基于点云密度的结构化道路边界增强检测方法

Enhanced Detection Method for Structured Road Edge Based on Point Clouds Density

为快速鲁棒地检测结构化道路边界,提出一种基于HDL-64E激光雷达点云密度的道路边界增强检测方法。通过建立虚拟雷达模型,利用点云密度特征,实现前景与背景分离,并利用随机采样一致性算法得到20m内的道路边界。为解决20~100m内道路边界点云稀疏、检测准确性下降的问题,提出利用光线切割模型对道路边界进行增强检测。在校园道路和城市快速路进行实验,道路边界检测率达到95%以上,有效检测距离可达70m以上,检测周期小于32ms。

For the fast and robust detection of structured road edge,an enhanced road edge detection method is proposed based on the density of point clouds from HDL-64 E lidar. By setting up virtual scan model and utiliazing the density features of point clouds,foreground is separated from background,and 20 meters of road edge is obtained by using random sample consensus algorithm. To solve the problem of poor detection accuracy of road edge within a distance of 20 and 100 meters due to sparse point clouds,a sheme is proposed to use ray cut model to conduct an enhanced detection on road edge. The results of experiments on both campus road and city express show that the effective detection rate of road edge reaches over 95% with a detectable distance longer than 70 m within a duration less than 32 ms.