road markings, when some other object is mistaken
for a road marking, a real road marking goes unde-
tected. Our proposed method does not detect strong
non-lane edges for two reasons. Firstly, estimating
the VP allows everything above the horizon to be ig-
nored, and so other vehicles or objects on the roadside
have much less influence. Secondly, instead of sim-
ply detecting edge lines, our method requires them to
be correctly paired, and so edges which do not corre-
spond to road markings are not detected.
Figure 10: Lane markings being tracked through occlusion.
In general, our method shows high detection ac-
curacy and good resistance to visual clutter. As Fig.
8 demonstrates, McDonald’s method copes less well
with such interference. In these respects, we fulfil the
first two of the criteria outlined in Section 1. Since we
track multiple road markings, our method also satis-
fies the third criteria. Fig. 10 shows tracking of lane
markings through occlusion by nearby vehicles.
We performed a quantitativeassessment of our VP
tracking by first manually determining the VP in 200
consecutive frames. Then we measured the distance
between the proposed method’s VP location and the
ground truth to have a mean of 2.0 pixels with a vari-
ance of 2.7. With the KF turned off, the mean was 4.4
pixels with a variance of 15.7.
Using a single thread on an Intel Core i5-660 CPU
at 3.33 GHz, our system can achieve an average per-
formance of 64 fps. The proposed method involves a
few parameters, for example, how long a road mark-
ing is tracked for before it is considered to be a lane
boundary, and how long one goes undetected before
it is dropped by the system. The values for all our
parameters have been empirically determined and re-
mained constant in all the experiments.
7 CONCLUSIONS
We described a novel variant of the HT for the detec-
tion and tracking of linear road markings, with several
novel refinements to the classical HT method, includ-
ing the use of two accumulators and the translation of
image coordinates to put their origin at the VP to fa-
cilitate more efficient analysis, such as ignoring road
markings which are not parallel to the lane. The VP
is also detected and tracked, and feeds back into the
HT process for subsequent frames. The method ex-
hibits real-time performance with spare capacity for
additional tasks.
ACKNOWLEDGEMENTS
The authors would like to thank Jaguar Land Rover
for their support and supply of videos.
REFERENCES
Cantoni, V., Lombardi, L., Porta, M., and Sicard, N. (2001).
Vanishing point detection: Representation analysis and
new approaches. In Proc. of ICIAP.
D’Cruz, C. and Jia Zou, J. (2007). Lane detection for driver
assistance and intelligent vehicle applications. In ISCIT
2007, pp. 1291–1296.
Godthelp, H., Milgram, P., and Blaauw, G. (1984). The de-
velopment of a time-related measure to describe driving
strategy. Human Factors: Journal of HFES.
Horne, J. A. and Reyner, L. A. (1995). Sleep related vehicle
accidents. BMJ, 310(6979):565–567.
Ieng, S.-S., Vrignon, J., Gruyer, D., and Aubert, D. (2005).
A new multi-lanes detection using multi-camera for ro-
bust vehicle location. In Proc. IEEE IV, pp. 700–705.
J¨ahne, B., Scharr, H., and K¨orkel, S. (1999). Handbook of
Computer Vision and Applications., pp. 125–152.
Jung, C. R. and Kelber, C. R. (2005). Lane following and
lane departure using a linear-parabolic model. Image and
Vision Computing, 23(13):1192–1202.
Lipski, C., Scholz, B., Berger, K., Linz, C., Stich, T., and
Magnor, M. (2008). A fast & robust approach to lane
marking detection & lane tracking. In SSIAI, pp. 57–60.
Loose, H. and Franke, U. (2010). B-spline-based road
model for 3d lane recognition. In Proc 13th Intelligent
Transportation Systems, pp. 91–98.
McDonald, J. (2001). Application of the HT to lane detec-
tion & following on high speed roads. In ISSC.
Mobileye
®
N.V. (2010). Lane Departure Warning.
http://www.mobileye.com.
NCSDR/NHTSA Expert Panel on Driver Fatigue and
Sleepiness (1998). Drowsy driving and automobile
crashes, Report No. DOT HS 808 707.
Rimini-Doering, M., Altmueller, T., Ladstaetter, U., and
Rossmeier, M. (2005). Effects of lane departure warning
on drowsy drivers performance and state in a simulator.
In 3rd Int. Driving Symp. on Human Factors in Driver
Assessment, Training, & Vehicle Design, pp. 88–95.
Voisin, V., Avila, M., Emile, B., Begot, S., and Bardet, J.-
C. (2005). Road markings detection and tracking using
Hough transform and Kalman filter. In Advanced Con-
cepts for Intelligent Vision Systems, pp. 76–83.
Wu, B.-F., Lin, C.-T., and Chen, Y.-L. (2009). Dynamic cal-
ibration and occlusion handling algorithms for lane track-
ing. In IEEE TIE, vol. 56, pp. 1757–1773.
ICPRAM 2012 - International Conference on Pattern Recognition Applications and Methods
536