摘要
随着无人驾驶、智慧城市建设技术的发展,能够高效、准确地提取城市道路标线尤为必要,本文根据道路标线在车载激光扫描点云空间场景中的强度特征以及几何形态,提出了一种联合布料模拟滤波(CSF)算法、强度特征图像以及边缘检测的道路标线提取新方法。首先利用CSF算法从原始车载点云数据中提取出地面点;其次将地面点转换生成强度特征图像,基于强度特征特性图像进行图像边缘检测以及连通分析提取的标线边缘;最后引入高斯混合模型(Gaussian mixed model)对标线候选点进行分类,剔除地面噪声点。为检验本文提出算法的效果,使用两种不同类型道路点云数据进行实验,结果表明,本文提出算法道路标线提取结果的平均完整率、准确率以及综合提取质量分别为92.5%、85.9%、89.0%,均优于对比模型,验证了本文提出算法的可靠性与优越性。
With the development of technologies for unmanned driving and smart city construction,it is necessary to efficiently and accurately extract urban road markings.This paper proposed a new method for extracting road markings based on the intensity features and geometric forms of road markings in spatial scenes of vehicle-mounted laser scanning point clouds.This method combined the cloth simulation filter(CSF)algorithm,intensity feature images,and edge detection.First,it utilized the CSF algorithm to extract ground points from the original vehicle-mounted point cloud data.Secondly,the ground points were converted into intensity feature images.Based on the intensity feature images,image edge detection was performed,and connectivity analysis was conducted to extract marked edges.Finally,the Gaussian mixed model was introduced to classify the candidate points for markings and eliminate ground noise points.To test the effectiveness of the algorithm proposed in this article,two different types of road point cloud data were used for experiments.The results show that the average completeness,accuracy,and comprehensive extraction quality of the road marking extraction results by the algorithm proposed in this article are 92.5%,85.9%,and 89.0%,respectively,which are better than those by the comparative model,verifying the reliability and superiority of the algorithm proposed in this paper.
作者
凌晓霞
蒋斌
LING Xiaoxia;JIANG Bin(Ningbo Institute of Surveying,Mapping and Remote Sensing Technology,Ningbo Zhejiang 315000,China;AVIC Anzhen(Zhejiang)Information Technology Company Limited,Ningbo Zhejiang 315000,China)
出处
《北京测绘》
2024年第3期343-348,共6页
Beijing Surveying and Mapping
关键词
车载激光点云
道路标线
布料模拟滤波算法
强度特征图像
边缘检测
vehicle-mounted laser point cloud
road markings
cloth simulation filter algorithm
intensity feature image
edge detection
