地球信息科学学报 ›› 2018, Vol. 20 ›› Issue (4): 462-470.doi: 10.12082/dqxxkx.2018.170470

• 全国激光雷达大会特约稿件 • 上一篇    下一篇

车载激光扫描数据中建筑物立面快速提取

邵磊1(), 董广军1, 于英1,2,3, 姚强强1, 张阿龙1   

  1. 1. 信息工程大学,郑州 450001
    2. 地理信息工程国家重点实验室,西安 710054
    3. 城市空间信息工程北京市重点实验室,北京 100000
  • 收稿日期:2017-10-09 修回日期:2017-12-23 出版日期:2018-04-20 发布日期:2018-04-20
  • 作者简介:

    作者简介:邵 磊(1992-),男,湖南常德人,硕士生,主要从事激光扫描数据的分类与目标提取研究。E-mail: shaoleixinda@163.com

  • 基金资助:
    国家自然科学基金项目(41501482);地理信息工程国家重点实验室开放研究基金(SKLGIE 2015-M-3-6);城市空间信息工程北京市重点实验室经费资助项目(2017203)

A Fast Method of Building Extraction from Mobile LiDAR Scanning Data

SHAO Lei1,*(), DONG Guangjun1, YU Ying1,2,3, YAO Qiangqiang1, ZHANG Along1   

  1. 1. Information Engineering University, Zhengzhou 450001, China
    2. State Key Laboratory of Geo-information Engineering, Xi′an 7100541, China;
    3. Beijing Key Laboratory of Urban Spatial Information Engineering, Beijing 100000, China
  • Received:2017-10-09 Revised:2017-12-23 Online:2018-04-20 Published:2018-04-20
  • Contact: SHAO Lei E-mail:shaoleixinda@163.com
  • Supported by:
    National Natural Science Foundation of China, No.41501482;State Key Laboratory of Geo-information Engineering, No.SKLGIE 2015-M-3-6;Beijing Key Laboratory of Urban Spatial Information Engineering, No.2017203.

摘要:

本文提出一种结合多种投影影像从车载激光扫描数据中提取建筑物的方法。该方法首先将点云数据投影到XOY平面,生成多种投影影像;然后结合建筑物几何语义特征,对已获取的投影影像进行几何约束与形态学运算,得到建筑物种子区域;在此基础上,通过设置高差阈值,在最高高程影像上进行建筑物种子区域的八邻域区域生长,得到建筑物区域;最后将影像上的建筑物区域反投影到三维空间,提取出建筑物目标。实验结果表明,该方法能有效提取点云数据中的建筑物立面,取得较高的正确率和完整率,且大大提高了计算效率。

关键词: 几何特征, 投影影像, 建筑物提取, 种子区域, 车载激光扫描系统, 形态学运算

Abstract:

The mobile LiDAR scanning system is a useful tool for getting the top information as well as fa?ade information of buildings, which makes it as primary means to obtain 3D city modeling infrastructure. The first step of 3D modeling is to extract the building data from the complex mobile point cloud data quickly and accurately. Therefore, it is of great significance to study a fast and effective method of building extraction from vehicle laser scanning data. The buildings in mobile laser scanning data has the characteristics of uneven distribution of point densities, lack of existence, and some of the buildings in the measured data are not strictly flat, the top data of the low building is not the fa?ade. In order to solve the problems discussed above, a method of building extraction in complex urban scenes from mobile LiDAR is proposed by using a variety of projection images. Firstly, the method projects the point cloud data into the XOY plane to produce a variety of projected images. Secondly, based on the geometric semantic features of the buildings, the geometric constraints and morphological calculations of the acquired projection images are processed to get the seed area of the building. On the basis of this seed area, the eight-neighborhood region of the building seed area is grown on the highest elevation image by setting the height difference threshold to obtain the building area. Lastly, the building area on the image is back-projected into three-dimensional space to extract the building objectives. Two data sets with different point densities and different scanning tools are used to verify the effectiveness of this method. Results show that this method has higher data processing efficiency than the existing three-dimensional extraction method because point cloud data is projected into the two-dimensional image and the geometrical features of the building are synthetically used in the process of building extraction. Using this method, both top surface and non-fa?ade buildings can be extracted precisely. In this paper, sub-regional growth methods solve the problem that it′s difficult to extract the buildings with incompleteness of cloud data which is caused by the blockage through the traditional projection method.

Key words: geometric features, projection images, building extraction, seed area, mobile LiDAR, morphological operations