摘要
现有车道识别方法易受道路上车辆和阴影的干扰,并且投影公式精度不高,不适用于摄像机安装俯仰角为0°的情况。为此对现有的投影公式进行了改进,在变换后的俯视图中使用圆曲线车道模型及基于密度的Hough变换进行车道识别。实验结果表明,该方法可以明显降低车辆、阴影对于车道识别的干扰,并且能够满足车道识别实时性的要求。
Conventional transform mapping formulas are not accurate for vehicular lane detection systems at zero CCD camera pitch angles, with the lane detection easily disturbed by the vehicles and shadows on the road. A lane detection approach based on improved transform mapping formulas was developed to solve these problems with the arc lane model and the density based Hough transform used in the transformed vertical view to detect the lane. Experimental results show that the method significantly reduces the disturbances due to the vehicles and shadows on the road and can provide real time lane detection.
出处
《清华大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2005年第11期1530-1533,共4页
Journal of Tsinghua University(Science and Technology)
基金
国家"十五"计划资助项目
关键词
机器视觉
车道识别
车道保持
投影变换
computer vision
lane detection
lane keeping transform mapping