Figure 4 gives an impression of the imagery from the different 
viewing angles. 
Figure 4: Example of nadir and oblique images 
  
  
  
  
  
  
At first the nadir looking strip was processed. Since the 
photogrammetric orientation of a single strip requires ground 
control points, 18 natural GCP's were collected along the 
avenue after the photo flight. Table 3 presents the triangulation 
results of the 34 nadir looking images. The accuracy of the 
GCP's was set to 4 cm in X, Y, Z. 
Table 3: Residuals [m] at 18 control points for AT Laage 
  
  
X Y Z 
Std. 0.07 0.06 0.03 
Max. 0.11 0.17 0.07 
Min. -0.13 -0.13 -0.06 
  
The georeferencing of the oblique images was more 
complicated, because tie points in neighbouring images could 
not be found automatically in the first instance. This is related to 
the fact, that the starting values of c and @ of the hand held 
images were unknown. After a manual definition of a minimum 
number tie points with a selected number of oblique images, a 
preliminary triangulation was conducted to obtain approximate 
angles in o and q. Thereafter the automatic tie point generation 
algorithm worked fine. Due to the apparent differences in the 
scale within the oblique images the precise determination of the 
ground control points was difficult. Nevertheless the results of 
the aerotriangulation of the oblique images were within 60 cm 
RMS at the GCP's. The most interesting aspect for the 
interpretation of the trees and other features along the street is 
the stereo view of the oblique images, because an 
orthorectification does not yield the full information. However 
the stereo interpretation with the Erdas Imagine Stereo Analyst 
1.4 of the imagery was only possible for small parts of the 
imagery, due to a software bug. 
3.3 Direct Georeferencing 
The comparison illustrates that there is still quite some manual 
labor necessary for georeferencing with PFIFF, which could be 
drastically reduced with a GPS/INS. Therefore a first test flight 
was conducted on the 7" of April 2004. A Litton 86 IMU of the 
Applanix 410" GPS/INS of DLR, Institute for Navigation was 
tightly mounted on the camera, see figure 5. The airborne 
  
' http://www.applanix.com/html/products/prod_airborn_tech 
410.html 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004 
L1/L2-Antenna (Sensor Systems S67-1575-76) was set on top 
of the aircraft, approximately above the camera. The lever arms 
between the camera, the IMU and the GPS-antenna were 
determined with a total station. For the determination of the 
boresite angles and information of the overall accuracy a small 
test field with 12 signalized targets was established. 
Figure 5: Litton 86-IMU mounted on Rollei db23 
  
The weather conditions of that day were miserable, with a 
cloudy sky and sporadic rainfall. Nevertheless a photo flight at 
altitudes of max. 650 m above ground could be realized, 
yielding a ground resolution of approx. 15 cm/pixel. The 1.5 h 
test flight consisted of two starts to obtain information of the 
stability of the GPS/INS solution. Therefore the test field was 
flown two times. Additionally to the test site a photogrammetric 
block with four strips and a total of 86 images was acquired. 
Also a strip of oblique stereo images was made. The flight took 
place within a 12 km range of the airport, where the GPS 
reference station was build up. In the first step an 
aerotriangulation of the test blocks was conducted to determine 
the bore site angles between the camera and the IMU. Based 
upon the initial values of the GPS/INS the tie points between 
the images were found automatically. With the position of the 
perspective centers set to be fixed, differences between the 
GPS/INS angles and the aerotriangulation were determined. To 
estimate the triangulation results without ground control 12 
signalised points were introduced first as check points and in the 
second step as GCP's. Table 4 presents the results of the first 
flight over the calibration site. 
Table 4: Residuals of aerotriangulation of the calibration site - 
with and without GCP's 
  
   
     
   
    
   
  
  
   
  
  
  
   
   
   
  
  
  
   
   
  
  
  
   
   
  
   
  
   
    
  
    
   
  
  
   
  
  
  
  
  
   
  
  
  
  
  
  
  
Residuals [m] at 12 check points for AT without GCP 
  
  
  
X Y Z 
Avg. 0.094 -0.014 0.630 
Std. 0.066 0.043 0.305 
Max. 0.223 0.041 1.408 
Min. 0.016 -0.114 0.019 
Residuals at 12 control points [m] for AT 
Std. 0.050 0.061 0.022 
Max. 0.105 0.089 0.060 
Min. -0.068 -0.108 -0.007 
  
    
Inter 
— 
Ther 
depe 
GPS, 
"The: 
relat 
App! 
Stall 
unde 
Addi 
com] 
on | 
pers] 
and | 
proc 
with 
even 
reco! 
were 
(resi 
centi 
posit 
Tabl 
| | 
that 
the [ 
The 
relia 
com 
univ 
com 
prov 
acqu 
is th 
have 
may 
reve 
phot 
For 
addi 
com