LiDAR辅助下利用超高分辨率影像提取建筑物轮廓方法  

Building Boundary Extraction Using Very High Resolution Images and LiDAR

作　　者：程亮[1] 龚健雅[2] 

机构地区：[1]南京大学地理信息科学系,江苏南京210093 [2]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079

出　　处：《测绘学报》2008年第3期391-393,399,共4页Acta Geodaetica et Cartographica Sinica

基　　金：国家973项目(2006CB701300)

摘　　要：以精确自动提取建筑物轮廓为目标,提出一种LiDAR辅助下利用超高分辨率影像进行轮廓提取的新方法。其要点分为4步:①预处理,检测LiDAR中建筑物点并分割成每栋建筑物的点集;②建立轮廓提取区,针对每栋建筑物做缓冲区和外接矩形,通过缓冲区过滤和外接矩形切割,建立轮廓提取感兴趣区域;③线段提取,借助LiDAR估算出建筑物概略主方向,并在该方向的约束下,自动、鲁棒地检测出建筑物的主方向和建筑物的线段;④轮廓筛选,基于LiDAR密度分析与Kmeans聚类动态筛选出精确轮廓。本方法所提取的建筑物轮廓定位精确、细节完好,轮廓提取准确率91%。In order to extract geometric precise and detailed building boundary, a new approach is presented integrating very high resolution images and LiDAR data. The process consists of a sequence of four steps. ① Pre-processing-identify segmented building points from raw LiDAR data. ② Create building images-af- ter creating a buffer and a bounding rectangle of a building according the segmented building points, filter the original image using the buffer and then cut the result image by the bounding rectangle.③ Line segments extraction-most buildings have perpendicular structure with two rectilinear axes. So an automatic and robust method on building principal orientations estimation is presented based on rough principal orientations constraint, thus improving the accuracy and robustness of line segments extraction. ④ Boundary selection-a selection strategy is proposed based on LiDAR data density analysis and Kmeans clustering. The results demon strated that the proposed approach determined building boundaries well.

关 键 词：LIDAR 超高分辨率影像 线段提取 轮廓筛选 

分 类 号：P237[天文地球—摄影测量与遥感]