基于改进的渐进多尺度数学形态学的激光雷达数据滤波方法  

Ladar Data Filtering Method Based on Improved Progressive Multi-Scale Mathematic Morphology

作　　者：赵明波[1,2] 何峻[1] 田军生[3] 付强[1] 

机构地区：[1]国防科技大学电子科学与工程学院自动目标识别重点实验室 [2]中国人民解放军93246部队 [3]空军装备研究院雷达与电子对抗研究所

出　　处：《光学学报》2013年第3期285-294,共10页Acta Optica Sinica

摘　　要：激光雷达点云数据的滤波处理是激光雷达数据处理的基础和至关重要的步骤,基于数学形态学的滤波算法应用广泛。针对现有基于数学形态学的滤波算法在处理包含大面积空白区域的点云数据时遇到的问题,提出了一种改进的渐进多尺度数学形态学滤波算法,通过改进形态开运算来处理空白区域。根据数学形态学滤波的基本原理,证明了改进形态开运算的可行性。详细阐述了所提算法的基本步骤及流程,并分析了其性能特点。利用仿真数据和公开测试数据,对所提算法进行了实验验证。实验结果表明,所提算法对存在大面积空白区域的点云数据具有良好的滤波性能;与其他典型滤波算法相比,滤波性能更优。Ladar point-cloud data filtering is basis and critical steps of the ladar data processing. Filtering algorithm based on mathematical morphology is applied widely. But the existing filtering algorithms have some problems when processing point-cloud data which contains some blank region of large area. To solve these problems, an improved progressive multi-scale mathematical morphology filtering algorithm is proposed. It deals with the blank region by improving morphological opening operation. According to basic principles of mathematical morphology, feasibility of improved morphological opening operation is proved. Then basic steps of the proposed algorithm are elaborated, and performance characteristics are analyzed. The proposed algorithm is verified experimentally by using simulation data and public test data. Experimental results show that the proposed algorithm can perfectly filter point-cloud data which contains some blank region of large area, and it has better filtering performance compared with other representative filtering algorithm in most cases.

关 键 词：遥感 激光雷达数据滤波 数学形态学 数字高程模型 点云数据 图像处理 

分 类 号：P228[天文地球—大地测量学与测量工程]