一种渐进加密三角网LIDAR点云滤波的改进算法

Advanced adaptive TIN filter for LIDAR point clouds data

传统滤波算法常常是针对某些特定的具有连续表面的区域来进行,带有一定的局限性。这样不能解决对诸如斜坡、密集植被等复杂地区进行真实准确地形提取的难题。本文提出了一种融合区域增长方法的LIDAR点云数据的滤波方法。该方法在对原始点云数据进行预处理的基础上,生成一个松散的TIN,通过迭代生成加密的TIN,其中通过采用两次区域增长的滤波方法,实现原始地形的提取。最后通过对复杂典型数据的对比实验验证了文中所提出的改进方法的有效性和准确性。

Most LIDAR filters aim at continuous terrain area, which cannot extract complicated terrain feature accurately, such as cliff or abrupt slope, slope with dense vegetation. This paper presents an advanced adaptive TIN filter to solve the problem. First, filter is based on a sparse TIN, and then progressively densified to LIDAR points. In each iteration, points within the edge and angle threshold are added to TIN, if the point meets certain criteria in relation to the triangle that contains it. At last, the region-growing method is employed to add points on cliff or abrupt slope into terrain. Experiment on typical complex terrain features proves the validity and precision of the method presented here.