MOR will 
hoto pro- 
espect to 
ages, the 
matching 
s. In sec- 
matching 
ions via 
| image 
image 
Fig. 4 demonstrates the configuration of 8 control points. 
The circles mark the location of the control points with re- 
spect to the image coordinate system (x,y) . The radius of 
the circles depends on the empirical sensitivity 8o;. Point 5, 
falsely located by the clustering procedure due to weak mo- 
del information, has been correctly eliminated by the robust 
estimation. The other points except point 1 ly near a straight 
line. Therefore point 1 is neccesary to stabilize this configura- 
tion. This is automatically found by the sensitivity analysis. 
Fig. 5 shows a configuration of 5 control points. Point 3, fal- 
sely located by the clustering procedure due to weak image 
information, has been correctly eliminated by the robust esti- 
mation. The other points ly in a bad geometric configuration. 
The point 4 near to the principle point, causes any 3 of the 4 
points to make the projection centre ly on the critical cylin- 
der. This is automatically found by the sensitivity analysis. 
Example 2: 
An other application of the matching procedures: 
  
Fig. 6 Roughly and manualy digitized object 
  
Fig. 7 Result of the matching procedure 
This is an example for using the matching procedure for se- 
miautomatically mapping. The object being of interest can 
be quickly and roughly digitized using a mouse cursor (cf. 
Fig. 6) which than is exactly fitted to the image (Fig. 7) 
using the matching procedure described in this paper. The 
observation equation for this approach are given by simply 
equating the end points of the extracted edges to the corre- 
sponding roughly digitized model edge according to eq. 12, 
but without using any projective function. The model points 
are treated as estimates. As the straight line extraction could 
run while the operator is digitizing, the fitting of the model 
to the image for this example takes less than 1 second on a 
SUN Sparc 2. 
Example 3 : 
Result of a control point location 
Image scale: 1 : 12 000, Pixelsize: 25 um = 30 cm. 
e 
     
Fig. 8 Result of the control point location 
ES. 4.1 c4 3 — zx 
SERA STE MS 
a I 
AUN A NE us dre A 
= NS me hn T. 
= ed ps S - - t Fah TM, -— 
ry NEA 77 és des LT 
Ue TON x A. 1 ~ uS à ‘ 
Pa ZE = RT 
— Y ~~ Lai hon — 
PU LZ zi ja | eA > E - 
rt I ou, Zo 
(^7. MEME SU at 
AS AE Y mI Fi A) Yi Nr 
a 
IAEA m n 7 ie — 
> > iv — a ~~ 
Le Vd. zs rE T | e 
cer rU t. per Ee d 
Fig. 9 Extracted straight line segments 
iz 
Zw 
# x 
A 
Sud 
hod 
s yt 
Xon 
Fig. 10 Cleaned set of matching candidates 
597