charou, E., 1999, *A New Methodology for qum à 
[1 Natural Grassland (NG) [HES Juniper (JP) Efficient Classification of Multispectral Satellite Images Using the pol 
; ; QUSE Neural Network Techniques", Neural Processing Letters, 9, pp. the ap 
Figure 3. RBFC classification result 35-43. vector 
Wang, F., 1990, "Improving Remote Sensing Image Analysis put 
Through Fuzzy Information representation,” Photogramm. Eng. TEE 
20 
Remote Sens., 56(8), pp. 1163-1169.