Multi-scale Morphology-preserved TIN Surface Modeling And Visualization For Virtual Globe

With the development of the earth observation technique, the storage, the representation and the analysis of massive global geospatial data have become an issue attracting wide concern in geographic information science. It is the trend to seamlessly integrate and represent multi-source heterogeneous geospatial data with virtual globes for their advantages comparing with traditional GIS (geographic information system). The capability in global geospatial information management and display made the virtual globes have strong potential in analyzing geographic problems and addressing the global environmental change issues. However, compared with the rapid development of the hardware and software systems, the accuracy problems in data processing and representation were rarely concerned in virtual globes. These problems made the power of the current virtual globes was mostly restricted to functions as a "three-dimensional (3D) geobrowser", which cannot be applied to further analysis.The multi-resolution modeling and visualization of global terrain data is a key component of rendering planet in virtual globes, which formed the basis for integrating and representing other geospatial information. For the multi-resolution terrain modeling, the traditional virtual globes usually use the simple interpolation methods to generate the regular square grid (RSG) based terrain pyramid. However, the obtained multi-resolution terrain always has accuracy uncertainty, and the surface morphology is distorted as a result of the loss of significant terrain parameters, which cannot be used for analysis purpose. For the visulization, it is a general way for virtual globes using Grid based geometry for global terrain data due to its easy to implement. However, the grid geometry is limited to contain points aligned to the grid only:break-lines and spot heights are not supported since they are generally not grid-aligned. In addition, the grid cannot represent detailed morphological features with a complex topological structure, and the unified grids also cause expensive redundancy in flat terrains. In light of these, the main works and contributions of this paper were presented as follows:(1) The core techniques for the generation of multi-resolution global terrain data and its visualization were reviewed. Based on these, the Discrete Global Grid Systems(DGGSs) and the spatial data model for organizing, indexing and visualizing the global multi-resolution TIN-based terrain data are discussed. Furthermore, the limitations of previous multi-resolution surface modeling and visualization methods are detailed.(2) The accurate measurement of terrain features is a key component in controlling the quality of multi-scale surface modeling. However, the existing methods used for terrain feature extraction often suffer from high sensitivity to terrain noise, which always propagate topological and semantic errors in multi-scale surface modeling. In view of this, a robust landform feature extraction method combining discrete curve evolution and skeleton construction technique is presented. The method can achieve highly accurate results and is robust. Moreover, the method can use various types of data as input, including the Grid DEMs, contours, and the LiDAR points.(3) The multi-scale data generated by the current virtual globes with off-line pyramid tended to lose significant features, and have low accuracy. As a result, the following analysis was unreliable. By introducing accuracy standards and cartography requirements, we proposed a global multi-scale TIN construction method considering topographic features. To solve the problem brought by the store reference of large-scale TIN data, we employed pyramid data model for data organization. Furthermore, an efficient virtual node (VN) method was used for data structure. Then, multi-scale TIN data can be subdivided seamlessly, and stored in the hierarchical database. The multi-scale TIN database was built under the guidance of accuracy standards and cartography requirements, with eligible accuracy for further analysis.(4) The Grid-based mode for global terrain rendering has the problem of low representation accuracy, the fixed structure, and unable to render or seamlessly integration of the irregular distributed geospatial data. To address these problems, we proposed a TIN-based mode for rendering global terrain data. We use the preprocessed offline TIN terrain pyramid to support the schedule and real-time rendering of out of core of multi-resolution TIN-based terrain data. According to the rules made by the processing step, we can use the keys of different data structures to request the corresponding data. With an organization of these data, the terrain meshes can then be generated for real-time rendering. To eliminate the cracks existed between different LODs, we proposed an real-time decoding method to stich the boundary triangles of TIN tiles with different resolutions.(5) In the virtual globe system, we have implemented the multi-scale modelling, pyramid organization and visual scence scheduling of LiDAR points cloud employing the proposed algorithm. Moreover, it is effective to use TIN structure for multi-scale coastal landforms rendering. It can reduce the data redundacy, while digitally express the tattered landforms, such as coastlines and reefs, with high preservation of the significant features. In addition, we implemented the volume modeling based on tetrahedral mesh and octree organization. On this basis, we discussed the representation and integration of geological data in the Virtual Globe.