| Among the data resources, network resources and computer resources, data always act as the most foundational and stirring element. The initialization of spatial data is the preparation of data for the requirement of the spatial analysis and according algorithms. Since spatial data have the features of complexity, huge quantity and dynamic, together with the dynamic of geographical processes, following difficulties are met with performing the initialization of spatial data based on vector data: some spatial data being difficult to express, the quantity of data being huge, not meeting the requirement of dynamic initialization and not fit for the need of continuous dynamic geographical analysis. These difficulties result in the complexity in spatial data expressing and the huge cost of storage and maintenance, also the conditions of data not prepared while needed, prepared but nor used and prepared but will not be used. All these will influence the constructure and developing of GIS.Spatial analysis has the functions of deeply analyzing the GIS data, predicting and managing the dynamic recesses. It is the distinct difference from any other MIS or IS. Spatial analysis is an important part of geographical analysis, however, some basic issues of it still need discussion: the integrality of spatial data conceptions, the quantity of spatial data, the basic spatial measurement of GIS, etc. They bring the bottleneck that blocking the developing of GIS. Based on above discussions, this dissertation will illustrate and fulfill the zero- initialization of spatial data with raster data structure, and it will perform network analysis with raster structure, including element abstraction, modeling, expressing, and will come up with the algorithms of common network analysis. The main aspects of this dissertation are listed as fellow:1. Research on the zero- initialization of GIS spatial data. It discusses the basic features and the models of spatial data. Beyond current topology and distance relations, it comes up with the location, neighborhood, nearness and influence conceptions of spatial data and consummates the descriptions of spatial relations. By analyzing the algorithms for creating derived data, it can be seen that the kinds and quantity together with the relations between them can be quite complex and substantive, there will be different ways of organizing and expressing of them according to different objectives, purposes and scales. The preparation of derived data will be vast in computing and storage, so it can not meet the need of dynamic geographical processes. Based on the computability of derived data,the feature of raster data that the whole research area is been structured with regular cells and related theory and technology in map algebra, the dissertation comes up with the idea of GIS spatial zero-initialization: while building a GIS, we may only initialize the primitive data and derive other data while need with map algebra algorithms. This will avoid above bottleneck and put GIS forward. The conception and basement of zero-initialization will be discussed in detail and the according raster algorithms operators will also be shown. Then it shows the need, fulfillment, possibility and significance of performing spatial data zero-initialization.2. The models and basic algorithms in network analysis. The traditional network models are based on graph theory. In this part, the dissertation systemically analyses the features of geographical networks, comparing related conceptions in graph theory, abstracts the basic elements and extended elements, consummating the elements abstraction and description of attributes. The net models based on graph theory are introduced. From the view of conceptual model, logical model and physical model, it discusses the base of modeling network. On the telling the difference of graph theory network and geographical network, it shows the limitation of modeling geographical network with graph theory model. There also difference between common GIS data model and geographical network model. It also introduces the extended network model based on object-relation database and features-based network model. It then casts some light on the relations between network analysis and the multi-representation of network. Common kinds of network analyses are discussed, including their content and fulfill field. It summaries the development shortest path algorithms by classifying them into problem kinds, network features and fulfillment technology. Lastly, this dissertation summarizes the development of shortest path algorithms in transport network.3. The basic issues in raster network analysis. Network analysis is considered to be done with vector data. On concerning the spatial data initialization and the dynamic features of the network, the raster based network analysis finds its way. In this part, the dissertation shows the difference between raster methods and vector methods, especially the benefits of using raster data. It discusses the creation of a cost surface and the searching strategy of the shortest path, as well as the topological operators, topological representation distance operators, algorithms of accumulated distance and basic models for representing 3 dimensional network. Then, it summarizes the deficiency of current network analysis in representing of network elements, the deficiency of network modeling, the elevation ofalgorithms and the content of network analysis. Based on the idea of zero- initialization, it brings the methods of spatial data representation, network modeling with raster data and organizing of geometric and attribute data. Map algebra is the base of raster map computation. The dissertation also illustrates the Tomlin map algebra and Hupeng map algebra, as well as the academic and technical function of them in raster network analysis.4. The algorithms of raster network analysis. It first discusses the conventional data requirement, then it shows the basic algorithms model - fire model. In this part of the dissertation, it comes up with the algorithms of connectivity analysis, path analysis, flow analysis, location-allocation analysis, and buffering based on network distance. The algorithms of path analysis includes shortest path analysis, optimal path analysis, Minimum Spanning Tree analysis as well as ESPO analysis. With the experimental transport data of Hangzhou city, the author verify these algorithms.5. Discusses on topics related to raster network analysis. It discusses the requirements of image resolution for analysis, including precision concerned and representation concerned requirements. Considering the influence of planar enforcement, it gives the models to model a non-planar network in the planar surface of raster GIS, gives the extended 8-neighbor-direnction model to represent connections directions. Based on the analysis of dynamic network models of continuously updated model and time interval model, the author proposes a trigger model to achieve better performance. Then the dissertation discusses the precision and efficiency of raster network analysis and feasibility of raster and vector network analysis. The spatial measurement is an important element in spatial analysis; the dissertation also discusses the importance of it and the relations of basic measurement and composed measurement, measurement and metadata standards and the selecting of measurement according to the size of researching area.6. Summarizes the work in this dissertation and gives some advises for further research which might improve the development and application of raster spatial analysis.The main creative work of the dissertation can be described as follow:1. The author tells the influence of spatial data initialization, shows the necessary of zero- initialization. Then he discuses the conception, methods and models of zero-initialization. He shows the relation of map algebra and zero- initialization and the strategy, feasibility and significance of spatial data zero- initialization.2. Consummates the abstraction of network elements. Besides the geometry, attribute and auxiliary elements, there still another important bur often neglected element, the restriction rules. The author discusses the abstraction of restriction rules, analyzes its functioning style and fulfills its function in algorithms. He also consummate the content of the elements' attribute.3. Consummates the representation of network with raster data. The author analysis the objections of using incremental linkage and incremental length in raster network analysis, gives the models based on zero-initialization, and proposes the extended 8-neighbor-direction model.4. Consummates the network model. The author analysis the objections of using graph theory for network modeling and the different requirement of common GIS data and geographical network data, then shows the special need for geographical network modeling. He proposes the model to model a non-planar network in the planar surface of raster and leveled network model.5. Gives the raster algorithms for common network analysis. Based on fire model, the author performs raster network analysis with the experimental transport data of Hangzhou city.5. Proposes a dynamic network analysis model, trigger model.6. Discusses the influence of spatial measurement to spatial analysis.
|
PDF Full Text Request