顾及轮廓特征的建筑物对象级三维变化检测方法  

3-D Change Detection Method of Buildings’ Object-Level Considering Contours Characteristics

作　　者：陈奎伊 胡翰[2] 丁雨淋 李晓明 卞雨凡 CHEN Kuiyi;HU Han;DING Yulin;LI Xiaoming;BIAN Yufan(Faculty of Geosciences and Environmental Engineering,Southwest Jiaotong University,Chengdu 611756,China;Department of Land Surveying and Geo-information,Hong Kong 999077,China;Research Institute for Smart Cities,School of Architecture and Urban Planning,Shenzhen University,Shenzhen 518060,China;Shenzhen Research Center of Digital City Engineering,Shenzhen 518034,China)

机构地区：[1]西南交通大学地球科学与环境工程学院,四川成都611756 [2]香港理工大学土地测量与地理资讯学系,香港999077 [3]深圳大学智慧城市研究院,广东深圳518060 [4]深圳市数字城市工程研究中心,广东深圳518034

出　　处：《地理信息世界》2019年第5期101-106,共6页Geomatics World

基　　金：四川省科技计划项目(2018SZ0314);国家自然科学基金项目(41631174,61602392)资助

摘　　要：建筑物对象级三维变化检测是城市三维建筑模型更新及违法建筑查违等应用的核心关键技术。针对现有变化检测方法的限制,创新性提出了顾及轮廓特征的建筑物对象级三维变化检测方法,充分利用倾斜实景三维模型来进行变化检测。首先通过高度分析提取建筑物的轮廓线,同时过滤掉植被和地形等无关因素的影响;然后将不同高度的轮廓进行拟合后进行对比检测,并构建三维包围盒。该方法针对建筑物进行变化检测,检测对象单一,语义明确,既避免了对差值DSM滤波时不够彻底,又准确地保留了建筑物的有效信息,同时筛选出了未发生对象级变化的建筑物,可以更高效地对其做进一步的结构级变化检测。Buildings' object-level 3-D change detection is the key technology of urban 3-D building models updating and illegal buildings inspection. In view of the limitations of existing change detection methods, this paper innovatively proposes a 3-D change detection method of buildings' object level, which considers the contour characteristics and makes full use of the oblique scene 3-D models to change detection. The first is to extract the contours of buildings by heightanalysis method and filter the influence of independent factors such as vegetation and terrain;then it is to fit the contours in different heights for comparative detection, and construct a 3-D box coating the buildings. This method performs change detection only on buildings with homogeneous object and clear semantics, not only avoids the incomplete filtering of DSM, but also accurately retains the effective information of buildings, moreover, buildings with no object-level changes are screened out, which can be more efficiently detected for further structural-level change.

关 键 词：变化检测 对象级 高度分析 轮廓提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]