Study On Technology And Application Of Qualitative Spatial Reasoning

RS and GPS provide technical support for collection of spatial information, andGIS creates the theoretical and technical architecture of spatial information processingand application from the aspect of quantity. However, spatial information is complex.Considering quantity, spatial information is mass. The kinds of spatial information notonly have numerical type, character type, but also have figure type and image type. Thecorrelation between spatial information is inherent. So, quantitative research of spatialinformation is not enough, and qualitative research is necessary. Qualitative spatial reasoning that is generated by combining qualitative reasoning,spatial reasoning and artificial intelligence (AI) has been a hotspot in the research of AI.Different from the researches of structure, action, function and their interconnections ofphysical system in qualitative physics, qualitative spatial reasoning analyzes andprocesses the human cognition of geometric spatialcommon knowledge. Different fromthe researches of modeling, descriptions, representations, relations' analysis andprocessing of spatial objects in spatial reasoning researching, qualitative spatialreasoning faces to the qualitative characters and qualitative relations of spatial objects.Qualitative spatial reasoning researches representation and process of human'squalitative cognition for spatial objects and their relations in geometric space. Spatial objects researched in this paper are points, lines and areas in 2-dimensionalspace, and contents are qualitative relations representation and reasoning betweenspatial objects. Qualitative relations researched are topological relations and directionrelations, and reasoning researches include combinative table reasoning and hierarchicalstructure reasoning. Based on the profound analysis for representation and reasoning ofqualitative spatial relations, this paper proceeds with some originality innovationresearches as follows:(1) Representation of spatial topological relations Based on the point-set topology and ontology, representation methods for spatialtopological relations are researched. That RCC model, n-intersection model andclassification invariant model have the same classification capability of topologicalrelations between 2-dimensional spatial areas is proved; That 4-intersection model isnecessary and sufficient for classifying topological relations between 2-dimensionalspatial areas is proved, then these 3 different methods are unified. Topological III重庆大学博士学位论文classification invariants between spatial objects with different types are present andproved, and research achievements given are consummated. A general representationmethod for intersection model of objects is present, and a base is established forachieving more strong capability of classifying topological relations. From the aspect of human cognition, HBM (Hierarchy-based Method), a newmethod for classifying topological relations, is present based on hierarchical cognition.The classification capability of HBM is not weaker than that of any model in RCCmodel, n-intersection model (4IM, 9IM, DE₄IM, DE₉IM), classification invariantmodel and CBM.(2) Representation of spatial direction relations Based on spatial projection model, D14, the representation model of directionrelations that satisfies human cognition, is present. Compared to D9 and D13, D14 hasmore strong representation capability of direction relations and the rationality ofcognition. D14 is more simple and convenient than D169. Based on 2D character model, character string model is present, and 14 rulesapplying 1-dimensional character projection sequence to determine spatial directionrelations are achieved. That character string model and D14 have the same classificationrepresentation capability of spatial direction relations.(3) Combinative table and reasoning technology A fusion method for spatial topolog...