PRODUCTION OF DSM/DTM IN URBAN AREAS:
ROLE AND INFLUENCE OF 3D VECTORS
C. Baillard
SIRADEL, 3 allée Adolphe Bobiére CS24343, 35043 Rennes, FRANCE
cbaillard@siradel.com
Commission III, WG III/2
KEY WORDS: DEM/DTM, Photogrammetry, Urban, Aerial, Vectors, Matching, Production, Performance
ABSTRACT:
Thanks to the evolution of new sensors and the development of efficient algorithms, the automatic production of DSM (Digital
Surface Models) and DTM (Digital Terrain Models) in urban areas has become possible. At SIRADEL, we have been interested ina
new approach using external 3D vectors and a pair of aerial stereo images. In this paper, our method for computing DSM and DTM
is briefly presented. Then this approach is compared to the traditional manual process in term of production cost and accuracy. In
particular, the role and the influence of 3D vector data is quantitatively assessed and analysed.
1. INTRODUCTION
1.1 Context
The use of 3D cartographic data has become very important for
many applications related to urban areas: telecommunication,
urbanism, estate agencies, communication, transport, tourism,
air analysis, flooding risk, etc. There is an increasing need for
accurate, realistic and affordable 3D digital data over cities. In
particular, the availability of urban DTM (Digital Terrain
Models) and DSM (Digital Surface Models) is a major concern
for most users.
Thanks to the recent evolution of new sensors (digital camera,
LIDAR data, high resolution satellites) and the development of
efficient algorithms, the automatic production of urban DSMs
including buildings and trees is now possible (Cord ef al., 1999;
Maas, 2001; Paparoditis and Maillet, 2001; Roux and Maitre,
2001; Fraser et al., 2002; Zinger er al, 2002). However the
automatic computation of 3D vector data is much more difficult
because of low-contrasted building contours, hidden areas and
complex-shaped buildings (Jung and Paparoditis, 2003). A
solution consists of using external 2D vector information
(Jibrini et al., 2002; Vosselman and Suveg, 2001). For most
applications requiring high quality vector data, a manual or a
semi-automatic intervention is necessary (Brenner, 2001;
Flamanc, 2003).
1.2 Industrial production: manual vs semi-automatic
This study focuses on the industrial production of urban
cartographic data. SIRADEL is a company that has produced
digital 3D data over a large number of European cities. These
data include reliable and accurate 3D vectors (buildings, water,
vegetation, road network, terrain breaklines, etc) and raster data
(DSM, DTM, land-use). Although initially dedicated to radio
planning applications, they can be used within many sectors.
The accuracy is below 1 meter, and the detail level is more than
sufficient for most applications.
Two different production lines coexist at SIRADEL. The first
one is based on manual photogrammetric capture. The benefits
of this method are the accuracy and the reliability of the
produced vectors, along with the availability and the low cost of
the source imagery. However it implies a high production cost
due to capture time. An alternative semi-automatic approach for
computing DSM and DTM has recently been studied, which
only requires a small set of 3D vectors. The input imagery is the
same one as in the manual approach: acquisition costs do not
increase and off-the-shelf image data can be used.
This paper presents a comparison of accuracy and production
cost related to both processes.
2. MANUAL PROCESS
2.1 Principle
A detailed set of 3D vectors is manually produced by
photogrammetric capture from stereo imagery. Raster data
(DTM, DSM) are then entirely derived from the captured
vectors (see Figure 1).
Vector capture
Planimetry vectors
Altimetry vectors
Y
Raster
Figure 1: Manual process
2.2 Source imagery
Aerial imagery is available at a scale between 1:15000 and
1:25000 (focal length 157mm), scanned at 14pm. The pixel size
is between 21 and 35cm. The stereo pairs have an overlap of
60% intra-band and 20% inter-band. The orientation of the
images is performed with an analytical stereo plotter. Examples
are given in Figures 8a, 8b and 8c.
International Archi
JET EIRE FE EUE
2.3 Photogramme
Vectors can be of 3
e Planimetr
buildings,
e Altimetry
describing
characteri
when ele
at intersec
These 3D vectors
depending on the ty
are captured at the
and the minimal he
is divided into inde
between them exce
superstructures, as
Each polygon is ass
roof point (roof r
analysis guarantees
topology (closed c
connected adjacent
Figure 2: E
24 Raster data cc
The DTM is compu
including roads, ra
aboveground elevat
the DTM produces
single elevation val
3. SE
3.1 Principle
e
E.
Planimetry
Altimetry
Roads ana
[s
Figu