Abstract: In essence, map data processing can be considered as a topological transformation, i.e.under continuous transformation (data processing) an invariance (structure feature) inherent in the data should be preserved. When a data processing system is enriched with topological relations and controlled by the later, the system's intelligence will be strengthened considerably. This causes the linear property in previous data processing to become non-linear (correlated). In accordance with the strength of the adjacency and incidence between the objects the topological relations can be classified into two major categories, determinative and fuzzy types. In principle, the former should be represented explicitly, while the later implicitly. In this paper a brief comparision between the two representations is made. It is emphasized that the implicit structure has some advantages, economy and flexibility. This paper notes that the quantity of topological information input manually should tend to minimum. The retrieval of geo-information can be classified into 2 groups, traditional and topological retrieval. The former deals with the object attributes only, e.g. the property oriented retrieval and location oriented one, the later is based on the spatial relationships (e.g.adjacency, incidence and inclusion) between the object sets. The essence of the indirect topological retrieval based on traditional one is that the explicitation of topological relations implicited in the data structure is realized through two inverted files called header index (attribute-oriented) and location index (raster or mesh oriented). At the end of this paper 3 typical means for indirect topological retrieval are introduced. Topological retrieval based on point set; Topological retrieval based on line set; Topological retrieval based on area set.