An Object-oriented Model For Map Data Multiple Representation Over Scale Space

The issue of multi-scale representations of spatial data is a hot research topic in GIS. Multi-scale representation database (MRDB) contributes to many GIS related fields, such as multi-scale spatial analysis, vector data progressive transmission over internet, data integration, and self-adaptive dynamic visualization and so on. At present, there are two strategies to construct a MRDB:static multiple versions and dynamic deriving. The former has some drawbacks, such as data redundancy, inconsistency and outdated; the latter seriously depends on the development of map generalization, which is a difficult problem. To remedy these shortages we present an object-oriented multi-scale spatial data model "lifespan model" from the viewpoint of data organization, which integrates static versions and dynamic operations together. The followings are the specific research contents of this study.1. The thesis gives a review on the research progress in multi-scale data model. Several representative research works are analyzed. They are GEODYSSEY, Map Cube Model, Stratified Map Spaces, Abstract Cell Complexes, MADS and VUEL. All of these models can be categorized into two classes:the hierarchy model (eg:the Map Cube Model and the Stratified Map Spaces Model) and the generalization deriving model (eg:GEODYSSEY、MADS. VUEL). Generally speaking, all these models either focus on the hierarchy or on the generalization. No one considers both of the hierarchy and generalization together. In fact, both of the representations and the scale transformations can be integrated under the framework of object-oriented. Meanwhile all of these models usually focus on the level of concept model or logical model or physical model. No one relates all of the three models and integrates them to form a system. Hence we think the future direction of multi-scale spatial data model is to integrate representations and scale transformations together, and join concept model, logical model and physical model.2. Scale is an important characteristic of spatial data. In this study, we discuss the concept of scale from three aspects:connotation, extension and category. From viewpoint of the cognition, we summarize the scale characteristics of spatial data including:scale effect, scale dependence, and scale invariance, which are the theory basis of scale transformation and multi-scale representation. From the viewpoint of representation, the multi-scale representations are not only the results of hierarchical spatial cognition and scale transformation, but also the tools to assist spatial cognition from coarse to fine. 3. From the viewpoint of object-oriented (00), we present a multi-scale data model "Lifespan Data Model", which integrates the representation and scale transformation together. According to the idea of 00, we model the representation as the attribute of object, and the scale transformation as the method of object respectively. Based on this model, we can dynamically derive any representation from the scaling of basic representation. Then the polymorphism of spatial data can be better represented and the problem of inconsistency and update can be avoided.4. From the viewpoint of data manipulation, we design a suit of scale transformation modes, including map generalization transformation, LOD transformation, morphing transformation and equal transformation. The basic idea is LOD transformation is to partition the full representation into series of details with different scale hierarchies, and then the representation at any scale can be accumulated with the details of corresponding levels. This mode has some favorable characteristics, such as little data volume, wide scale span, operator integration and so on. The morphing transformation is based on the idea of shape interpolation. It can be used to drive smooth and consecutive representations. Within the framework of 00, all of these modes are integrated by the technologies of aggregation and inheritance. Consequently, the efficiency and flexibility of the lifespan model are greatly improved.5. From the viewpoint of data organization, we present a graph based multi-scale data structure "evolution graph". For the evolution graph, the vertex denotes one or sets of representation, and the edge denotes the scale transformation operation relating two consecutive representations. Based on the properties of representations, we summarize four types of vertex:solid vertex, fake vertex, detail vertex and complex vertex. Based on the properties of scale transformation we summarize four types of edges:generalization edge, LOD edge, morphing edge and equal edge. The graph with different vertexes and edges can denote different scale transformation modes. A series of ordered vertexes and edges directly reflect the complex scale evolution process of spatial data over a large scale space. Finally the evolution graph extends the traditional scale point oriented static representations to the scale space oriented dynamic representations.6. Based on the Domap software platform, we develop a lot of experiments to verify the feasibility of the model and algorithms.