CEE A IE PDA FETT
IN wn
EEE
PASAT LEB AY
the pointings in y-direction tending to be somewhat less accurate
than in x-direction of the comparator. Targeted object points were
measured with pointing accuracy about 0.1 um less than réseau cros-
ses. This holds true for all photography excepting targets and cros-
ses of RMK flights in 1:11,000 which were measured by a rather un-
Skilled person (m, - t 1.6 um; m, - t 1.8 um). These differences
show up in the results of spatial resections for single photo-
graphs as well as in simultaneous bundle adjustments of several
strips.
5. MODES OF COMPUTATIONS
5.1 GENERAL REMARKS
A partial field calibration of & photogrammetric system may be (0
obtained among others by describing the deviation from the mathema-
tical model adopted for analytical photogrammetric handling of the
physical process by additional parameters. These polynomials of para-
meters can be the more complex the more observations are at hand.
The parameters of the polynomials should be neither correlated between
themselves nor should they with any parameters of the analytical mathe-
matical models. This implies a certain pattern of observationals. It
seems at least to be doubtful whether there exists any worsening ef-
fect of none-significant parameters.
Attention must be paid to the amount of control necessary for an
unbiased calibration result. It has been shown elsewhere (Mauelshagen
1976) that with at least four strips (e.c. strips 1 and 2 of the Rheidt
Test Area, each flown at 1:11,000 scale in opposite directions) cover-
ing the Test Area, four fully controlled points at the corners of the
area and one at the centre may suffice. This holds true for normal
flying attitude and longitudinal lap of 60 %. It is very essential that
the rest of targets need not be control points, and that is the bulk of
targets needed. They only serve for spatial resection of homologues €
rays, thus strengthening the whole geometry.
Partial field ealibration started at the Rheidt Test Area in the
mid sixties with spatial resection of single photographs and a poste-
riori computation of regression polynomials. For this procedure the
whole lot of targeted points had to be known with full geodetic con-
trol. After installation of the bundle adjustment programm BOBUE
(Mauelshagen 1976), which allowed for additional parameters and full
inversion of normal equations, bundle adjustment was introduced for
calibration purposes.
Further developments of BOBUE enable us to choose additional para-
meters to a wide extent at will. Besides that the three points of any
group of points may be treated as one point in a rigid adjustment
(Ellenbeck 1977). Hence the number of unknowns is reduced remarkably.
