基于车载序列影像的高建筑密度城市三维建筑物重建方法

项目名称： 基于车载序列影像的高建筑密度城市三维建筑物重建方法

项目编号： No.41201471

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 地理学

项目作者： 田一翔

作者单位： 同济大学

项目金额： 25万元

中文摘要： 移动测量系统是满足空间信息数据采集加工在成本和更新速度两个方面需求的典型技术代表。针对高建筑密度的大型城市建筑物重建中的难题，本申请基于移动测量系统获取的多视序列影像立体像对，充分利用序列影像间高重叠度、较小的强度变化、尺度变化的特点，通过大量的冗余信息确定同名特征，从而确保提取点、线特征的可靠性。通过分析建筑物与城市地理环境间的关系，结合建筑物数字化平面图，将目标的识别与重建相结合，实现建筑物的快速识别。在常见建筑物模型知识的指导下重建建筑物多面体模型，从而避免遮挡、纹理缺乏等因素对特征匹配准确度和稳定性的影响，获得建筑物自身及与城市地理环境拓扑关系正确的几何模型。再通过线段约束下的密集匹配方法重建建筑物细节信息，提高重建模型的几何精度。该研究将形成一套基于移动测量系统的三维建筑物可靠重建方案，为城市，特别是高建筑密度大型城市的三维模型的构建和大型基础设施的变化监测提供坚实的基础。

中文关键词： 序列影像；建筑物重建；特征提取；信息融合；拓扑重建

英文摘要： The mobile mapping system is a fast and economic spatial information acquisition technique. A high percentage of overlap between adjacent images, together with fewer photometric changes in images, are the main advantages of using stereo image sequences as the source data for object reconstruction. The high overlap of images in image sequences leads to a high redundancy of observations for each feature. Corresponding features in image space can therefore be matched or tracked more easily and reliably than that matched from a few images. In order to solve the building reconstruction problems of the big cities with high building density, this application analyzes the geometrical and topological relations between buildings and their relations with city environment. The building recognition can therefore be solved by fusing information from three dimensional features, building ground plane and topological relation between buildings and city environment. The knowledge of generic building structures can be implemented as rules and constraints, which provide essential guidance for recovering building's main structures and get rid of occlusions and lack of texture. Dense matching with the edge constraints is then applied to recover details on the building facades. This study will form a reliable method based on mobile ma

英文关键词： image sequence；building reconstruction；feature extraction；information fusion；topological reconstraction