ICESat-2机载试验点云滤波及植被高度反演

发布时间：2018-07-05 18:56

本文选题：ICESat- + MABEL　；参考：《遥感学报》2014年06期

【摘要】：新一代星载激光雷达卫星ICESat-2将采用多波束微脉冲光子计数技术,并进行高程剖面式的对地观测。由于该点云数据具有背景噪声大、密度低并呈线状分布等特点,传统的点云滤波算法并不适用,研究新的点云滤波算法十分必要。本文以ICESat-2的机载模拟器MABEL数据为例,首先介绍了微脉冲光子计数激光雷达的基本原理和数据特点,并针对高程剖面点云提出基于局部距离统计和最小二乘局部曲线拟合的点云滤波算法;然后,对美国加利福尼亚州Sierras-Forest地区MABEL试验中532 nm通道的光子点云进行滤波处理,并利用识别的地面点插值得到3 m分辨率的线状DEM,进而估算了该区域美国云杉的平均树高;最后,对该滤波算法进行精度评价,并分析了误差来源及其对DEM精度和树高反演精度的影响。结果表明:(1)该算法整体精度达97.6%,能有效剔除绝大部分噪声点且对地形起伏具有较强的自适应能力;(2)误分噪声点影响了滤波过程中局部地形的拟合,而滤波过程中的分类误差将降低DEM和树高反演的精度。
[Abstract]:A new generation of spaceborne lidar satellite ICESat-2 will adopt multi-beam micro-pulse photon counting technique and carry out elevation profile Earth observation. Because the point cloud data has the characteristics of large background noise, low density and linear distribution, the traditional point cloud filtering algorithm is not suitable, so it is necessary to study the new point cloud filtering algorithm. Taking the data of MABEL, an airborne simulator of ICESat-2, as an example, this paper first introduces the basic principle and data characteristics of micro-pulse photon counting lidar. A point cloud filtering algorithm based on local distance statistics and least square local curve fitting is proposed for elevation profile point cloud, and then the photonic point cloud of 532nm channel in MABEL experiment in Sierras-Forest region of California is filtered. The linear DEM with 3 m resolution is obtained by interpolating the identified ground points, and the average tree height of the spruce in this region is estimated. Finally, the accuracy of the filtering algorithm is evaluated. The error source and its influence on Dem accuracy and tree height inversion accuracy are analyzed. The results show that: (1) the overall accuracy of the algorithm is 97.6, which can effectively eliminate most of the noise points and has a strong adaptive ability to terrain fluctuation; (2) the false division of noise points affects the local terrain fitting in the filtering process. The classification error in filtering process will reduce the accuracy of Dem and tree height inversion.
【作者单位】：中国科学院遥感与数字地球研究所
【基金】：国家自然科学基金重大国际合作研究项目(编号:41120114001) 中国科学院百人计划专项
【分类号】：TN958.98

【参考文献】