International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
order to increase user’s productivity it is important to 
automatically recognize and measure the fiducial marks. In LPS 
a new module has been implemented to recognize and measure 
the fiducial marks automatically and precisely. Instead of 
creating a fiducial template database, user has only to 
approximately measure one fiducial mark at the beginning. 
Then the software will go through each image in the whole 
project and find all the fiducial marks automatically. The 
underlying algorithm is a least square template matching. 
Therefore the fiducial marks can be found very accurately. 
  
  
Figure 3: Automatic Fiducial Measuremert. (a) Manually 
measure one fiducial mark, (b) Automatically measured 
fiducial Mark with least square template matching, (c) All 
automatically measured fiducial marks in a strip. 
An example is shown in Fig. 3. By using some optimization 
procedures such as dynamic template size determination, robust 
fiducial location estimation and hierarchical search, the 
automatic fiducial measurement process is also very fast. Each 
image can be finished just in a few seconds with a normal 
desktop PC. 
3 AUTOMATIC POINT MEASUREMENT 
Automatic point measurement (APM) includes automatic 
assistance of ground control point measurement, automatic tie 
point collection over entire block or sub-block and automatic 
tie point transfer from one image or sub-block to another. The 
APM procedure of LPS includes block connection to set up the 
relationship between images, feature extraction and matching, 
and robust gross error checking and least square correlation. 
The APM algorithms are independent of sensor type; therefore 
it works not only with images from aerial cameras, but also 
with images from digital cameras, non-metric cameras and 
satellite sensors. Furthermore it is also tolerant of large scale 
and rotation variations. Therefore the LPS APM can not only 
be used to connect image blocks for triangulation, it can also be 
used for various registration purposes, e.g. automatic image 
warping. The example shown in Fig. 4 demonstrates one of 
these applications. The left image is a registered aerial 
photograph; the right image is a SPOT image without any 
orientation information. These two images have a significant 
orientation difference and a scale difference about 1: 5.5. They 
have only a small overlap. The LPS APM can find tie points for 
this application. The wer only needs to measure two tie points 
manually and approximately in order to give the software 
839 
  
Figure 4a: A registered aerial image (left) and a 
raw SPOT image (right) with 2 mannal tie noints 
  
Figure 4b: Automatically found tie points 
disnlaved on their corresnondine imaces 
nia aU art 
  
    
  
     
  
    
   
ny 
editis 
  
E nd 
Pj 
L| 
         
3 i 
" 
THAN od 
   
Figure 4c: Enlarged view of three found tie 
noints in their original seale and orientation 
some initial information about the scale and rotation, then the 
software will find many accurate tie points for these two 
images automatically. 
4 BLOCK TRIANGULATION 
LPS has a core triangulation package and an add-on 
triangulation module called ORIMA. The core triangulation can 
handle block triangulation for frame camera geomdry, orbital 
pushbroom geometry and generic sensor geometry. Self- 
calibration and robust gross error detection are available in both 
packages. ORIMA is an extensive triangulation package with 
stereo point measurement, rigorous GPS/IMU support, 
advanced error handling and rich residual analysis tools. 
  
Figure 5: ORIMA stereo point measurement interface 
and oranhic noint and imaoe connection disnlav