AUTOMATED MEASUREMENT OF GROUND CONTROL OBJECTS IN LARGE SCALE AERIAL PHOTOGRAPHS 
Bjarke Mgller Pedersen 
Department of Development and Planning 
Laboratory for Photogrammetry and Surveying 
Aalborg University 
DK-9220 Aalborg 
Denmark 
bmp@i4.auc.dk 
Commision Ill, Working Group 2 
KEY WORDS: Digital photogrammetry, exterior orientation, existing digital map, target template, matching. 
ABSTRACT 
The automated determination of the exterior orientation by using map data is tested in 1:5,000 digital aerial photographs. The 
applied method is area based matching with target templates derived from object types in large scale digital maps. Results 
comparable to manual measurements are achieved. Problems may occur in case of many similar looking nearby objects. 
Blunders can, however, be detected by robust estimation. If these results can be verified by further tests, the method may be 
profitably applied to map revision, orthophoto production, and thematic mapping. 
KURZFASSUNG 
Die automatische Bestimmung der äußeren Orientierung wird mittels Daten einer topographischen Datenbank und digitalen 
Luftbildern im MaBtab 1:5000 ausgefürt. Die angewandte Metode ist eine fláchenbasierte Zuordnung von Bildausschnitten 
und Mustermatrizen, welche von verschiedenen Objekttypen in groBimafstábigen Karten abgeleitet wurden. Die Ergebnisse 
sind mit denen bei manueller Messung vergleichbar. Probleme können im Falle von vielen ähnlichen und in der Nähe gelegenen 
Objekte auftreten. Grobe Fehler lassen sich jedoch mittels robuster Ausgleichung aufdecken. Sofern diese Resultate sich in 
weiteren Tests bestätigen, kann die Orientierungsmethode bei der Kartenergänzung, Orthophotoherstellung und thematische 
Kartierung mit Vorteil benutzt werden. 
RESUME 
La définition automatique de l'orientation extérieure en utilisant les données de carte a été éprouvée dans une photographie 
aérienne digitale de l'échelle 1:5000. La méthode utilisée est l'assortiment basé sur surfaces acvec un garbarit de but qui dérive 
des types d'objects dans des cartes digitales à grande échelle. Les résultats peuvent se comparer aux opérations de mesurage 
manuelle. S'il y a beacoup d'objets ressamblants à proximité cela peut poser un problème. Il est quand méme possible de 
détecter les erreurs en faisant un ajustage robuste. Si les conclusions peuvent étre vérifer par plus d'épreuves, cette méthode 
peut étre utilisée avec avantage pour les révisions des cartes, pour le production des orthophotos et la cartographie thématique. 
1 INTRODUCTION points. Vosselman and Haala [Vosselman & Haala, 1992] 
solves the detection problem by a relational matching method 
using structural descriptions of large linear features like roads 
and rivers. Giilch [Giilch, 1994] solves the pointing problem 
for signalized points using region segmentation and active 
contour models. Heikkinen [Heikkinen, 1994] matches linear 
The interior orientation by recognizing the pose of the digital features derived from a GIS to lines detected in an operator- 
image and measuring the fiducials is now fully automated by selected window. 
means of matching well known target templates and image E i : s : 
patches [Schickler, 1995]. The relative orientation, aerotri- A very different and promising line of research tries to Inte- 
angulation and DEM compilation processes have been auto- B'ête several sensor systems like GPS and Inertial Navigation 
mated by matching corresponding image patches or features, Systems. Results reported in [Schwarz, 1995] are accuracies 
using highly redundant measurements [Hellwich et al., 1994], of 0.1-0.2 gon for the rotation angles and 1-2 metres for the 
Digital image processing has made available to photogram- 
metry a number of techniques for automatic measurements, 
and in recent years the potential of automating important 
photogrammetric processes has been demonstrated. 
[Tsingas, 1091], [Kryzstek, 1995]. position, using a low-cost system. 
More difficult are the semantic processes of exterior/absolute As of now, none of these approaches to automate the ex- 
orientation and map compilation. Although remark- terior/absolute orientation solve both the detection and the 
able results have been achieved in specific scenes, e.g. pointing problem without restricting constraints. Some hy- 
[Polis et al., 1995], map compilation has a long way to go brid solutions may well be possible, but are still not avail- 
before the process is truly automated. able. There is a tendency towards using larger structures like 
houses, roads and rivers. Large structures are not so sen- 
sitive to different kinds of 'noise' like vegetation, shadows, 
occlusions and differences in the background. 
As for the exterior/absolute orientation, there are some pre- 
sented approaches to solve the task. The approach presented 
by Fórstner [Fórstner, 1988] and Schickler [Schickler, 1992] 
uses 3-D wireframe control points, but depends on a con- In digital photogrammetry, there is a new strategy for auto- 
trol point data base established by the Survey Department of matic measurements, which is very different from the strategy 
Nordrhein-Westfalen in Bonn, consisting mainly of roof gable of analytical measurements, e.g. [Ackermann, 1995]: 
633 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996