一种顾及地形复杂度的LiDAR点云多尺度滤波方法

A multi-scale filtering method for LiDAR point clouds considering terrain complexity

针对复杂地形区域的机载LiDAR数据滤波方法中自适应阈值设置问题,根据地形多尺度效应,提出一种自适应阈值的机载LiDAR点云多尺度滤波方法。该方法采用影像金字塔策略按分辨率从高至低逐级构建LiDAR点云分层格网,滤波过程则从最大尺度格网(顶层格网,最低分辨率)开始,采用局部统计分析的方法自适应地确定高差阈值,同时结合薄板样条内插出下层各格网控制点的高程值,直至最底层格网完成原始激光点云滤波。通过我国某山区城市复杂地形的LiDAR数据实验表明顾及地形复杂度的LiDAR点云多尺度滤波方法能够快速有效地提取高精度DEM,能够满足实际生产需求。

Focus on the problem of adaptive threshold determining in filtering of airborne LiDAR in complexity region, a multi-scale filtering approach in which the threshold for point filtering is determined adaptively by the multi-scale effect is reported in this paper. Firstly, hierarchical grids of the LiDAR point clouds are created from high resolution to low resolution as the image pyramid. Secondly, the filtering pro- cedure is executed iteratively at each level in a top-down fashion. The adaptive threshold is determined by the local statistical analysis in local regions, and then the interpolated surface is estimated with thin plate spline based on the filtered control point set at the previous level. This progress is iteratively executed until the last level grid is finished, which means the filtering of the LiDAR point clouds is achieved. Accuracy assessment results in a hilly city area indicate that the proposed multi-scale filtering method which consider the complexity of terrain can produce the high quality DEM effectively, which can he used in application.