International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
If V (1, J) represents the elevation value of the grid point G (4, J), 
dx and dy are coordinate increments in axis X and axis Y 
respectively from P (x, y) to its down left corner G (I, J), AX = 
dx / d, AY = dy / d (d is the grid size of the GRID), elevation 
value of a stochastic point P can be formulated like this: 
HP (x, y))=(1- 02)*(1- A Y)*V(LJ)- AX*(1)- AY) VQ MJ) O- 
2X AYE WLI yr AX 4 FEV] J) 
As the elévation value of each grid point (e.g. G (4. J)) is 
available in DEM data of GRID type and each random point 
with known coordinate values could be located in a certain grid 
as illustrated in fig.14, elevation computations for each sample 
point could be carried out using above formula, then. 
3.5 Output results 
A delimited format of text file has been outputted to help record 
the output results of the algorithm, as shown in following fig.15, 
  
  
   
  
#7) Sect.tut - IPE ; 2 
ILE) ale) al) HERC d 
1:3008088.61,2596710.93,—2.5,538007.0 — 
29,520027.02,2596550.55,—0.88,53090 
530057 .07 ,2596538.3,0.5,53|0054.23, 
530087 .1,2596536.95,90.76,530084.21 
.5380118.3,25965n9.5,9.19,530115.28 
530158.33,259655^7.25,—1.51,530155. 
20178.36,259655hn.96,—2.16,530175.2 
92908 .36,2596542 .64,-2-16,53602605 .22 
.35,530238.37,2596550.27,—2,530235 
.4h6,520268.37,2596537.88,—2.02,5390 
57,.75,30298.54,2596535.48,—2.08,53029 . 
uf 
ad - 
Fig.15 output meshes in formatted text file 
  
    
  
  
  
  
    
   
  
i zr xi 
  
Each text line has numerically expressed one section along the 
river, which is composed of delimited (X, Y, Z) values of each 
sampling points in sequence, further transformations could be 
carried out on this formatted text file, too, e.g. transformed file 
could directly be imported into TecPlot9.0 and corresponding 
meshes will be displayed vividly in pictures. 
4. EVALUATIONS AND CONCLUSIONS 
£90000 
650000 
600000 
(Uni t:meter) 
330000 
  
  
^ 
2600000 p= 
  
530000 
2590000 p= 
25830000 f= 
2570000 f= 
2560000 f= 
2550000 
Fig.16 output meshes of the whole river trunk (contrasted with Fig.1) 
As a good contrast to fig.1, output meshes of the whole river 
trunk in above fig.16 have been capable of preserving boundary 
characteristics of the river fairly well, so it comes to a 
conclusion that the algorithm is well boundary-fitted as required. 
Following pictures in fig.17 give us a brief look at the output 
meshes of the automatic generation algorithm in some 
representative parts of the river, as have been magnified to 
show it clearly, each picture could only give us part view of the 
whole river in 2D. 
  
  
EER 
A 
  
  
  
  
  
  
  
  
  
  
bare 
  
17 (b) vertical part 
  
  
17(c) at the corner 17 (d) gradient parts 
Fig.17 output meshes of the automatic generation algorithm in 
representative areas of the river 
As can be conclude from fig.17 (a) to (d), structure of the 
output meshes is steady and distribution of all the meshes is 
auto-adaptive to the variation of the river way in 2D, 
unfavourable phenomena such as “overlapping” and “wresting” 
could be eliminated efficiently. 
  
Fig.18 output meshes of a certain part of the river in 3D 
  
Fig.19 original 3D spatial distribution of the same part 
Experiments based on random sampling method have been 
carried out to evaluate the precision of this algorithm especially 
on elevation interpolation; average error of the elevation values 
in centimetres does exist in these sampling experiments, in 
which the spatial resolution of the DEM is 30 meters or even 
larger, however, exceptional error in decimetres seldom 
appears, as illustrated in fig.18 and fig.19 by comparison, 
though these two pictures are local snapshots of different 
software (TecPlot9.0 and ArcGIS 3D Analyst (GIS By 
ESRI"M 2004)), respectively, both the grid size (d = 30 meters) 
and the average interval between those sampling points (60 
meters or so)in each section are large in very number, spatial 
distributions of same part of the river are very much alike as 
456 
  
Internatic 
MC, Hen LO 
displayed 
summariz 
landscape 
volume-f 
characteri 
though it 
spatial re 
amount c 
section, i 
be applie 
analysis : 
higher sp 
each sect 
Here is a 
and the e 
file of the 
bytes) in 
input cen 
with 192 
meters ar 
length of 
sections | 
and statis 
configure 
Tabl 
By contr 
by hand « 
days bef 
Referen« 
Andres / 
models 
TRANSA 
324. 
Chen Gu 
scale im: 
101-102 
Kwon, J 
parallel | 
Image Pi 
Lin Chac 
body-fitt 
vol.12, N 
Prades-N 
images. 
Processi 
Ren Lili: 
extractio 
Water Sc 
Tang Yi 
body-fitt