机载LiDAR点云数据的多级滤波方法

Multilevel Method for Airborne LiDAR Point Clouds Data Filtering

激光雷达数据滤波是获取高精度数字高程模型的重要手段和前提条件。为了实现对激光雷达数据的准确滤波,在分析部分经典算法特点的基础上,提出一种机载Li DAR点云数据的多级滤波算法。该方法首先通过格网化分建立数据索引,在局部区域利用均值限差法得到可靠的初始地面种子点,求解拟合曲面系数,在降低Ⅱ类误差的前提下,利用曲面拟合算法进行多次迭代滤波,然后使用三角形角度限制法判断剩余点,最终得到较为准确的滤波结果。基于ISPRS的数据实验表明,与经典算法相比,所提方法在Ⅱ类误差较低的情况下总体滤波精度高,能够得到可靠的滤波结果,具有较高的可行性。

LiDAR data filtering is an indispensable means and precondition for obtaining high-precision digital elevation model. Based on analyzing the characteristics of some classical algorithms, a multi-level filtering algorithm for airborne LiDAR point cloud data is proposed. Firstly, the data index is established by grid points in this method. Then a reliable initial ground seed point is obtained by the mean limit-difference method in the local area, and the coefficient of fitting surface is solved. Then on the premise of reducing the class II error, the surface fitting algorithm is used to implement multiple-iteration filtering and the triangle angle restriction method is then used to judge the remaining points, thus more accurate filtering results are obtained. Finally, the data experiments based on ISPRS indicate that compared with the classical algorithm, the proposed method has higher accuracy and better reliability, and this method could acquire reliable filtering results and is of fairly high feasibility.