Introduction
Close range photogrammetry with its short object distances provides co-
ordinate accuracies that are often difficult to match by conventional survey-
ing means. To provide control with higher accuracy can therefore be rather
difficult, time consuming and expensive.
Rather than utilizing the traditional two step approach of separately
adjusting the surveying measurements to obtain control point coordinates
which are subsequently used in the photogrammetric adjustment, GEBAT provides
a means of simultaneously adjusting both types of measurements.
This has the advantage of fully utilizing geodetic observations, which
by themselves may not be sufficient for coordinate determination.
A computer program named GEBAT (General Bundle Adjustment Triangulation)
was developed by the authors at the University of New Brunswick (El-Hakim,
1979; El-Hakim & Faig, 1981) and later modified at the National Research
Council (El-Hakim, 1982). It is designed to achieve the highest possible
accuracy by reducing the effect of systematic errors, which is particularly
critical when using non-metric cameras. The photogrammetric accuracy is ex-
pressed in the same way as ground surveying accuracy.
The Program GEBAT-V
This program is designed for close-range applications where a non-metric
camera is employed. Derived from the general GEBAT system, the geodetic ob-
servations are restricted to spatial distances and. height differences, while
self calibration is arranged in a photo-variant mode (i.e. each photograph
has its own calibration parameters (calibrated principal distance, principal
point coordinates plus 8 parameters of a harmonic function)). The "V" :
designates the photo "variant" mode. Details of mathematical models and
solution are given in (El-Hakim, 1979). In addition, gross error detection is
applied using the data snooping approach. The program computes the redundancy
number for each observation and applies a statistical test as described in
(El-Hakim, 1981). Finally, the variance-covariance matrix and error ellip-
soid for each adjusted object point is computed.
The critical value for the data snooping test and the confidence level
for the error ellipsoid form part of the input to be provided by the user.
Furthermore, any information available about the camera-station parameters
as well as ground coordinates of any point (not necessarily a control point)
can be utilized in the program. Suitable weight must be introduced for such
information. The program allows the weighting of each station parameter (or
even fixing it) and of each coordinate of each object point.
The photogrammetric mathematical model is the self-calibration bundle
adjustment model:
(x, -X.)m, jKY,-Y )my) * (Z,-Z )m,,
xA x, Hav m -f
XX my + (0-1 Im, + (Zu-Z IM,
(1)
and
CaS" Om, + 2-2 0m,
ef
-y +
YA “OY, vs
U Img
145