The paper presents an approach for the reclassification and generalization of land-use information from topographic information. Based on a given transformation matrix describing the transition from topographic data to land-use data, a semantic and geometry based generalization of too small features for the target scale is performed. The challenges of the problem are as follows: (1) identification and reclassification of heterogeneous feature classes by local interpretation, (2) presence of concave, narrow or very elongated features, (3) processing of very large data sets. The approach is composed of several steps consisting of aggregation, feature partitioning, identification of mixed feature classes and simplification of feature outlines. The workflow will be presented with examples for generating CORINE Land Cover (CLC) features from German Authoritative Topographic Cartographic Information System (ATKIS) data for the whole are of Germany. The results will be discussed in detail, including runtimes as well as dependency of the result on the parameter setting.