摘要
针对道路场景中常见的由阴影、地面积水、刹车印记等因素引起的车道线对比度低和车道线不明显问题,提出一种基于Bayes后验概率的鲁棒性车道线检测算法.该算法在二次曲线道路形状模型基础上,在图像平面中引入中心车道线模型描述左右车道线,结合图像信息和车道线先验概率构造后验概率函数,将车道线检测问题引申为最大后验概率问题,并在优化过程中将具有参数少、结构简单和收敛速度快等特点的粒子群优化算法与具有全局搜索特性和不易早熟等特点的免疫克隆策略相融合,提高车道线检测的鲁棒性.实验结果验证了上述算法在车道线检测过程中对阴影等不良因素的鲁棒性.
In order to solve such problems as the low lane contrast and invisible lane in the road scene caused by shadow,water stain and brake trace,a robust lane detection approach based on Bayes posterior probability was proposed.The center lane model was introduced into image plane to describe the left and right lanes based on the quadratic road model.Through combining the image data with the prior probability of real lane to construct the posterior probability function,the problem of lane detection could be exten...
出处
《沈阳工业大学学报》
EI
CAS
2010年第6期679-686,共8页
Journal of Shenyang University of Technology
基金
教育部博士点新教师基金资助项目(200802861061)
江苏省交通厅科技研究计划项目(08X09)
关键词
车道线检测
中心车道线
后验概率
先验概率
似然函数
鲁棒性
免疫克隆
粒子群优化
lane detection
center lane
posterior probability
prior probability
likelihood function
robust
immune clone
particle swarm optimization
