with the 
ss the in 
excluding 
a transfer 
1 h. This 
from the 
k all 82 
h, which 
hages per 
time into 
mages of 
the large 
ction and 
dicated in 
ige of the 
erface of 
lock. But 
npared to 
ibuted to 
uses less 
ared to 
ers this 1s 
re is still 
ngulation 
indicates 
/here the 
from our 
  
     
  
  
  
  
  
  
  
  
  
  
  
  
AT processing step mode | St. Gallen Zug 
1 | Preparation for AT (data files) manual 20h 4.0 h 
2 | Interior Orientation manual 2.0h 2.0h 
3 | Automatic Point Measurement batch 23h 13h 
4 | Interactive Point Measurement manual 10h 50h 
5 | Blunder Detect & Solve manual 15h 2.0h 
6 | Measurement of ground control points manual 8.0h 2.0h 
7 | Simultaneous Solve and Re-measurements | manual 8.0h 7.0h 
8 | Final GPS supported block adjustment manual 3.0h 3.0h 
Total] 50.3h 26.3h 
  
  
  
  
Table 4: Processing steps for AT and elapsed time 
4. ASSESSMENT OF HATS 
The quality of the results and the efficiency of the 
whole triangulation process are dependent on the 
algorithm used for the measurements. In our 
investigations the following aspects caused problems 
for the correlation algorithm: 
e Extreme height differences in the images resp. 
block 
e Strips with different flight dates (vegetation 
changes in summer) 
e Shadows from early morning flights (bad quality 
terrain representation) - 
e Densily forested areas and lakes 
To improve HATS with respect to speed, precision, 
robustness and user friendliness we suggest the 
following software improvements which are 
summarized below: 
(1) The superimposition of the tie point pattern in 
any image/model of the whole block with the 
possibility of interactive placement of tie points in 
anticipation of problem areas which will occur in 
APM will improve the success rate of measured 
points. 
(2) The use of an existing DTM in APM speeds up 
the APM process and increases the robustness 
significantely, so that less IPM will be required 
afterwards. We have in Switzerland for example a 
hectare-raster DTM covering the whole country, 
which could be used. 
(3) The use of GPS camera station data in HATS 
should be supported more efficiently. 
(4) An ideal improvement of increasing speed after 
APM would be to measure the tie points only in their 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
two nadir images across strip direction and the 
program then is capable of measuring all other related 
combinations of this point automatically. 
(5) The implementation of an image matching 
technique (Gruen, 1985), which uses besides two shift 
parameters also two shears and scales, improves the 
precision of the measurements slightly. A small 
drawback of a slightly reduced speed should be 
neglected. 
Comparing the latest HATS with an earlier version of 
the orientation tool in SOCET Set it must be stated 
that the graphical user interface and the additional 
implementated options increased the userfriendliness 
significantly. In Kersten and Stallmann (1995) the 
SOCET Set version 2.4 was compared to an 
experimental software package and no significant 
differences were found between both software 
packages. Both systems used a semi-automatic 
measurement mode, but the image matching algorithm 
in the experimental system performed 10% more 
accuratly than the correlation algorithm of SOCET 
Set. Today, the system from Helava provides much 
more automation in data processing and the 
advantages of HATS are: 
e Automatic measurement mode using an user 
defined tie point pattern 
* Quasi online blunder detection for eliminating 
gross errors 
e Point remeasurement capability after residual 
checking 
As a drawback of the overall system it must be noted 
that the build-up of a graphical user interface or the 
redisplay of images on the extraction monitor is often 
too slow (up to 20 seconds during the measurement