![](bg1.png)
A CAMERA AUTO-CALIBRATION ALGORITHM FOR REALTIME
ROAD TRAFFIC ANALYSIS
Juan Carlos Tocino Diaz, Quentin Houben, Jacek Czyz, Olivier Debeir and Nadine Warz
´
ee
LISA, Universte Libre de Bruxelles, Avenue Franklin Roosevelt 50 CP165/57, Brussels, Belgium
Keywords:
Camera calibration, Road lane markings, Visual traffic analysis.
Abstract:
This paper presents a new mono-camera system for traffic surveillance. It uses an original algorithm to ob-
tain automatically a calibration pattern from road lane markings. Movement detection is done with a Σ − ∆
background estimation which is a non linear method of background substraction based on comparison and
elementary increment/decrement. Foreground and calibration data obtained allow to determine vehicles speed
in an efficient manner. Finally, a new method to estimate the height of vehicles is presented.
1 INTRODUCTION
Road traffic is increasing each year. Understanding
its characteristics is very helpful to motorway admin-
istrators to cope with this growth, and achieve regula-
tions. Among traffic characteristics, user behaviours
and classes of vehicles are the most relevant.
Until a few years ago, the main measurement tool
for traffic analysis was the magnetic inductive loop.
That kind of sensors has serious drawbacks: it is ex-
pensive to install and maintain. Indeed, it needs to be
placed inside the road, provoking traffic disruption.
Furthermore, it is unable to detect slow or stationary
vehicles, being not accurate for stop and go situations.
On the other side, video sensing is a good solu-
tion since it is inexpensive, easy to install and able
to cover a wide area. Furthemore, it has little traffic
disruption during installation or maintenance. Finally,
video analysis allows monitoring many variables such
as traffic speed, vehicle count, vehicle class and road
state.
Most of the existing video solutions are based on
mono-camera systems. A state of the art can be found
in (Kastrinaki et al., 2003). Among all the methods,
background methods are the most used since they de-
mand small computational power and are simple to
program. In such method, a residual image is ob-
tained from substracting the background from the cur-
rent image (jun Tan et al., 2007). Other solutions
use tracked features, see (Dickinson et al., 1989) and
(Hogg et al., 1984).
In this work, a new mono-camera method for traf-
fic analysis is presented. It uses an original algorithm
to obtain automatically a calibration pattern from road
lane markings. Movement detection is done with
a Σ − ∆ background estimation. It is a non linear
method of background substraction based on compar-
ison and elementary increment/decrement (Manzan-
era, 2008). Foreground and calibration data obtained
allow to determine vehicles speed in an efficient man-
ner. Finally, a new method to estimate the height of
vehicles is presented.
This paper is organized as follows: section 2
presents the auto-calibration algorithm. Methods
used to estimate vehicle characteristics are discussed
in section 3. Experiment results are depicted in sec-
tion 4 and finaly, section 5 ends this paper with the
conclusion.
2 CAMERA
AUTO-CALIBRATION
Road traffic analysis needs a calibrated camera. How-
ever, a camera calibration is performed by observing
a calibration object whose geometry in 3D-space is
known with good precision. In order to avoid the use
of special object of known geometry such as chess-
board, which implies to stop road traffic during op-
eration, the adopted solution is to form a calibration
pattern from the road lane markings. To do that, two
parameters are necessary: the lane length and width.
Based on these values, the algorithm presented in this
626
Carlos Tocino Diaz J., Houben Q., Czyz J., Debeir O. and Warzée N. (2009).
A CAMERA AUTO-CALIBRATION ALGORITHM FOR REALTIME ROAD TRAFFIC ANALYSIS.
In Proceedings of the Fourth International Conference on Computer Vision Theory and Applications, pages 626-631
DOI: 10.5220/0001803906260631
Copyright
c
SciTePress