树木遮挡下的机载Lidar点云建筑物轮廓提取

发布时间：2018-03-20 06:13

本文选题：建筑物轮廓　切入点：建筑物轮廓规则化　出处：《西南交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：建筑物是城市重要的组成,是"数字城市"不可缺少的组成部分。从机载Lidar(Light Detection And Ranging)点云中提取建筑物是城市建模的关键问题之一,而建筑物的轮廓则是表达建筑物的关键信息。目前建筑物轮廓提取的研究大都是针对完整的建筑物点云。针对由于相邻高大树木等对建筑物的遮挡,使得建筑物点云存在部分缺失,建筑物轮廓不完整的情况,目前还没有有效的轮廓提取方法。本文针对建筑物被遮挡的情况下准确提取建筑物轮廓开展两方面研究:1)滤波,将原始Lidar点云中,地面点和非地面点分开;2)建筑物轮廓提取,从非地面点云中提取建筑物点,并提取建筑物轮廓。首先在现有的偏度平衡滤波(SKF)算法的基础上,利用局部拟合高差代替点的高程,提出基于高差的偏度平衡滤波(SKF-HD)算法。该算法保持了对高大地物提取效果的同时,提高了低矮地物的提取效果,且显著提高了地形起伏区域的适应性。三组不同地形、不同区域的实验结果表明该算法能够更好的提取低矮地物、适用于不同程度的起伏地形。与偏度平衡滤波算法相比较,提出的算法在平坦区域、地形起伏较小区域和地形显著起伏区域的总体精度分别增大4.8%、5.1%和13.3%。在不同地形条件下,尤其是在地形显著起伏区域,新算法的滤波精度得到了明显的提高。然后改进MBR算法,针对被遮挡的规则多边形建筑物提出多级最小外接矩形(MMBR)算法,准确的提取建筑物的轮廓。该方法不仅可以准确得到建筑物没有遮挡部位的轮廓的同时还能得到准确的被遮挡部位的轮廓。三组不同遮挡情况、不同形状的实验区域的实验结果表明该算法能够准确的得到被遮挡区域的建筑物轮廓。与直角约束的迭代最小二乘(HLSPC)算法相比较,提出的算法在规则矩形建筑物、L型复杂多边形建筑物和复杂多边形建筑物中都能准确的提取建筑物轮廓,且Vd值分别减小31.8%、14.3%和12.5%。在不同遮挡情况下,MMBR算法的精度都较高。
[Abstract]:Building is an important part of a city and an indispensable part of "digital city". Extracting buildings from airborne Lidar(Light Detection And moving cloud is one of the key problems in city modeling. The outline of the building is the key information to express the building. At present, the research of extracting the outline of the building is mostly aimed at the complete point cloud of the building. There is no effective method to extract the building contour because of the partial absence of the building point cloud and the incomplete outline of the building. In this paper, when the building is occluded, two aspects of the research: 1) filtering are carried out to accurately extract the building contour. The building contour is extracted from the original Lidar point cloud, the ground point and the non-ground point are separated, and the building contour is extracted from the non-ground point cloud. Firstly, based on the existing skewness balance filtering algorithm, the structure contour is extracted from the non-ground point cloud. By using local fitting height difference instead of elevation, a skewness balanced filter SKF-HD algorithm based on height difference is proposed, which not only keeps the extraction effect of tall ground objects, but also improves the extraction effect of low ground objects. The experimental results of three groups of different terrain and different regions show that the algorithm can extract low ground objects better, and is suitable for different degree of undulating terrain. The overall accuracy of the proposed algorithm increases by 4.8% and 13.3% respectively in the flat region, the small relief area and the significant relief area. Under different terrain conditions, especially in the terrain significant undulating area, the proposed algorithm increases the accuracy of the proposed algorithm by 4.8% and 13.3%, respectively, under different terrain conditions, especially in the region with significant topographic relief. The filtering accuracy of the new algorithm is improved obviously, and then the MBR algorithm is improved, and the multilevel minimum outer rectangle MMBR algorithm is proposed for the occluded regular polygon building. This method can not only accurately get the contour of the unoccluded part of the building, but also get the contour of the occluded part. The experimental results of different shapes show that the proposed algorithm can accurately obtain the building contour of the occluded region, which is compared with the iterative least squares (HLSPC) algorithm with rectangular constraints. The proposed algorithm can accurately extract the building contours in regular rectangular buildings with L-shaped complex polygon and complex polygonal buildings, and the V _ d value decreases by 31.8- 14.3% and 12.5% respectively. The accuracy of MMBR algorithm is higher under different occlusion conditions.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】

中国期刊全文数据库前10条

1 李云帆;谭德宝;高广;刘瑞;;双阈值Alpha Shapes算法提取点云建筑物轮廓研究[J];长江科学院院报;2016年11期

2 李亮;王成;李世华;习晓环;;基于机载LiDAR数据的建筑屋顶点云提取方法[J];中国科学院大学学报;2016年04期

3 孙蒙;顾和和;;联立偏度与峰度变化曲线的机载LiDAR点云滤波方法[J];测绘科学技术学报;2016年01期

4 程效军;程小龙;胡敏捷;郭王;张立朔;;融合航空影像和LIDAR点云的建筑物探测及轮廓提取[J];中国激光;2016年05期

5 董保根;马洪超;车森;解龙根;何乔;;LiDAR点云支持下地物精细分类的实现方法[J];遥感技术与应用;2016年01期

6 胡永杰;程朋根;陈晓勇;李荣;辛俊伟;;改进的参数统计机载雷达点云滤波方法[J];测绘科学;2015年11期

7 韩元利;;网格总分并行式Delaunay三角网建模方法[J];测绘学报;2015年06期

8 尚大帅;周勃;王江涛;付浩;;一套基于虚拟格网的城区LiDAR点云数据滤波流程[J];测绘技术装备;2015年02期

9 林金彦;邹时林;胡永杰;;基于渐进形态学开运算和偏度平衡法的LiDAR数据滤波方法[J];安徽农业科学;2015年16期

10 李乐林;江万寿;郭程方;;基于区域多次回波点密度分析的城区LiDAR建筑物提取[J];国土资源遥感;2015年02期