基于倾斜影像点云的建筑物提取算法研究

Research on Building Extraction Algorithm Based on Oblique Image Point Clouds

多视倾斜影像密集匹配后能够生成海量点云数据,但数据本身缺乏有效的建筑物分类信息.针对此问题,提出一种基于倾斜影像点云的建筑物提取算法.首先对三维点云进行去噪处理和植被的剔除,将点云进行空间格网分区后降维到二维平面,通过赋予二维平面格网内每个点一定的权值,进而对格网进行特征值重采样.然后对二维平面进行图像形态学处理,利用骨架提取算法找到建筑物的轮廓,将该轮廓和三维点云进行融合后重新整饰建筑物边缘,提取出最终的建筑物点云.两组实验结果表明,该方法提取的建筑物点云轮廓清晰,建筑物立面提取较好,该算法具有较好的稳健性.

Multi-view oblique images can generate massive point cloud data after intensive matching,but the datalack effective building classification information. Aiming at this problem,this paper proposes a building extractionalgorithm based on oblique image point clouds. First,performing the noise reduction on the 3D point clouds and theremoving the vegetation points,after the point clouds are partitioned by the spatial grids,the point clouds arereduced to the 2D plane. The grid’s eigenvalues are resampled by assigning a certain weight to each point in the 2D plane grids. Then the two-dimensional plane is processed by image morphology and by skeleton extraction algorithm tofind the outline of the building. After merging the contour with the three-dimensional point clouds,the edges of thebuilding are re-polished to extract the final building point clouds. The experimental results of the two groups show thatthe point clouds of the building extracted by this method are clear and the building facades are well extracted.The algorithm has good robustness.