文献类型：期刊文献

中文题名：机载激光雷达和高光谱组合系统的亚热带森林估测遥感试验

英文题名：The remote sensing experiment on airborne LiDAR and hyperspectral integrated system for subtropical forest estimation

作者：刘清旺[1] 谭炳香[1] 胡凯龙[1,2] 樊雪[1] 李增元[1] 庞勇[1] 李世明[1]

第一作者：刘清旺

机构：[1]中国林业科学研究院资源信息研究所、林业遥感与信息技术重点开放性实验室,北京100091;[2]中国矿业大学(北京)地球科学与测绘工程学院,北京100083

年份：2016

卷号：26

期号：3

起止页码：264-274

中文期刊名：高技术通讯

外文期刊名：Chinese High Technology Letters

收录：CSTPCD;;Scopus;北大核心:【北大核心2014】；CSCD:【CSCD_E2015_2016】；

基金：863计划(2013AA12A302);国家自然科学基金(41201334);国家科技支撑计划(2012BAH34B02)资助项目

语种：中文

中文关键词：森林高度;优势树种;激光雷达(LiDAR);高光谱;分类

外文关键词：forest height; dominant tree species; light detection and ranging（LiDAR）; hyperspectual; classification

分类号：S787.2

摘要：为了提高森林的类型识别及生物物理参数反演精度,采用国产机载激光雷达和高光谱组合系统(ALHIS),选择湖北典型亚热带森林开展了航空遥感试验,获取了试验区激光雷达点云、高光谱和CCD影像数据,提取了森林高度和优势树种类别信息。对数据的分析表明,激光雷达林分平均高的估测精度达到90.67%,激光雷达估测平均高与地面实测胸径加权平均高之间显著相关(R2=0.73,RMSE=1.29m)。按照优势树种分类结果进行统计,发现马尾松、栓皮栎和其它树种的林分平均高分别为9.62m、9.30m、8.79m,不同树种之间的林分平均高相差不大。高光谱优势树种识别总体精度达到82.00%(Kappa=0.70),试验区森林和非森林面积所占比例分别为60.01%和39.99%,马尾松、栓皮栎和其它树种面积在森林中所占比例分别为59.77%、24.99%和15.23%。试验证明,ALHIS能够同时获取高分辨率的植被遥感特征数据,以用于森林制图、优势树种/树种组识别、碳储量估算及生态环境建模等研究。
To improve the accuracy of forest＇s type extraction and biophysical parameters inversion,an aviation experiment on the typical subtropical forest area in Hubei was conducted by using the Airborne Light detection and ranging and Hyperspectral Integrated System（ ALHIS）,and acquired the point cloud data and the hyperspectral and CCD（ Charge Couple Device） images. The forest heights were extracted and the dominate tree species were identified by using these data. The estimation accuracy of average height reached 90. 67% at stand level. The correlation between the average height estimated by using the light detection and ranging（ Li DAR） and the average height of field measurements weighted by DBH（ diameter at breast height） was significant（ R2= 0. 73,RMSE = 1. 29m）.According to the dominant tree species classification,the average heights of Pinusmassoniana Lamb.,Quercusvariabilis Bl. and other species were 9. 62 m,9. 30 m and 8. 79 m,respectively. The variation between different species was not significant. The classification accuracy of dominant tree species using hyperspectual image was 82. 00%（ Kappa = 0. 70）. The proportions of the forest area and the non-forest area were 60. 01% and 39. 99% respectively. The proportions of the areas of Pinusmassoniana Lamb.,Quercusvariabilis Bl. and other species were 59. 77%,24. 99% and 15. 23%,respectively. The experiment shows that the ALHIS can acquire high resolution remote sensing data describing vegetation characteristics for forest mapping,dominant tree species / group species recognition,carbon estimation,ecological environment modeling,etc.