Demand for up-to-date geographic data is increasing due to the fast growth of Geographic Information System (GIS). The process of updating GIS database can be subdivided into three steps. In the first step, changes of the landscape must be detected. In the second step, different data sources must be used to add further attributes. In the last step, the changes with all additional information have to be stored into a GIS database. In this paper, a new approach of change detection for updating GIS database by using remote sensing image is presented. This approach of change detection is based on post-classification comparison and spatial reasoning. Firstly, two different classification methods of satellite imagery, one of pixel-based classification and other of object-oriented classification, have been used. The already existing GIS data are used for deriving the training areas for the classification. The fusion of the results of classification methods yields a good comparison of classification to realizing the change detection between the remote sensing data and GIS data. Secondly, a qualitative approach using spatial reasoning system (Region Connection Calculus RCC8) to representing GIS data and satellite image content in addition to the quantitative information is performed. After a step of segmentation of the satellite image, the spatial reasoning system RCC8 is used to extract and identify the spatial relations (topological and orientation) between the GIS objects and the objects of the results of segmentation. The topological relation describes how the boundaries of two objects relate to each other, while the orientation relation describes where the objects are placed relative to one another. The fusion of the results of post-classification comparison and the spatial reasoning system can detect changes at the level of thematic data, geometric and spatial relations between objects in satellite image and the GIS data. Finally, the changes with all information are returned and stored into the GIS database. This approach is tested with Corine Land Cover (CLC) database and SPOT satellite images.