k-means 32 12
k-means 32 8
k-means 10 12
k-means 16 12
k-means 16 8
kd-tree 500
k-means 10 8 -
kd-tree 150 ^ k-means
kd-tree 50 |. 8& kd-tree
552 456.8 0
(a) comparison of build times for 9000 features in minutes
100 r— 450
oh |
I
=. 80 1
iS 1100 E
5 ur
= 60 | =
8 bus
3 150 §
«s 40 2
s ——328]|'
20 0
200 400 600 800 1,000
# of recursively searched leaves
(c) hierachical trees with k-means clustering
100
= E
Ta ©
> 9
9 1 60 =
5 5
8 =
"s 540 $
20 _
800
# of recursively searched leaves
(b) kd-tree with randomized trees
100 o 150
A &-— 10 12 ae 2 |
_. 80f{—r— 1612 ntl | =z
E = 3212 ONE 1100 .&
> / E
S 60 1 =
5 w7 oret. =
= , a ET 90 8
40 = Se” 9
&^
|
20 i i =i)
200 400 600 800 1,000
# of recursively searched leaves
(d) hierachical trees with k-means clustering
Figure 8: Search precision vs. speed (dashed lines) for 9 different trees (number of features for all plots: 9000). Angle noise was set
to Tangle = 1.5? , position noise to opos = 0.05m. The accuracy is related to the precision at this particular noise level for the linear
search. Parameters: k-means branching factor and number of iterations, kd-tree number of trees.
Figure 9: All three fisheye images overlaid with model edges after correct self-localization with the real image sequence. Complete
sequence see http: //www2.ipf.kit.edu/Projekte/3DTracks/.
Micusik, B., 2004. Two-view geometry of omnidirectional cameras. PhD
thesis, Czech Technical University.
Micu&ik, B. and Pajdla, T., 2006. Structure from motion with wide cir-
cular field of view cameras. Pattern Analysis and Machine Intelligence,
IEEE Transactions on 28(7), pp. 1135-1149.
Muja, M. and Lowe, D. G., 2009. Fast approximate nearest neighbors
with automatic algorithm configuration. In: International Conference
on Computer Vision Theory and Application (VISSAPP'09), INSTICC
Press, pp. 331-340.
Ottlik, A. and Nagel, H.-H., 2007. Initialization of model-based vehi-
cle tracking in video sequences of inner-city intersections. International
Journal of Computer Vision 80(2), pp. 211-225.
Puig, L., Bermudez, J., Sturm, P. and Guerrero, J., 2012. Calibration of
omnidirectional cameras in practice: A comparison of methods. Com-
puter Vision and Image Understanding 116(1), pp. 120 — 137.
Reitmayr, G. and Drummond, T., 2006. Going out: Robust model-based
tracking for outdoor augmented reality. In: Proceedings of the Interna-
tional Symposium on Mixed and Augmented Reality (ISMAR).
Scaramuzza, D., Martinelli, A. and Siegwart, R., 2006. A flexible tech-
nique for accurate omnidirectional camera calibration and structure from
motion. In: Computer Vision Systems, 2006 ICVS'06. IEEE Interna-
tional Conference on, IEEE, pp. 45-45.
Simon, G., 2011. Tracking-by-synthesis using point features and pyra-
midal blurring. In: International Symposium on Mixed and Augmented
Reality (ISMAR), pp. 85-92.
Steger, C., 2002. Occlusion, clutter, and illumination invariant object
recognition. International Archives of Photogrammetry Remote Sensing
and Spatial Information Sciences 34(3/A), pp. 345-350.
Svoboda, T., 1998. Motion estimation using central panoramic cameras.
Ulrich, M., Wiedemann, C. and Steger, C., 2009. Cad-based recognition
of 3d objects in monocular images. In: IEEE International Conference on
Robotics and Automation, Kobe, Japan, pp. 1191-1198.
Weinmann, M., Wursthorn, S. and Jutzi, B., 2011. Semi-automatic image-
based co-registration of range imaging data with different characteris-
tics. In: PIA11 - Photogrammetric Image Analysis, Vol. 38 Part 3/W22,
pp. 119-124.
Wuest, H., Vial, F. and Stricker, D., 2005. Adaptive line tracking with
multiple hypotheses for augmented reality. In: Fourth IEEE and ACM
International Symposium on Mixed and Augmented Reality (IBMAR).
Wuest, H., Wientapper, F. and Stricker, D., 2007. Adaptable model-
based tracking using analysis-by-synthesis techniques. In: 12th Interna-
tional Conference on Computer Analysis of Images and Patterns (CAIP),
LNCS, Vol. 4673, Springer Berlin/Heidelberg, Wien, pp. 20-27.
48
