摘要
针对道路边界图像提取方法的点云数据预处理不佳,导致方法存在提取效果差的问题,提出基于车载激光点云的道路边界图像自动提取方法。根据不规则三角网构建自适应滤波算法,分区域滤波处理点云数据,获取内插点云的高程数值,从而配准边界正射影像与点云数据;采用道路轮廓曲线方位与能量函数,构建边界自动提取模型,结合最邻近分类算法划分点云数据类别和提取特征,通过引入贝塞尔曲线优化拟合点云数据,实现道路边界图像自动提取。实验结果表明:采用Velodyne HDL-64车载三维激光扫描设备获取的基础数据,本文提取精度高于96%,滤波最大误差为1.3%;与传统方法相比,本文提取精度提高了5%,滤波最大误差降低了0.6%以上,验证了该方法提取精度、降低误差的有效性,以及自动提取效果高、连续性好。
As the poor preprocessing of point cloud data in the road boundary image extraction method leads to the poor extraction effect of the method,an automatic extraction method of road boundary image based on vehicle-mounted laser point cloud is proposed.An adaptive filtering algorithm is constructed based on the irregular triangulation network,the point cloud data is processed by regional filtering,and the elevation value of the interpolated point cloud is obtained,thereby registering boundary orthoimage and point cloud data.The automatic boundary extraction model is constructed by using the orientation and energy function of the road contour curve,and combined with the K-nearest neighbors classification algorithm to classify point cloud data and extract features.The automatic extraction of road boundary image is realized by introducing the Bezier curve to optimize fitting point cloud data.Velodyne HDL-64 vehicle-mounted 3D laser scanning equipment is used to obtain basic data.The extraction accuracy of the method is higher than 96%,and the maximum filtering error is 1.3%.Compared with the comparison method,the extraction accuracy of the method is increased by 5%and the maximum filtering error is reduced by more than 0.6%,which shows that the method effectively improves the extraction accuracy,reduces the error,and has a high automatic extraction effect and good continuity.
作者
王玉
WANG Yu(School of Information and Finance,Xuancheng Vocational and Technical College,Xuancheng 242000,Anhui,China)
出处
《上海电机学院学报》
2023年第1期45-50,共6页
Journal of Shanghai Dianji University
关键词
车载激光点云
道路边界图像提取
不规则三角网
贝塞尔曲线
云数据配准
vehicle-mounted laser point cloud
road boundary image extraction
irregular triangulation
Bezier curve
cloud data registration
