![](bg6.png)
Table 1: RMSE: single vehicles.
scenario x
w
[m] y
w
[m] d
w
[m]
scenario 1 0.68 4.18 4.28
scenario 2 3.72 0.98 4.05
overall performance 1.89 2.90 4.19
Table 2: RMSE: fusion of two vehicles.
Method x
w
[m] y
w
[m] d
w
[m]
scenario 1
CF 0.35 0.55 0.69
CI 0.35 0.55 0.69
CU 2.87 2.67 3.94
scenario 2
CF 0.15 1.00 1.01
CI 0.14 1.07 1.08
CU 0.15 3.72 3.73
overall performance on both scenarios
CF 0.25 0.78 0.85
CI 0.24 0.81 0.89
CU 1.51 3.19 3.83
second scenario the CF outperforms the other algo-
rithms. In contrast to the results presented in (Matzka
and Altendorfer, 2008) the CU has the worst perfor-
mance in all scenarios. Based on these results it is
proposed to use the CF as a general fusion method,
since the algorithm delivers precise results in all sce-
narios. One could imagine a combination of CF and
CI to get the best results in all situations. It is not sur-
prising that the CF and CI deliver similar good results
as long as the position vectors r
w
of the detections are
relatively accurate. For each vehicle the lateral po-
sitions of the detections are very accurate. Thus one
could get a fused result of two cameras by calculating
the intersection of the two rays on which the detec-
tions are located. This is similar to stereo vision. The
fused position vector of CF and CI is close to the re-
sult that would be obtained by this ray intersection.
The main difference of CI and CU is determined by
the fused covariance.
This leads to the scenario where CU could out-
perform CF and CI. Inconsistent states are obtained if
lateral position errors occur due to erroneous data of
the detection algorithm or an erroneous vehicle po-
sition. These states can then be handled by a CU
algorithm. It would make sense to include an algo-
rithm that can detect where inconsistent states occur
and then change the fusion algorithm to CU.
ACKNOWLEDGEMENTS
The developed system is part of the Active Safety Car
project which is co-funded by the European Union
and the federal state NRW.
REFERENCES
Bar-Shalom, Y. and Blair, W. D. (2000). Multitarget-
Multisensor Tracking: Applications and Advances.
Artech House Inc, Norwood, USA.
Hartley, R. and Zisserman, A. (2003). Multi View Geometry
in Computer Vision (2nd Edition). Cambridge Univer-
sity Press, Cambridge, United Kingdom.
Kaempchen, N. (2007). Feature-level fusion of laser scan-
ner and video data for advanced driver assistance sys-
tems. Technical report, Fakultaet fuer Ingenieurwis-
senschaften und Informatik, Universitaet Ulm.
Matzka, S. and Altendorfer, R. (2008). A comparison of
track-to-track fusion algorithms for automotive sensor
fusion. In Proc. of International Conference on Mul-
tisensor and Integration for Intelligent Systems , 2008
IEEE, pages 189–194.
Merwe, R. V. D. and Wan, E. (2003). Sigma-point kalman
filters for probabilistic inference in dynamic state-
space models. In In Proceedings of the Workshop on
Advances in Machine Learning.
Meuter, M., Iurgel, U., Park, S.-B., and Kummert, A.
(2008). The unscented kalman filter for pedestrian
tracking from a moving host. In Proc. of Intelligent
Vehicles Symposium, 2008 IEEE, pages 37–42.
Nunn, C., Kummert, A., Muller, D., Meuter, M., and
Muller-Schneiders, S. (2009). An improved adaboost
learning scheme using lda features for object recogni-
tion. In Proc. of Intelligent Transportation Systems,
2009. ITSC ’09. 12th International IEEE Conference
on, pages 1–6.
Ponsa, D. and Lopez, A. (2007). Vehicle trajectory es-
timation based on monocular vision. In Marti, J.,
Benedi, J., Mendonca, A., and Serrat, J., editors, Pat-
tern Recognition and Image Analysis, volume 4477 of
Lecture Notes in Computer Science, pages 587–594.
Springer Berlin / Heidelberg.
Ponsa, D., Lopez, A., Lumbreras, F., Serrat, J., and Graf,
T. (2005). 3d vehicle sensor based on monocular vi-
sion. In Proceedings of the 8th International IEEE
Conference on Intelligent Transportation Systems, Vi-
enna, Austria.
Smith, D. and Singh, S. (2006). Approaches to multisen-
sor data fusion in target tracking: A survey. Knowl-
edge and Data Engineering, IEEE Transactions on,
18(12):1696 –1710.
Smith, R. C. and Cheeseman, P. (1986). On the representa-
tion and estimation of spatial uncertainly. Int. J. Rob.
Res., 5(4):56–68.
VISAPP 2011 - International Conference on Computer Vision Theory and Applications
312