the final accuracy 
> control and check 
le of such type of 
  
ntrol point 
this project were: 
n NATO standard 
scale photography, 
ing basements and 
s images and their 
> photogrammetric 
h ArcGIS, which 
oresponding RPC 
Evolution projects 
the figure 4. 
  
   
Orthophoto 
Feature Collection 
  
Cancel | 
ing the possible 
ition) 
tage has a major 
ction and all data 
to this part. For 
y in relation to a 
combinations of 
o measurement of 
jints with known 
provement of the 
more observation 
field load. The 
rvation equations 
carity conditions. 
ormly distributed 
es (in PUWG-92 
ol points and the 
1alysis, The RMS 
(terior orientation 
1 0f 3D model in 
on, using of RPC 
coefficients only, and in addition, including 1 or 5 control 
points (CP), supported also by 5 or 17 tie points (TP). When 
orientation of images has been executed only with RPC 
coefficients (in Summit Evolution), only the approximate 
orientation of 3D model was achieved and its stereo viewing 
was disturbed. The RMS for X, Y,Z calculated from differences 
between field coordinates of check points and the corresponding 
points measured on such oriented 3D photogrammetric model 
was: RMS[X]=16,6m, RMS[Y]-2,7m and RMS[Z]-47,4m. 
Exterior orientation by use RPC coefficients only, was also 
executed with remote sensing PCI Geomatica software and the 
RMS (only for XY) were: RMS (X) 217,3 m RMS (Y) 23,9 m. 
As it can be seen from the above, the accuracy of RPC 
orientation by both systems is similar, however the advantage of 
the photogrammetric approach is reconstruction of 3D model 
and stereo measurement not only X and Y but also Z coordinate. 
The results of other cases (executed in stereo Summit 
Evolution system), with inclusion of 1 or 5 control points only 
and in addition to 5 control, 10 and 17 tie points (stereo 
measured), are presented in table 1. 
Tab. 1 RMS (X,Y and Z) for check points in 3D models 
reconstructed with RPC only, and in addition, with 1 or 5 
control points supported by 5 or 17 tie points. 
  
  
  
  
  
  
  
  
  
  
  
Ne es RMS | RMS | RMS | RMS 
; (X) (Y) (XY) (Z) 
de CP) [m] | [m] [m] [m] 
check (CHP) 
RPC Only 
Teo" 166 | 27 16,8 47,4 
1CP 
OE 0,9 0,8 12 2,2 
5 CP 
er 0,7 0,7 1,0 10 
5CP+5TP 
er Sd 0,5 0,6 0,8 0,8 
5CP+17TP 
m T. 0,4 0,5 0,7 0,6 
  
  
As it can be seen from table 1, for the model reconstructed 
only by RPC coefficients, the RMS (XY) and (Z) are very large 
and also the very high affinity can be observed between X and 
Y coordinates. Inclusion of 1 or 5 control points (CP) for 
external orientation, has improved significantly the accuracy of 
3D model reconstruction. The RMS evaluated for the horizontal 
and vertical coordinates of 17 check points was decreased 14 or 
17 times for XY and 22 or 47 times for Z respectively, 
regarding to accuracy of orientation by RPC coefficients only. 
Stereo-measuring of tie points, in addition to a few control 
points, gives the next accuracy improvement. As it is shown in 
table 1, more tie points involved better accuracy achieved. 
Using for orientation 5 control points and in addition 5 or 17 tie 
points, can decrease the RMS(XY) decreased to 0,8m or 0,7m 
and RMS(Z) to 0,8 or 0,6m, respectively. The advantage of tie 
points inclusion is not only accuracy improvement but their 
stereo measurement which allows choosing any type of natural 
terrain point, even of very poor quality. 
In the experiment, all the measurements were performed on 
the photogrammetric station (Summit Evolution) integrated 
online with ArcGIS system (platform). This solution has 
allowed transferring directly all measured data directly to GIS 
database and has given a number of possibilities such as 
snapping, attributing objects and as a result creating 2D/3D 
databases online. Databases in ArcGIS provide various formats 
of the files. The creation of the shape files and definition of the 
coordinate system is possible in the ArcCatalog module. In this 
project, a geobase was created with separate shape files for 
points, lines, polygons, multi-points and multi-patch objects. All 
of those formats can store height information, but only multi- 
patch objects can deal with 3D data (such as blocks). 
For registration and edition of the measured data, ArcMap 
module is used. Connection of Summit Evolution with ArcMap 
is supported by three DAT/EM Tools (Capture, Drawing and 
Editing - fig. 5), which allow to determine a type of digitized 
object and its edition, information on elements, registration and 
edition of DTM points, creation of offsets and switching 
between platforms of Summit Evolution and ArcMap. 
  
  
  
  
  
  
  
  
  
Fig.5 DAT/EM Capture Tools in ArcMap (Manual of SE). 
In the process of orientation, coordinates of the ground 
points (control and check) archived in established GIS database 
were used. All these coordinates were transferred to Summit 
Evolution auxiliary file using the instruction 
Edit->Control File from tools set of Summit Evolution. 
After orientation of stereo images, the 3D model was formed 
and  stereo-digitization of all 3D data requested for 
reconstruction of buildings (edges of the roofs), has been 
executed and then transferred directly to the selected layers of 
ArcMap database. Stereo digitized roofs of buildings in the 
shape format are presented in figure 6. 
  
    
$doteotehru f9Qoute 
n 
À 
  
  
  
Fig. 6 Sterco digitized roofs of buildings in the shape format 
The ArcMap database has contained also the separate layers, 
with such data as DTM and ground footprints of buildings, 
which were available from other sources. These data, together 
with 3D data of building roofs, were useful in the next process 
of 3D modelling of buildings, performed in Google SketchUp 
programme. This intuitional and user-friendly application 
hinges on two basic components — edges and surfaces. The most 
important tool - Push/Pull was used on the imported file with 
115