transformation.
Orientation parameters are preferably computed by resection. When two or
more pictures have got their orientation parameters, object co-ordinates of
points are computed by intersection.
At least three control points are required to do a space resection. An ad-
ditional point is required to avoid a false solution. Number of required
points may be reduced by including approximate values of orientation angles.
If there are too few control points, approximate co-ordinates may be used.
These may also be obtained by intersection from other pictures. Some iter-
ations of resection and intersection will improve the co-ordinate approxima-
tions and orientation parameters, and make them useful as initial values for
the bundle adjustment.
When object control is given mainly by auxiliary data, it may be impossible
to compute the preliminary values by the resection and intersection proce-
dure only. In this case models of two or more pictures are formed by rela-
tive orientation. The models are transformed to object system by orthogonal
transformation on three or more points.
During each type of computation the computer may be told to select different
combinations of available data, and to point out significant variations in
the results. The operator may interactively make further selections of data,
thus detecting and isolating gross errors.
Erroneous data may be excluded from further use, or manual corrections may
be applied, such as changing point identifiers to correct identification er-
rors. It is also possible to go back to the monocomparator and digitize some
points anew before proceeding.
Bundle adjustment.
In the bundle adjustment, a simultaneous solution of all unknown co-ordinates
and orientation parameters may require lots of space and time in a computer.
This may cause difficulties even on large computers. A variety of procedures
have been tried in order to solve this problems. The present program makes
use of repeated intersections and resections, alternatively regarding orien-
tation parameters and point co-ordinates as fixed values. The program does
not require much space, since not more than six unknowns are solved simultan-
eously. In most cases it will also save computer time, even if several re-
petitions are required. Acceleration factors may be applied to the correc-
tions in order to speed up the convergence of the solution.
Processing and presentation.
Nominal data may be transferred to the data base from an existing construc-
tion data base, or it may be obtained from the drawings. Varying conditions
may then be checked out, based on the discrepancies between measured and
nominal co-ordinates. The results are displayed by graphical output as well
as in deviation tables.
The graphical presentation is very important for an unambiguous interpreta-
tion of the result. Depending on the current problem, varying presentation
forms may be used. Deviations may be displayed as magnified vectors or as
258