南京林业大学学报(自然科学版) ›› 2025, Vol. 49 ›› Issue (2): 185-193.doi: 10.12302/j.issn.1000-2006.202306010

• 研究论文 • 上一篇    下一篇

基于QSM的落叶松一级枝条数量提取与模型构建

彭文越1,2(), 贾炜玮1,3,*(), 王帆1, 李鑫1,2, 李丹丹1   

  1. 1.东北林业大学林学院,黑龙江 哈尔滨 150040
    2.吉林省林业调查规划院,吉林 长春 130607
    3.东北林业大学森林生态系统可持续经营教育部重点实验室,黑龙江 哈尔滨 150040
  • 收稿日期:2023-06-13 接受日期:2023-09-21 出版日期:2025-03-30 发布日期:2025-03-28
  • 通讯作者: *贾炜玮(jiaww2002@163.com),教授。
  • 作者简介:

    彭文越(1246584738@qq.com)。

  • 基金资助:
    国家重点研发计划(2023YFDZ200802);中央高校基本科研业务专项资金项目(2572019CP08)

Extraction and construction of a QSM-based model of first-order branches of Larix gmelinii plantations

PENG Wenyue1,2(), JIA Weiwei1,3,*(), WANG Fan1, LI Xin1,2, LI Dandan1   

  1. 1. School of Forestry, Northeast Forestry University, Harbin 150040, China
    2. Forestry Investigation and Planning Institute of Jilin Province, Changchun 130607, China
    3. Key Laboratory of Sustainable Forest Ecosystem Management-Ministry of Education, Northeast Forestry University, Harbin 150040, China
  • Received:2023-06-13 Accepted:2023-09-21 Online:2025-03-30 Published:2025-03-28

摘要:

【目的】基于地基激光雷达(terrestrial laser scanning,TLS)点云数据,应用定量结构模型(quantitative structural model,QSM)以全自动的方式提取落叶松不同相对着枝深度处一级枝条的数量,构建一级枝条密度的线性混合预测模型,为落叶松冠层研究提供理论基础。【方法】基于30株落叶松(Larix gmelinii)TLS数据,应用QSM算法获取树木结构参数,将提取值与实测值进行回归分析,探究点云数据建模的准确性。采用点云分层方法探究不同相对着枝深度处一级枝条数量的提取精度。以Poisson回归模型作为基础的计数模型,建立基于样木效应的一级枝条密度最优混合模型,并对模型进行评价。【结果】30株落叶松一级枝条数量的平均提取精度为80.71%,均方根误差(RMSE)为6.9594。不同冠层的一级枝条数量提取结果存在差异,相对着枝深度范围在(0.7,0.8]的提取效果最好,平均精度为87.78%。一级枝条密度的最优模型为基于样木效应的混合模型;枝条密度的最优模型为基于相对着枝深度的自然对数[ln(RDINC)]、相对着枝深度的平方(RDINC2)、高径比(HT/DBH)3个随机效应参数的线性混合模型(R2=0.745 4、均方根误差为0.229)。【结论】基于地基激光雷达扫描数据,通过定量结构模型获取单木的分枝结构参数,该方法具有适用性和可靠性。基于样木效应建立落叶松一级枝条的密度混合模型,不仅可以反映冠层内一级枝条的分布密度变化,还可以预测树冠整体的生长趋势。

关键词: 落叶松, 点云数据, 一级枝条数量, 定量结构模型(QSM), 线性混合模型

Abstract:

【Objective】The quantitative structural model (QSM) algorithm was used to extract the number of first-order branches at different relative branch depths using a fully automatic strategy based on terrestrial laser scanning (TLS) point cloud data. A linear mixed prediction model of first-order branch density was constructed to provide a theoretical basis for research studies on Larix gmelinii plantation canopies. 【Method】 The TLS data of 30 L. gmelinii plantations and the QSM algorithm were used to obtain parameters pertaining to tree structure, and the extracted and measured values were subjected to regression analysis for exploring the accuracy of modeling based on point cloud data. The extraction accuracy of the method used for determining the number of first-order branches at different relative branch depths was assessed using the point cloud layering method. The optimal mixing model of the first-order branch density of the sample tree effect was constructed using the Poisson regression model, and the model was evaluated. 【Result】 The average extraction accuracy of the model constructed based on 30 L. gmelinii plantation branches was 80.71%, and the RMSE was 6.959 4. There were differences among the number of first-order branches extracted from different canopies, and the best results were obtained when the relative branch depth ranged from 0.7 to 0.8. The average accuracy was 87.78%, and a mixed model based on the sample tree effect was found to be optimal for determining the first-order branch density. The optimal model of branch density was a linear mixed model based on three random effect parameters, namely natural logarithm of the relative branching depth [ln(RDINC)], square of the relative branching depth (RDINC2), and ratio of tree height to breast diameter [HT/DBH], and the values of R2 and RMSE were 0.745 4 and 0.229, respectively. 【Conclusion】 The parameters pertaining to the branch structure of individual trees were obtained based on the ground-based laser radar scanning data, using the quantitative structure model that was applicable or reliable. Based on the effects observed in the sample wood, the density mixed model of the first-order branches of L. gmelinii plantations not only reflects the changes in the distribution density of the primary branches in the canopy, but can also predict the overall growth trend of the crown.

Key words: Larix gmelinii, point cloud data, number of first-order branches, quantitative structural model (QSM), linear mixed model

中图分类号: