In order to plan behavior-based navigation strategies, a mobile robot needs information about the connectivity of the space, and about those geometric properties that allow the selection of the adequate behavior in each area. We propose to represent this information by a topology-based map, a graph that encodes the topological structure of the free space in the environment, in which nodes represent large open spaces and arcs represent passages between them. In this paper, we show how to extract a topology-based map from a fuzzy occupancy grid using image processing techniques, and how to use this map for planning behavioral navigation. We illustrate our technique by giving an example of office navigation on a real robot.