![](bg8.png)
timates are introduced in the EKF, which slows down
even further to find the correct pedestrian speed.
A critical analysis concerns the number of colli-
sion predictions that the system can generate from
the moment of the pedestrian’s appearance to the
collision. As we can see in Figure 7, the number
of collision predictions from our system is consider-
ably higher than the system with the YOLO detector,
which indicates that our system has a higher chance
of predicting a collision before the collision happens.
5 CONCLUSIONS
We propose an approach to locate near and distant
pedestrians based on a multi-window detector. We
also propose a filtering strategy that has made it pos-
sible to reduce the number of false positives in our
multi-window detector. We integrated this detector to
a complete based-vision PCP system running on the
vehicle. By combining detectors with different win-
dows, we can outperform accuracy from individual
detectors and even the YOLO-based detector. We also
proposed the synthetic collision scenarios that permit-
ted evidencing quality improvements in our collision
prediction system due to higher processing rates.
We will further seek precision improvements to
pedestrian detection using the multi-window strategy
and the collision prediction assessment strategy to
support multiple pedestrians in future work.
ACKNOWLEDGEMENTS
We would like to thank Coordination for the Improve-
ment of Higher Education Personnel (CAPES) for
their financial support.
REFERENCES
Bar-Shalom, Y., Li, X. R., and Kirubarajan, T. (2004). Esti-
mation with applications to tracking and navigation:
theory algorithms and software. John Wiley & Sons.
Benenson, R., Omran, M., Hosang, J., and Schiele, B.
(2014). Ten years of pedestrian detection, what have
we learned? In European Conference on Computer
Vision, pages 613–627. Springer.
Cafiso, S., Di Graziano, A., and Pappalardo, G. (2017). In-
vehicle stereo vision system for identification of traffic
conflicts between bus and pedestrian. Journal of traf-
fic and transportation engineering (English edition),
4(1):3–13.
Dalal, N. and Triggs, B. (2005). Histograms of oriented
gradients for human detection. In 2005 IEEE com-
puter society conference on computer vision and pat-
tern recognition (CVPR’05), volume 1, pages 886–
893. IEEE.
Dosovitskiy, A., Ros, G., Codevilla, F., Lopez, A., and
Koltun, V. (2017). CARLA: An open urban driving
simulator. In Proceedings of the 1st Annual Confer-
ence on Robot Learning, pages 1–16.
Haas, R. E., Bhattacharjee, S., and M
¨
oller, D. P. (2020).
Advanced driver assistance systems. In Smart Tech-
nologies, pages 345–371. Springer.
Hartley, R. and Zisserman, A. (2003). Multiple View Geom-
etry in Computer Vision. Cambridge University Press,
New York, NY, USA, 2 edition.
Helali, A., Ameur, H., G
´
orriz, J., Ram
´
ırez, J., and Maaref,
H. (2020). Hardware implementation of real-time
pedestrian detection system. Neural Computing and
Applications, pages 1–13.
Hernandez-Juarez, D., Chac
´
on, A., Espinosa, A., V
´
azquez,
D., Moure, J. C., and L
´
opez, A. M. (2016). Embedded
real-time stereo estimation via semi-global matching
on the gpu. Procedia Computer Science, 80:143–153.
Hirschmuller, H. (2008). Stereo processing by semiglobal
matching and mutual information. IEEE Transac-
tions on Pattern Analysis and Machine Intelligence,
30(2):328–341.
Jurecki, R. S. and Sta
´
nczyk, T. L. (2014). Driver reaction
time to lateral entering pedestrian in a simulated crash
traffic situation. Transportation research part F: traf-
fic psychology and behaviour, 27:22–36.
Keller, C. G., Enzweiler, M., and Gavrila, D. M. (2011).
A new benchmark for stereo-based pedestrian detec-
tion. In 2011 IEEE Intelligent Vehicles Symposium
(IV), pages 691–696. IEEE.
Krizhevsky, A., Sutskever, I., and Hinton, G. E. (2012). Im-
agenet classification with deep convolutional neural
networks. In Advances in neural information process-
ing systems, pages 1097–1105.
Li, Y., Zheng, Y., Morys, B., Pan, S., Wang, J., and Li,
K. (2020). Threat assessment techniques in intelli-
gent vehicles: A comparative survey. IEEE Intelligent
Transportation Systems Magazine.
Nguyen, D. T., Nguyen, T. N., Kim, H., and Lee, H.-J.
(2019). A high-throughput and power-efficient fpga
implementation of yolo cnn for object detection. IEEE
Transactions on Very Large Scale Integration (VLSI)
Systems, 27(8):1861–1873.
Organization, W. H. (2018). Global status report on road
safety.
Redmon, J. (2013–2016). Darknet: Open source neural net-
works in c. http://pjreddie.com/darknet/.
Redmon, J. and Farhadi, A. (2018). Yolov3: An incremental
improvement. arXiv preprint arXiv:1804.02767.
Schneider, N. and Gavrila, D. M. (2013). Pedestrian path
prediction with recursive bayesian filters: A compara-
tive study. In German Conference on Pattern Recog-
nition, pages 174–183. Springer.
VISAPP 2021 - 16th International Conference on Computer Vision Theory and Applications
676