Hee Ju Park
£0 a ©, KE natch podras ^ j
lige an inae 3 line aus enge 1
image 1 image 2
Figure 5. Suspicious corresponding point and line relation Figure 6. Angle due to average slope in height
Instead of Xx j We can use x which is modified by angle a. The Angle c may be related with relief displacement, and
.also may not be related with relief displacement. So we define C,; as follow
C.= SO sv. a
sl] 3 x. es v. eexp(q) cl] n1]
i j
And we reject the line in which the expression cov) is less than a certain threshold, for example «0.6,
d
S. =, eexpQ)
Furthermore we reject the pairs which have less overlapping points than a threshold, for example «10.
Then we define the parameter Strength of matching SM, for line pairs as follows.
exp(-abs(dx) / 2sdv(dx C, *w(i, J; (I, k
Y p( (dx) | (ax)) (IW, XC, - Y. di et J A ))
SM; = t t[((GG, mytd(jny2 x | o ^ Add (18)
; 1 ]
N ! : #X X : ;
am] +LdG, m) +d(j.n)]/2 e ev F[(dG.h)-rd(j.k)y/2
(m,n): a pair of matched points
h, k: neighbouring possible candidates for corresponding lines near line i, j
Wp: weight for neighbouring matched point. For example Wp=7
C ,; line shape similarity coefficient between line i and line j
Cor : line shape similarity coefficient between line h and line k
wi, j;(h,k)) is like a co-operative coefficient in [Gruen et al, 1998], which is related with the probability when a
possible corresponding pair (i, j) and its neighbouring pair (h, k) exist at the same time. In this context we define it as
follows.
If the relation between i, h and j, k is reversed, yw, j; (h,k)) = -1.
And if i, h is connected and j is overlapping with both i and h, then w(i, j:(h,k))=1
And if j, k is connected and 1 is overlapping with both j, k then w(i, j;(h k))=1.
And if i, h is connected and j, k is connected then w(i, j;(h,k))=1.
exp(—abs (dx))
1+[d(i,h)+d(j,k)]/2
If we have a set of strength of matching parameters { SM 1, } for lines in the left image, then we can easily get another
Other case ,(j. RUSSES
set of strength of matching parameters { SM 2, } for lines in the right image. {SM1, } and {SM2, } are symmetric.
When we calculate sM 2,» We only need to find a symmetric value corresponding to each sy. in the list of { sm)
and then assign it.
The next step is to find the matching pair using { SM 1, } and { SM2,}- For each i in SM1, we determine the pair which
has the highest Strength of Matching value. One line can have two or more corresponding lines. So we choose all
possible pairs with highest SMs which are not overlapping each other. We apply the same process with ( $M2, )- Ifa
702 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
line
rest
pl
resu
[He
was
set |
not
use
can
was
line
In tl
solv
In tl
The
for «
One
pair
one
The
betv
wor
eval
stud
REI
Foet
Circ
Gru
Phot
Gru
Ima;
Han
Heit
Lab.
Hen:
116¢
Mas
and