%0 Journal Article
%T 像素级和目标级耦合的三维建筑物变化检测方法<br>Pixel-Level and Object-Level Combined 3D Building Change Detection Method
%A 张志华
%A 张璐琪
%A 丁鹏辉
%A 朱文嘉
%A 丁晓龙
%A 杨必胜
%J Geomatics Science and Technology
%P 216-224
%@ 2329-7239
%D 2023
%I Hans Publishing
%R 10.12677/GST.2023.113024
%X 城市建筑物三维变化检测可以服务于城市精细化管理、基础数据库更新以及灾害评估。随着城市的快速发展，城市建筑物变化类型更加复杂，现有变化检测方法难以满足需求。本文提出了一种像素级与目标级耦合的三维建筑物变化检测方法。本方法首先将两时相机载激光点云格网化并利用对应格网高度差异定位像素级变化区域；然后利用机载激光点云生成建筑物目标；最后联合像素级和目标级的变化信息，基于监督学习的方法判断建筑物目标的变化类型。利用本文方法在机载激光点云数据集上实验并进行定量评价，召回率和准确率分别为90.3%和84.8%。实验结果表明提出的方法可以对复杂城市场景建筑物变化做出精准定位及类型判断，并可以应用于违建发现。<br />
The 3D building change detection can be used for urban refinement management, basic database updating and disaster assessment. With the rapid development of cities, the change types of urban buildings have become more complex, and the existing change detection methods can hardly meet the requirements. In this paper, a pixel-level and object-level combined change detection method is proposed. Firstly, the height difference obtained by gridding the two temporal airborne point clouds is used to locate the pixel-level change area; then the building point cloud is used to generate building objects; finally, the change information at the pixel-level and object-level is combined and the change type is determined based on a supervised learning method. Using the proposed method to detect building changes in the airborne point cloud dataset, the recall and accuracy were 90.3% and 84.8% respectively. The experimental results show that the proposed method can accurately determine the building change types in complex urban areas and can be applied to illegal building detection.
%K 机载激光点云，变化检测，违建发现，监督学习；Airborne LiDAR Data
%K Change Detection
%K Illegal Construction Discovery
%K Supervised Learning
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=68482