Ying Chen 
  
x 1 1 1 
a = Sa. * Ax, fr. ep mir. ra, | 
i 1 
Hig = Ye. * Ax, X ria) ; (8) 
& 1 
MES M, (6. tAy ya i) 
izl J 
  
Where, Al,Axand Ay could be soluted by chain code; x and y indicate centralized coordinates. 
The scale moment invariant, which has been normalized, can be expressed as: 
Hp 
Toa 7, pq 
Hoo 
Le. 
Hp 
Then the following scale and rotation invariant line moments can be derived to be, 
À = ho + My | 
9, 7 (7b, MY +47; 
Wei Wen and A.Lozzi[1993] also presented two important invariant line moments. They could be used as matching 
primitives. Considering scale invariance and change them to: 
Jb 
+ ——— 
2 
J^ 
2 
On ax 
min 
a 
2 
ó 
2 ©) 
2 MATCHING AND CONSTRAINTS 
In order to obtain a correct matching between different sensor images (reference image b and real-time image r), 
following vector spatial distances would be selected. And the measurement criterion is that the total of these vector 
spatial distances is minimum. 
  
  
  
  
> D, (i,j)< MIN (10) 
k=] 
lp G0) — $7 (| 
DG Ds ; 
i pas ai 
Gran BD DU] 
D 2 (4, ) = f 
m ple + 42 
ur Ol uli) = dou 67) 
D. (i, ) = 
p-b: 0s $.I.CD 
  
(11) 
where: Parameters i and j indicate the ith and the jth edge in image b and r respectively. 
  
180 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.