Experiences with the Digital Photogrammetric Program Package Orpheus ... 
641 
window, bilinear resampling is used to display the image window, so that it is possible to measure well-defined points with an 
accuracy better than 1 pixel. A point is measured by pressing the left mouse button either in the overview or in the zoom window. 
There are several modes of point measurement which differ by the way the point identifier is incremented. These modes correspond 
to plotting modes such as measuring all points in one image or transferring points from one image to the other ones. In the “point 
transfer” mode, epipolar lines can be used to support point identification in the other images. Lines can be defined to connect a point 
pair by snapping both end points or by a special point measurement mode connecting the points in the order they are measured. In 
figure 1, a window representing a GESTALT is open. The GESTALT has no graphical representation, but the points contained in it 
are highlighted in all open images. Thus, GESTALT 3003 in figure 1 is a vertical straight line corresponding to the edge of the 
building. No homologous points have to be found: the parameters of the straight line (the coefficients of the polynomial function in 
equation 2) are determined by intersection of bundles of rays in hybrid adjustment. By measuring/deleting a point in this situation, 
the point will also be inserted to/deleted from all open GESTALTs. 
tsKKlDlaial 
gKN Diala! 
last digitised point: 
312 026 
Statistics.. 
Y: -550.16 
image ptopetties 
Measu/ement Mode 
last digitised point. 
G Identity Point 
G Delete Line 
<* Digitise Point G Move Point 
G Line G Closed Line 
G Delete Point 
G Snap Line 
r~ Preview 
P Epipolat lines 
Digitisation Mode 
G Identify G Digitise 
Current Point T 
I Transformation.. 
X. 778.47 
Y: 472.28 
Dose this window 
identifier of next point to be digitised 
[Si Î : Photo 
Figure 1: Interactive measurement of points and lines using ORPHEUS 
2.3 Work flow for the generation of 3D photo models 
The work flow for the generation of a 3D photo model using ORPHEUS basically consists of four stages: (1) data import, (2) 
orientation of the images, (3) creating the geometrical model, and (4), mapping the texture to the geometrical model to create the 3D 
Photo model. In principle, an arbitrary number of images taken with an arbitrary number of cameras can be used in this work flow. 
Data import comprises the import of control points from ASCII files, the import of images (JPEG or TIFF files), the generation of 
image pyramids for the visualisation of large images and the import of camera data from calibration protocols. For the latter purpose, 
the polynomial coefficients describing camera distortion have to be estimated from the distortion tables provided by the camera 
manufacturer. 
Orientation of the images: For scanned analogue images, the fiducial marks have to be measured to reconstruct inner orientation. 
For computing outer orientation, both control and tie points have to be measured in the images, and GESTALT observations can be 
defined. Approximate values for the rotation angles have to be provided either by an automatic spatial resection using at least four 
points of known object co-ordinates, or by applying a graphical tool for positioning the axes of the camera co-ordinate systems of all 
photos in a ground plot of the object co-ordinate systems. If self-calibration is to be applied, approximate values for the focal lengths 
of the cameras involved have to be provided. After that, hybrid photogrammetric adjustment is used to determine all unknown 
mapping parameters and the object co-ordinates of all control and tie points. In this context, the user has every freedom with respect 
to the control information which is to be used to define the geodetic datum of the block. Control points can be used as “observed 
object co-ordinates”, but it is also possible to directly use theodolite measurements instead of control points, or to use GESTALTs for 
modelling plumb/horizontal lines to level the block and import distance measurements for determining the scale. A residual plot in 
the images can be used to correct the errors detected by robust estimation. The stochastic model of adjustment can be checked by 
variance component analysis. It is possible to mix self-calibration of amateur cameras with metric cameras. At any instance, the