基于无人机激光雷达点云的单木生物量估测

Single tree biomass estimation based on UAV LiDARpoint cloud

【目的】为提高森林单木生物量估测精度和效率,本研究基于无人机激光雷达技术对哈尔滨城市林业示范基地的水曲柳、樟子松样地进行点云数据获取及单木生物量估测。【方法】通过优化算法对获取的点云数据进行树高、冠幅等单木结构参数的提取;然后基于改进的凸包算法获取树冠体积、树冠投影面积等树冠因子。最后将上述获得的单木结构参数引入传统CAR生物量模型中,建立基于无人机激光雷达点云数据的单木生物量模型。【结果】1)基于点云数据提取的单木结构参数与实测数值间的相关性较好。其中水曲柳样地平均冠幅和树高值的决定系数R^(2)分别为0.82和0.86,而樟子松样地平均冠幅和树高的决定系数R^(2)分别为0.80和0.84。2)通过与国家林业局颁布的水曲柳、樟子松生物量异速生长方程进行对比得出,当引入树高、冠幅、树冠投影面积和树冠体积作为CAR模型参数时构建的生物量模型拟合效果最优,R^(2)分别为0.83、0.79,相应的均方根误差RMSE分别为18.912和8.120 kg/株。通过最优生物量模型评价指标可以看出,两块样地生物量模型的总相对误差SRE分别为-0.541%和0.311%,平均相对误差MRE分别为0.014%和0.020%以内,而平均相对误差绝对值MARE分别为9.19%和6.95%。3)当引入树冠体积作为变量时,生物量模型的精度明显提高。相比于树高、冠幅作为变量的模型,树冠体积的引入使得水曲柳、樟子松生物量模型的R^(2)分别提高了0.076和0.060,RMSE分别下降6.759和1.386 kg/株。【结论】本研究说明无人机激光雷达点云数据能够通过结合其提取的单木结构参数对森林单木生物量进行估测研究,并能取得较好的拟合优度和较高的预测精度。

【Objective】To improve the estimation accuracy and efficiency of forest single tree biomass,based on UAV LiDAR technology,point cloud data acquisition and single tree biomass estimation were conducted on the Fraxinus mandshurica and Pinus sylvestris plots in the Harbin Urban Forestry Demonstration Base in this study.【Method】The single tree structure parameters such as tree height and crown width were extracted from the obtained point cloud data through an optimization algorithm.Then based on improved convex hull algorithm to obtain crown factors such as crown volume and crown projection area.Finally,the obtained tree height,crown width,crown volume,crown projection area were introduced into the traditional CAR biomass model to establish a single tree biomass model based on UAV LiDAR point cloud data.【Result】1)The correlation between the parameters of the single tree structure extracted from the point cloud data and the measured valuesis good.The determination coefficients R^(2) of the average crown width and tree height of Fraxinus mandshurica were 0.82 and 0.86,respectively,while the determination coefficients R^(2) of the average crown width and tree height of Pinus sylvestris were 0.80 and 0.84,respectively.2)By comparing with the allometric growth equations of Fraxinus mandshurica and Pinus sylvestris biomass promulgated by the State Forestry Administration,the biomass model fitting effect was optimum when the tree height,crown width,crown projected area and crown volume were introduced as CAR model parameters,and the R^(2) was 0.83 and 0.79,respectively,and the corresponding RMSE(root mean square error)was 29.141 kg/tree and 8.120 kg/tree,respectively.It can be seen from the evaluation index of the optimal biomass model that the SRE(sum of relative error)of the two plot biomass models was-0.541%and 0.311%,respectively,and the MRE(mean relative error)was 0.014%and 0.02%,respectively,and the MREA(mean relative error absolute value)was 9.19%and 6.95%,respectively.3)When crown volume was introduced as a variable,the accuracy of biomass model was significantly improved.Compared with the model with tree height and crown width as variables,the introduction of crown volume increased the R^(2) of the Fraxinus mandshurica and Pinus sylvestris biomass models by 0.076 and 0.060,and the RMSE decreased by 6.759 kg/tree and 1.386 kg/tree,respectively.【Conclusion】This study shows that the UAV LiDAR point cloud datacan be used to estimate the forest single tree biomass by combining the extracted single tree structure parameters,and can obtain a better fit goodness and high prediction accuracy.