Research On The Classification And Fusion Of Geographic Information Based On Semantic Understanding

With the advent of the ambient information age, the ubiquitous geospatial information is changing our life and understanding way. Much progress has been made in interdisciplinary applications of geospatial information constantly. A new people-oriented,user-centric, real-time and dynamic map demand is emerging, which highlights the combinations of human-computer interaction of geographic information and knowledge sharing. However, there remain many issues to be solved. First, the experts from various application domains should take into account how to strengthen the understanding the same geographical concept, and solve the problem of semantic intersection and overlap during the process of classification among the basic geographic information; Second, due to the different data standards and incompatible terminologies for expressing spatial information in geographic information science, it is easy to produce semantic heterogeneity. How should we carry out multi-source semantic integration of geospatial information, and reduce the complexity of the problem when the geographical concepts have increased rapidly. Finally, we should consider how to determine the semantic relationships among different geographical concepts in the process of geospatial semantic classification and integration.In response to the above problem, we mainly focus on the semantic classification of the fundamental geographic information in China and multi-source geospatial semantic fusion research in this thesis. We also introduced several theories to enhance the related research such as formal concept analysis, granular computing and description logics. Meanwhile, we implement the hierarchical semantic categorization of the fundamental geographic information from GB/T13923-2006in China, and discuss the granularity problem of multi-source semantic integration in relation to geospatial information, additionally, the semantic reasoning problems based on Description Logic have been explored.The main contents of this thesis could be summarized as follows:Firstly, we review the main progress of the current geospatial semantic classification, semantic integration and geospatial related applications. It is discovered that some researchers in geospatial semantics domains pay less attention to the standardization of geographic information classification, semantic granularity and geographical concept matching theory based on non-numerical measure, then, the overall research framework and main content in this thesis are presented.Secondly, this thesis reviews Geo-Ontology, Description Logic and OWL, respectively to discuss the formal representation of geospatial semantics, As a top-level geographic ontology framework, Basic Formal Ontology (BFO) is used to build a Chinese administrative division ontology according to the fundamental geographic information in China, at the same time, OWL is used to represent the relevant geographic concepts of the fundamental geographic information.At the same time, due to semantic heterogeneity including overlapping semantic, cross-pattern semantic and the same name of different geographic concepts, some semantic descriptions in the other domains, in relation to the fundamental geographic information, are introduced to eliminate the heterogeneity problem such as overlapping semantic and cross-pattern semantic of the standard categorization from GB/T13923-2006in China. A method is proposed to explore the hierarchical semantic categorization of the fundamental geographic information based on Formal Concept Analysis. First, we extracted the semantic of geographical entity, and construct the fused concept lattice, we then implemented Lattice-tree transforming method of generating the new hierarchical semantic classification of hydrological domain, which has reduced the direct transitive relation of intents in geographic concepts. Finally, the method eliminates the overlapping concept and crossing semantic among the original classification system in order to build a new classification level, which is an extension and improvement for the current classification system by using Formal Concept Analysis.Then, a geographic semantic fusion method based on Granular Theory is presented to deal with the complexities of multi-source geospatial data semantic integration. On one hand, we apply a fuzzy equivalence granular method to divide geographical concepts. On the other hand, an entropy-based weighted concept lattice is proposed as a granular method to reduce geospatial semantics from the point of view of the intent reduction. Experiments are further conducted by combing fundamental geographic information data and spatial data in the hydraulic engineering domain in China. We carry out a series of comparative experiments of the geospatial integration to explore the experiment effects under different threshold conditions. In part, the presented method is a useful attempt for multi-source geospatial semantic integration, which is in line with the scale cognition toward the geographical world, and takes into account the importance of different geographical conceptsAt last, there still exist some problems such as a calculation error to be solved in view of similarity matching theory of traditional research. A description logic-based method is proposed to detect semantic relations between concepts and roles of different geographic objects. Experiments have been conducted to carry out logical reasoning tests by combining fundamental geographic information data from GB/T13923-2006and land use data from GB/T21010-2007in China. The results indicate that the proposed method has solved the semantic conflicts and overlapping problems during the matching process of the concepts and roles, in a way, and possess the logic accuracy for completely rule-based semantic reasoning, furthermore, the method is provided with the general application value for the formal reason rules regardless of the thinking subject of the specific application.This study focus on problems solving involved geospatial semantic classification and integration, which might contribute to geographic information sharing and interoperability. The presented theory framework and technological practice in this thesis, may play an increasingly important role for enriching the theoretical system of geospatial semantics for next-generation spatial databases and geographic information services network. Additionally, some theories and methods in this thesis, as a brand-new perspective, have reduced the complexity of semantic classification and integration of multi-source geospatial data to same certain extent, which is conducive to solving the similar problems for the related domains.