‘est Hungary
ce resection and in
t'y complex and the
inimally necessary
he space resection,
rientation elements
alue following the
-error and it can be
thod or the robust
tions, which needs
space resection in
1 initial values and
t a unique solution
ind the differences
of the gross error
1ethod.
ve solution we can
give approximate
accuracy which is
iterative process
netry)
rs is not so easy if
s errors. The least
ss errors to other
> robust estimators
ber of points with
| avoid the above
International Archives of the Photogrammetry, Remote Sensing
1.2 The procedure
First I give a direct solution for the space resection based on
three control points (chapter 2.1). Than I explain how to get the
adjusted values of unknowns using the Jacobian Mean Theorem
(Gleinsvik, 1967), if the number of control points is more than
three (chapter 2.2). And finally I present the basic formulas
necessary to detect the points with gross error (chapter 2.3).
2. DIRECT ANALYTICAL METHODS
2.1 Space resection without adjustment
As it is seen on Figure 1. we have a tetrahedron with a,b,c
sides. The ABC triangle is known and formed by the control
points, The 4'B'C'P tetrahedron including the ¢, B,y is also
known after the measurement of image points. The goal is to
determine the outer orientation elements (@,@,K and
Xo»Yo>Zo). It is wise to first calculate only the projection
center coordinates and after this the rotation angles can be
calculated with well known direct equations.
M D
i
A
V
e Y«
T
& nt s
pe
\ >
AC
7 \
A:
uU
n
^
-— — — n9 HÀ— Lor
f T mee f
| v
o HE
ctt
E p
Figure 1. Space resection based on three control points
Let’s derive equations for the a,b,c sides, since using these
values we can calculate the P projection center coordinates
using the well known distance equations from the coordinate
geometry.
If we take the a side as a basic distance the sides of b. and C
will differ only with an n scale factor, so in this case we have
only two unknowns ( a and 7 ).
We can setup three independent equations for the triangles
AABC , ABCP and AACP using the cosine-theorem:
and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
2 2 5
d'-a -(a-ny -2a’ncosa
(2)
e? z (a- ny. *« (a- my. — 2a? nmcos f
f =(a-m} +a —2a^mcosy
We can eliminate the side a from the equations reducing the
three equations into one forth-degree equation (Jancso, 1994):
Win' + Won’ - Wa -WnW,-0 (3)
where W, - f(d.e,f,a, fy)
After solving the equation (3) we can calculate the unknown
sides:
d
a= + FA SATT (4)
l+n° —2ncosa
b=a-n
e -f ea hl
C-mm-
2(acosy —bcos B)
Now we can calculate the projection center coordinates using
the distance equations:
Xp XV ANY) HZ. Zu
FEN HN ET 6
c =(X, made)! *(Y, 7 Val *(, -Zoy
The solution of (5) is :
I? T [3 ; [2 >
7. = -Y,E4Xv,-4u.w, SE EVEN MW LEN EYE Aw,
ep -—
= :Z,>0
Qu, 2, 2u, , (6)
AX mk -kT,
Yo mk RE,
where Hnc obe Wade k, , parameters are functions of
coordinates of the A, B,C control points and the sides of
a,b,c (Jancso, 1994).
Finally we can calculate the rotation angles of @,w,x from the
rotation matrix with the well-known direct equations (Hirvonen,
1964):
he a (Hs
Rely, 1, fs (7)
In fo Tay
and
pz -arctg(r,/r..)
Q = arcsin(r,, ) (8)
x = —arctg(n,/r,,)