The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
this paper we store the data block by block divided by regular 
regions. The adjacent regions store the common vertexes and 
triangular meshes as virtual vertexes and edges. Every block has 
its serial number. Then we can evoke the data stored in the 
RDBMS through the index of the serial number. These could 
build a more efficient database based on the triangulated 
irregular network model and will be helpful for the integration 
of the multi-scaled DBMS. 
Quaternary Triangular Mesh (QTM for short) is a proper 
strategy to store, index and visualize the data. The surface of the 
earth can be divided and subdivided by the triangular mesh as 
the following Fig (Figure 2 and Figure 3). 
Figure2. The Global Division and Subdivision of the triangular 
mesh 
Figure3. The Hierarchical Subdivision of the triangular mesh 
According to the approach of global division and subdivision 
above, the data of the QTM method can be organized as the 
Figure 4 illustrates. The left figure is when the direction of the 
triangular mesh on the ellipsoidal is up while the right one is the 
opposite. So every block has it own index. 
Figure4. Data organization based on the hierarchical subdivision 
4 INTEGRATION AND UPDATE 
Due to the storage strategy above, the data is stored in the 
DBMS block by block. So when it comes to the visualization, a 
logical seamless database could be formed including the 
geometry information and property information. Because the 
adjacent regions store the common vertexes and triangular 
meshes as virtual vertexes and edges, we can use these virtual 
vertexes and edges with the divide and conquer algorithm to 
finish the integration between the adjacent regions. 
Figure 5 illustrates a case. 
A and B are the TIN-DEM products used to integrate a seamless 
TIN. A1/B8, A7/B1, A2/B9, A6/B4, A3/B10, A4/B11, A5/B12 
are the common vertexes of the two blocks. When the any one 
of them is first chosen, the other one will search the common 
vertexes and edges and find the polygon that consists of these 
common vertexes and edges. In the polygon it will use LOP to 
optimize the triangular meshes. Then the integrated TIN is 
generated. 
Figure5. Integration between two regions 
We can build the pyramid of TIN-DEM automatically by 
selecting the sample points at certain grade, so are the layers of 
existing multi-scale seamless TIN DEMs. Thus, the update of 
the database can be finished in the large scale TIN-DEM. The 
other scale TIN-DEMs could update correspondingly. When 
some new geographical feature points or edges are appended, it 
uses the incremental algorithm to update of the data. 
5 EXPERIMENT AND CONCLUSIONS 
There is an experiment to test the strategy. The proposed 
strategy was implemented in C++ language on Windows XP 
operation system platform. Figure 6 is the data that selected to 
construct the TIN. Figure 7 shows the constructed triangular 
meshes. Figure 8 is the visualization system base on the 
strategy.