中南林业科技大学学报2025，Vol.45Issue(8)：29-40,12.DOI:10.14067/j.cnki.1673-923x.2025.08.004
基于立木拓扑特征的ULS-TLS点云融合算法
Based on the standing tree topological features of ULS-TLS point cloud fusion algorithm:taking Ginkgo plantation as an example
摘要
Abstract
[Objective]This study constructed a fusion algorithm for topological characteristics of standing trees(TCS)based on tree positions,attempting to register and fuse point cloud data from different platforms to extract high-precision single-tree forestry measurement factors.[Method]Leafed and leafless scans were conducted on flat and sloped terrains of a ginkgo plantation.TLS point cloud data served as source data,while ULS point cloud data was used as target data.Ground and canopy topological features were provided.The KNN algorithm was applied in a defined search radius to find neighboring tree points,constructing triangulated irregular network(TIN)pairs.A similarity matrix(TSM)was built by comparing the angle and area of each triangle in the TIN pairs,and matching"tree-tree"pairs were identified for initial coarse registration.Final fine registration was achieved using ICP algorithm.Mensuration factors for single trees were extracted from platform and fused point clouds,with accuracy evaluation.[Result]1)After fine registration and fusion of point cloud datasets coarsely registered using the TCS algorithm,the RMSE(reg)values were all less than 0.170 m.During the leaf-off period,the RMSE(reg)errors for flat and sloped terrain decreased by 0.237 m and 0.445 m,respectively.During the leaf-on period,the RMSE(reg)errors for flat and sloped terrain decreased by 0.046 m and 0.170 m,respectively.The errors for flat terrain were consistently lower than those for sloped terrain.During the leaf-off period,the RMSE(reg)for flat terrain(0.088 m)decreased by 39.7%compared to the leaf-on period(0.146 m),while the decrease for sloped terrain was 42.6%.This confirmed that the registration accuracy during the leaf-off period was superior to that during the leaf-on period.Additionally,the fusion results for the same tree showed improved integration of the crown and trunk,with a further reduction in the offset between the trunk and crown sections;2)The R2 values for DBH extraction from fused point cloud ranged from 0.944 to 0.992,and RMSE values were 1.734 cm to 2.108 cm,showing little difference from the extraction results of TLS point cloud;3)Tree height extraction from fused point clouds achieved optimal results,with R2 ranged from 0.825 to 0.902 and RMSE ranged from 0.995 to 1.840 m,better than TLS and ULS point clouds;4)The Crown width extraction from fused point clouds was superior,with R2 ranged from 0.817 to 0.861 and RMSE ranged from 0.963 to 1.334 m.[Conclusion]The accuracy of mensuration factors extracted from fused point clouds verifies the algorithm's applicability and superiority in the study area,offering a new approach for ULS-TLS laser scanning in forest surveys under varying conditions.关键词
激光雷达/点云融合算法/单木测树因子提取/银杏人工林Key words
LiDAR/point cloud fusion algorithm/extraction of mensuration factors for single tree/Ginkgo plantation分类
农业科技引用本文复制引用
李纪霖,孙圆,纪北京,张忻慧,温小荣,刘玉华,余鹏飞..基于立木拓扑特征的ULS-TLS点云融合算法[J].中南林业科技大学学报,2025,45(8):29-40,12.基金项目
江苏省研究生科研与实践创新计划项目(SJCX23_0345). （SJCX23_0345）
