Abstract:

In the Three Gorges Reservoir area, there covers thick soils and flourish vegetation on the top of the rocks. The Three Gorges Reservoir area is an area which possesses flourish vegetation and poor lithology information. So it is difficult to classify and recognize the lithology in the Three Gorges Reservoir area, and on this aspect there are still no mature methods. In this paper lithology analysis is conducted focusing on the Three Gorges Reservoir area. Badong County in the Three Gorges Reservoir area is chosen as the study area. The ETM+ remote sensing image shot in May, 2000 is adopted to establish 17 classification factors on the aspects of texture, spectrum and vegetation covering. The remote sensing image is piled up with the geological map; and 1 101 sample points are chosen. Decision tree algorithm C4.5 is adopted to mine the interpretation knowledge of lithology in Badong County. The study precisions of the trees before and after pruning are 96.6% and 95.9% respectively. The precision of the rule extraction is 93.1%; and the confidence values of the rules are very high. By knowledge driving and rule matching, intelligent litology classification is conducted with the classification precision of 90.11%. The classification results are compared with ones of the other 6 methods (IsoData, K-Means, Mahalanobis Distance, Maximum Likelihood, Minimum Distance and Parallelepiped). The experimental results have shown that the decision tree method possesses the best classification results and the highest precision; and it is obviously superior to the other 6 methods.

Key words: remote sensing, lithology classification, intelligence, Three Gorges