solve the real problem. Here one observes from (4.7) and (4.9), that a good
artificial approximation of the variance covariance matrix of the constrained
coordinates is achieved by:
-1
I 0 e aH 0 f T HH 447
1 (4.10)
ett I | 0 c-2 "ll g I
in which all matrices H are composed using c = 1.
(4.10) is elaborated into:
/ 11 12
e M S Un
k k 1 J
21 -22 21. 11 12: ^
C iH C 2 4 1 Fi: f H
aX ¥ Xk (4.11)
( 11 12
Ey S
k k
21 22 -22
eH e ah + Le - S a
L k k k
So the artificial variance covariance matrix for a constrained block implies an
extension of the artificial matrix of the known points to the new points X? as well
(using the parameter value c j from the known coordinates) and in addition the
-22
x
orthoganalized matrix (c 2 K c 1 H to the new points x2,
X X (cel)
k
The extension of the matrix in case of more densification steps (for instance
11 11
G, itself has already a structure such as in (4.11) in stead of c 1 ), has
been proven to be possible and results in: X
( 11 12 13
C iH C 1 C i"
X X X
21 22 -22 23 -23
C 1 C a mic 2. C 12H C 1 *(o a! C 12H
x x X X X x X
31 32 32 33 -31 =33
H H - H H - H - H
e C 1 le à C 1° e «e. )? +48 3 2)
X X X x x x X X x
(4.12)
- 342 -