Representation, reasoning and similar matching for detailed topological relations with DTString

Most current topological relation models cannot capture the details of spatial relations; thus, complex spatial relations cannot be distinguished using these models. Under certain circumstances, only detailed topological relations can satisfy users' demand. To resolve this problem, this study proposes a new topological relation model named DTString that describes the full details of the topological relation between two regions by a boundary string. DTString is proved to be a JEPD (Jointly Exhaustive and Pairwise Disjoint) and atomic relation model. Thus, undividable topological relations can be captured with it, and this characterization makes it suitable for modeling complex topological relations. Furthermore, DTString-based reasoning algorithms are investigated; they are more efficient than algorithms based on geometric calculations because they use purely string-based calculations. Finally, methods for retrieving similar geometrical structures are investigated by combining topological relations with shape and distribution features. Experiment results show that DTString outperforms existing topological models. DTString and the proposed similarity measurements can be potentially applied to content-based image retrieval, spatial query and CAD, etc.