In: Paparoditis N., Pierrot-Deseilligny M.. Mallet C. Tournaire O. (Eds). 1APRS. Vol. XXXVIII. Part ЗА - Saint-Mandé, France. September 1-3, 2010 
6. CONCLUSION 
In this paper, we have applied a powerful MCS to combine 
statistical and neural classifiers based on the FMV. To test the 
algorithm, three different classifiers based on four datasets of 
different sensor and scene characteristics were applied. The 
results showed an improvement in terms of overall 
classification accuracy and omission and commission errors of 
individual classes. Average overall accuracies of individual 
algorithms were 96.75%, 95.9% and 93.7% for SVMs, SOM 
and CTs respectively whereas the proposed fusion algorithm 
gives an accuracy of 97.85% which is an improvement of 
around 1.1%. This is an enhancement considering the limited 
room for improvement beyond 96.9% accuracy achieved with 
the SVMs, and that the data are most likely subject to other 
errors in image acquisition and image to lidar geographic 
registration, as well as errors in filtering of lidar point clouds. 
On the other hand, the average commission and omission errors 
have been reduced compared to the best single classifier. A 
comparison of the results with some of the existing fusion rules 
such as Maximum Rule (MR) and Weighted Sum (WS), 
demonstrates that the proposed fusion algorithm gives the best 
results. The computational cost involved in implementing the 
combined classifiers based on the FMV method is much higher 
than that of MR and WS methods. However, the processing 
time could be reduced by splitting large test areas into smaller 
parts, processing each part separately and combining the results 
later. For example, dividing the Fairfield test area, which is 
4km* in area, into four equal parts can saved more than 85 % of 
processing time (from 390 s to 58 s). The results in this paper 
demonstrate the overall advantages of the proposed fusion 
algorithm for combining multiple classifiers. 
REFERENCES 
Alimohammadi, A., H.R. Rabiei, and P.Z. Firouzabadi, 2004. 
A new approach for modeling uncertainty in remote sensing 
change detection process. Proceedings of the 12th 
International Conference on Geoinformatics - Geospatial 
Information Research: Bridging the Pacific and Atlantic, 07- 
09 June 2004, University of Gavle, Sweden, pp. 503-508. 
Benediktsson, J.A.. Chanussot, J. and Fauvel, M., 2007. 
Multiple Classifier Systems in Remote Sensing: From Basics 
to Recent Developments: MCS 2007, LNCS 4472, (M. 
Haindl, .1. Kittler, and F. Roli, editors), Springer Verlag, 
Berlin 2007, pp. 501-512. 
Breiman, L., Friedman, .1., Olshen, R., and Stone, C. J. 
(editors), 1984. Classification and Regression Trees, 
Chapman & Hall, New York, 358 p. 
Forstner, W., and Gulch, E., 1987. A fast operator for 
detection and precise location of distinct points, Comers and 
Centres of Circular Features. Proceedings of the ISPRS 1987 
Intercommission Workshop on Fast Processing of 
Photogrammetric Data, 2-4 June 1987, Interlaken, 
Switzerland, pp. 281-305. 
Giacinto, G., and F. Roli, 2001. Design of effective neural 
network ensembles for image classification purposes. Image 
Vision and Computing Journal, 19(9-10):669—707. 
Hao, H., C.L. Liu, and H. Sako, 2003. Comparison of Genetic 
Algorithm and sequential search methods for classifier subset 
selection. Proceeding of the seventh International Conference 
on document Analysis and Recognition (ICDAR), 3-6 August 
2003, Edinburgh. Scotland, pp. 765-769. 
Herrera. F. and Verdegay, J. L., 1996. A linguistic decision 
process in group decision making. Group Decision 
Negotiation, 5, pp. 165-176. 
Hugo, C., Capao, L., Fernando, B. and Mario, C, 2006. Meris 
Based Land Cover Classification with Self-Organizing Maps: 
preliminary results, Proceedings of the 2nd EARSeL SIG 
Workshop on Land Use & Land Cover, 28 - 30 September 
2006, Bonn. Germany, unpaginated CD-ROM. 
Kanellopoulos, I., G. Wilkinson. F. Roli, and J. Austin 
(editors), 1997. Neurocomputation in Remote Sensing Data 
Analysis. Springer, Berlin. 
Kohonen, T., 2001. Self-Organizing Maps. Third Edition, 
Springer, New York. 
Lam. L.. and CY. Suen, 1997. Application of majority voting 
to pattern recognition: an analysis of its behaviour and 
performance. IEEE Transactions on Systems, Man, and 
Cybernetics, 27(5): 553-568. 
Le, A.C., V.N. Huynh, A. Shimazu, and Y. Nakamori, 2007. 
Combining classifiers for word sense disambiguation based on 
Dempster-Shafer theory and OWA operators. Data & 
Knowledge Engineering, 63 (2):381-396. 
Mountrakis, G„ R. Watts, L. Luo, and J. Wang, 2009. 
Developing Collaborative Classifiers using an Expert-based 
Model. Photogrammetric Engineering and Remote Sensing, 
75(7):831-844. 
Salah, M., Trinder, J. and Shaker, A., 2009. Evaluation of the 
self-organizing map classifier for building detection from lidar 
data and multispectral aerial images. Journal of Spatial 
Science, 54, pp. 15-34. 
Vapnik, V., 1979. Estimation of Dependences Based on 
Empirical Data [in Russian]. Nauka, Moscow, 1979. (English 
translation: Springer Verlag, New York, 1982). 
Yager, R.R., 1998. On ordered weighted averaging 
aggregation operators in multicriteria decision making. IEEE 
Transactions on Systems. Man, and Cybernetics, 18, pp. 183- 
190. 
Yu-Chang, T. and C. Kun-Shan, 2009. An adaptive 
thresholding multiple classifiers system for remote sensing 
image classification. Photogrammetry Engineering and 
Remote Sensing, 75(6):679-687. 
Zadeh, L. A., 1983. A computational approach to fuzzy 
quantifiers in natural languages. Computers and Mathematics 
with Applications, 9, pp. 149-184. 
ACKNOWLEDGEMENTS 
The authors wish to acknowledge AAMHatch for the provision 
of the UNSW and Fairfield datasets, the Land Property 
Management Authority (LPMA), NSW. Australia for Bathurst 
data sets and the Toposys GmbH, Germany for the Memmingen 
datasets.