摘要
利用点云数据空间分布特征和回波强度信息,结合局部均值变点统计方法,提出了一种用于激光雷达数据帧的车道标线识别算法。该算法首先基于车载激光雷达采集的道路周围环境点云数据中高程信息进行滤波,确定可行驶区域。然后利用局部均值变点统计对可行驶区域点云数据中的回波强度值进行标记提取,即车道标线点云数据粗提取。最后基于EM(最大期望)方法聚类,从而完整、准确地识别车道标线。实验结果表明,该算法不仅能够准确定位可行驶区域,进而可以实现车道标线的自动提取;而且有效抑制了道路周围环境对车道标线识别的干扰,验证了算法的有效性。
Based on the information of spatial distribution and reflection intensity of the point cloud data,an effective algorithm is proposed to identify lane markings from 3D LIDAR data frames using the local mean change point statistics. The algorithm firstly determines the driving region by filtering the elevation information of the surrounding environment point cloud data collected by the on-vehicle laser radar. Then,the reflection intensity in the point cloud data of the driving region is marked and extracted using the local mean change point statistic,rough extraction of lane markings data. Finally,the significant points are clustered based on EM（ maximum expectation）method,thus the lane markings is fully and accurately identified. The experimental results show that the algorithm can not only locate the driving region accurately and extract lane markings automatically,but also restrain the influence of the surrounding road environment on lane markings identification effectively, and verify the effectiveness of the algorithm.
出处
《科学技术与工程》
北大核心
2017年第16期87-92,共6页
Science Technology and Engineering
基金
国家自然科学基金(61074140
61573009
51508315)
汽车安全与节能国家重点实验室开放基金(KF16232)
山东省自然科学基金(ZR2014FM027)
山东省社会科学规划研究项目(14CGLJ27)
山东省高等学校科技计划(J15LB07)资助
关键词
智能车
激光雷达
均值变点统计
车道标线识别
intelligent vehicle
laser radar
mean change point statistic
lane markings identification
