中文摘要：

针对建筑物立面自动提取难题,根据多角度航空倾斜影像密集匹配生成的三维点云,提出了一种顾及点云几何及颜色信息的建筑物立面点云自动提取方法。首先通过计算三维点云中各点对应的法向量,并根据法向量方向进行粗略的分割;然后根据各点的颜色信息剔除植被点云;在此基础上剔除少量地面点,通过基于聚类分析的后处理获取立面点云。利用2组数据进行了实验,结果表明,本文方法能够自动提取出建筑物的立面,立面的完整性和正确性都大于90%,准确率大于83%,为后续立面的重建提供了基础。

英文摘要：

Aimed at tackling the difficuhy in extracting facade from point cloud derived from dense matching of multi- angle airborne oblique images, this paper proposes a facade extraction method considering point cloud geometrical and color information. This method calculates normal vector for each point in the point cloud, and then uses the orientation of the normal vector for initial facade segmentation. After that, color information is used to remove point cloud about vegetation. On the basis of the initial result, the remaining ground points are removed. Finally, clustering analysis is used to refine the result, and facade point cloud can be obtained. Two groups of data sets are used for experiments, and the results reveal that the proposed method can automatically extract the facade of the building, with the completeness and correctness up to 90%, and the accuracy is more than 83%, thus providing the basis for the subsequent facade reconstruction.