Over the past decade, automated systems dedicated to geopositioning have been the object of considerable development. Despite the success of these systems for many applications, they cannot be directly applied to qualitative descriptions of space. The research presented in this paper introduces a visibility and constraint-based approach whose objective is to locate an observer from the verbal description of his/her surroundings. The geopositioning process is formally supported by a constraint-satisfaction algorithm. Preliminary experiments are applied to the description of environmental scenes. Key words: Landscape perception, place descriptions, scene-finding approach, geopositioning.