of the camera, which contains the roll w , the pitch and the swing ae.
The instantaneous orientation parameters are functions of the orienta-
tion parameters in the update points Pj; represented by the vector p,
and the terrain coordinates of the DEM-points Pss represented by the
vector k.
At first approximate values of these unknowns are determined and in-
serted into the equations (1). The differences between the computed
image coordinates Xi N and Y; N° which are functions of p and k, and
the correlated image coordinates Xi N and Yin lead to the obser-
vation equations > ,
Vo dn = FN NTuR D PROPER PSN
(2)
Yi NY 4,N 7 Fy (Pg) 7 Yi
For the absolute orientation simple error equations with known control
points can be added.
By least squares adjustment the most probable values of the orientation
parameters p and the DEM-coordinates k are found.
The observation equations (2) are linearized by truncation of Taylor
series. The normal equations are set up and solved in two steps. The
unknowns k are eliminated and the reduced normal equations, which are
of band structure, are solved in a direct way. This means, that the co-
efficients of the normal equations are determined directly from the ob-
servations. In the second step the unknowns k are computed by substi-
tution of the orientation parameters p.
3. ACCURACY OF THE STRIP MODEL
It must be emphasized that the DEM model strip is form-invariant and
stable even without any control points. Without them it could be com-
puted in any scale and orientation within a local system. The insertion
of measured orientation parameters and/or control points establishes
the absolute orientation and possibly improves the accuracy of the
strip. The standard errors of the DEM-coordinates k can be expressed by
h 5 1
OF =O) Sy Tena V em a Kk. (3)
The least-squares adjustment yields the standard error O7, which can
be replaced and expressed by the actual image standard error. It depends
on the accuracy of calibration of the camera, the image correlation
errors and the errors of interpolation (of the orientation parameters)
between the update points (Fig. 6). This portion 03 of the image error
o is a function of the interpolation error Ip? the interpolation
error is the difference between the interpolated parameter between ad-
jacent update points and its exact value. This interpolation error can
be reduced to a few microns by suitable means, for example stabilized
- 350 -