ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002
SCENE CONSTRAINTS FOR DIRECT SINGLE IMAGE ORIENTATION WITH
SELFDIAGNOSIS
Mirko Appel? and Wolfgang Förstner®
* Siemens AG, Corporate Technology, Otto-Hahn-Ring 6, 81730 München, Germany — mirko.appel@siemens.com*
? Institut für Photogrammetrie, Universität Bonn, Nussallee 15, 53115 Bonn, Germany — wf@ipb.uni-bonn.de
KEY WORDS: Orientation from Points and Lines, Industrial Application, Projective Geometry, Maximum Likelihood
Estimation, 3D Image Map Registration
ABSTRACT
In this paper we present a new method for single image orientation using an orthographic drawing or map of the scene.
Environments which are dominated by man made objects, such as industrial facilities or urban scenes, are very rich of
vertical and horizontal structures. These scene constraints reflect in symbols in an associated drawing. For example,
vertical lines in the scene are usually marked as points in a drawing. The resulting orientation may be used in augmented
reality systems or for initiating a subsequent bundle adjustment of all available images.
In this paper we propose to use such scene constraints taken from a drawing to estimate the camera orientation. We use
observed vertical lines, horizontal lines, and points to estimate the projection matrix P of the image. We describe the
constraints in terms of projective geometry which makes them straightforward and very transparent. In contrast to the
work of (Bondyfalat et al., 2001), we give a direct solution for P without using the fundamental matrix between image
and map as we do not need parallelity constraints between lines in a vertical plane other than for horizontal lines, nor
observed perpendicular lines.
We present both a direct solution for P and a statistically optimal, iterative solution, which takes the uncertainties of
the contraints and the observations in the image and the drawing into account. It is a simplifying modification of the
eigenvalue method of (Matei and Meer, 1997). The method allows to evaluate the results statistically, namely to verify
the used projection model and the assumed statistical properties of the measured image and map quantities and to validate
the achieved accuracy of the estimated projection matrix P.
To demonstrate the feasibility of the approach, we present results of the application of our method to both synthetic data
and real scenes in industrial environment. Statistical tests show the performance and prove the rigour of the new method.
1 INTRODUCTION in many industries. These documents are created during
the design process, and they are used and completed by
builders. Furthermore, drawings are referred to on a daily
basis for maintenance and update of buildings and facili-
ibration. This problem has been well investigated in the ties. It is therefore quite advantageous to use these doc-
past by many researchers (Faugeras, 1993, Kanatani, 1996, uments. In practice, many have taken advantage of such
Klette et al., 1998, Faugeras and Luong, 2001). Most of documents to find some reference points in order to regis-
these methods use point and line correspondences between ter virtual and real world coordinates.
the images and/or calibration patterns to estimate the cam-
era's intrinsic and extrinsic parameters. However, camera
calibration may be more reliable and easier to carry out
if further scene constraints are taken into account. Scenes
which are dominated by man made objects are usually very
rich of such constraints. For example, urban scenes or
scenes in industrial environment contain lots of vertical
and horizontal structures (see Fig. 1). Many works in com-
puter vision and photogrammetry literature exploit these
constraints by using vanishing points for recovery of cam-
era orientation (Caprile and Torre, 1990, Wang and Tsai,
1991, Youcai and Haralick, 1999, van den Heuvel, 1999).
Here, we propose to take advantage of horizontal and ver-
tical structures in conjunction with a map or drawing of
the scene. It is important to note that very often such maps
or drawings are readily available. Drawings are with no
doubt the most important and commonly used documents
Many tasks in computer vision and photogrammetry re-
quire prior estimation of a camera's orientation and cal-
In this paper we aim at providing methods for direct sin-
gle view orientation using these commonly available doc-
uments. We use vertical lines, horizontal lines, and points
to estimate the projection matrix P. We describe the con-
straints in terms of projective geometry which makes them
straightforward and very transparent. In contrast to the
work of (Bondyfalat et al., 2001), we give a direct solu-
tion for P without using the fundamental matrix between
image and map as we do not need parallelity constraints
between lines in a vertical plane other than for horizontal
lines.
We present both a direct solution for P and a statistically
optimal, iterative solution, which takes the uncertainties of
the contraints and the observations in the image into ac-
count. It is a simplifying modification of the eigenvalue
method of (Matei and Meer, 1997). The method allows to
*This work was done while M. Appel was with the Institut für Pho- — €Valuate the results statistically, namely to verify the used
togrammetrie, Universität Bonn projection model and the assumed statistical properties of
Fig
soci
occi
the :
the
achi
The
use
valu
all c
The
intri
sent
tern
pro
opti
Ex
sect
Firs
titie
ric €
tecl
Not
met
spa
mat
mo;
lett