基于移动多线激光雷达扫描的树冠叶面积估计方法

Crown Leaf Area Estimation Based on Mobile Multi-layer LiDAR Scanning

摘要:

移动单线激光雷达（Laser detection and ranging，LiDAR）扫描（Mobile single-layer LiDAR scanning，MSLS）树冠叶面积估计方法使用单一视角的单线激光雷达采集树冠点云数据，获取的冠层信息不够全面，限制了树冠叶面积估计精度。本文提出一种基于移动多线LiDAR扫描（Mobile multi-layer LiDAR scanning，MMLS）的树冠叶面积估计方法，使用多线LiDAR从多个视角采集树冠点云数据，提升树冠叶面积估计精度。首先，将多线LiDAR采集的点云数据变换到世界坐标系下，通过感兴趣区域（Region of interest，ROI）提取出树冠点云。然后，提出一种MMLS树冠点云融合方法，逐个融合单个激光器采集的树冠点云，设置距离阈值删除重复点，添加新点。最后，构建MMLS空间分辨率网格，建立基于树冠网格面积的树冠叶面积估计模型。实验使用VLP-16型多线LiDAR传感器搭建MMLS系统，设置1、1.5 m 2个测量距离和间隔45°的8个测量角度对6个具有不同冠层密度的树冠进行数据采集，共得到96个树冠样本。采用本文方法，树冠叶面积线性估计模型的均方根误差（Root mean squared error，RMSE）为0.104 1 m2，比MSLS模型降低0.057 8 m2，决定系数R2为0.952 6，比MSLS模型提高0.067 5。实验结果表明，本文方法通过多线LiDAR多视角树冠点云数据采集、MMLS树冠点云融合和空间分辨率网格构建，有效提升了树冠叶面积估计精度。

Abstract:

The mobile single-layer laser detection and ranging (LiDAR) scanning (MSLS) method for estimating tree crown leaf area used a single-layer LiDAR sensor with a single perspective to collect point cloud data on crowns, but the obtained canopy information was not comprehensive enough, which limited the accuracy of crown leaf area estimation. A crown leaf area estimation method was proposed based on mobile multi-layer LiDAR scanning (MMLS), which used a multi-layer LiDAR sensor to collect crown point cloud data from multiple perspectives and improve the accuracy of crown leaf area estimation. Firstly, the point cloud data collected by multi-layer LiDAR was transformed into the world coordinate system. The crown point cloud was extracted through the region of interest (ROI). Then, an MMLS crown point cloud fusion method was proposed, which fused the crown point clouds collected by a single laser one by one, set a distance threshold to remove duplicate points, and added new points. Finally, an MMLS spatial resolution grid was constructed, and a crown leaf area estimation model was established based on the crown grid area. The experiment used a multi-layer LiDAR sensor VLP-16 to build an MMLS system. Two measurement distances of 1 m and 1.5 m, and eight measurement angles with an increment of 45°, were set to collect data from six tree crowns with different canopy densities. A total of 96 tree crown samples were obtained. Using the proposed method, the root mean square error (RMSE) of the linear estimation model for crown leaf area was 0.104 1 m2, which was 0.057 8 m2 lower than that of the MSLS model, and the coefficient of determination R2was 0.952 6, which was 0.067 5 higher than that of the MSLS model. The experimental results showed that the proposed method can effectively improve the accuracy of crown leaf area estimation through multi-layer LiDAR multi-perspective crown point cloud data collection, MMLS crown point cloud fusion, and spatial-resolution grid construction.