基于雷达点云的果树快速识别与分割方法研究  

Rapid identification and segmentation method of fruit trees based on radar point cloud

作　　者：董鑫 杜新武[1] 余秀芳 韩静鸽 周利明[2,3] 冉小琴[2] DONG Xin;DU Xinwu;YU Xiufang;HAN Jingge;ZHOU Liming;RAN Xiaoqin(College of Agricultural Equipment Engineering,Henan University of Science and Technology,Luoyang Henan 471000,China;Chinese Academy of Agricultural Mechanization Sciences Group Co.,Ltd.,Beijing 100083,China;State Key Laboratory of Agricultural Equipment Technology,Beijing 100083,China;Xuchang Vocational and Technical College,Xuchang Henan 461000,China)

机构地区：[1]河南科技大学农业装备工程学院,河南洛阳471000 [2]中国农业机械化科学研究院集团有限公司,北京100083 [3]农业装备技术全国重点实验室,北京100083 [4]许昌职业技术学院,河南许昌461000

出　　处：《农业工程》2024年第8期27-33,共7页AGRICULTURAL ENGINEERING

基　　金：中国机械工业集团有限公司青年科技基金项目(QNJJ-PY-2022-31)。

摘　　要：在果园自动化作业生产中,果树的精准识别是果园精准作业的重要前提。针对在复杂果园环境中单株果树难以精准快速识别的问题,基于激光雷达三维点云技术对果树识别进行研究,提出了一种精准、快速、高自动化的果树识别方法及装置。构建多维度传感融合识别装置对果树进行特征点云数据扫描,建立果树点云数据集,通过点云质量及有效角度滤波进行数据筛选。利用边界特征点识别算法对单株果树边界特征数据进行快速提取,利用边界区域特征构建果树有效扫描数据。该方法可适用于单株果树轮廓、特征的快速识别。为了验证该果树识别方法准确性,进行了果树树冠轮廓测量精度试验,以及果树群组识别试验,结果表明,该方法能够对果树进行有效识别,并且物理指标综合测量精度>97.125%,可为果园精准作业提供有效依据。Accurate identification of fruit trees in production of orchard automation is an important prerequisite for orchard accurate operation.In view of problem that it is difficult to accurately and quickly identify a single fruit tree in a complex orchard environment,fruit tree recognition was studied based on lidar 3D point cloud technology,and an accurate,fast and highly automated fruit tree identification method and device was proposed.By constructing a multi-dimensional sensing fusion identification device,characteristic point cloud data of fruit trees was scanned,and point cloud dataset of fruit trees was established,and the data was screened by point cloud quality and effective angle filtering.Boundary feature point recognition algorithm was used to quickly extract boundary feature data of a single fruit tree,and effective scanning data of fruit trees was constructed by using boundary area features.It can be applied to rapid identification of contours and characteristics of individual fruit trees.In order to verify accuracy of fruit tree identification method,fruit tree canopy contour measurement accuracy test and fruit tree group identification test were carried out,and results showed that this method could effectively identify fruit trees,and comprehensive measurement accuracy of physical indicators was greater than 97.125%,which could provide an effective basis for accurate operation of orchard.

关 键 词：果园精准作业 激光雷达 果树识别 数据融合 传感器 

分 类 号：S24[农业科学—农业电气化与自动化]