A LONG RANGE PHOTOGRAMMETRIC METHOD
WITH ORTHOGONAL PROJECTION MODEL
Tetsu ONO*, Shin-ichi AKAMATSU*, Susumu HATTORI**
*Graduate School of Engineering, Kyoto University, Yoshida-Honmachi, Sakyo-ku, Kyoto 606-8501, JAPAN
ono@jf.gee.kyoto-u.ac.jp, akamatsu@info.gee.kyoto-u.ac.jp
**Faculty of Engineering, Fukuyama University, 1 Sanzo, Gakuen-cho, Fukuyama, 729-0292, JAPAN
hattori@fuip.fukuyama-u.ac.jp
KEY WORDS: photogrammetry, vision metrology, orthogonal projection, orientation model, long range observation
ABSTRACT
This paper describes a method appropriate for 3-D measurement from a long distance with a digital camera mounting
a super telescopic lens. Long distance observation with telescopic lens camera is a effective method for displacement
measurement such as small movement of top of a large cliff. However the conventional orientation method with the
central perspective model widely used in close range photogrammetry is unstable in this case because of weak condition.
In this study, an alternative model called the orthogonal projection model is applied. This model, which derived from
the affine projection model with a constraint of orthogonality, is simple and better adapted to long distance observation.
Furthermore, it is a great advantage that the orthogonal projection model does not require initial values of orientation
parameters. In this study, the geometric characteristics of the orthogonal projection model were clarified by various
simulations, and also the effectiveness of the method was verified by long distance field tests. Focal length of telescopic
lens used in the field test was 400mm and distance to objects was over 100m. RMSE between the adjusted results with a
camera and those with a totalstation was about 2.8mm. The accuracy with the proposed method is more than twice higher
than that with the conventional method.
1 INTRODUCTION
In recent years digital close range photogrammetry has be-
come applied to various objects in various fields. However
it has not widely enough spread in fields of civil engineer-
ing and construction. In construction work, displacements
of cliff faces or construction materials have to be precisely
observed for disaster prevention, but ordinarily the target
sites range over vast areas and moreover arrangement of
view points is restricted. For example, movement of un-
stable rocks on top of large cliff has to be monitored to an
accuracy of a few millimeters at ground points over 100m
distant in some cases.
Close range photogrammetric techniques using camera with
wide-angle lens are not available in these situations. Ap-
proaching the target objects is dangerous and may be im-
possible. It is not uncommon that fixing a deal of reflective
targets on the objects is not allowed. In many cases, there is
no alternative but to observe the object shape without clear
marks from a distance. By using a telescopic lens cam-
era, optical resolution can be kept high enough in spite of
long distance observation. As focal length increased, how-
ever, view angle becomes smaller and geometric condition
becomes worse. What is worse, available control points
may be not many enough and may be ill-placed. Under
these circumstances, the central perspective model gener-
ally used in close range photogrammetry is hardly applica-
ble because of ill-posed problems.
Scaled orthographic projection models are widely used in
computer vision to model the imaging process(Ullman 1979;
Huang and Lee 1989; Tomasi and Kanade 1992; Shapiro et
al 1995; Xu and Sugimoto 1999). They are mainly used for
calculation of an approximation to perspective projection
1010
model, but it should be noted that these models are stable in
ill-conditions. Some papers in computer vision categorize
them to the following camera models; the affine model, the
weak perspective model and the para-perspective model.
The authors contrived the orthogonal projection model (Ono
2002), which belongs to the weak perspective model in the
sense of camera model, but the basic concept is quite dif-
ferent from that of computer vision. The main aim of 3-D
image analysis in computer vision is efficient estimation
of motion. The weak perspective model is not a rigorous
model but a approximate one in the real world. On the
other hand, the orthogonal projection model is directly de-
rived from the central perspective model without approx-
imation. This model can provide an accurate solution by
itself.
2 ORTHOGONAL PROJECTION MODEL
2.1 Basic concept
The conceptual diagram of orthogonal projection model is
illustrated in Figure 1. The procedures of orthogonal pro-
jection model consist of the weak perspective projection
and the projection transformation from the central perspec-
tive images.
From another viewpoint, this can be considered as a pro-
gressive model of the affine projection model, which Okamoto
(1992,1998) proposed for long range observation.
2.2 Derivation of model equations
If the lens distortions and the shift of the principal points
are negligible, the central perspective model is expressed
Inte
Fig
mc
as:
By
SCI
wl
Tt
thi
TI
of