Research On Knowledge Representation And Reasoning Mechanism For Spatial Relation Reasoning

Our living world is a spatial environment. Spatial reasoning is an essential activity forrecognizing world. Spatial reasoning has a widely application perspectives in GIS and somerelated areas, so it becomes an important aspect of GIS elementary theory. For a long time,people are exploring automatic spatial reasoning method, it means that compute must has humanspatial sense, spatial cognition, spatial representation, logical reasoning, learning andcommunication ability in spatial environment, this is why spatial reasoning is more difficultythan common reasoning. With respect to the research of spatial reasoning, spatial relationreasoning is a main content of spatial reasoning.Although some achievements have been achieved in spatial relation reasoning research area, itstill stays at the exploring stage, and there still exist many unsolved problems. On the whole, itlacks the methodology guidance and standardisation on spatial relation reasoning research. Forsome basic issues on spatial relation reasoning, such as the research contents and thecharacteristics of spatial relation reasoning, there has not been a common sense. It still lacksdeep and systematic research on knowledge representation and reasoning mechanism for spatialrelation reasoning. The realisation technology, the application model development for spatialrelation reasoning and so on all need further research.To solve the problems in spatial relation reasoning research, this paper will focus on creating theautomatic reasoning model of spatial relation. By integrating point set topology, algebra, settheory with the research achievements on ontology, knowledge engineer, artificial intelligenceand so on, spatial relation reasoning research is expanded from four aspects. They areformalisation representation of spatial relation, the basic reasoning model of spatial relation, thereasoning mechanism of spatial relation, and the way finding spatial relation reasoningapplication model. From the engineering point of view, the basic theory and technology issueson spatial relation reasoning are explored, and the automatic reasoning model of spatial relationreasoning is created. This research aims at getting an integrated framework of spatial relationreasoning, and providing effective technologies for practical application of spatial reasoningproblems.The main work is as follows: Introducing the ontology theory as a tool for spatial relation concepts abstracting andcognitive representation primitive analyzing. Starting from the spatial relation reasoningontology issues, some basic concepts and questions of spatial relation reasoning which mustclarify are analyzed. From the cognition difference of ontology primitive for spatial relationformal representation, the different spatial relation representation method is analyzed, andthe classification method of spatial relation representation based on the ontology primitiveis proposed. According to the standardization of spatial relation formalizationrepresentation ability, the research achievements in qualitative formalization representationmethods of topological, cardinal direction, and distance relations are summarized, the mainlimitations of the current spatial relation representation method is pointed out.Analyzing and Comparing several existing spatial relation reasoning method systematically,including calculus logical reasoning, reasoning based on composition table, algebraicreasoning, and production reasoning. After that, a suitable spatial relation reasoning methodintegrating algebraic reasoning, reasoning based on composition table, and productionreasoning is pointed out.To overcome the limitations of topological relation, cardinal direction relation, and distancerelation formalization representation, the significance of multi-level and multi-scalerepresentation spatial relation is explored, so the idea of multi-level and multi-scalerepresentation and reasoning spatial relation is proposed. According to that orderinginformation is less constraining than metric information and ordering information yieldsstronger constraints than topological information, an total multi-level and multi-scalerepresentation framework of spatial relation is constructed. After that, the role of ontologyin knowledge representation and reasoning of spatial relation is explored. A generic coreontology for spatial relation reasoning applications (SCR-Ontology) is constructed. Thenwe extend the core ontology at different level. According to the idea of multi-level andmulti-scale representation spatial relation, spatial relation is represented at different leveland scale. Based on the core ontology, we construct a multi-level and multi-scalerepresentation model for integrated representation of disjoint, topological, cardinal directionand distance relation.With respect to the disjoint relation representation and reasoning, the scale concept ofdisjoint relation representation is given, and the multi-scale representation model ofdisjoint relation based on the ontology is constructed.With respect to the topological relation representation and reasoning, the scale conceptof topological relation representation is given. Based on the SCR-Ontology,topological relation representation model at different level and scale is constructed. Weextend the SCR-Ontology with dimension, separation, and metric at different level, itarrives the aim at multi-level and multi-scale representation topological relation. Onthe basis of this, some spatial relation representation and reasoning algorithms forcomposed objects and objects with holes are given.With respect to the cardinal direction relation representation and reasoning, the scaleconcept of cardinal direction relation representation is given. The three-levelrepresenting and reasoning schema for cardinal directions between objects with point,line and area as reference are proposed. Each schema is divided into two parts again,the first one is the primary cardinal direction representation and reasoning model, thesecond one is the detailed representation and reasoning model with topological anddistance relation to represent and deduce more particularly. So it overcomes thelimitations of previous cardinal directions models using quite crude approximations inthe form of objects' abstract generalization points or their minimum boundingrectangles. Enlarging the cardinal direction representation and reasoning, representingand reasoning cardinal directions more accurately. Then the multi-level andmulti-scale cardinal direction relation representation and reasoning model based on theontology is constructed.With respect to the distance relation representation and reasoning, giving the scaleconcept of distance relation representation. Analyzing two different distance conceptsof Euclid distance and Voronoi distance. Constructing the multi-scale distance relationrepresentation model.On the basis of this, an integrating and unifying spatial relation qualitative formalizationrepresentation and reasoning model based on the ontology is constructed, it can representand deduce disjoint relation, topological relation, cardinal direction relation, and distancerelation. So it realizes the logical representation unification of spatial relation, and providesreuse tool model for spatial relation reasoning application.With respect to spatial relation reasoning mechanism, an integrated reasoning methodcombining algebra, composition table, and production method is presented. It can takeadvantages of each reasoning method. For improving the efficiency of spatial relation, themulti-level reasoning method for spatial reasoning is also proposed. Finally, multi-levelreasoning method for way finding spatial relation reasoning is studied.In this paper, the theory research and application of spatial relation reasoning is combined.The way finding spatial relation reasoning in large-scale space is analyzed, On the basis ofthis, the limitations of Kuipers way finding spatial relation reasoning model is explored, forexample, the spatial relation storing problem and so on. The method of dynamic deducingspatial relation in way finding by Voronoi is put forward. By using adjacent relation impliedin Voronoi, it can deduce various adjacent relations between object and path, object andplace;it can also conclude order relations among places on the path. For realizing wayfinding spatial relation reasoning, the 4-level schema is proposed. It includes spatialrelation representation basic model based on the SCR-ontology, spatial relation compositionbased reasoning model based on the ontology, spatial relation representation and reasoningextended model based on the ontology model, and the way finding spatial relationreasoning model based on the ontology. Each with its own ontology, each with its ownmathematical foundation, and each abstracted from the levels below it. Further more, themulti-level way finding spatial relation reasoning method is presented, and the way findingspatial relation reasoning model based on ontology is designed. Finally, based on COMtechnology, partly demonstrate the spatial relation composition based reasoning model andway finding spatial relation reasoning model.