relations among contour lines and then to realize raster 
to vector conversion. It is easy to find the whole end 
points and their stepped segments recorded in Freeman 
code. So we define a morphological window rather than 
rectangle window such as in [5] that contains end points 
and is surrounded by connected contour lines and their 
normal lines. Generally, the window contain the matched 
pair of end points and their relationship (direction, 
distance) can be determined easily. Then the matched 
pair of points can be inferred from knowledge. 
As you know, contour lines couldn't cross each other, 
and their elevations are determined. Elevations of 
contour lines on a monotonous slope is increased one by 
one between a pair of summit and sink. So the end 
points appear in pairs in a morphological window; after 
connecting two end points belong to the same contour, 
the window is divided into two parts, each part contain 
other end points in pairs also. After gap connecting, the 
contour elevation is determined no matter from what 
direction to infer. 
4.4 Contour Line Raster To Vector Conversion 
Contour map raster to vector conversion includes 
geomophological points and lines determining, 
monotonous slope division, contour elevation deduction 
and discrete points selection. 
Let X represents a contour map after gap connecting, 
X* is background of X, SK( X^) is skeleton of X“ ‚then 
SK(X°)= X°O{L;} (7) 
delete short arcs of SK( X“), then 
SK, (X°) = SK(X*)O{E,} (8) 
Thus, saddle points set S4 can be determined as 
32 
8 -[Jek,o»eo» (9) 
i=1 
here Qj represents three-cross points set in 8-connect, 
€ refers to dilation operation. Then the area A contain 
summit and sink can be determined as 
A z (SK,QX^)* OG (H3; X (10) 
where O refers to erosion operation; and summit and 
sink points set S2 as 
S, 2 AO{D;} (11) 
where structure elements {Dj} as 
0 9 
0 $O I 
0 0 
i=1,2,...8 rotate T as soon as i increase 1, and 
Qus;sod 
0 m 
0:00 
i=1,2,...8 rotate as soon as i increase 1. 
Then, a monotonous slope can be determined as a area 
between two geomophological points (saddle, summit or 
sink). Provide p €(S,US,), 
W=p®{H}[NW(S US,)) <2] 
W'-WetHyx* 
where N(W(S, US,)) refers to the cross points between 
W and (S,US,), thereafter the contour map is divided 
(12) 
, , , 
into monotonous slopes (W, ,W, ,--.,W, ). 
Extracting one of contour line on a monotonous slope 
relies on geomorphologic points also. Provide A be a 
monotonous slope, geomorphologic point p eA( XS US), 
A, = p@{H};,X° (13) 
then the nearest contour line of slope A from p is 
A, 2 (4) 9 H)f)X (14) 
replace p with (A, UA,) ,repeat 
A, -(AUA)O (I; X* 
A, 7 (A, {HD NX 
until 
(A, UA )0 UD; X* & A (16) 
where “=" represents images in left and right are the 
same. Therefore, the contour is extracted one by one. 
(15) 
Due to terrain rise and fall, the contour elevation 
monotonously changed near saddle. In order to inference 
contour elevation, all of summits and sinks must be 
found; their elevations are deduced from control points 
elevation annotation. As a provided condition, there is 
one and only one saddle between two near summit 
constrainly. The inference method insists of two steps: 
firstly to determine the number of contour lines of a 
monotonous slope between one point of summit or sink 
and one point of saddle, and to set them a sequence in 
the order of increase, that means, to set a sequence for 
all the summits or sinks to infer elevation; secondly to 
infer elevations of contours on a slope from summit or 
sink to saddle, meanwhile this saddle is no longer valid in 
the next inference. 
The vectorlized contour is stored in the form of discrete 
points. The density of discrete points is determined with 
the terrain roughness and interpolation precision[7]. In 
order to describe the rise and fall of the terrain, three 
neighbor points keep up a slope, that is, the distance 
from the middle point to the line connected by the other 
two points should be more than a threshold which relates 
to precision. 
5. DEM CONSTRUCTION AND 
DATA RESTORATION 
The method of generating TIN to represent DEM is 
described as follows: 
5.1. Homotopic Sequential Thinning For The Skeleton. 
Let X eZ, the skeleton S(X) can be described as 
follows: 
SK(X) = X © {L;} (17) 
where "O" represents to thickening operation. 
532 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996 
  
5.2. 
Acc 
skel 
corr 
of p 
diffe 
con: 
5.3. 
Gec 
In ( 
poin 
mea 
poin 
trian 
5.4. 
Ordi 
sides 
repre 
data 
meth 
netw 
poly 
topol 
    
SS 
E 
@