FROM MANUAL TO AUTOMATIC STEREOPLOTTING :
EVALUATION OF DIFFERENT ROAD NETWORK CAPTURE PROCESSES
Sylvain AIRAULT*, Olivier JAMET*, Frédéric LEYMARIE**
* Laboratoire MATIS
Institut Géographique National
2, avenue Pasteur
F-94160 Saint-Mandé(FRANCE)
E-mail : airault,jamet 9 matis.ign.fr
** Département S2IG2
SYSECA
66, avenue Pierre Brossolette
F-92247 Malakoff (FRANCE)
E-mail : frederic.leymarie 9 syseca.thomson.fr
XVIII ISPRS Congress
Commission III
KEYWORDS : digital photogrammetry, object recognition, aerial image, road detection, user interface.
ABSTRACT :
Within the framework of the French topographic database capture, IGN is looking for solutions related to the photogrammetric data
capture automation with the double goal to reduce the production costs and to reduce the data availability delay. Since digital
stereoplotters have been introduced in the production lines, it is now possible to integrate automatic or semi-automatic tools in the
traditional data capture process. One of the most advanced project in the MATIS laboratory concerns the road network automatic
extraction. We are working at the same time on very ambitious approaches, looking for a fully-automatic interpretation of the image
and on semi-automatic approaches, looking for a short term efficiency. Corresponding to this second axis, this paper concerns the
different semi-automatic solutions which can be built around a road following algorithm to increase the efficiency of such kind of
approaches. It presents also what means will be used to evaluate the contribution of these different solutions in a production context.
RÉSUMÉ :
Dans le contexte de la saisie de la Base de données topographiques, l’IGN s’intéresse à des solutions concernant l’automatisation de
la saisie photogrammétrique avec le double but de reduire à la fois les coûts de production et le délai de disponibilité des données sur
tout le territoire. Dans la mesure où des appareils de restitution numériques ont été introduits dans les chaînes de production, il est
désormais possible d’intégrer des outils automatiques ou semi-automatiques dans les processus traditionnels de saisie. L’un des
projet les plus avancés du laboratoire MATIS concerne l’extraction automatique du réseau routier. Nous travaillons parallèlement sur
des approches très ambitieuses, visant une interprétation complètement automatique de l’image, et sur des approches semi-
automatiques, visant une opérationnalité à court terme. Correspondant à ce deuxième axe, ce papier présente différentes solutions
semi-automatiques qui peuvent être construites autour d’un algorithme de suivi de routes pour tenter d’augmenter l’efficacité de ce
type d'approches. Il présente également quels moyens seront mis en oeuvre pour évaluer l'apport de ces solutions dans un contexte
opérationnel.
1. INTRODUCTION
Within the framework of the French Topographic DataBase!
production, IGN? is looking for solutions related to the
photogrammetric data capture automation with the double goal
to reduce both data acquisition costs and data availability delay.
Since digital stereoplotters have been introduced in the
production lines, it is possible to integrate automatic or semi-
automatic tools in the traditional data capture process.
1.1. The French Topographic Database
In 1990, IGN began the production of the topographic database
which is going to replace the traditional topographic 1:25000
maps as topographic reference on the whole French territory.
The semantic content of this database is very close to the one of
traditional maps with a better geometric accuracy (in the range
of one meter r.m.s.) and 3-D coordinates for all the objects. Data
acquisition consists of a photogrammetric data capture from
aerial photographs at the scale of 1:30000. The need to reduce
the data availability delay incited IGN to invest in digital
techniques (Jamet, 19952) : a research program was started up in
image analysis to provide semi-automatic helps to the plotting
: BDTopo®.
Institut Géographique National (FRANCE).
14
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
and softcopy stereoplotters have been acquired in the production
lines. The digital photogrammetric platform which has been
chosen is named SAPHIR? Géo and is developed by SYSECA.
SAPHIR Géo is a solution for high-accuracy softcopy image
exploitation system which integrates a comprehensive
photogrammetric and image processing package with an object-
oriented GIS able to manage on-line geo-referenced objects.
1.2. The semi-automatic approach
The MATIS" laboratory is the research laboratory of IGN which
works on the questions related to the topographic database data
capture automation. One of the most advanced project in our
laboratory concerns the road network automatic extraction. We
are working at the same time on very ambitious approaches,
looking for a fully-automatic interpretation of the image and on
semi-automatic approaches, looking for a short term efficiency.
Fully-automatic systems have often been presented in the
literature for the extraction of man-made objects from aerial
images but no system has been shown to work both accurately
and completely with a variety of images and objects (Hsieh,
1995). In front of the difficult problem of the automatic
3 System for the Automation of Photo-Interpretation and Remote
sensing data analysis.
^ Méthodes d'Analyse et de Traitement d'Images pour la Stéréo-
restitution.
interpretatior
automatic ap]
contribution
(Heipke, 199
We propose
(Airault, 1°
publications.
seems to be
solution if th
only when tl
model used
automatic te
(predictable
parameters, [
on its exhau:
results of oui
automatic ex
modest acco
enhancement
landscape ch
of the effici
automatic we
time taken by
the user’s de
1.3. How to i
To enhance
the automatic
One solution
which would
network grap
e fully-autc
e interactiv
Figure 1:
If fully-autor
road networl
from now to
low error rat
part of the ro
visual contro|
2.7
The semi-au
algorithms us
The main ori,
step and the
complete pr
successive) s!
