dc.description.abstract | Developing real-time automotive driver assistance systems to alert
drivers about driving environments and possible collision with other vehicles
has attracted much attention lately. In order to achieve such a system, we use
a camera mounted on a vehicle to capture road scenes for lane and preceding
vehicle detection.
The proposed system consists of twelve functions: lane detection,
multiple lane estimation, classification of solid/dashed lane marks,
classification of yellow/white lane marks, classification of single/double lane
marks, vehicle direction estimation, vehicle lateral offset estimation, lane
departure warning, preceding vehicles detection, distance estimation for
preceding vehicles, brake light detection, and turn signal detection.
In this thesis, the lane marks are detected by searching the best-fitted
parameters of a defined lane model on the image, and we verify the results by
the width of lane marks and the difference of the slopes and the intercepts in
consecutive image frames. Preceding vehicle detection is divided into two
independent parts based on daytime and bad weather conditions. In the
daytime condition, we utilize cast shadow, left/right borders of a vehicle to
detect preceding vehicles and use ratio of the road width and vehicle width,
symmetry, and variance to verify vehicle region. In bad weather condition
such as rainy day, we generate several templates based on the tires and the
bottom of vehicles. According to template matching, we can find out the
vehicle candidates and then use ratio of the road width and vehicle width,
symmetry, and variance to verify.
In the experiments, the proposed methods are evaluated on several
different weather conditions such as sunny day, misty day, dusky day, cloudy
day, rainy day, and night. The average vehicle detected rate is 94.5 % and
average processing time for executing all twelve functions is about 0.045
second run on a general PC with Intel® Pentium® D 3.0GHz CPU. From the
experimental results, we find that the proposed approaches can stably detect
or track the lanes and preceding vehicles in real time. | en_US |