Abstract
this paper puts forward several optimizing methods in building Triangulated Irregular network in traditional ways according to the questions appeared in actual projects. These methods can improve the efficiency of application based on Digital Terrain Model.

Terrain information is different from land use, type of soil, geological unit and so on. So commonly it is explained as a continuously varying surface which cannot be simulated approximately by a choropleth map. So in order to implement terrain analysis, it is an efficient way that simulates the configuration of the earth's surface with a group of digital data representing locations on the land surface - Digital Terrain Model (DTM).

DTM can be used for all kinds of route designing, engineering evaluation and production of sectional profiles, etc. it has brought great benefits in many applications such as analysis of land use, planning and management, disaster prediction and the like.

There are many ways to build DTM. Triangulated Irregular Network (TIN) is the simplest and the most applicable one among methods directly using dispersing data.

I. The necessity of fast building TIN
As the functions mentioned above, building DTM with original observations can be applied to surveying-engineering or other fields that can be used to implement a series of applications such as interpolating contours, calculating cut and fill, describing profiles, etc. Building TIN is the foundation and key of this kind of method, its quality and efficiency can affect subsequent works. It takes long time in writing and running programs because of the big amount of computation in building TIN and complex data structure.

In addition, the more advanced the means of obtaining data it uses, the more numbers of data it collects. As a result, the speed of building. TIN will get shower and slower. Optimization of a building TIN is the urgent affair so as to improve the program efficiency.

II. The principle of building TIN
On concerning of mathematics, what we will do is searching for an extending point C, which makes the vertex angle c's degree to be the maximum of all values larger than the threshold. Point C is generated from side AB to become a extending point. At the same time the point C and the point M must be located to different sides of AB (see figure 1). How to find the extending point rapidly becomes the key process, it will spend almost 90 percent of time in building TIN. We may divide data into some rectangle blocks in order to optimize building TIN.


Figure 1
III. The data blocking in building TIN
Data blocking is the most efficient way to shorten the time of building TIN. So it can be adopted widely. As we usually know, the original observation data are disorderly, and the distribution of data is irregular. But when searching for an extending point, the final result is only related to the surrounding data points. In order to find the needed data points from the abundant original data, we must divide these original data into some blocks. There're many ways of data blocking, and different data structures, so we should be use different data blocking ways. In general there are three ways of data blocking as the following:

1. data blocking with overlapping
   to divide the whole area into several blocks with overlapping parts could be satisfy the continuity of the data (see figure 2). This method is suitable to meet the conditions of building network. When we build TIN on different blocks, the relationship and data structure are very complex because there're overlapping areas within each other. So it is hard to complete building TIN. Moreover, the size of the block and the overlapped area are not easy to define, thus we need to do some adjustment base on data distribution mode (sparse or dense). Under particular conditions, some TINs could be missed. If the data distribution is uniform, the efficiency of building network will be higher. So this method is adopted in some cases.
   
   
   Figure 2
   
2. the method of non-overlap single area management.
   To divide the whole area into some little areas which are neighboring and non-overlap (see figure 3).
   
   
   Figure 3
   When we use this method to build net, there're no need to consider the effects of the surrounding area. We should record the outmost net border when we build the net of each area. After all blocks are built, it connects the borders of neighboring areas, so we get the TIN of the whole area. Because we deal with each area separately when building the net, the data structure is very simple, and the efficiency of building net is also very high, so it's relatively easy to handle. But because of lacking consideration of the surrounding areas when building the net, it makes the TIN at neighboring areas is not the best one (doesn't build the net based on the biggest angle principle) thus some long and narrow triangles will appear (See figure 4). If there's no need for high quality but for high efficiency, we can use this method. But on the other side, we should adopt blocks with suitable size, or may miss some TINs or create too many long and narrow triangles.
   
   
   Figure 4
   
3. the method of non-overlap multi-area disposal According to some problems of the former tow kinds of methods, we put forward to the method of non-overlap multi-area disposal. Because there is not overlap among areas, the data structure is simple and it is convenient when we use this method t build net. Multi-area disposal method avoids the appearance of the long and narrow triangles in neighboring areas, and has little effect by the size of blocks. It gets a better combination between the complexity of data structure and the calculating quantity, so it's been widely used. Now we will describe this method of building net.