Mining Topological Relationship Patterns from Spatiotemporal Databases

Mining topological relationship patterns involve three aspects. First one is the discovery of geometric relationships like disjoint, cover, intersection and overlap between every pair of spatiotemporal objects. Second one is tracking the change of such relationships with time from spatiotemporal databases. Third one is mining the topological relationship patterns. Spatiotemporal databases deal with changes to spatial objects with time. The applications in this domain process spatial, temporal and attribute data elements to find the evolution of spatial objects and changes in their topological relationships with time. These advanced database applications require storing, management and processing of complex spatiotemporal data. In this paper we discuss a model-view-controller based architecture of the system, the design of spatiotemporal database and methodology for mining spatiotemporal topological relationship patterns. Prototype implementation of the system is carried out on top of open source object relational spatial database management system called postgresql and postgis. The algorithms are experimented on historical cadastral datasets that are created using OpenJump. The resulting topological relationship patterns are presented.