基于激光雷达与RGB相机融合的玉米作物行检测算法研究
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

基金项目:

国家重点研发计划项目(2022YFD2002001)和智能农业动力装备全国重点实验室开放项目(SKLIAPE2023012)


Maize Crop Row Detection Algorithm Based on Fusion of LiDAR and RGB Camera
Author:
Affiliation:

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:

    针对单一传感器在面对复杂田间环境适应性差的问题,本文提出了一种基于固态激光雷达(LiDAR)与RGB相机融合的玉米作物行检测方法。首先,研究了固态激光雷达和RGB相机联合标定方法,同步获取玉米作物行图像和点云数据并进行数据预处理。然后,将预处理后的图像数据和点云数据融合,实现点云“着色”,基于点云“着色”提出聚类感兴趣密度区域算法。利用“着色”点云完成聚类,并结合作物种植农艺标准(行距),分别验证点云信息和颜色信息的可用性,能够选择最优信息完成作物行感兴趣区域聚类。最后,通过划分点云水平条带的方式确定目标点云的特征点聚类区域,取作物行特征点,并利用最小二乘法拟合作物行检测线。仅需调整行距参数,算法可实现全生命周期的作物行检测,利用正常工况下玉米苗期、前期、中期和后期数据开展算法验证,作物行中心线平均误差不大于1.781°,准确率不小于92.69%,平均耗时不超过102.7 ms。此外,为验证算法鲁棒性,开展了复杂农田背景环境,如高杂草背景、断行、苗期杂草高度与玉米高度相近以及玉米完全封行4种工况作物行检测,算法平均误差不大于1.935°,准确率不小于91.94%,平均耗时不超过108.3 ms。通过讨论阐述了基于点云“着色”开展作物行中心线提取的优越性,本文算法可为作物行中心线可靠检测提供参考。

    Abstract:

    In response to the poor adaptability of a single sensor in facing complex field environments, a maize crop row detection method was proposed based on the fusion of solid-state LiDAR and RGB camera. Firstly, a joint calibration method for solid-state LiDAR and RGB camera was studied to simultaneously acquire maize crop row images and point cloud data for data preprocessing. Next, the preprocessed image data and point cloud data were fused to achieve point cloud “coloring”, and a clustering algorithm based on point cloud “coloring” for detecting regions of interest was proposed. The clustering was done by using the “colored” point cloud, and the availability of both point cloud and color information was separately validated based on crop planting agronomic standards (row spacing) to cluster the regions of interest effectively. Finally, by dividing the point cloud into horizontal strips, the feature points of the target point cloud were clustered to identify crop row feature points, and a crop row detection line was fitted by using the least squares method. By adjusting only the row spacing parameter, the algorithm can achieve crop row detection throughout the crop lifecycle. The algorithm’s performance was verified by using data from maize seedling, early, mid, and late stages under normal conditions, with average centerline error not more than 1.781°, accuracy not less than 92.69%, and average processing time not more than 102.7 ms. Furthermore, to test the algorithm’s robustness, crop row detections under four challenging conditions, including high weed background, missing rows, weed height similar to maize height, and completely closed rows, were conducted in complex agricultural field backgrounds. The algorithm showed an average error of not more than 1.935°, accuracy not less than 91.94%, and an average processing time not more than 108.3 ms. Discussions highlighted the superiority of using point cloud “coloring” for extracting crop row centerline, providing a reliable approach for detecting crop row centerlines.

    参考文献
    相似文献
    引证文献
引用本文

江庆,安东,韩华宇,刘京辉,郭延超,陈黎卿,杨洋.基于激光雷达与RGB相机融合的玉米作物行检测算法研究[J].农业机械学报,2024,55(10):263-274. JIANG Qing, AN Dong, HAN Huayu, LIU Jinghui, GUO Yanchao, CHEN Liqing, YANG Yang. Maize Crop Row Detection Algorithm Based on Fusion of LiDAR and RGB Camera[J]. Transactions of the Chinese Society for Agricultural Machinery,2024,55(10):263-274.

复制
分享
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:2023-12-25
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2024-10-10
  • 出版日期:
文章二维码