an image straight 
ces in the object 
node to some leaf 
ble mapping. Then, 
‘ect mapping. Each 
tisfy the following 
raint, the rigidity 
ional distance and, 
an image straight 
e are homologous, 
id to a unique 
other words, each 
is used in some 
search space in the 
ration reduces the 
gidity constraint is 
el relating straight 
jbject spaces. One 
| image and object 
mmaselli and Tozzi 
rvations are straight 
ization process. The 
igm called matching 
om another similar 
ating (Faugeras and 
ght feature, the aim 
| number of object 
e matching process. 
rastically reduced. 
elational Distance: 
| at this step. The 
plied if the node of 
fies the uniqueness 
ormalised relational 
e similarity between 
fined in the range [0; 
| from unit and label 
s. In an ideal case, if 
is zero, the node 
patible. In practical 
:ssary a threshold. 
osis is based on 
e need of redundant 
tment, only after the 
asible the application 
would be to apply a 
g. Kalman filtering 
At this moment, only 
nted. 
revious sections was 
n of the method, an 
simulated. Random 
‘he spatial view of the 
/ed in figure 2. 
table 1. One image 
na 1996 
  
41 
  
2 
  
  
12 
10 
  
  
3 
  
  
4 
Figure 2 - Spatial view of the simulated straight features. 
space straight feature is defined by angular (a) and linear 
(b) parameters. As already mentioned, random errors 
were introduced in these parameters. The standard 
deviations of a and b parameters are O , and O,, 
respectively. One object space straight feature is defined 
by a fixed point (Xi, Y1, Z1) and a normalised vector (Vx, 
Vy, Vz). The latest line of table 1 shows the elements of 
exterior orientation (KX, @, @, Xo, Yo, Zo) and the focal 
length used to simulate the image space straight 
features. 
A summary of obtained results is presented in table 2. 
This table shows that all correspondences were found, 
which can be denoted by f= {(lo, Mo), 
«4, (l12, M12)}. As 
expected, the normalised relational distances (GDN) are 
zero. This is because the image and object space 
relational descriptions do not have any discrepancy. The 
Table 1 - Simulated data. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
No Image Straight Features (I) Object Straight Features (M) 
a, b c, oy X1 Y1 Z1 Vx Vy Vz 
(x107) (mm) x10% | (x10°m) (m) (m) (m) 
0 -33 160.72 4 112 500 3500 100 1 0 0 
1 -38 -160.74 17 225 500 2000 100 0 -1 0 
2 -40 -160.70 157 2921 500 500 100 1 0 0 
3 276 -160.71 39 224 2000 500 100 1 0 0 
4 178 160.72 157 2920 3500 500 100 0 1 0 
5 187 160.71 17 125 3500 1500 100 0 1 0 
6 -188 107.18 157 2923 3000 3000 100 1 0 0 
7 18 107.16 10 140 500 3000 100 1 0 0 
8 -36 -107.13 6 116 1000 500 100 0 1 0 
9 105 -107.13 10 112 1000 1000 100 1 0 0 
10 -230 107.11 17 225 3000 500 100 0 1 0 
11 106 107.14 39 562 3000 2500 100 0 1 0 
12 -46 53.59 6 117 1000 2500 100 1 0 0 
  
  
  
  
  
  
  
  
  
  
  
  
K=0°; $ 0*; (0-0*; X9-2000m; Y,-2000m; Z;-1500m; focal distance - 150mm 
Üimage straight features represented by x=a*.y + b*, because of indeterminations in the representation y= a.x + b, when the 
feature is closely vertical. 
  
ht Features 
lh 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
  
Features (I 
Va(x1 
-12 
27 
147 
-169 
-229 
-238 
168 
-38 
19 
-23 
185 
-151 
47 
x1 
Table 2 - Matchi 
results. 
    
mm 
8 
11 
3 
10 
BD 
8 
-28 
A 
-10 
A 
27 
2 
-8 
K 6"; $ 9-15"; (07 4"; X7 1999.891m; Y.- 1999.835m; Z;7 1499.961m 
0,26, 0,-9' 0,=9, 0, =0119m; 0, =0.110m; 0, =0.038m 
0? 21; 67 - 0.808785 
133 
Object Straight 
Features (M 
Mo 
M4 
Ma 
Ms 
Me 
M; 
Ms 
M1 
M 
M4 
© 
D 
z 
OJ|JO|O|JO|O|/OJOJO/OJOJOJOIO