摘要
基于障碍物栅格图的道路边界检测目前主要通过对道路区域或符合边界特征的种子点进行生长来实现.但这些算法易受路内障碍物的干扰,其主要原因在于道路形态信息的缺失,包括道路趋势信息和道路宽度分布信息.为此,本文通过对障碍物栅格图进行距离变换获得道路的趋势信息,同时通过构建道路的宽度分布直方图来获取道路宽度分布信息.在此基础上,完成道路边界种子点的提取,接着通过区域生长获取完整的道路边界点,最后进行二次曲线拟合得到完整的道路边界形状.实验证明,本文方法对车流量较大环境下的道路边界提取具有较高的鲁棒性和准确性.
The road curb detection based on obstacle grid map is implemented mainly through the growing of road area or the seed points satisfying the features of road curb. However, these algorithms are prone to be affected by the obstacles inside the road area because of the absence of the road shape information, such as road trend information and road width distribution information. To overcome this problem, distance transformation is firstly performed on obstacle grid map to acquire the trend information, followed by the construction of width distribution histogram to express road width distribution information.Based on these, road curb seed points are extracted, the full road curb points are obtained through region growth, and the road curb shape is acquired through quadratic curve fitting. The experiment proves that the algorithm presented performs relatively more robust and accurate for the road curb extraction in relative busier road.
出处
《机器人》
EI
CSCD
北大核心
2016年第3期322-328,共7页
Robot
基金
国家自然科学基金重大研究计划重点项目(91120307)
国家自然科学基金重大研究计划集成项目(91320301)
国家自然科学基金重大研究计划培育项目(91420104)
