THE EXTERIOR ORIENTATION OF DIGITAL IMAGES 
BY ROAD MATCHING 
FAYEZ SHAHIN & KURT NOVAK 
Dept. of Geodetic Science & Surveying 
The Ohio State University 
Columbus, OH 43210, USA 
Commission II, Working Group 1. 
KEY WORDS: B-splines, linear feature extraction, relational matching, automatic orienta- 
tion. 
ABSTRACT 
The ultimate goal of digital photogrammetry is to produce maps automatically from digital 
images. For that purpose, the exterior orientation of the photos is required. It establishes the 
correspondence between the image and ground coordinates. Most approaches only address the 
relative orientation of the images. This paper presents a general procedure for the automation 
of the exterior orientation, which can be used for real time mapping applications. This proce- 
dure takes advantage of a mobile mapping system (GPSVan) developed and implemented at 
The Ohio State University’s Center for Mapping. Various sensors collect information about the 
environment of highways from a moving van. A satellite (GPS) receiver mounted on the van 
determines the road alignment in the ground coordinate system. These alignments are math- 
ematically represented by cubic B-splines to serve as a 3-D model of the roads on the ground. 
Digital images covering the same area are processed to find roads automatically. The extracted 
roads are also represented by cubic B-splines to serve as a model of the roads in image space. 
By applying relational matching and tree search methods, the best match between the roads 
in the digital images and their corresponding 3-D model in object space can be found. Thus, 
the correspondence problem can be solved and the computation of the exterior orientataion 
is possible. This approach is extremely efficient for orienting satellite images and small scale 
aerial photos. Therefore, it has a great potential for real time mapping applications. 
1. INTRODUCTION 
conjugate points or features in overlapping 
images. Several algorithms and techniques, 
  
  
In recent years it has been shown that sev- 
eral important photogrammetric tasks, like 
the relative orientation of images, the aerial 
triangulation, and the derivation of digital el- 
evation models, can be automated with very 
little or no human operator intervention. The 
main concern of solving these tasks automat- 
ically is to establish correspondence between 
174 
such as area based matching and feature 
based matching, have been used successfully 
to solve the correspondence problem between 
overlapping images [Schenk et al. 1991, Ack- 
erman 1991]. 
Another group of tasks, including the exte- 
rior orientation of images, the geo-registration 
  
of images 
ages, have 
tomation. 
essary in 
has to be 
spondence 
the object 
orientatio! 
to determ 
trol point: 
the image 
difficult pi 
ing the 3- 
points, wh 
age repres 
Is 1n rastel 
In photog: 
dence bet: 
tablished 
trol point 
cases, suck 
aerial pho! 
üfy contrc 
of the ima 
procedure 
by matchi: 
on the gro: 
ficient for 
scale aeria 
tential for 
During th 
ping syste 
mented by 
Ohio Stat 
tion about 
a driving 
on the veh 
alignment 
vak 1991]. 
model of 1 
images of 
extract ro: 
roads on t] 
respondent 
and image 
matically.