Abstract
Abstract:
This paper proposes and validates a real-time onroad
vehicle detection system, which uses a single camera for
the purpose of intelligent driver assistance. A three-step vehicle
detection framework is presented to detect and track the target
vehicle within an image. In the first step, probable vehicle
locations are hypothesized using pattern recognition. The vehicle
candidates are then verified in the hypothesis verification step. In
this step, lane detection is used to filter vehicle candidates that are
not within the lane region of interest. In the final step tracking
and online learning are implemented to optimize the detection
algorithm during misdetection and temporary occlusion. Good
detection performance and accuracy was observed in highway
driving environments with minimal shadows.