为了准确重建正反面叶片叶绿素3维分布, 利用高光谱激光雷达, 采集了不同生长状态的绿萝叶片与植株的空间-光谱域点云数据, 设计了一种基于分类预测的重建方法。通过偏最小二乘回归构建叶片正面与反面光谱的叶绿素含量预测模型, 采用光谱自适应阈值选择方法实现植株点云中叶片正反面的分类, 并根据类别标签选择模型计算叶绿素含量, 重建植株的叶绿素3维分布。结果表明, 该方法得到的植株叶绿素3维分布更接近真实值, 决定系数达到0.69, 均方根误差为4.97。这一结果可为植物表型研究提供新的数据基础和理论方法。

In order to accurately reconstruct the 3-D chlorophyll distribution of adaxial and abaxial leaves, the leaves and plants of epipremnum aureum in different growth states were selected as the experimental samples, whose spatial-spectral point cloud data were acquired by hyperspectral light detection and ranging (LiDAR), and a reconstruction method based on classification and prediction was designed. Chlorophyll content prediction models of adaxial and abaxial leaves were constructed by the partial least squares regression modeling method, to calculate the chlorophyll of different leaves, adaxial and abaxial leaves were classified with the spectral adaptive threshold selection method, and then the plant 3-D distribution of chlorophyll was reconstructed. The results show that this reconstruction method can obtain the predicted value with a coefficient of determination 0.69 and a root mean square error 4.97, which is closer to the true value. This result will provide a basis and theoretical approach for plant phenotypes research.