基于3D LiDAR感知的大田花生长势信息获取
CSTR:
作者:
作者单位:

华南农业大学工程学院,广州 510642

作者简介:

胡炼,E-mail:lianhu@scau.edu.cn

通讯作者:

何杰,E-mail:hooget@scau.edu.cn

中图分类号:

S565.2;S24

基金项目:

国家现代农业产业技术体系建设专项(CARS-14);特定高校学科建设专项(2023B10564002)


Obtaining growth information of peanut in fields based on 3D LiDAR perception
Author:
Affiliation:

College of Agriculture, South China Agricultural University, Guangzhou 510642, China

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    为实现大田环境下快速准确获取作物株高、冠幅等长势信息,以花生为研究对象,采用3D LiDAR感知技术获取大田花生点云数据,经配准、去噪等处理,构建三维点云模型;基于KD-TREE的点云植株分割算法对大田花生植株点云数据进行单株分割,采用体素网格法估计植株体积、旋转卡壳法提取株高和最大冠幅,获取花生长势信息;在花生种植试验区域分别采集3个不同时期花生植株得到点云数据,采用3D LiDAR感知技术开展花生单株分割和提取株高与最大冠幅的验证试验,考察长势信息获取精度,采用召回率与精确率对结果进行精度评价。结果显示,大田花生单株分割的召回率与精确率均可达84%以上,表明该方法应用于大田花生点云数据分割具有较好的准确性与完整性;将提取的花生株高与最大冠幅等参数与人工测量值进行对比,3个时期植株高度的平均绝对百分比误差分别为6.271%、4.368%和4.986%,最大冠幅的平均绝对百分比误差分别为7.114%、5.606%和4.541%,株高均方根误差分别为0.010、0.015和0.027 m,最大冠幅均方根误差分别为0.011 、0.020和0.021 m;株高数据线性回归决定系数分别为0.888、0.951和0.842,最大冠幅数据线性回归决定系数分别为0.934、0.932和0.927,表明使用点云测量可实现花生表型参数高精度、非破坏性提取。

    Abstract:

    Peanut was used to rapidly and accurately obtain crop growth information including plant height and canopy width in the field.3D LiDAR perception technology was used to obtain the point cloud data of peanut in the field.A 3D point cloud model was constructed through registration, denoising, and other processing of the point cloud data.The point cloud plant segmentation method based on KD-TREE was used to segment the point cloud data of individual peanut plants.The convex hull algorithm was used to estimate plant volume and the rotating caliper method was used to extract plant height and maximum canopy width to obtain the growth information of peanut.Point cloud data of peanut plants at three different stages of growth were collected in a peanut planting area.The method proposed was used to carry out verification tests for the segmentation of individual peanut plant and the extraction of plant height and maximum canopy width.The accuracy of obtaining growth information was investigated.The accuracy of the results was evaluated with recall rate and precision rate.The results showed that the recall rate and precision rate of the segmentation of individual peanut plant in the field was over 85%, indicating that the method proposed has good accuracy and completeness for segmenting point cloud data of peanut in the field.The extracted parameters including plant height and maximum canopy width of peanut were compared with those of manual measurements.The average absolute percentage error of plant height and maximum canopy width at three different stages of growth was 6.271%, 4.368%, 4.986%, and 7.114%, 5.606%, 4.541%, with the root mean square error of 0.010, 0.015, 0.027 m, and 0.011, 0.020, 0.021 m, respectively.The linear regression determination coefficient of plant height and maximum canopy width was 0.888, 0.951, 0.842, and 0.934, 0.932, 0.927, respectively.It is indicated that the use of point cloud measurement can achieve high-precision and non-destructive extraction of phenotypic parameters for peanut in the field.It will provide important technical support for the cultivation and breeding of peanut.

    参考文献
    相似文献
    引证文献
引用本文

胡炼,刘于轩,臧英,何杰,汪沛,黄俊威,黄培奎,赵润茂.基于3D LiDAR感知的大田花生长势信息获取[J].华中农业大学学报,2025,44(4):102-112

复制
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-05-01
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2025-07-24
  • 出版日期:
文章二维码