运用激光雷达数据的单木树冠提取算法对帽儿山林场单木参数估测的影响  

Effect of Three Individual Tree Crown Delineation Algorithms on Estimating Individual Tree Parameters with LiDAR Data

作　　者：李岩[1] 史泽林 程坤 甄贞[1] Li Yan;Shi Zelin;Cheng Kun;Zhen Zhen(Northeast Forestry University,Harbin 150040,P.R.China)

机构地区：[1]东北林业大学

出　　处：《东北林业大学学报》2019年第11期59-65,共7页Journal of Northeast Forestry University

基　　金：东北林业大学大学生创新创业训练计划项目(201710225281)

摘　　要：选取黑龙江省尚志市帽儿山实验林场的10块高郁闭度的天然次生林样地为研究对象,利用激光雷达数据构建的冠层高度模型(CHM),分别使用分水岭分割算法、区域生长法和区域的分层横截面分析(RHCSA)3种方法提取单木位置、树高和冠幅信息。利用手动勾绘的单木树顶和树冠作为参考数据进行精度检验(包括单木树冠提取精度和单木参数估测精度检验),探索不同单木提取算法估测单木参数的可行性。结果表明:RHCSA算法对单木树冠提取的总体精度为83.64%,区域生长算法总体精度为75.19%,分水岭算法总体精度为68.65%;对于单木参数估测,区域生长法与分水岭算法的单木定位精度高(RMSE为1.12 m),RHCSA算法得到最高的树高与冠幅提取精度高(RMSE分别为0.62、1.11 m)。因此,RHCSA算法更适用于帽儿山林场单木树冠提取与单木参数的估测。With ten plots of natural secondary forest with high canopy density in Maoershan Experimental Forest Farm in Shangzhi, Heilongjiang Province, by LiDAR-derived canopy height model(CHM), we applied watershed segmentation, region growing and region-based hierarchical cross-section analysis(RHCSA) to delineate individual tree crowns and estimate tree positions, tree heights and crown diameters. In order to explore the feasibility of estimating individual tree parameters by different individual tree crown delineation algorithms, manually delineated treetops and crowns were used for accuracy assessment including the assessments of crown delineation and parameter estimation. The RHCSA algorithm obtained the highest overall accuracy of crown delineation(83.59%), followed by region growing algorithm(74.39%) and watershed algorithm(68.65%). For parameter estimation assessment, region growing method and watershed segmentation obtained the highest accuracy of individual tree location(RMSE: 1.12 m), and RHCSA algorithm obtained the highest accuracy of estimating tree height and crown diameter(RMSE: 0.62 m and 1.11 m, respectively). Generally, RHCSA algorithm is more suitable for delineating individual tree crowns and estimating individual tree parameters in Maoershan Forest Farm comparing to watershed segmentation and region growing.

关 键 词：树冠 分水岭分割算法 区域生长法 区域的分层横截面分析算法 激光雷达 

分 类 号：S758.4[农业科学—森林经理学]