ICESat-2(Ice, Cloud, and land Elevation Satellite-2)是美国NASA(National Aeronautics and Space Administration)在2018年发射的激光测高卫星,其上搭载的激光测高系统ATLAS(Advanced Topographic Laser Altimeter System)采用微脉冲多波束光子计数激光雷达系统,因其低能耗、高探测灵敏度、高重复频率的特性极大改善了沿轨采样密度,但也使获取的数据中包含大量的噪声,如何有效实现光子点云去噪分类成为后续应用的关键,也是当前研究的热点和难点,为此本文提出一种基于卷积神经网络的光子点云去噪和分类算法。首先将光子点云按照沿轨和高程方向划分格网,去除明显的噪声光子,并将每个粗信号光子点栅格化为影像;然后基于少量样本构建的卷积神经网络分类模型实现光子点云精去噪和分类;最后利用机载激光雷达数据进行验证,并与ATL08产品的去噪分类结果进行对比。结果表明,对于裸地和森林区域,卷积神经网络算法均能有效去除噪声光子,特别对于森林区域,可同时实现去噪和分类;其中,裸地区域地表计算的R2和RMSE分别为1.0和0.72 m,森林区域地表和树冠计算的R2分别为1.0和0.70, RMSE分别为1.11 m和4.99 m。本文利用深度学习算法实现光子点云去噪分类,在裸地和森林区域均取得了较好的结果,为后续光子点云数据处理提供了参考。 相似文献
Glaciers are one of the most important land covers in alpine regions and especially sensitive to global climate change. Remote sensing has proved to be the best method of investigating the extent of glacial variations in remote mountainous areas. Using Landsat thematic mapping (TM) and multi-spectral-scanner (MSS) images from Mt. Qomolangma (Everest) National Nature Preserve (QNNP), central high Himalayas for 1976, 1988 and 2006, we derived glacial extent for these three periods. A combination of object-oriented image interpretation methods, expert knowledge rules and field surveys were employed. Results showed that (1) the glacial area in 2006 was 2710.17 ± 0.011 km2 (about 7.41% of the whole study area), and located mainly to the south and between 4700 m to 6800 m above sea level; (2) from 1976 to 2006, glaciers reduced by 501.91 ± 0.035 km2 and glacial lakes expanded by 36.88 ± 0.035 km2; the rate of glacier retreat was higher in sub-basins on the southern slopes (16.79%) of the Himalayas than on the northern slopes (14.40%); most glaciers retreated, and mainly occurred at an elevation of 4700–6400 m, and the estimated upper limit of the retreat zone is between 6600 m and 6700 m; (3) increase in temperature and decrease in precipitation over the study period are the key factors driving retreat. 相似文献