Abstract: In order to build digital orchards and improve the level of intelligent management, this study extracted peach tree structural parameters based on MLS LiDAR point cloud. Firstly, a backpack equipped multiplatform LiDAR was used to collect the point cloud data of peach orchard during leaf development, and an improved KMeans clustering algorithm was used to segment a single peach tree point cloud. Then, some branches with holes were sampled to obtain high density point cloud data of peach branches. Finally, the quantitative structure model (QSM) of peach tree was reconstructed by using cylinder fitting with different diameters, and five structural parameters of peach tree were extracted. The results showed that this method could  achieve accurate 3D model reconstruction of peach trees. The determination coefficients of crown width, plant height, trunk diameter, length of primary and secondary branches and measured values extracted by point cloud 3D reconstruction were 0.779, 0.939, 0.978, 0.965, 0.986, respectively, the root mean square errors were 0.280 m, 0.076 m, 0.003 m, 0.066 m, 0.068 m, respectively, and the average relative error was 8.6%, 2.5%, 3.2%, 2.6%, 8.4%, respectively. The research results can provide data support for the intelligent management of peach orchard.

Key words: LiDAR point cloud, peach tree, structural parameters,  , quantitative model, peach tree segmentation

摘要： 为构建数字化果园并提高智能化管理水平，探索基于MLS LiDAR提取桃树结构参数的方法。使用背包搭载多平台激光雷达采集展叶期桃园点云数据，采用改进KMeans聚类算法分割单棵桃树点云；对部分存在空洞的枝条点云上采样，得到较高密度枝条点云数据；使用不同直径的圆柱拟合重建桃树定量结构模型(QSM)，提取桃树5项结构参数。结果表明：该方法能实现桃树精准三维模型重建，重建后提取的冠幅值、株高、主干直径、一二级枝条长度与实测值决定系数分别为0.779、0.939、0.978、0.965、0.986，均方根误差分别为0.280 m、0.076 m、0.003 m、0.066 m、0.068 m；平均相对误差为8.6%、2.5%、3.2%、2.6%、8.4%。研究结果可为桃园智能化管理提供数据支撑。

关键词: LiDAR点云, 桃树, 结构参数, 定量模型, 桃树分割