文献类型：期刊文献

中文题名：基于地基激光雷达的亚热带森林单木胸径与树高提取

英文题名：Individual Tree DBH and Height Estimation Using Terrestrial Laser Scanning(TLS) in A Subtropical Forest

作者：刘鲁霞[1] 庞勇[1] 李增元[1]

第一作者：刘鲁霞

通信作者：Pang, Yong

机构：[1]中国林业科学研究院资源信息研究所

年份：2016

卷号：52

期号：2

起止页码：26-37

中文期刊名：林业科学

外文期刊名：Scientia Silvae Sinicae

收录：CSTPCD;;EI(收录号：20161602274998);Scopus(收录号：2-s2.0-85044301403);北大核心:【北大核心2014】；CSCD:【CSCD2015_2016】；

基金：国家高技术研究发展计划(863计划)项目(2012AA12A306);国家重点基础研究发展计划(973计划)项目(2013CB733404;2013CB733406)

语种：中文

中文关键词：地基激光雷达;亚热带森林;胸径;树高;胸高断面积

外文关键词：terrestrial laser scanning; subtropical forest; DBH; tree height; basal area of breast-height

分类号：S757

摘要：【目的】以云南省普洱市天然林与杉木人工林为研究对象,针对云南省山区森林树种繁多、林下灌木草本茂密的林分环境,根据森林中树木的形态特征,利用地基激光雷达(TLS)扫描数据提取样地尺度单木胸径与树高,为森林调查工作提供参考。【方法】将获取的多站地基激光雷达扫描数据分为多站拼接及单站2种分析方式,采用Hough变换算法及树干的形态特征对样地内单木进行识别与胸径提取,根据树干生长方向及单木在垂直方向上的分布提取树高。【结果】1)对于多站拼接数据,即使在林分条件最为复杂的原始林,单木识别率仍可达到81%;对于单站数据,随着扫描距离增加,单木识别率降低,实际操作时单站布设比多站拼接简单;2)多站拼接胸径及胸高断面积估测结果更接近于样地真实值,多个单站平均结果比只使用一站扫描数据提取的结果更加适合估测样地胸径及胸高断面积,半径10 m比半径5 m及15 m范围内数据更加适合估测样地胸径及胸高断面积;3)天然林单木树高估测结果为R^2=0.77,RMSE=1.46 m;人工林单木树高估测结果为R^2=0.94,RMSE=0.96 m。【结论】本研究根据树干垂直向特征,设置的一系列参数可以剔除Hough变换算法在非树干处的识别圆,可提高单木识别及胸径、树高的估测精度。受扫描站布设及林分条件影响,人工林的估测结果好于天然林。多站拼接相比单站扫描更加接近于样地实测结果,多个单站平均更能代表样地实际情况,只用一站数据具有一定的偶然性。
【Objective】Based on the trunk＇s shape and terrestrial laser scanning（ TLS） data,the DBH,height and location of individual tree were estimated in the mountainous forests including natural forest stands and Cunninghamia plantation stands of Yunnan Province. 【Method】DBH and height of individual tree were extracted by merged and single station TLS data. Location and DBH of individual tree were detected and extracted by applying a Hough transform algorithm coupled with trunk＇s shape. Then,tree height was estimated via trunk＇s direction and vertical distribution of canopy. 【Result】1） Based on multiple stations TLS data,the accuracy of tree identification was about 81% in the natural forest stands,with complicated stand structures and compositions. For the single station TLS data,the accuracy of tree identification decreased with the increasing TLS data area. The acquisition of single station TLS data was much easier than that of multiple stations. 2） Mosaicked multiple stations data provided higher accuracies of DBH and basal area of breastheight estimation,as compared to single station data. It was better to use the averages of multiple stations data to derive DBH and basal area of breast-height than single station data. It was more suitable to use data collected within a radius of10 m to estimate DBH and basal area of breast-height than those collected within 5 m and 15 m radii. 3） The tree height estimation（ R^2= 0. 94,RMSE = 0. 96 m） of the plantations was more accurate than that（ R^2= 0. 77,RMSE = 1. 46 m） of the natural forests. 【Conclusion 】According to the feature along trunk,the most detected circle out of trunk could be deleted. This improved the estimating accuracies of tree detection,DBH and tree height. Locating single trunks andestimating DBH and height of individual trees were greatly dependent upon stand structural conditions and distributions of scanning stations. The merged multi scan TLS data extracted the most accurate result compared with averaged multi scan and single scan TLS data. The single scan TLS data showed occasional result.