The responsibility of a vision-based lane departure warning (LDW) system is
to alert a driver of an unintended lane departure. Because these systems
solely rely on the vision sensor's ability to detect the lane markings on the
roadway, these systems are extremely sensitive to the roadway conditions.
When a vehicle's LDW system fails to detect lane markers on the roadway, it
loses its ability to alert the driver of an unintended lane departure. The goal of
this research is to use GPS combined with inertial sensors and a high accuracy
map to assist a vision-based LDW system. GPS navigation systems
are available in many automobiles, along with automotive-grade inertial
sensors.
The low accuracy of a typical GPS receiver found in an automotive navigation
system is largely attributed to a position error. This error is too large to allow
the GPS receiver to locate a vehicle in a particular lane on a roadway.
A method to measure this error using a vision-based LDW system, together
with a high-accuracy map, is presented in this paper. With the error known,
the accuracy of the GPS receiver is increased to a high-enough level to
localize the vehicle on a particular lane. Next, a method fusing GPS/inertial
navigation sensor/vision and a high-accuracy map for highway lane tracking is
presented.
This method provides a backup lateral offset measurement that can be used
for LDW when the LDW vision system loses track of the lane markings.
