基于单目视觉的智能车路口实时定位方法

Method of Real Time Intersection Location Based on Monocular Vision for Intelligent Vehicle

为解决智能车在路口的高精度实时定位问题,基于单目视觉,提出一种路口实时定位方法。对需要定位的路口进行编号并建立路口场景特征库,采用路口场景识别的方法进行路口粗定位,通过路口图像中的停止线检测与测距以及车道线检测计算车辆航向角和偏移距离,综合距离、航向角和偏移距离进行位置坐标计算。在真实道路环境的路口测试结果表明,提出的定位方法具有精度高、实时性好、鲁棒性强的优点,适合于智能车路口视觉导航。

In order to solve the problem of high accuracy real-time location of intelligent vehicles at the intersection,this paper proposes a real-time intersection location method based on monocular vision. The intersections that need to be located are numbered and an intersection scene feature library is created. The intersection scene recognition method is used for rough location. Through the stop-line detection and ranging of intersection images as well as the lane detection,the course angle and offset distance of the vehicle are calculated. Coordinate position is calculated by synthetic distance,heading angle and offset distance. The intersection test results in real road environment show that the proposed method has the advantages of high accuracy, real-time performance and robustness. It is suitable for the visual navigation of intelligent vehicles at the intersection.