levation Models for
Images by Digital
| Archives of
Vol 28, No.4.
is of Surveying in
Tang Emperors’
Vienna, Commision
ing von Sicherheit
ung bei Einsatz der
und Luftbildwesen,
-anguage), product
)duct information,
r Gelandemodelle
ung. Deutsche
. 418
nzeption und Bear-
jen. Wissenschaft-
lessungswesen der
tik des sechsten
ad Archaeologicam
. Abb. 13-14.
AT: Frankfurt (1983-
a 1996
ON CRITICAL CONFIGURATIONS OF PROJECTIVE STEREO CORRELATION
Gerhard Brandstátter
Professor, Institute of Applied Geodesy and Photogrammetry
Technical University Graz, Austria
Commission Ill
KEY WORDS: Projective photogrammetry, stereo correlation, critical configurations
ABSTRACT:
Stereocorrelation in projective photogrammetry is performed by means of eight irregularly distributed points and
yields eight linear equations for the determination of
configurations are characterized by vanishing of the determinant of those eight equations, wherefrom critical
point distributions may be derived. Of course, the well-known critical configuration of five orientation points will
belong to it but with respect to eight points, additional conditions of singularity will arise. The following
considerations contain investigations on this subject using projective relations in homogeneous vector notation.
KURZFASSUNG:
the components of the correlation matrix. Critical
Stereokorrelation der projektiven Photogrammetrie erfolgt mittels acht unregelmäßig verteilter Punkte und
beruht auf acht linearen Gleichungen für die Bestimmung der Komponenten der Korrelationsmatrix. Kritische
Anordnungen sind charakterisiert durch das Verschwinden der Determinante jener acht Gleichungen, woraus
die kritischen Punktanordnungen abzuleiten sind. Natürlich wird auch die wohlbekannte kritische Konfiguration
der fünf Orientierungspunkte dazugehören, bei acht Punkten müssen aber noch weitere Singularitäts-
bedingenen auftreten. Die nachstehenden Ausführungen enthalten einige Entwicklungen zu diesem Problem
unter Verwendung projektiver Beziehungen in homogener Schreibweise.
0. INTRODUCTION
Algebroprojective photogrammetry deals with pro-
jective images of completely unknown interior
orientation. The coordinate system of such images
cannot be referred to a more or less orthogonal and
isometric external coordinate system, but to an
(affine) internal system defined by suitable object
points which are projected to homologous points in
the images. In this case, stereo correlation results in
the determination of eight parameters, consisting of
the usual set of relative orientation and an additional
set of three parameters related to an interior affine
coordinate system defined by three non-collinear
homologous image points (Brandstätter 1991). These
two affine systems are corresponding images of the
first coordinate plane of a tridimensional affine
coordinate system in the object space and produce
most uncomplicated projective transformations.
The parameters of stereo correlation are the eight
significant components of a 3x3-matrix C, the core of
the homogeneous coplanarity condition. Applying this
condition to eight pairs of homologous points, a
system of 8 linear equations results wherefrom the
components mentioned above may be computed. The
solution depends on the regularity of its matrix of
International Arch
77
ives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
coefficients exclusively composed of the homologous
affine image coordinates and hence on the spatial
distribution of the corresponding object points in the
tridimensional model space. The matrix will become
singular if the positions of those points of correlation
correspond with a critical configuration. The intention
of the following treatise is, to discover this critical
situation from the geometric relations and to find out
a strategy to avoid it. For this aim the following
symbols in homogeneous notation are used:
homogeneouse affine image
coordinates
affine coordinates of the
object space
Yo center of projection
M= {my} j- 0.12
k=0,123
P= {ow} j=k=0,1,2,3 regular projective matrix
C= {ci}, i=j=0,12 matrix of correlation
A = {an} |=-M=3...7 matrix of coefficients
Q = {au} j=k =0,1,2,3 matrix of a quadric surface
local stretching coefficient
u! - (1,u,,u;)
y! -(1y« Yo. Y3)
singular projective matrix
M eere SEN Bonner anne