ibul 2004
jints
jints
timate.
ve show
s image
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
Figure 11: Results (detail).
6 CONCLUSION
We have proposed an approach for pose estimation using points
and segments features for a special case : range and digital im-
age recording. Points allow to constrain geometry, filling gaps
in region of space where there are too few segments. Thus, even
though segments are geometrically very constraining, to ensure
a high quality of the geometrical determination, points and seg-
ments have both to be considered. It has been shown that very
few observations can bring reasonable results.
Segments should be used with cautiousness. Indeed, rectilinear
features are straight to a certain extent. The longer the segments
the higher the risk of reality being far from straightness hypothe-
ses. Anyway a very easy solution to this problem is to split linear
features in smaller parts.
Error propagation will enable to predict if an a priori expected ac-
curacy on pose estimation is reached. Error will be also estimated
on primitives to qualify 3D reconstruction.
This approach can be easily extended to solve 3D to 3D registra-
tion, simulating image projection from station point of view. It
has already been extended to aerial and terrestrial bundle adjust-
ment (Jung and Boldo, 2004).
Further evaluations should be carried out. First of all, we should
look to ground control primitives which do not play any role dur-
ing compensation. Besides, we will compare more precisely re-
sults provided by correlation points. We would like also to exam-
ine distortions in scanner data.
Further improvements should be done towards automation. Ap-
proximate solution could be recovered automatically by vanish-
ing points detection in the image (Van den Heuvel, 1998) and by
finding horizontal and vertical directions in points cloud.
To improve reliability of adjustment, residuals should be com-
pared to a threshold at each iteration. This threshold should be
tuned thanks to expected precision of pose estimation, in relation
with primitives variance. Through process, segments and points
would be disabled or activated, depending on their residuals.
Finally, further extensions should pay attention for spatial repar-
tition of matched primitives. Spatial repartition should take into
account the fact that points and segments have not the same geo-
metric influence on the global system.
REFERENCES
Deriche, R., 1987. Using canny's criteria to derive a recursively
implemented optimal edge detector. IJCV 1(2), pp. 167-187.
Dhome, M., Richetin, M., La Preste, J. and Rives, G., 1989. De-
termination of the attitude of 3-d objects from a single perspective
view. PAMI 11(12), pp. 1265-1278.
Fórstner, W., 2004. Uncertainty and projective geometry. In:
E. Bayro-Corrochano (ed.), Handbook of Computational Geom-
etry for Pattern Recognition, Computer Vision, Neurocomputing
and Robotics, Springer.
Habib, A., 1999. Aerial triangulation using point and linear fea-
tures. In: ISPRSGIS99, pp. 137-142.
Hanek, R., Navab, N. and Appel, M., 1999. Yet another method
for pose estimation: A probabilistic approach using points, lines
and cylinders. In: CVPR99, pp. II: 544—550.
Haralick, R., Lee, C., Ottenberg, K. and Nolle, M., 1994. Review
and analysis of solutions of the 3-point perspective pose estima-
tion problem. 1JCV 13(3), pp. 331-356.
Harris, C. and Stephens, M., 1988. A combined corner and edge
detector. In: Alvey88, pp. 147—152.
Hartley, R., 1997. In defense of the eight-point algorithm. PAMI
19(6), pp. 580—593.
Hartley, R. and Zisserman, A., 2000. Multiple view geometry in
computer vision. In: Cambridge.
Jung, F. and Boldo, D., 2004. Bundle adjustment and incidence
of linear features on the accuracy of external calibration parame-
ters. In: International Archives of Photogrammetry and Remote
Sensing.
Kumar, R. and Hanson, A., 1994. Robust methods for estimating
pose and a sensitivity analysis. CVGIP 60(3), pp. 313-342.
Taillandier, F. and Deriche, R., 2002. Reconstruction of 3D linear
primitives from multiple views for urban areas modelisation. In:
PCV'02, Vol. 34:3B, pp. 267-272.
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. 32, pp. 632-639.
Van den Heuvel, FE. A., 1999. A line-photogrammetric mathe-
matical model for the reconstruction of polyhedral objects. In:
Videometrics VI, Vol. 3641, pp. 60-71.
Wang, Z. and Jepson, A., 1994. A new closed-form solution for
absolute orientation. In: CVPR94, pp. 129-134.
