摘要
为解决果园果树冠层信息提取易受光照、雨雾、风等环境因素影响。该研究采用毫米波雷达,结合果园全天候作业特征,提取株高、冠幅、体积等冠层信息参数。基于毫米波雷达点云参数特性,提出基于可变轴椭球模型的自适应DBSCAN与Alpha-shape相结合算法,通过自适应聚类分割与三维重构,解决了传统算法需要人为判断并输入邻域半径Eps、邻域密度阈值Minpts、滚动球半径α三个全局参数的问题,并模拟全天气候环境提取果树冠层信息。与人工测量结果相比,株高、冠幅、冠层体积提取结果的均方根误差RMSE分别为2.99 cm、2.44 cm、0.07 m^(3),平均相对误差MRE分别为3.38%、4.11%、12.82%,决定系数R2分别为0.89、0.91、0.57,果树冠层信息提取结果真实可靠。全天气候环境下不同光照强度、喷雾量对冠层信息提取结果无显著影响,不同风速下经归一化处理,拟合建立二段式函数,消除了风速对冠幅和冠层体积提取结果的影响。该研究提出的基于可变轴椭球模型的自适应DBSCAN与Alpha-shape相结合算法能够满足毫米波雷达全天候环境下精准提取果树冠层信息的需求,对实现果园精准管理与作业具有参考意义。
An accurate and rapid detection of the orchard canopy is easily interfered by the environmental factors,such as the sunlight,rain,fog,and wind.It is often required to extract the canopy information under all-weather environment.This study was conducted on the extraction of the orchard canopy information using millimeter-wave radar.A multi-module collaborative data acquisition was also established to realize the multi-source data fusion.The STM32F407ZGT6 microcontroller was equipped with the adjustable-speed diaphragm pumps,ring-shaped atomizing nozzles,and axial fans,in order to simulate the all-weather environment.The point cloud data was performed on the fusion and preprocessing,according to the point cloud parameter of the millimeter-wave radar.A combined algorithm of the adaptive DBSCAN and Alpha-shape was proposed using the variable-axis ellipsoid model.The conventional algorithms were often required the manual judgment and the input of three global parameters,such as the neighborhood radius Eps,neighborhood density threshold Minpts,and rolling ball radius α.The plant height,crown width,and canopy volume were measured to extract the values after adaptive clustering segmentation and three-dimensional reconstruction.The accuracy of the extraction was evaluated on the target canopy information under allweather climate conditions.The three-dimensional reconstruction of the orchard canopy shared the stronger adaptability,compared with the conventional DBSCAN,Alpha-shape and the single adaptive algorithm.The best performance was achieved in the three-dimensional reconstruction.Compared with the measurement,the root mean square error(RMSE)of the stem height,canopy width,and canopy volume extraction was 2.99 cm,2.44 cm,and 0.07 m^(3),respectively.The average relative error(MRE)was 3.38%,4.11%,and 12.82%,respectively,and the determination coefficient(R²)was 0.89,0.91,and 0.57,respectively,indicating the reliable extraction of the orchard canopy information.There was no significance in the extraction between the spray volumes and the orchard canopy information under different illuminations in all-weather environment.Different wind speeds shared no significance on the extraction of the plant height.But there was the significance on the extraction of the canopy width and volume.The detection of the dynamic targets by millimeter-wave radar were more sensitive than that of the static ones.The multi-frame data was detected by the millimeter-wave radar,because the disturbance of the axial fan was caused the branches and leaves to randomly sway.More point clouds of the canopy boundary were collected to increase the extraction values of the fruit crown width and canopy volume.The extraction information of the crown width and canopy volume were normalized under different wind speeds.A two-stage function was established to remove the influence of the wind speed on the extraction.All-weather information acquisition was also realized to accurately extract the canopy parameters under harsh orchard environments,such as illumination,rain,fog,and wind.It is of great significance for the precise operations in orchards.
作者
蒋寅龙
杨洲
解明坤
段洁利
金莫辉
付函
徐兴
JIANG Yinlong;YANG Zhou;XIE Mingkun;DUAN Jieli;JIN Mohui;FU Han;XU Xing(College of Engineering,South China Agricultural University,Guangzhou 510642,China;School of Mechanical Engineering,Guangdong Ocean University,Zhanjiang 524088,China)
出处
《农业工程学报》
北大核心
2026年第1期181-190,共10页
Transactions of the Chinese Society of Agricultural Engineering
基金
广州市重点研发计划项目(2024B03J1355)
国家现代农业产业技术体系建设专项(CARS-31)
广东省基础与应用基础研究基金项目(2023A1515110797)。
关键词
农业机械
果树
冠层信息提取
毫米波雷达
全天候能力
agricultural machinery
fruit tree
canopy information extraction
millimeter wave radar
all-weather capabilities
