^ Part B3. Istanbul 2004
nding Masks for a pair of
1 two orthoimages by FFT
n matrix problem between
ssible homologous planes
o normals vectors of the
e system.
und axis 7 with an angle
| passing from the internal
ngent ground system (rota-
to the vehicle motion with
and vanishing points).
on passing from the ground
ce system of the 3D planes
the rotation matrix passing
| by :
d (8)
rix by :
t+dt)
atl (9)
=
Ti 3 (t--dt)
ln’ jean ll
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
So we get the desired values of rotations by computing :
A = 3 5071
Q(t)-+ (t-+dt) =. T )n — ground : 25 : ni, — ground (10)
As a result, we get the desired values of relative shift and rota-
tions that are the desired parameters of approximate solution to
initialise our photogrammetric bundle.
4 ESTIMATING ROBUST AND ACCURATE TIE POINTS AND
SEGMENTS
We have up to now estimated the approximate relative orientation
of the platform and DFMs expressed in the platform system for
each capture. This approximate relative orientation is necessary
to initialise the bundle adjustment in order to linearise the bun-
dle system. Moreover both depth and pose information provide
a very good predictor to restrict image-matching search space an
consequently find very robust matches by image correlation that
can be directly inserted as accurate tie point and segments mea-
surements in the photogrammetric bundle of stereo pairs to pro-
vide high-precision orientation parameters of the images : rota-
tion matrix giving the angular orientation of the second couple in
the first, and three translational displacements (7;, T, T.) giv-
ing the direction of the inter-couple translation.
The 3D relative position and orientation of images in object space
will be determined by a Global Multi-Cameras Bundle Adjust-
ment "icono-triangulation" that integrates measures from the
images (intra-stereo tie points and segments), measures from
GPS with their uncertainties, ground control features : horizontal
and vertical vanishing lines and accurate road marks. The mathe-
matical model applied for sequential MMS images orientation is
based on least-squares adjustment (Jung and Boldo, 2004).
5 CONCLUSION
We have presented an original way of estimating pose from im-
ages of a stereo rig in the case of planar scenes as encountered
in urban areas. Stereo rigs provide a metric and a scale in the
pose estimation problem. They also hugely decrease the robust-
ness of tie point estimation which is in general the weakness of
target tracking algorithms in image sequences. The 2D ill-posed
matching is transformed in a well 3D matching problem. We
do not need to have short baseline in the motion, our system is
wide baseline efficient thus we do not need to have such impor-
tant frame rates as in video sequences.
Stereo rigs with short baselines also provide very good sur-
face models to initialize a fine multiview surface reconstruction
scheme after a bundle adjustement of all images acquired by the
system has been performed.
REFERENCES
Baillard, C., 1997. Analyse d'images aériennes stéréo pour la
restitution 3D en milieu urbain. PhD thesis, ENST, Laboratoire
MATIS.
Bentrah, O., Paparoditis, N. and Pierrot-Deseilligny, M., 2004.
Stereopolis: An image based urban environments modeling sys-
tem. In: MMT 2004. The 4th International Symposium on Mo-
bile Mapping Technology, Kunming, China.
Deriche, R., 1987. Using canny's criteria to derive a recursively
implemented optimal edge detector. In: International Journal of
Computer Vision, Vol. Inumber 2, pp. 167-187.
Deriche, R., Vaillant, R and Faugeras, O., 1992. In theory and
applications of image analysis, chapter from noisy edges points
to 3D reconstruction of a scene : A robust approach and its un-
certainty analysis. In: World Scientific. Series in Machine Per-
ception and Artificial Intelligence, pp. 71-79.
Fischler, M. A. and Bolles, R. C., 1981. Random sample con-
sensus: A paradigm for model fitting with applications to image
analysis and automated cartography. In: Comm. Assoc. Comp.
Mach, Vol. 24number 6, pp. 381—395.
Jung, F. and Boldo, D., 2004. Bundle adjustment and incidence of
linear features on the accuracy of external calibration parameters.
In: Proc. ISPRS, Istambul, 2004, to appear.
Shufelt, J. A., 1999. Performance evaluation and analysis of
vanishing point detection techniques. In: IEEE Transactions
On Pattern Analysis and Machine Intelligence, Vol. 21 number 3,
pp. 282-288.
Taillandier, F. and Deriche, R., 2002. Reconstruction of 3D linear
primitives from multiple views for urban areas modelisation. In:
Proceedings PCV02, Vol. B.
Van den Heuvel, F. A., 1998. Vanishing point detection for archi-
tectural photogrammetry. In: H. Chikatsu and E. S. Editors (eds),
International Archives of Photogrammetry and Remote Sensing,
Vol. 32number 5, pp. 652—659.
