Information Technology Reference
In-Depth Information
A Template Matching and Ellipse Modeling
Approach to Detecting Lane Markers
Amol Borkar, Monson Hayes, and Mark T. Smith
Georgia Institute of Technology, Atlanta, GA, USA
{amol,mhh3}@gatech.edu
Kungliga Tekniska Hogskolan, Stockholm, Sweden
{msmith}@kth.se
Abstract.
Lane detection is an important element of most driver as-
sistance applications. A new lane detection technique that is able to
withstand some of the common issues like illumination changes, surface
irregularities, scattered shadows, and presence of neighboring vehicles is
presented in this paper. At first, inverse perspective mapping and color
space conversion is performed on the input image. Then, the images are
cross-correlated with a collection of predefined templates to find can-
didate lane regions. These regions then undergo connected components
analysis, morphological operations, and elliptical projections to approx-
imate positions of the lane markers. The implementation of the Kalman
filter enables tracking lane markers on curved roads while RANSAC helps
improve estimates by eliminating outliers. Finally, a new method for cal-
culating errors between the detected lane markers and ground truth is
presented. The developed system showed good performance when tested
with real-world driving videos containing variations in illumination, road
surface, and trac conditions.
Keywords:
Lane Detection and Lane Keeping, Template Matching,
Driver Assistance Systems, Advanced Vehicle Safety Systems.
1
Introduction
Driver safety has always been an area of interest to automotive research. With
the advancement of semiconductor design, powerful electronic devices with small
footprints are starting to appear in many vehicles. These devices are capable of
performing various tasks to assist the driver of an automobile paving the way
for Driver Assistance (DA) systems.
One of the many task performed by such a DA system is Lane Departure
Warning (LDW). In LDW, the positions of lane markers around the host vehicle
are continuously monitored to determine if a lane change is imminent with the
help of exogenous inputs like steering angle, commuting speed, and rate of lane
marker movement. Consequently, a vital component of LDW is lane detection
which is described as a problem of locating painted white or yellow markings on
the road surface. In vision based lane detectors, a camera mounted under the
rear-view mirror is used to acquire data for lane detection.
Search WWH ::
Custom Search