LIDAR-based vehicle tracking for a virtual mirror

摘要

Low visibility conditions and blind zones induce hazards to driving and have been shown to be a cause of many automobile crashes every year. In order to reduce these effects, the virtual mirror, a computer-generated display that uses a high-accuracy DGPS system coupled with an onboard geo-spatial database, was used to render the view that the driver would see in a mirror-like display. In order to correctly display and integrate the information (lane boundaries, etc.) from the database with the sensed objects, a real-time method of locating and tracking nearby vehicles was developed. This paper outlines the algorithms and equipment used to achieve real-time vehicle tracking based on a scanning laser range sensor, and the experiments performed to determine the accuracy of the virtual mirror's ability to correctly display moving vehicles in the sensor's field of view.