基于自适应和多尺度数学形态学的点云数据滤波方法

An Adaptive and Multi-Scale Mathematic Morphological Filter for Point Cloud Data Filtering

数学形态学滤波是从激光雷达(LIDAR)点云数据中识别地面点、创建数字高程模型的一种重要方法。在分析现有滤波方法的优劣性以及数学形态学滤波方法存在的问题的基础上,提出一种改进的、多尺度的、具有一定自适应性的数学形态学滤波算法。该方法通过构建一个粗糙的地面不规则三角网(TIN)来确定实验区域的地形特征,从而选择恰当的地形坡度参数,用于多尺度数学形态学滤波中高差阈值的计算,将每次迭代前后高差小于阈值的点划分为地面点,同时保留地面点的高程值用于下次的迭代计算。实验结果表明该方法能够有效识别地面点和地物点,并且保留地形的细节信息。

Mathematic morphological filter is one of the most important methods to detect ground points from LIDAR data and build the DEM. On the basis of the analysis of the advantage and disadvantage of the existing filtering methods and the problems of mathematic morphological filter, an improved, adaptive and multi-scale mathematic morphological filter was proposed. At first, a coarse TIN was built to confirm the ground feature of test area, and the slope parameter of ground was determined by the TIN. Then, the slope parameter was used for the computation of elevation difference threshold. If the elevation difference of point between previous and this filtered terrain surface was lower than the elevation difference threshold, it was classified to be ground point, otherwise it was classified to be non-ground point. At the same time, the height value of the ground point was reserved to the next iteration. The result of test shows that this filter could remove most non-ground points effectively, and reserve the detail information of the ground.