Deformation monitoring of structures is a common application and one of the major tasks in surveying engineering, but there is a great challenge for the coordinate datum unification of multitemporal point cloud data. To address this problem, an automatic detection method of the virtual reference datum in multitemporal point cloud is proposed in this article. To obtain the corresponding grids of multitemporal point cloud, an appropriate coarse registration algorithm is adopted to approximately transform the multitemporal data into a unified coordinate system. Then, the stable areas are extracted based on the probability density functions similarity of the corresponding grids, which are defined as the virtual reference datum. Furthermore, an improved 3D normal distribution transform algorithm considering the cell boundaries and iteratively selecting an appropriate cell size is constructed to achieve fine registration of the virtual reference datum. Finally, the coordinate unification of the multitemporal point cloud is implemented according to the transformation parameters of the virtual reference datum. The proposed method is tested on a landslide point cloud, which is captured by static terrestrial laser scanning. The virtual reference datum extraction accuracy of the two-temporal landslide point cloud captured on the same day is 96%, and the coordinate unification accuracy is 3 mm. The experimental results demonstrate that the proposed method is effective in coordinate datum unification of multitemporal point cloud.

中文翻译：

---

一种用于变形监测的多时相点云虚拟参考基准自动检测方法

结构变形监测是测绘工程中的常见应用和主要任务之一，但多时相点云数据的坐标基准统一面临巨大挑战。针对这一问题，本文提出了一种多时相点云中虚拟参考基准的自动检测方法。为获得多时相点云对应的网格，采用适当的粗配准算法将多时相数据近似变换为统一坐标系。然后，根据对应网格的概率密度函数相似度提取稳定区域，定义为虚拟参考基准。此外，构建了一种改进的 3D 正态分布变换算法，考虑单元边界并迭代选择合适的单元大小，以实现虚拟参考基准的精细配准。最后，根据虚拟参考基准的变换参数实现多时相点云的坐标统一。所提出的方法在由静态地面激光扫描捕获的滑坡点云上进行了测试。同一天捕获的双时相滑坡点云虚拟参考基准提取精度为96%，坐标统一精度为3 mm。实验结果表明，该方法在多时相点云坐标基准统一中是有效的。