2-3-2 INTERPOLATION WITH POINT BY POINT 
It is necessary to select new reference points 
while a point is inserted by this method. The 
principle of this method is shown in Figure 4. 
At the first, we define the effective area for 
DTM sampling that means what area will be 
sampled or not. For deciding the interval of 
interpolation, all intersections of each scan 
line and defined area should be determined, 
before interpolation is processing . As 
mentioned above, there are N reference points 
to be selected in neighborhood of sampling 
point. Then suitable interpolation model is 
used according to accuracy criterion, e.g. 
weighted moving average method, bilinear 
interpolation, two-order polynomials and linear 
least squares prediction etc. Although the 
result of surface modelling is more feasible, 
description of terrain features is still worse 
due to grid DTM. And it is time-consuming 
especially as the number of reference points is 
increasing. 
  
  
  
  
  
Figure 4 The principle of interpolation with 
point by point 
2-4 CASE STUDY 
The case study is based on a topographic map 
used for the design of a golf court. The: shape 
of study area is irregular. There are total 
514 random points in the area which is gained 
by digitizing contour lines. The Delauney 
triangulated network for a set of 514 random 
points is shown in Figure 5. Figure 6 is the 
original contour map of the study area. The 
polygon of the area is constituted with 166 
vertices. Then a 65x65 grid DTM is generated 
by the methods mentioned above, i.e. TIN and 
point by point methods. Heights outside the 
study area are given by negative values and are 
not processed. Fig. 7 shows the contour map 
which is the result of TIN method and Fig. 8 
shows the result of point by point method with 
least squares prediction. 
  
Figure 5 The Delaunay triangulated network for 
a set of 514 points 
904 
Figure 6 
  
Figure 7 The contour map obtained from the 
result of TIN 
  
Figure 8 The contour map obtained from the 
result of point by point with least 
squares prediction method 
III. EXTRACTION OF TERRAIN FEATURES FROM DTM 
The extraction of terrain features from DTM can 
be done by either raster or vector methods. In 
fact, extracting terrain features and display 
the result by raster method is simpler than 
vector method. ' Moreover, raster display has 
better effect in landscape visualization. 
Slope gradient and aspect in each grid can be 
computated with DTM data[13). Feature lines of 
terrain, such as ridges and valleys, are 
the important factors of describing the 
characteristics of terrain. They are helpful 
for interpretation of streams and watersheds 
and can be used as reference features for image 
matching [ 16, 3, 6 JJ. For example, a 3x3 
operating mask is used to decide ridges. If 
the central pixel P contents with following 
three criterions, then pixel P is a ridge: 
1. Height of pixel P is not the lowest when 
it compare to heights of neighboring 8 
points, see Figure 9. 
^ rrt Oui OT OM ocr ee 0D 4 H 
Au rub lale — Baa ial ah Sede Du —