Abstract: The purpose of this paper is to overcome the problem that the uniform classification model cannot be used to improve the accuracy of land use classification effectively, and to provide a method with high precision and efficiency to extract qualitative and quantitative information of land use/cover. In this paper, two extraction methods, i.e., the hierarchical information extraction method and the method based on knowledge and rule, are combined. The hierarchical information extraction method can create a cleaner extracting environment when extracting specific types of land use/cover. The method based on knowledge and rule can give full play to the role of priori knowledge and experience, and reference diverse assistant information comprehensively. The purpose of combining the two methods is to further improve the classification accuracy. Taking the metropolitan area of Nanjing as an example, a set of land use information hierarchical extraction process is designed and put into effect. In the extraction process, TM image spectrum information and the temporal and spatial characteristics of land use in study area are fully considered. Using different ways and data such as water extraction index model, Ratio Resident-area Index (RRI), NDVI, DEM data and the urban built-up area boundary, we extracted eight types of land use/cover, i.e., water, urban build-up area, rural build-up area, mining land, bare land, woodland, farmland, and urban green land. Specially, we constructed a build-up and quarry discrimination index (BQDI) to efficiently distinguish build-up area from mining/bare land, and make full use of elevation and slope information to extract woodland from vegetation coverage area. In addition, the urban build-up area boundary plays an important role in distinguishing urban build-up area and rural build-up area, farmland and urban green land, mining land and bare land. The results showed that the overall accuracy (OA) reached 88.67%, and the Kappa index reached 0.85 when extracting the land use/cover information of 2012 in study area. Besides, the overall accuracy on other years' information extraction were all higher than 87%, and the Kappa index were all more than 0.84. We concluded that the hierarchical land use information extraction method based on knowledge and rule possesses high precision and wide applicability. The method has a good reference value to land use/cover extraction practice in other regions. Further research should be undertaken to increase the degree of automation in information extraction process, the efficiency and accuracy of extraction in the future.

Key words: knowledge and rule, land use/cover, hierarchical extraction, the metropolitan area of Nanjing