基于激光雷达的温室茄子三维重建与表型测量  

3D reconstruction and phenotypic measurement of greenhouse eggplant based on LiDAR

作　　者：韩笑宇 张波 赵秀丰 董鑫[2] 白皓然[1] 贾天怡 姜宏宇 王瑞雪 HAN Xiaoyu;ZHANG Bo;ZHAO Xiufeng;DONG Xin;BAI Haoran;JIA Tianyi;JIANG Hongyu;WANG Ruixue(Qingdao Agricultural University,Qingdao Shandong 266109,China;Chinese Academy of Agricultural Mechanization Science Group Co.,Ltd.,State Key Laboratory of Agricultural Equipment Technology,Beijing 100083,China;Tongyu County Promotion Station of Agricultural Mechanization Technology,Baicheng Jilin 137200,China;Nongan County Service Master Station of Agricultural Machinery Management,Changchun Jilin 130212,China;Chinese Machinery Cloud Agriculture(Jilin)Co.,Ltd.,Siping Jilin 136000,China)

机构地区：[1]青岛农业大学,山东青岛266109 [2]中国农业机械化科学研究院集团有限公司,农业装备技术全国重点实验室,北京100083 [3]通榆县农业机械化技术推广站,吉林白城137200 [4]农安县农业机械管理服务总站,吉林长春130212 [5]中机云农业(吉林)有限公司,吉林四平136000

出　　处：《农业工程》2025年第1期63-68,共6页AGRICULTURAL ENGINEERING

基　　金：国家重点研发计划项目(2022YFD2002304);中国机械工业集团有限公司青年基金项目(QNJJ-PY-2022-31)。

摘　　要：针对传统手动提取作物表型的方式存在效率低、精度差等缺点,提出一种基于三维重建获取茄子植株表型信息的方法。使用激光雷达获取茄子点云信息并进行处理,构建了基于半径滤波和统计滤波单株点云提取,DBSCAN叶片点云聚类的茄子植株点云三维重建方法。研发了基于三维重建激光雷达点云的温室茄子株高、冠幅、叶面积和叶倾角的生理结构表型参数的建模方法,基于单株点云最小包围盒计算株高和冠幅,基于三角面片化计算叶面积,基于高频法向量计算叶倾角。试验结果表明,提出的基于三维重建提取茄子植株表型参数的方法能够快速精准地获取茄子植株的表型信息,可以用于作物表型参数的在线检测,具有较好的实用性。Aiming at shortcomings of traditional manual extraction of crop phenotypes,such as inefficiency and poor accuracy,a meth-od based on 3D reconstruction to obtain phenotypic information of eggplant plants was proposed.Using LiDAR to obtain eggplant point cloud information and process,a 3D reconstruction method of eggplant plant point cloud based on radius filtering and statistical filtering single-plant point cloud extraction,and DBSCAN leaf point cloud clustering was constructed.A modelling method for physiological structural phenotypic parameters of greenhouse eggplant plant height,crown breadth,leaf area and leaf inclination based on 3D recon-struction of LiDAR point cloud was developed,and plant height and crown breadth were calculated based on the minimum boundig box of single plant point cloud,leaf area based on triangulation,and leaf inclination based on high-frequency normal vector.Experimental results proved that proposed method for extracting phenotypic parameters of eggplant plants based on 3D reconstruction could obtain phen-otypic information of eggplant plants quickly and accurately,and could be used for online detection of phenotypic parameters of crop,with great practicality.

关 键 词：茄子 三维重建 表型测量 激光雷达 点云处理 

分 类 号：S625[农业科学—园艺学]