EN中文

地球信息科学学报 >

2017 , Vol. 19 >Issue 7: 861 - 871

DOI: https://doi.org/10.3724/SP.J.1047.2017.00861

地球信息科学理论与方法

车载激光扫描数据路坎点云提取方法

  • 罗海峰 ,
  • 方莉娜 , * ,
  • 陈崇成
展开
  • 福州大学 地理空间信息技术国家地方联合工程研究中心,空间数据挖掘与信息共享教育部重点实验室,福建省空间信息工程研究中心,福州 350002
*通讯作者：方莉娜(1983- ),女,博士,助理研究员,研究方向为车载激光扫描数据的道路环境特征感知。E-mail:

作者简介：罗海峰（1990- ）,男,硕士生,研究方向为园林与生态景观设计规划技术。E-mail:

收稿日期: 2016-08-22

  要求修回日期: 2016-12-27

  网络出版日期: 2017-07-10

基金资助

国家自然科学基金青年科学基金项目（41501493）

福建省科技计划重点项目（2015H0015）

福建省教育厅中青年教师科研项目（JAT160078）

中国博士后面上科学基金项目

收起

Curb Point Clouds Extraction from Vehicle-Borne Laser Scanning Data

  • LUO Haifeng ,
  • FANG Lina , * ,
  • CHEN Chongcheng
Expand
  • National Engineering Research Centre of Geospatial Information Technology, Key Laboratory of Spatial Data Mining and Information Sharing of Ministry of Education, Spatial Information Research Center of Fujian Province, Fuzhou University, Fuzhou 350002, China
*Corresponding author: FANG Lina, E-mail:

Received date: 2016-08-22

  Request revised date: 2016-12-27

  Online published: 2017-07-10

Copyright

《地球信息科学学报》编辑部 所有
Fold

摘要

车载激光扫描系统能够快速准确地获取街道环境的点云数据,但由于扫描点云的点密度高、数据量大、空间分布不均匀、地物相互遮挡及城市街道环境复杂等特点,难以直接从原始点云数据中提取出路坎点云。本文首先通过分析路坎点云的空间分布特征和局部几何特征,构建包含相对高程、法向量方向、多尺度高程差及多尺度高程方差的点云特征向量;然后,采用SVM提取城市街道环境车载激光扫描数据中的路坎点云,并对提取结果进行聚类去噪,优化路坎点云。最后,通过Street Mapper 360系统和Lynx Mobile Mapper V100 系统采集的4份不同城市街道环境车载激光扫描数据对本文方法进行验证,其中路坎点云提取结果的完整度均超过了94.99%、准确度均超过91.88%、精度亦均达到了90.55%以上。实验结果表明,本文方法能够精确地提取复杂城市街道环境中规则或不规则的路坎点云,且具有较强的稳健性,适用于各类复杂的城市街道环境。

关键词： 车载激光点云; 支持向量机（SVM）; 多尺度点云特征; 点云特征向量; 路坎点云提取

本文引用格式

罗海峰 , 方莉娜 , 陈崇成 . 车载激光扫描数据路坎点云提取方法[J]. 地球信息科学学报, 2017 , 19(7) : 861 -871 . DOI: 10.3724/SP.J.1047.2017.00861

Abstract

Vehicle-borne laser scanning system can quickly and accurately obtain 3D point cloud data of the street scene by multiple sensors and data processing technology. Extracting street curb point clouds from vehicle-borne laser scanning data is important for many applications such as road panning and maintenance, but it is difficult to directly extract curb point clouds from the original data due to the high point density, large amount of data, uneven spatial distribution of point clouds, object obscuring each other and complex urban street scenes. A novel algorithm for extracting curb point clouds from vehicle-borne laser scanning data based on Support Vector Machine （SVM） is proposed in this paper. The algorithm is described as followings: firstly, the original data is thinned and segmented into a series of point cloud blocks to improve efficiency. Secondly, a point clouds feature vector is constructed including the relative elevation, normal vector direction, multi-scale height difference and height variance, through the analysis of spatial distribution and local geometric features of curb point clouds. Thirdly, a new method is proposed to refine the point cloud feature of the border of ground and object on the ground such as point clouds of the bottom of tree, building facade, fence, and so on, which avoids errors in dividing the border point clouds into curb point clouds. Fourthly, the obtained training samples are processed by artificial choice. Radial Basis Function （RBF） is selected as kernel. Particle Swarm Optimization （PSO） is used to optimize penalty factor C and kernel function parameter γ, then the feature vector is taken as the input to train SVM. The obtained SVM is used to extract curb point clouds from original data. Finally, cluster analysis is performed on the extracted results for eliminating noisy point clouds when the number of point clouds in cluster is less than threshold. Experiments were undertaken to evaluate the validities of the proposed algorithm with four different street scene datasets acquired by Street Mapper 360 System and Lynx Mobile Mapper System, respectively. The completeness of the results of curb point clouds extraction is over 94.99%, correctness is above 91.88%, and quality is above 90.55%. This proves that the proposed algorithm not only has high accuracy, but also has strong robustness to extract curb point clouds with regular or irregular shape from complex urban street scenes of vehicle-borne laser scanning data.

Key words： vehicle-borne LiDAR; support vector machine （SVM）; multi-scale point clouds feature; point clouds feature vector; curb point clouds extraction

1 引言

车载激光扫描系统作为近年来迅速发展的高新测绘技术,能够快速精确地获取道路及道路两侧地物的三维空间信息[1]。车载激光扫描数据具有点密度高、数据量大、空间分布不均匀、地物相互遮挡及场景复杂等特点[2],相比于车载激光扫描系统硬件的快速发展,其数据处理方法仍较为滞后。目前,国内外专家学者主要围绕着车载激光点云的高程阈值[3]、投影密度[4-5]、特征图像[6]、法向量[7]等特征进行地物点云分类,提取建筑物立面、车辆、树木、杆状地物等目标,而针对复杂城市街道环境中路坎点云提取的相关研究较少。
目前车载激光扫描数据道路提取的相关研究中,一些学者将车载激光扫描数据转化成二维的特征图像,然后结合图像处理技术提取道路。其中,Jaakkola等[8]将车载激光扫描数据生成强度特征图像和高程特征图像,然后利用图像处理算子检测路坎位置。Boyko等[9]利用二维的路网数据将车载激光扫描数据分段,并转化成基于高程的特征图像,然后采用Snake模型提取图像中路坎的位置。Serna等[10]将车载激光扫描数据投影成距离图像,距离扫描仪近的点设置较大像素值,结合路坎的高差和长度约束条件,利用数学形态学的算法分割距离图像,进而提取道路边界。此类方法虽然具有较高的效率,但均依赖大量的人工经验阈值,提取精度受限于阈值的选择,而位置误差易受特征图像分辨率的影响。一些学者根据不同类型地物在扫描线上具有不同的空间分布特征,提出基于扫描线的道路点云提取方法。如Zhang[11]在每条扫描线中利用高斯低通滤波器提取路面点云,然后基于Hough变换检测垂直分布的点作为道路边界点,该方法仅适用于垂直形态的路坎。方莉娜等[12]采用移动窗口法先提取扫描线上的地面点云,然后根据路坎与路面之间的高程、坡度、点密度的变化提取路坎点云,该方法同样依赖大量经验阈值,且提取精度受窗口大小影响。Guan等[13]结合轨迹辅助数据划分数据块,将数据块投影至道路横截面,提取横截面上的主点,根据主点高差、坡度检测路坎位置,由于路坎点不一定是主点,因此这种基于主点检测路坎位置的策略也将会造成一定的路坎位置误差。
一些学者根据道路点云在一定邻域范围中具有相似的形态特征,提出基于聚类分析的方法检测路坎。如闫利等[7]基于扫描点云的法向量采用模糊聚类的方法提取道路;Zhou等[14]通过检测高程差异提取道路边界,利用KNN算法对路坎点云进行聚类,并结合长度、宽度信息优化聚类结果,这类仅采用了单一特征进行聚类分析的方法,难以应用于复杂的街道环境中。Ibrahim等[15]首先利用点密度分割地面点,然后结合地面点的高程离散度、法方向及高程梯度等信息,采用高斯滤波器提取路坎,该方法虽然克服了单一特征聚类的缺陷,但仍需依赖大量人工经验阈值的选择。Zhan等[16]通过计算每个点的法向量、法向量残差、离散度,利用SVM将点云数据分为地面、斜坡及树木等;Varney等[17]利用地物点云的体积特征作为SVM的输入,将点云数据分为地面、植被、建筑物等,这类基于SVM的点云分类方法虽能有效地提取地面点云,但未能精细准确地提取出路坎点云。针对以上问题,本文结合SVM应用于目标提取的优势,首次将其引入至车载激光扫描数据路坎点云提取的算法研究中,通过分析车载激光扫描数据中路坎点云的空间分布特征和局部几何特征,构建包含相对高程、法向量方向、多尺度高程差及多尺度高程方差共六维的点云特征向量;利用SVM初步提取路坎点云,并通过聚类去噪优化提取结果;最终实现基于SVM的车载激光扫描数据路坎点云提取方法。在保持较高提取精度的同时有效提高了车载激光扫描数据路坎点云提取方法的稳健性,能够适用于各类复杂的城市街道环境,并且无需依赖辅助数据。

2 基于SVM的路坎点云提取方法

本方法首先对原始车载激光扫描数据进行抽稀和分段处理;然后结合点云的空间分布特征和局部几何特征构建包含相对高程、法向量方向、多尺度高程差及多尺度高程方差共六维的点云特征向量,采用SVM初步提取路坎点云;最后对提取结果进行聚类去噪,从而实现车载激光扫描数据路坎点云的提取。该方法的具体流程如图1所示。
Fig. 1 Flow chart of curb points extraction from vehicle-borne laser scanning data

图1 路坎点云提取流程

2.1 数据预处理

由于车载激光点云的数据量庞大,本文对原始车载激光扫描数据进行抽稀和分段处理提高数据处理的效率。同时基于车载激光扫描系统作业方式的特点,能够准确获取各类地物在竖直方向上的空间分布信息,但在地势起伏较大的区域,地物点云的绝对高程并不能准确地表征各类地物的空间分布特征,因此本文采用地物点云的相对高程表征地物的空间分布特征。首先搜索目标点 P x 所在街道横断面上的所有点,记为点集 P i ,如图2所示;然后根据式（1）计算目标点的相对高程 H x r 。
H min a = min ( H P i a ) H x r = H x a - H min a （1）
式中： H P i a 为点集 P i 的绝对高程值; H min a 为点集 P i 中最小绝对高程值; H x a 为目标点的绝对高程值。
Fig. 2 The distribution of street profile point clouds

图2 街道横截面点云分布

利用相对高程作为路坎点云提取的特征之一,在后续路坎点云提取中能够有效克服地形起伏的影响,准确地将提取结果约束在路坎的高度。如图3所示,由于沿街道方向的地势存在较大的起伏,导致图3（a）中街道两端同为路面点云的绝对高程差近达10 m,经过高程改正预处理后该区域中街道两端路面的相对高程差仅为0.08 m（图3（b））,因此点云的相对高程能够更加准确地表征地物在竖直空间的分布特征。
Fig. 3 The spatial distribution of absolute and relative height of point clouds

图3 点云绝对高程和相对高程的空间分布

2.2 构建点云特征向量

原始车载激光扫描数据中,虽然每个离散点都具有空间坐标信息,但并无关联,且单独一个离散点无法描述被扫描物体的整体或局部信息,因此本文针对路坎点云的提取,构建点云邻域内的法向量方向、多尺度高程差及多尺度高程方差作为描述被扫描物体的局部几何特征。
2.2.1 法向量方向
在城市街道环境车载激光点云中,不同类型的地物,其点云法向量方向存在明显差异。例如,路面点云法向量方向为竖直方向,建筑物立面点云法向量方向一般情况为水平方向,树木枝叶点云法向量方向则无规律、朝向各方向等,因此点云的法向量方向信息可以作为目标点云识别分类的特征。本文定义目标点云法向量方向（Normal Direction,ND）为目标点云的法向量与坐标系Z轴的夹角,当法向量朝向Z轴正值方向时,计算法向量与向量（0,0,1）的夹角,当法向量朝向Z轴负值方向时,计算法向量与向量（0,0,-1）的夹角。如图4所示的城市 街道点云不同类型地物的法向量方向分布状况中,路坎点云法向量方向分布与路面、树木及建筑物立面等点云的法向量方向具有较明显的区分性,因此将点云的法向量方向作为提取路坎点云的特征 之一。
Fig. 4 The distribution of normal vector direction of street point clouds

图4 街道点云法向量方向分布

2.2.2 局部高程特征
在城市街道车载激光点云中,不同类型地物的局部点云的高程差（ ΔH ）和高程方差（ σ 2 ）存在明显差异,如局部路面点云高程变化较小,则其高程差较小,高程方差也较小;相比较于路面点云,路坎点云高程略有变化,则路坎点云具有一定的高程差,且其高程方差也略大于路面点云的高程方差;而建筑物、树木等地物在局部范围内高程值变化较大,则其高程差较大,高程方差也远大于路面和路坎,因此局部高程差和高程方差可作为本文提取路坎点云的特征。
根据式（2）、（3）分别计算目标点邻域内点集 P i 的高程差 Δ H x r 和高程方差 σ x 2 。
Δ H x r = max H p i r - min H p i r （2）
式中： H p i r 为目标点邻域内点集 P i 的相对高程。
σ x 2 = ∑ i = 1 N H p i r - H ̅ r 2 N （3）
式中： H ̅ r 为目标点邻域内的平均相对高程;N为点集 P i 中点的数量。图5（a）为局部高程差的分布,图5（b）为局部高程方差分布。
Fig. 5 The distribution of local elevation features

图5 局部高程特征分布

由于目标点的局部高程特征是由其一定搜索半径邻域内的点集决定的,不同的搜索半径将会导致不同地物具有不同的局部高程特征。本文针对路坎点云,采用图6所示的2种大小不同的搜索半径 r 1 和 r 2 （ r 1 、 r 2 均大于路坎高度）,其中 r 1 < r 2 。在这2种尺度下,路坎点云的高程差基本保持不变,而在较小尺度下可能与路坎点云具有相似的高程差的建筑物斜面点云或树冠的散乱点云等的高程差将随着邻域搜索半径的扩大而增大。同样在这2种尺度下,路面点云的高程方差随着邻域搜索半径的扩大基本保持不变;路坎点云的高程方差随着邻域搜索半径的扩大而减小;而建筑物点云及树木点云的高程方差则随着邻域半径的扩大而增大。利用不同尺度（不同邻域搜索半径）下不同目标点云所表现的局部高程特征变化趋势,将多尺度的高程差和多尺度的高程方差作为提取路坎点云的特征,能够克服单一尺度下不同地物具有相同的局部高程特征,从而影响路坎点云的提取结果。
Fig. 6 Multi-scale neighborhood of the target points

图6 目标点的多尺度邻域

2.2.3 交界点云特征细化
在城市街道环境车载激光点云中,如建筑物立面底部或栅栏底部与路面交界处的局部点云形态特征与路坎较为相似,且相对高程也较为近似,难以进行区分,如图7所示的法向量方向分布。如直接利用这些特征提取路坎点云,将会导致建筑物底部或栅栏底部与地面交界处的点云错分为路坎点云,且在后续聚类去噪的路坎点云优化处理中难以将其剔除。
Fig. 7 The distribution of normal direction of junction point clouds

图7 交界点云法向量方向分布

针对以上问题,本文首先基于法向量方向特征值计算交界点云系数 α 区分交界点和非交界点：
α = e N D max - N D x e N D x - N D min （4）
式中： N D min 为搜索半径 r 1 邻域内法向量方向最小值; N D max 为邻域内法向量方向最大值; N D x 为目标点的法向量方向。若 α < 1 时,目标点标记为交界候选点。若 α ≥ 1 时,目标点标记为非交界候选点。
如图8所示,蓝色部分为交界候选点集 P c ,黄色部分为非交界候选点集 P nonc 。
Fig. 8 The results of junction point clouds segmentation

图8 交界点云划分结果

在交界候选点集 P c 中,包含路坎点云和所有地物底部点云等,然后采用搜索半径 r 1 邻域内点集特征值的平均值作为目标点的特征值,一方面使得路坎点云的各特征值均保持较为一致的数值范围,另一方面将各地物底部点云特征值向数值正方向拉伸;在非交界候选点集 P nonc 中则采用搜索半径 r 1 邻域内点集特征值的最小值作为目标点的特征值,将非交界候选点集中与地物底部邻接点云的特征值向数值负方向收缩,从而实现交界点云特征细化。图9为交界处点云法向量方向特征细化结果。
Fig. 9 The results of normal direction refinement of junction point clouds

图9 交界点云法向量方向细化结果

最后构建包括相对高程及交界点云特征细化后的法向量方向、两个不同尺度的局部高程差和局部高程方差组成的点云特征向量F,如式（5）所示。
F = H r , ND , Δ H 1 , Δ H 2 , σ 1 2 , σ 2 2 （5）

2.3 基于SVM提取路坎点云

支持向量机（SVM）作为机器学习领域极为重要的发展成果之一,在解决非线性、高维模式识别等问题中表现出许多独有的优势[18]。本文通过人工提取少数路坎与非路坎点云作为训练样本,将上文所得的车载激光点云特征向量F作为SVM的输入特征,选择应用最广泛的径向基函数（Radial Basis Function,RBF）作为核函数,采用粒子群优化算法（Particle Swarm Optimization,PSO）进行惩罚因子C及核函数参数 γ 的寻优,然后通过SVM的训练与分类,实现车载激光扫描数据中路坎点云的初步提取。由于提取结果中仍存在着大量散乱的噪点,如图10（a）所示,本文通过对提取结果进行聚类分析,设定聚类阈值N（点云数量）,当聚类簇中点云数量小于阈值N时,将该簇点云视为噪点剔除,最终得到路坎点云如图10（b）所示。
Fig. 10 The extraction results of curb points

图10 路坎点云提取结果

3 实验与分析

本文实验采用Cloud Compare软件对原始点云数据进行抽稀和分段处理,其余实验在Matlab R2010b平台下实现,SVM采用的是Faruto先生在台湾大学林智仁教授设计的开源库LIBSVM的基础上开发的LIBSVM加强工具箱,其中包含本文采用的SVM模型参数寻优的粒子群优化算法的辅助函数。

3.1 实验数据

实验共选用4份不同的城市街道环境车载激光扫描数据,每份原始数据均包含大量的树木、栅栏、车辆、交通指示牌及建筑物等地物。其中图11（a）数据1是由Street Mapper 360系统采集的长约197 m的城市街道扫描点云数据,其特点是存在较大的地势起伏,且点云密度较大,共有14 568 794个数据点;图11（b）数据2为长约175 m的城市街道扫描点云数据,其主要特点是街道中只存在一边的路坎,且部分路坎被灌木遮盖,同时在道路中存在一个圆形转盘,其共有8 546 051个数据点;图11（c）数据3为长约694 m的城市街道数据,该数据主要特点是点云密度较小、场景非常复杂,在道路中央存在大量形状各异且非规则的隔离带等,其共有7 209 113个数据点;图11（d）数据4为长约570 m的城市街道点云,其特点为街道比较狭窄,路坎被大量停放在道路两旁的车辆遮挡,造成路坎点云缺失严重,该数据共有15 481 507个点,数据2、3、4均由Lynx Mobile Mapper V100 系统采集。
Fig. 11 Vehicle-borne laser scanning data of urban street environment

图11 城市街道环境车载激光扫描数据

3.2 实验结果

实验中,数据1的点云抽稀两点间最小距离阈值为0.05 m,数据2、3、4中的点云抽稀两点间最小距离阈值为0.02 m,邻域半径 r 1 = 0.3 m , r 2 = 0.5 m 经过多次试验比较,设置粒子群优化算法中的点群半径为0.2 m,种群数量为20,最大、最小惩罚因子分别为100和0.01,最大、最小核函数参数 γ 亦分别为100和0.01,最大迭代次数为100,学习因子分别为 c 1 = 1.5 、 c 2 = 1.7 。根据以上设置的参数值,得到4份实验数据路坎点云提取结果,如图12-15所示。其中,图12（a）、13（a）、14（a）、15（a）分别为4份实验数据基于SVM提取的路坎点云初步结果;图12（b）、13（b）、14（b）、15（b）分别为4份实验数据初步提取路坎点云的聚类结果;通过多次试验比较设定数据1、数据2、数据3、数据4的聚类阈值分别为150、200、300、100,最终得到路坎点云如图12（c）、13（c）、14（c）、15（c）所示。
Fig. 12 Experimental results of data 1

图12 数据1实验结果

Fig. 13 Experimental results of data 2

图13 数据2实验结果

通过路坎点云提取结果可以看出,本文方法能够准确地提取各种复杂街道环境车载激光扫描数据中的路坎点云,且对于道路中转弯处、圆形转盘、规则或非规则花坛等的路坎点云均能准确提取。但仍存在一些错误提取的路坎点,这主要是因为在复杂的街道环境中不可避免地存在与路坎分布极为相似的地物,如图14（c）道路中央花坛中成排低矮的灌木,其激光扫描点云与路坎点云的空间分布特征和局部几何特征均极为相似,导致路坎点云的错分现象。同时还存在着一些漏提取的路坎点,从图15（c）可以看出,路坎点云提取结果中存在着部分路坎漏提的现象,这主要是由于原始数据中的路坎被遮挡较为严重,其次是因为在聚类去噪过程中,部分路坎点云聚类簇中点云数量小于设定的阈值被错误当成噪点剔除。
Fig. 14 Experimental results of data 3

图14 数据3实验结果

Fig. 15 Experimental results of data 4

图15 数据4实验结果

3.3 实验结果分析

实验中处理数据1、2、3、4所消耗的时间分别为684、720、1296和1512 s。由于原始车载激光扫描数据中没有提供路坎或道路边界等参考数据,因此本文无法直接对实验结果进行精度评价。本文通过与Cloud Compare软件交互的方式测量出实验结果中正确提取的路坎长度TP,即实际中存在且被正确提取出来的路坎;错误提取的路坎长度FP,即所提取出的路坎在实际中并不存在;未被提取的路坎长度FN,即在实际中存在的路坎,但没有被提取出来的部分,分别取多次测量结果的均值作为最终实验结果,如表1所示。同时,采用目前道路边界提取与识别中应用较广泛的客观评价指标完整度（Completeness）、准确度（Correctness）和精度（Quality）进行结果评价,如表2所示。
Tab. 1 The results of curb point clouds extraction

表1 路坎点云提取结果

实验结果/m 实验数据
数据1 数据2 数据3 数据4
TP 409.12 215.73 2344.19 491.88
FP 21.22 10.75 207.23 20.15
FN 13.12 11.39 37.32 13.99
Tab. 2 The accuracy of extraction results of curb point clouds

表2 路坎点云提取精度

评价指标
（%）		 实验数据
数据1 数据2 数据3 数据4
完整度Com 96.89 94.99 98.43 97.23
准确度Cor 95.07 95.25 91.88 96.06
精度Qua 92.26 90.69 90.55 93.51
Completeness = TP / TP + FN （6）
Correctness = TP / TP + FP （7）
Quality = TP / TP + FP + FN （8）
通过表2中4份实验数据的路坎点云提取结果评价指标可以看出,采用本文方法提取路坎点云结果的完整度均超过了94.99%、准确度均超过91.88%、精度亦均达到了90.55%以上。由于原始数据3中路坎点云被遮挡较少,其初步提取结果中路坎点云均较为完整,在聚类去噪中,其聚类簇点云个数均大于阈值,不会被错误剔除,因此该数据结果具有较高的完整度;而该数据实验结果准确度较低是由于在道路中央的花坛中存在大量成排的几何形状与路坎极为相似的低矮灌木,实验将这些点云错分为路坎点云导致该数据实验结果的准确度较低。

4 结论

本文通过分析车载激光扫描数据中路坎点云的空间分布特征和局部几何特征,并对交界处点云特征进行细化,构建包含相对高程、法向量方向、多尺度高程差及多尺度高程方差的点云特征向量,然后利用SVM初步提取路坎点云,最后对提取结果进行聚类去噪优化。并通过4份不同城市街道环境的车载激光扫描数据对本文方法进行验证,实验结果的3个评价指标的最小值均超过了90%,表明本文方法能够精确地提取复杂城市街道环境中规则或不规则的路坎点云,且具有较强的稳健性,适用于各类复杂的城市街道环境。但由于车载激光扫描数据中路坎点云容易被其他地物遮挡,根据本文的方法直接从点云数据中提取出的路坎点云将缺失被遮挡部分的路坎点云;同时笔者通过对比实验选取特征,因此本文构建的特征向量并非最优,实验结果中仍存在少部分错分的路坎点云。在后续的学习工作中结合这些问题进一步研究,将缺失的路坎点云进行正确的补充,以提高整个道路边界的完整性,且对目前的特征向量进行优化,进一步提高路坎点云提取结果的精度。

The authors have declared that no competing interests exist.

参考文献

原文顺序 | 文献年度倒序 | 文中引用次数倒序
[1]
Tan B, Zhong R.Objects classification with vehicle-borne laser scanning data[J]. Journal of Remote Sensing, 2012,16(1):50-66.The vehicle-borne laser scanning (VLS) technology can quickly acquire three-dimensional information of earth surface with high precision. As a new technique of data collection acquisition, it has been gradually applied to Geographic Information System (GIS) industry. The classification on laser scanning points (LSP) is the premise and key step to the feature extraction and the model building of urban objects. This paper presents a classification method mainly suitable for the urban’s typical objects. Firstly, it extracts the LSP of the ground surface through taking three items into account, including characteristics of data collection, the vehicle's GPS trajectory, and the slope relationship between the adjacent LSP in the same scanning line. Secondly, as for the remaining LSP, the object division should be completed first through the methods of Grid and Segmentation, and then performs the objects classification by computing several statistical indicators of the spatial characteristics. Finally, this paper takes the Sanya city as the study area, to verify the effectiveness of this method. The result shows that, this classification method successfully classifies the categories of the ground, buildings, trees and street lamps from the LSP.

DOI

[2]
陈昌鸣,向煜,龙川.基于车载激光雷达的行道树提取研究[J].北京测绘,2015(1):18-21.对车载激光雷达点云数据进行研究,提出了一种分层投影的方法提取行道树,该方法对点云数据进行分层投影,利用不同高程点云投影后表现出的不同特征,采用面向对象的方法对行道树进行提取,最后叠加分析得到行道树点云。实验证明,该方法能够比较完整的提取出行道树点云。

DOI

[ Chen C M, Xiang Y, Long C.The research of street trees extraction based on vehicle-borne LiDAR[J]. Beijing Surveying and Mapping, 2015,1:18-21. ]

[3]
吴芬芳,李清泉,熊卿.基于车载激光扫描数据的目标分类方法[J].测绘科学,2007,32(4):75-77.

[ Wu F F, Li Q Q, Xiong Q.Target classification method based on vehicle-borne laser scanning data[J]. Science of Surveying and Mapping, 2007,32(4):75-77. ]

[4]
李必军,方志祥,任娟.从激光扫描数据中进行建筑物特征提取研究[J].武汉大学学报·信息科学版,2003,28(1):65-70.提出了一套基于建筑物几何特征的信息挖掘方案 ,并具体介绍了研究中采用的一些原理、方法与实用算法 ,从而可以直接从激光扫描数据中提取建筑物的平面外轮廓信息。利用本方案对车载激光扫描系统获取的城市建筑物试验数据进行了具体的处理 ,并给出了三维建模和可视化表达结果。

DOI

[ Li B J, Fang Z X, Ren J.Extraction of building's feature from laser scanning data[J]. Geomatics and Information Science of Wuhan University, 2003,28(1):62-70. ]

[5]
Shi W Z, Wu F F.Feature extraction and modeling of urban building from vehicle- borne laser scanning data[J]. Mccarthy, 2004.Abstract Laser scanning is playing an important role in an acquisition of the 3D spatial objects in urban because of real-time and high efficiency than traditional methods. The research presented in this paper is focused on segmentation of range image, especially captured by vehicle-borne laser scanning systems, which is the crucial technology for extracting objects from laser scanned data. Here, a method, named as Density of Projected Points (DoPP) is proposed for the range image segmentation. With such a process, it is possible for feature extraction, object classification and modeling without other auxiliary data. An experimental study is conducted to demonstrated the feasibility of the proposed solution on urban building object extraction.

[6]
杨必胜,魏征,李清泉,等.面向车载激光扫描点云快速分类的点云特征图像生成方法[J].测绘学报,2010,39(5):540-545.以车载激光点云数据为研究对象,提出一种适合于其快速分类与目标提取的点云特征图像生成方 法。首先将扫描区域进行平面规则格网投影,通过分析格网内部点云的空间分布特征（平面距离、高程差异、点密集程度等）确定激光扫描点的定权,从而生成车载 激光扫描点云的特征图像。利用生成的点云特征图像,可采用阈值分割、轮廓提取与跟踪等手段提取图像分割的建筑物目标的边界,从而确定边界内部点云数据,实 现目标分类与提取。本文以Optech公司的车载激光扫描数据为试验对象,验证本文提出方法的可行性和实用性。

DOI

[ Yang B S, Wei Z, Li Q Q, et al.Point cloud feature image generation method for rapid classification of vehicle-borne laser scanning point cloud[J]. Acta Geodaetica et Cartographica Sinica, 2010,39(5):540-545. ]

[7]
闫利,张毅.基于法向量模糊聚类的道路面点云数据滤波[J].武汉大学学报·信息科学版,2007,32(12):1119-1122.

[ Yan L, Zhang Y.Road points cloud filtering based on fuzzy clustering of normal vectors[J]. Geomatics and Information Science of Wuhan University, 2007,32(12):1119-1122. ]

[8]
Jaakkola A, Hyyppä J, Hyyppä H, et al.Retrieval algorithms for road surface modelling using laser-based mobile mapping[J]. Sensors, 2008,8(9):5238-5249.Automated processing of the data provided by a laser-based mobile mapping system will be a necessity due to the huge amount of data produced. In the future, vehicle-based laser scanning, here called mobile mapping, should see considerable use for road environment modelling. Since the geometry of the scanning and point density is different from airborne laser scanning, new algorithms are needed for information extraction. In this paper, we propose automatic methods for classifying the road marking and kerbstone points and modelling the road surface as a triangulated irregular network. On the basis of experimental tests, the mean classification accuracies obtained using automatic method for lines, zebra crossings and kerbstones were 80.6%, 92.3% and 79.7%, respectively.

DOI PMID

[9]
Boyko A, Funkhouser T.Extracting roads from dense point clouds in large scale urban environment[J]. ISPRS Journal of Photogrammetry & Remote Sensing, 2011,66(6):S2-S12.This paper describes a method for extracting roads from a large scale unstructured 3D point cloud of an urban environment consisting of many superimposed scans taken at different times. Given a road map and a point cloud, our system automatically separates road surfaces from the rest of the point cloud. Starting with an approximate map of the road network given in the form of 2D intersection locations connected by polylines, we first produce a 3D representation of the map by optimizing Cardinal splines to minimize the distances to points of the cloud under continuity constraints. We then divide the road network into independent patches, making it feasible to process a large point cloud with a small in-memory working set. For each patch, we fit a 2D active contour to an attractor function with peaks at small vertical discontinuities to predict the locations of curbs. Finally, we output a set of labeled points, where points lying within the active contour are tagged as oad and the others are not. During experiments with a LIDAR point set containing almost a billion points spread over six square kilometers of a city center, our method provides 86% correctness and 94% completeness.

DOI

[10]
Serna A, Marcotegui B.Urban accessibility diagnosis from mobile laser scanning data[J]. ISPRS Journal of Photogrammetry & Remote Sensing, 2013,84(84):23-32.In this paper we present an approach for automatic analysis of urban accessibility using 3D point clouds. Our approach is based on range images and it consists in two main steps: urban objects segmentation and curbs detection. Both of them are required for accessibility diagnosis and itinerary planning. Our method automatically segments facades and urban objects using two hypotheses: facades are the highest vertical structures in the scene and objects are bumps on the ground on the range image. The segmentation result is used to build an urban obstacle map. After that, the gradient is computed on the ground range image. Curb candidates are selected using height and geodesic features. Then, nearby curbs are reconnected using B zier curves. Finally, accessibility is defined based on geometrical features and accessibility standards. Our methodology is tested on two MLS databases from Paris (France) and Enschede (The Netherlands). Our experiments show that our method has good detection rates, is fast and presents few false alarms. Our method outperforms other works reported in the literature on the same databases.

DOI

[11]
Zhang W.LIDAR-based road and road-edge detection[J]. IEEE Intelligent Vehicles Symposium, 2010,46(6):845-848.In this paper, a LIDAR-based road and road-edge detection method is proposed to identify road regions and road-edges, which is an essential component of autonomous vehicles. LIDAR range data is decomposed into signals in elevation and signals projected on the ground plane. First, the elevation-based signals are processed by filtering techniques to identify the road candidate region, and by pattern recognition techniques to determine whether the candidate region is a road segment. Then, the line representation of the projected signals on the ground plane is identified and compared to a simple road model in the top-down view to determine whether the candidate region is a road segment with its road-edges. The proposed method provides fast processing speed and reliable detection performance of road and road-edge detection. The proposed framework has been verified through the DARPA Urban Challenge to show its robustness and efficiency on the winning entry Boss vehicle.

DOI

[12]
方莉娜,杨必胜.车载激光扫描数据的结构化道路自动提取方法[J].测绘学报,2013,42(2):260-267.车载激光扫描系统获取的结构化道路环境（城市道路/高速公路）点云数据量大、目标复杂,难以有效提取出道路的点云。本文通过分析扫描线上激光点云的空间分布和统计特征,提出一种适用于结构化道路环境的道路点云自动提取方法。在Optech公司提供的两份车载激光扫描点云数据中,道路提取结果的完整率、准确率、提取质量相应地超过94.92%、95.80%和91.13%。通过定量和定性的试验分析,该方法不仅适应于有固定道路宽度的规则道路提取,同样适用于无固定宽度的非规则道路提取。

[ Fang L N, Yang B S.Automated extracting structural roads from mobile laser scanning point clouds[J]. Acta Geodaetica et Cartographica Sinica, 2013,42(2):260-267. ]

[13]
Guan H, Li J, Yu Y, et al.Automated road information extraction from mobile laser scanning data[J]. IEEE Transactions on Intelligent Transportation Systems, 2015,16(1):194-205.This paper presents a survey of literature about road feature extraction, giving a detailed description of a Mobile Laser Scanning (MLS) system (RIEGL VMX-450) for transportation-related applications. This paper describes the development of automated algorithms for extracting road features (road surfaces, road markings, and pavement cracks) from MLS point cloud data. The proposed road surface extraction algorithm detects road curbs from a set of profiles that are sliced along vehicle trajectory data. Based on segmented road surface points, we create Geo-Referenced Feature (GRF) images and develop two algorithms, respectively, for extracting the following: 1) road markings with high retroreflectivity and 2) cracks containing low contrast with their surroundings, low signal-to-noise ratio, and poor continuity. A comprehensive comparison illustrates satisfactory performance of the proposed algorithms and concludes that MLS is a reliable and cost-effective alternative for rapid road inspection.

DOI

[14]
Zhou L, Vosselman G.Mapping curbstones in airborne and mobile laser scanning data[J]. International Journal of Applied Earth Observation and Geoinformation, 2012, 18(18): 293-304.The high point densities obtained by today's laser scanning systems enable the extraction of various features which are traditionally mapped by photogrammetry or land surveying. While significant progress has been made in the extraction of buildings and trees from dense point clouds, little research has been performed on the extraction of roads. In this paper it is analysed to what extent road sides can be mapped in point clouds of high point density. In urban areas curbstones are often used to separate the road surface from the adjacent pavement. These curbstones are mapped in a three step procedure. First, the locations with small height jumps near the terrain surface are detected. Second, midpoints of high and low points on either side of the height jump are generated, put in a sequence to obtain a polygonal chain describing the approximate curbstone location. A sigmoidal function is then fitted to all points near the polygonal chain to increase the accuracy. Third, small gaps between nearby and collinear line segments are closed. GPS measurements were taken to analyse the performance of the road side detection. The analysis showed that the completeness in airborne laser scanning (ALS) data varying between 53% and 92% is higher than that in mobile laser scanning (MLS) data ranging from 54% to 83%, depending on the amount of parked cars occluding the curbstones. The RMS value in the comparison with the GPS points measured from ground survey was 0.11m in ALS data and 0.06m in MLS data, respectively.

DOI

[15]
Ibrahim S, Lichti D. Curb-based street floor extraction from mobile terrestrial LIDAR point cloud[J]. ISPRS-International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2012, XXXIX-B5:193-198.Mobile terrestrial laser scanners (MTLS) produce huge 3D point clouds describing the terrestrial surface, from which objects like different street furniture can be generated. Extraction and modelling of the street curb and the street floor from MTLS point clouds is important for many applications such as right-of-way asset inventory, road maintenance and city planning. The proposed pipeline for the curb and street floor extraction consists of a sequence of five steps: organizing the 3D point cloud and nearest neighbour search; 3D density-based segmentation to segment the ground; morphological analysis to refine out the ground segment; derivative of Gaussian filtering to detect the curb; solving the travelling salesman problem to form a closed polygon of the curb and point-inpolygon test to extract the street floor. Two mobile laser scanning datasets of different scenes are tested with the proposed pipeline. The results of the extracted curb and street floor are evaluated based on a truth data. The obtained detection rates for the extracted street floor for the datasets are 95% and 96.53%. This study presents a novel approach to the detection and extraction of the road curb and the street floor from unorganized 3D point clouds captured by MTLS. It utilizes only the 3D coordinates of the point cloud.

DOI

[16]
Zhan Q, Yu L.Objects classification from laser scanning data based on multi-class support vector machine[C]. International Conference on Remote Sensing, Environment and Transportation Engineering, 2011:520-523.

[17]
Varney N M, Asari V K.Volumetrie features for object region classification in 3D LiDAR point clouds[C]. Applied Imagery Pattern Recognition Workshop. IEEE, 2014:1-6.

[18]
张琴,蔡勇.支持向量学习机在点云去噪中的应用[J].计算机技术与发展,2011(6):85-88.在解决非线性、高维模式识别以及小样本等问题中,支持向量机表现出许多独有的优势。提出将支持向量机学习分类方法应用于点云去噪中,能够稳定地进行机器学习,训练得到判别模型,快速、准确地识别出噪声点与非噪声点。通过对小样本数据的统计学习,能够推广到大规模数据中去进行结果的预测估计。用SVM对点云数据样本进行学习训练、测试,识别分类,从而达到去噪光顺的目的。实验表明,此方法在有效去除噪声的同时能较完整地保留点云数据信息。

DOI

[ Zhan Q, Cai Y.Application of support vector machine in point clouds denoising[J]. Computer Technology and Development, 2011,6:85-88. ]

Options
文章导航

/

〈 〉