A morphology-based method for building change detection using multi-temporal airborne LiDAR data
A morphology-based method for building change detection using multi-temporal airborne LiDAR data
复制标题
DOI：
10.1080/2150704x.2017.1402384
复制
发表时间：
2018-02
影响因子：
2.3
通讯作者：
Yuanzhe Xi;Qingli Luo
中科院分区：
文献类型：
--
作者：
Yuanzhe Xi;Qingli Luo
摘要建筑物变化检测在土地管理和城市规划中具有重要作用。提出了一种利用多时相机载光探测和测距(LiDAR)数据对建筑物变化进行自动检测的方法。该方法利用不同时间采集的两组点云数据，利用不规则三角网(TIN)滤波和支持向量机(SVM)将土地利用分为地面、建筑物和植被三种类型。之后，应用形态算法将变化的建筑物与树木区分开来，大大减少了识别的误差率。利用规则分析方法，将变化的建筑对象分为新建、较高和拆除三种类型。案例研究区位于沈阳市刘家村，中国。这两组LiDAR测试数据分别是由徕卡ALS 60(3分/平方公里)和ALS 80(15分/平方公里)从2014年9月和2017年2月采集的。现场调查证明，识别出的新建、拆除和较高变化建筑的准确率分别为95.4%、92.9%和100%。该方法具有较高的准确性和可靠性，为建筑物变化检测提供了参考。
ABSTRACT The detection of building changes plays an important role for land management and urban planning. This study proposes an automatic method that applies the morphology-based method for detection of building changes with multi-temporal airborne Light Detection and Ranging (LiDAR) data. In this method, three types of land use are classified as ground, building and vegetation by Triangulated Irregular Network (TIN) filter and Support Vector Machine (SVM) with the two sets of point clouds acquired at different times. Thereafter, the morphological algorithm is applied to distinguish the changed buildings from trees, which significantly reduces the identification of errors. The changed building objects are classified as three types: ‘newly built’, ‘taller’ and ‘demolished’, with the use of rule analysis. The case study area is located in Liujia village, Shenyang, China. The two sets of LiDAR test data were acquired by Leica ALS 60 (3 points/km2) and ALS 80 (15 points/km2) from September 2014 and February 2017, respectively. Field investigation proves that the accuracy of the identified changed buildings of ‘newly built’, ‘demolished’ and ‘taller’ are 95.4%, 92.9% and 100%. The proposed method has the advantages of high accuracy and reliability, which will provide a reference for the detection of building changes.