International Archives of the Photogrammetry, Remote Sensing and Spatial [Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
Photo control processes, possible improvements: 
- . Concerning: Automated control of  photo-center 
accuracy and overlap between neighboring photos. 
Suggestion: lt is evident that GPS has already 
minimized the number of photo-center errors. Use of 
INS/IMU could improve positioning even more and 
allow less attention to the photo-center control 
- . Concerning: Visual control of the appearance of 
photographic end-products. The production line from 
2004 will no longer include photographic end-products. 
But the in-house control procedures still demand contact 
copies, to compensate for loss in detail in the scanned 
images of 21 micron. 
Suggestion: Need for contact copies disappear with a 
better scan solution or by usages of digital cameras 
- . Concerning: Manual random check of scanned images. 
Suggestion: Automatic control of all images should be 
implemented. Future use of digital cameras might make 
control of scanned images unnecessary 
Object control processes, possible improvements: 
- . Concerning: Control of object geometry is currently in 
2D, by polynomial rectified photos 
Suggestion: Controlling all objects in all three 
dimensions is essential, and can be solved by: 
- Use of photogrammetric transformation of vector 
data (X, Y, Z) into raw image 
- Use of orthophotos 
- Concerning: Control of object data according to 
specifications, is currently by “hard-coded” routines. 
Suggestion: Specification rules should be archived and 
accessed in a database. This would also make it possible 
to control object data according to their setup 
specification 
- . Concerning: Current control of data completeness is 
performed visually by polynomial rectified images. 
Suggestion: Automatic change detection or identical 
accuracy and better correlation all around the image, 
between image-objects and vector-objects is essential - 
and can be solved by: 
- Use of photogrammetric transformation of vector 
data (X, Y, Z) into raw image 
- Use of orthophotos. 
- . Automatic detection and classification of objects / 
changes in a image (Olsen, 2004) 
2.4.2 Optimizing, control processes. It was recognized from 
this evaluation process (2003), that object-geometry control 
did have highest priority for development. A tool for 
photogrammetric transformation of vector data to raw image 
was therefore decided, as a new control procedure. 
3. CONTROL OF 3D-DATA IN 2D-ENVIRONMENT 
To illustrate the continuous development of the control 
procedure used for TOPIODK, the chronicle steps of 
optimizing object-geometry control are listed: 
- In the database setup phase: random checks in 3D 
- J Overall 2D-control by polynomial rectified images 
- In the database update phase: overall 3D control, by 
photogrammetric transformation of objects (X, Y, Z) 
into the raw image 
754 
3.1 Photogrammetric transformation of vector data 
The developed method for photogrammetric transformation 
of vector data is called the “Vector Transformation" module 
(VT module). It is implemented by approved photogram- 
metric methodologies but is for practical and economical 
reasons designed to work on PCs without stereoscopic view. 
The VT module projects the vector data photogrammetrically 
on top of the raw image, accounting for camera properties 
and map-object coordinates (X, Y, Z). 
The goal of the method is to develop a tool for the control of 
geometric accuracy of individual map-objects in 3D so that: 
- . Map-objects that are registered correctly in X, Y and Z — 
will be displayed (draped) exactly on top of their 
appearances in the raw image (as good as the 
photogrammetric parameters allows). 
- . The accuracy of the draping will be the same in the 
whole image, meaning no need of shifting vector data in 
the image periphery to match their appearances (as is 
needed for polynomial rectified images — fig.4). 
- When single map-objects are displaced from their 
position in the image, something is wrong in that 
specific object. 
- . When all map-objects are displaced from their position 
in the image, some fundamental error has occurred. 
Error can be from the VT-parameter-calculation or from 
the model orientation used by contractors for the object 
extraction. Both can and must be checked. 
Integrating the technique into production in KMS was done 
with guidance from a research-project of the methodology 
done by Jon C. Olsen (KMS). Functions were integrated in 
the production and control software Mapcheck, and 
facilitated with automation features. 
3.2 Vector Transformation methodology 
The Vector Transformation module is built on the 
fundamental equations of photogrammetry. Parameters for 
the transformation are calculated through the knowledge of 
inner orientation (IO) and rotation (D) of the image. Stating 
that coherence between vector map-objects in geographical 
coordinates and their coordinates in image pixels, are 
calculated for each image, by information about: 
- Camera calibration 
- Fiducial coordinates 
- A number of well distributed ground control points. 
The inner orientation is found through an affin transformation 
of the six variables, for the coherence between fiducials 
measured in image and data from the camera calibration, 
solved with a least square calculation. 
The rotation matrix (D) is evaluated by : 
DX’= M (X-Xo) 
D Y! « M(Y-Yy) 
D -c = M (Z-Zy) 
c 
Where: M. 
h 
h- (ZZ) 
Intern 
and: 
which 
The rc 
ofan 
measu 
locatic 
rotatic 
point 
transf 
in the 
33 V 
The V 
as pa 
TOP 
3.3.1 
deteri 
Figure 
control 
betwee