bul 2004
and the
en suc-
dle any
conical
lection
ces two
ets are:
couple
lersects
nes T.
ect be-
1zation
is com-
1e mid-
can be
nd 77^);
line D,
P and
defined
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004
e two line segments 51 and S5 (resp. 51 and 55); The
intersection of both interpretation planes P; and P:
defines a line D; that should lie within 7. Because of
numerical errors, D; and D» are not coplanar. Here
again, a mean-squares estimation of 7l is computed
from the selected segments.
In the computed solution R 4, vector Z. is the normal vec-
tor of 7L, P. — £40 70, and Y, = norm(CP, A Z,).
Figure 6: Coplanar features geometry.
The selected features may be some surface details that do
not necessarily belong to the 3-D model. The operator is
not required to explicitly indicate which set of features he
will select. This can easily be guessed afterwards. In the
same order of idea, the features are matched automatically.
For the three-points set, all the possible combinations are
tested. The distance between the interpretation lines of
each couple is accumulated and the lightest combination
is selected. For the two-segments set, an equivalent test is
performed to select the combination that provides the near-
est lines to coplanarity.
Figure 7: Coplanar features example.
Figure 7 example displays the selected features (a) and the
matched object on the computed solution (b). Then suc-
cessive interactive actions are:
- 1) translation within 7T (ec),
- 2) rotation around Z, (d).
3.6 Parallel lines configuration
This configuration relies on the selection of the projection
in a single image of two lines assumed to be parallel in the
207
3-D space. It provides a direction which is used to facili-
tate the registration task. Only two rotation parameters are
fixed.
Figure 8: Parallel lines geometry.
The selected segments S4 and S» define two lines which
intersect at point P,,. In projective geometry, this point
Py, is known as a vanishing point where all lines with a
common direction meet. This direction is given by vector
GP... The computed solution is :
2, norm(CPæ)
Ra : P, = Pr
—
= norm(CP, AZ)
I
Figure 9: Parallel lines example.
Figure 9 example displays the selected features (a) and the
matched object on the computed solution (b). Then suc-
cessive interactive actions are:
- 1) parallel translation to the image plane 7 (c),
- 2) rotation around Z, (d),
- 3) translation along the interpretation line of P, (e).