Intermodal Mobility Assistance for Megacities - Smart Information Systems: Computational Intelligence for Real-Life Applications - page 360

flow analysis model on the UCSD dataset, and noted a performance of 94.90 %
average accuracy over four trials. This is the demonstration that our holistic video
analysis approach is very promising.
We plan to extend the current video feature set based on dense optical flow vectors
by integrating more sophisticated motion descriptors such as vehicle crowd density
using background subtraction and/or optical flow methods, and trajectory-based
features by tracking sampled points in the optical flow fields.
The task of finding the best possible way through the city with a focus on integrat-
ing the gathered traffic information has been addressed by the PlainRoutingBean .
This component works with the A*-algorithm on a Digraph representing the streets
of Berlin. The response time of this component (i.e., the time required by the Plain-
RoutingBean to receive a request, calculate a route, and return the result) always
lies below 400 ms, and between 50 and 100 ms in most of the cases. This good per-
formance is achieved at the expense of considerable memory consumption. It takes
around 3 Gbyte of RAM in order to run smooth. Reducing this memory consumption
is one possible suggestion for future optimization. We also expect that the response
time can be reduced by several milliseconds. In the current implementation, only
little effort has been put into performance optimization, as the attention was centered
on the development of the functionality itself. This could also be addressed through
future improvements.
The RealtimeTrafficBean currently supports live traffic data only. A mechanism
for storing and loading traffic situations is planned. This enables the components to
plan future routes with a realistic view on the traffic situation for the given time.
In addition, integrating more interfaces for different service providers into this bean
(e.g., real-time data from cars) is another possibility.
Acknowledgments This work is funded by the Federal Ministry of Education and Research
( BMBF ) under funding reference number 01IS12049.
References
1. A. Albiol, J.M. Mossi, Video-based traffic queue length estimation, in 2011 IEEE International
Conference on Computer Vision Workshops ( ICCV Workshops ), (IEEE, 2011), pp. 1928-1932
2. M.J. Black, P. Anandan, The Robust estimation of multiple motions: parametric and piecewise-
smooth flow fields. Comput. Vis. Image Underst. 63 (1), 75-104 (1996)
3. N. Buch, J. Orwell, S.A. Velastin, Detection and classification of vehicles for urban traffic
scenes, in 5th International Conference on Visual Information Engineering, 2008. VIE 2008 ,
(IET, 2008), pp. 182-187
4. N. Buch, S. A Velastin, J. Orwell, A review of computer vision techniques for the analysis of
urban traffic. IEEE Trans. Intell. Transp. Syst. 12 (3), 920-939 (2011)
5. A.B. Chan, N. Vasconcelos, Probabilistic kernels for the classification of auto-regressive visual
processes, in IEEE Computer Society Conference on Computer Vision and Pattern Recognition,
CVPR 2005 , vol. 1 (IEEE, 2005), pp. 846-851
6. K.G. Derpanis, R.P. Wildes, Classification of traffic video based on a spatiotemporal orientation
analysis, in 2011 IEEE Workshop on Applications of Computer Vision ( WACV ) (IEEE, 2011),
pp. 606-613