h the 
them 
las to 
nised 
inted, 
y for 
Within 
| only 
when 
Ss and 
his is 
large 
n-line 
ar the 
to be 
e the 
in the 
> final 
| and 
whole 
veral 
irface 
base 
|, this 
juires 
ology 
pe a 
esign 
vill be 
tions 
ns SO 
ration 
| are 
maps 
data 
f the 
and 
the 
s will 
| are 
were 
se of 
e of 
as to 
itions 
Selective sampling is supported with functions for 
measurement of randomly located spot heights and 
characteristic terrain lines. 
Systematic sampling can be done using profile or 
grid measurements. Once profile/grid parameters are 
set, upon each redraw of actual data window, theoretical 
locations for profile/grid points are marked on the 
screen. Of course, systematic sampling can be 
performed also by proper movements of stereoplotter 
handwheels to accomplish nearly regular distance 
between points during grid measurement. 
Progressive sampling is based on analysis of TIN 
DTM generated from the points measured. This concept 
has already been presented in several papers (Bill, 
1986; Mann, 1988; Reinhardt, 1988). After the TIN is 
created and the surface calculated, curvature of each 
side if triangle is tested, and if it exceeds the specified 
limit (based on the required height accuracy) additional 
measurement is suggested. Locations of these 
measurements are marked on the screen until 
densification step is finished. 
Contour measurement is supported with several 
criteria for automatic on-line selection of contour points 
during dynamic contour digitizaton. These are: distance, 
time, and curvature (tube) criteria. Combination of these 
criteria is also possible. 
Standard functions for interactive data manipulations 
and editing are developed: browse, delete, undelete and 
attribute changes. All operations are supported by on- 
line help and designed in a same manner as the rest of 
the MapSoft functions in order to make training and 
work as easy as possible. Several workstations 
connected in a computer network can simultaneously 
use and update data stored within the common data 
base. 
4.3. DTM creation 
Irregularly distributed height points and other height 
information are dominant in case of data acquisition on 
analogue stereoplotter which cannot drive measuring 
mark on the specified position. DTM modelling based on 
TIN (Triangulated Irregular Network) has therefore been 
chosen, since it provides very efficient and simple 
processing of such data. 
Triangulation: Several algorithms for the TIN 
generation have been implemented and tested, each of 
which respects terrain lines. These are: continuous 
development of the TIN (McCullagh, 1980), radial sweep 
algorithm (Mirante, 1982) and point insertion algorithm 
(Reinhardt, 1988). Each of these algorithms result in 
Delaunay triangulation with well known properties, 
constrained by terrain lines. The best results are 
achieved with point insertion algorithm and it will be 
briefly described. 
Initial triangulation consisting of four imaginary vertices 
located outside active DTM area is created. After that 
terrain lines are processed sequentally. If the line length 
exceeds specified value, then the line is divided. The 
heights for the new points are interpolated by linear 
interpolation from the end line points. Each line segment 
is subsequently processed as a separate line. For each 
line end points are first inserted into the TIN. Insertion of 
a point is done by finding the existing TIN triangle that 
the point falls in. The triangle is after that divided into 
three new triangles. This is followed by local TIN 
optimization which is performed until Delaunay criterion 
is satisfied (Sibson, 1978). When the both line points are 
inserted, triangulation is again modified to ensure that 
the line forms edge in triangulation. Whenever possible, 
swapping of the alternative diagonals for the 
quadrilateral consisted of the two TIN triangles with the 
common edge is performed. Otherwise, new point is 
inserted into the TIN and its height is interpolated using 
linear interpolation from the end points of the line. 
After processing of terrain lines, single height points are 
inserted into triangulation using the same algorithm. 
Using this algorithm it is also possible to obtain the TIN 
with the minimum sum of all distances. To achieve this, 
only minor modification of the swapping algorithm for 
local triangle optimization is necessary. Spatial triangle 
sorting for faster manipulations is perfomed after 
triangulation is finally completed. 
Surface creation is done by finite elements method. 
Three methods for interpolation and terrain surface 
presentation are supported: linear, third order and fifth 
order polynomial interpolation. The simplest and fastest 
solution is linear interpolation. In that case terrain 
representation is done with triangular facets. The 
calculated surface is continuous but not smooth. This is 
primarily used for initial checks of the data. 
For high quality terrain modelling triangular surface 
patches are represented by high order polynomials. 
Third or fifth order polynomials can be used optionally. 
In each TIN node, partial derivatives are estimated using 
information on neighbouring nodes (Akima, 1974, 1978). 
If fifth order polynomial is chosen, then first and second 
derivatives are calculated. This yields five derivatives for 
each TIN node. In case of third order polynomial, only 
three values for first derivatives are calculated. Where 
breakline exists, two or more sets of derivatives are 
calculated for each node. 
Terms of the polynomial are calculated from the heights 
and derivatives of the three triangle points (Barnhil, 
1981; Sekulovié, 1984). The result after calculation and 
connection of all triangular polynomial patches is a 
continuous and smooth terrain surface. The only 
exception is across breaklines where smoothing is 
intentionally avoided. 
209 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996