ine- 
ible 
our 
urs 
ing 
tain 
and 
ther 
vere 
ited 
spot 
eful 
ross 
tour 
ight 
and 
data 
  
tours 
| are 
  
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
terrain modeling, verification and editing. The software that 
has met all of the above requirements was Surflng. This 
software is part of the MapSoft GIS software. MapSoft is GIS 
software with extensive support to large scale mapping, 
specifically tailored for handling cadastral and topographic 
surveying maps. Various surveying data acquisition techniques 
are supported, as well as all the spatial data analysis functions 
required by surveyors. Software keeps the data within standard 
RDBMS using geo-relational data modeling approach, so the 
project size is practically unlimited. All the software was 
developed by the researchers from the Institute of geodesy, 
Faculty of Civil Engineering, University of Belgrade. It was 
planned to use MapSoft as a basic environment for the later 
development of special software for management with final 
country wide DTM data. 
Surfing is a complete solution for DTM data processing and 
analysis. It can perfectly operate autonomously or as a MapSoft 
module with high level of integration with other modules. As 
all the other MapSoft's modules it is developed by using 
Visual C++ and other advanced software development tools. It 
is a standard Windows application with rich and intuitive user 
interface and fast graphics. 
  
Figure 3. Surfing application for DTM analysis 
DTM modeling provided by Surflng is based on TIN 
(Triangulated Irregular Network). Constrained (optionally 
conforming) Delaunay triangulation is used as a basis for 
terrain surface reconstruction. Breaklines and other terrain 
form lines are fully supported. Terrain modeling is 
accomplished by using standard linear interpolation over 
triangles and also by using bicubic surface patches over 
triangles. The second method provides much better results. All 
standard DTM analysis functions are provided: height 
interpolation, profile and crossection interpolation, contour 
interpolation, volume calculations, 3D terrain visualization, 
data conversions, etc. 
\ 
5. DATA VERIFICATION AND ERROR DETECTION 
Considering large amount of data and lot of work done by 
operators during different stages of initial data capture and 
data digitization it is normal to expect errors to appear. At this 
stage of project realization, special attention is dedicated to the 
problem of gross errors detection and elimination. Some 
procedures are designed specially for these purposes. All of 
653 
them were 
environment. 
implemented using Surflng DTM software 
Firstly, considering that contour interval was already known — 
it is less than 10m, it is possible to mark all the triangles with 
height difference larger than some specified value (usually 
10m). Such triangles are indicating possible errors in contour 
heights. Some of these triangles are the consequence of 
missing contours. The reason could be the situation similar to 
the one illustrated on Figure 1. Each of these cases should be 
carefully examined and corrected. 
Also, all the triangles where the difference between triangle 
face and bicubic surface patch is larger then some specified 
threshold are also marked. This usually indicates that there is 
no enough height data that could provide accurate terrain 
surface reconstruction, regardless of method of interpolation. 
Majority of these triangles belong to areas where horizontal 
triangles are located, i.e. triangles with all three vertices 
belonging to the same contour, or to contours with the same 
height. These areas are not treated at this stage, since they are 
going to be subject of the routines for automatic terrain form 
lines detection. Other cases are examined and corrected if 
needed. 
Another method for error detection is interpolation of heights 
on points with known heights. Such points are available on 
black map layer in a form of spot heights. These points are 
digitized manually. It is possible to use these points as control 
points for DTM, providing that they were not included in 
process of building DTM. Considerable errors were detected 
using this method. Most of them were related to errors during 
initial data capture and map production. 
Visual DTM data verification is also used during this stage. It 
is done by comparing interpolated contours and digitized ones, 
and also by 3D visualization of DTM. Surflng supports second 
option by rendering large DTM data sets in near real-time. 
This is accomplished by using oct-tree DTM data (triangles) 
indexing and frustum view culling for fast selection of triangles 
that has to be rendered. 
All of these data verification options are accompanied by 
simple correction of detected errors. Data editing is done 
directly on data that are kept within MapSoft’s database. 
6. TERRAIN SURFACE RECONSTRUCTION 
Procedures described in the previous chapter were primarily 
designed with the purpose of detecting and correcting large, 
i.e. obvious errors. Definite verification and quality assessment 
of acquired DTM is possible only after final terrain surface is 
reconstructed. A lot of methods have been developed for 
generating high quality DTM. Some of these methods are 
general-purpose (linear prediction, finite element method) and 
some are specially designed or modified to use contour data in 
the best way. It is widely accepted that the best results are 
achieved by using linear prediction since it is based on 
sophisticated statistical analysis of data. However, even for this 
method, there are certain issues that have to be taken care of 
when contour data are used (Heitzinger, 2001).