建筑物边界约束的航空影像DSM优化

Optimization of Aerial DSM with Building Boundary Constraint

针对航空影像密集匹配生成点云数据边界模糊的问题,提出了一种基于DSM灰度影像矢量边界与DEM无约束D-三角网嵌套生成具有精确边界的建筑物表面模型的方法。通过逐点内插法建立实验区点云数据的DSM深度影像图;根据计算机视觉中的边缘检测算子,提取深度影像中建筑物的准确边界;建立DEM的无约束D-三角网,将准确建筑物边界作为硬边界嵌入三角网中,最终将建筑物三角网和地面点三角网拼合,生成"纯净"建筑物表面模型。实验结果表明,优化后的建筑物高度和平面信息无精度损失,该方法有较强实用性。

To solve the problem that building boundary of DSM interpolated from points cloud generated by dense matching of aerial images was indistinct, this paper proposed a new method with the nestification between vectorial boundary based on depth-image of DSM and D-triangulation of DEM to generate Building DSM（BDSM）. Depth-image of experimental area DSM was built through point by point interpolation of DSM. Building boundary in it was extracted by edge detector in field of computer vision. Then nest it into constrained D-triangulation of DEM as hard edge. Finally, combine building with DEM D- triangulation and get a building DSM with clear boundary. Experimental results show that there is no loss of accuracy comparing with before and after optimization in terms of height and plain information.