基于几何模型的绿萝叶片外部表型参数三维估测  

Three-dimensional Estimation of Money Plant Leaf External Phenotypic Parameters Based on Geometric Model

作　　者：徐焕良[1,2] 马仕航 王浩云[1,2] 胡华东 殷佳来 车建华[1] XU Huanliang;MA Shihang;WANG Haoyun;HU Huadong;YIN Jialai;CHE Jianhua(College of Information Science and Technology,Nanjing Agricultural University,Nanjing 210095,China;Postdoctoral Mobile Station of Agricultural Engineering,Nanjing Agricultural University,Nanjing 210031,China)

机构地区：[1]南京农业大学信息科技学院,南京210095 [2]南京农业大学农业工程博士后流动站,南京210031

出　　处：《农业机械学报》2020年第12期220-228,共9页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家级大学生创新创业训练计划项目(201910307072Z);中央高校基本科研业务费专项基金项目(KYZ201914、KJQN201732);国家自然科学基金项目(31601545);江苏省重点研发计划项目(BE2016803)。

摘　　要：为快速高效获取叶类植物叶片的外部表型参数、掌握植株生长状况,以绿萝叶片为研究对象,提出一种基于几何模型的叶长、叶宽与叶面积的三维估测方法。利用微软Kinect V2相机,自80 cm高度垂直位姿获取绿萝叶片局部点云,并进行直通滤波去噪与包围盒精简等预处理,测量得到点云外形参数,输入预先建立的SAE网络分类预测得到几何模型参数,并基于曲面参数方程建立叶片几何模型。采用粒子群优化算法计算几何模型离散点云和局部点云间的空间距离,进行空间匹配,利用遗传算法求解最优匹配模型的内部模型参数,输出最优匹配模型的叶长、叶宽与叶面积作为估测结果。实验共采集150片绿萝叶片的局部点云数据,将估测结果和真实值进行数学统计与线性回归分析,得出叶长、叶宽与叶面积估测的平均误差分别为0.46 cm、0.41 cm和3.42 cm^2,叶长估测R^2和RMSE分别为0.88和0.52 cm,叶宽R^2和RMSE分别为0.88和0.52 cm,叶面积R^2和RMSE分别为0.95和3.60 cm^2。实验表明,该方法对于绿萝叶片外形参数的估测效果较好,具有较高实用价值。In order to obtain the external phenotypic parameters of the leaves and grasp the growth status of the plants quickly and efficiently,a three-dimensional estimation method of leaf length,leaf width and leaf area was proposed based on a geometric model by using the leaves of money plant.The Microsoft Kinect V2 camera was used to obtain the local point cloud of the leaf from the 80 cm height vertical pose and perform preprocessing such as pass-through filtering,denoising and simplification of the bounding box.The shape parameters of the point cloud were measured,and the pre-established SAE network classification prediction was used to obtain the geometric model parameters.The geometric model of the blade was established based on the surface parameter equation.The particle swarm optimization algorithm was used to calculate the spatial distance between the discrete point cloud and the local point cloud of the geometric model for spatial matching.The genetic algorithm was used to solve the internal model parameters of the optimal matching model,and the leaf length,leaf width and leaf area of the optimal matching model were output,were used as the estimation result.A total of 150 point cloud data were collected from the experiments.The estimated results and real values were analyzed by mathematical statistics and linear regression analysis.The average errors of the estimated leaf length,leaf width,and leaf area were 0.46 cm and 0.41 cm and 3.42 cm^2,respectively.The R^2 and RMSE of estimated leaf length were 0.88 and 0.52 cm,the R^2 and RMSE of leaf width were 0.88 and 0.52 cm,and the R^2 and RMSE of leaf area were 0.95 and 3.60 cm^2,respectively.It can be known from the experimental results that this method had good estimation effect on the shape parameters of money plant leaves,and it had high practical value.

关 键 词：绿萝叶片 表型参数 三维点云 几何模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] S126[自动化与计算机技术—计算机科学与技术]