Extraction of crop canopy features and decision-making for variable spraying based on unmanned aerial vehicle LiDAR data 基于无人机 LiDAR 数据的作物冠层特征提取与变量喷施决策
Shaoyong Luo ^("a "){ }^{\text {a }}, Sheng Wen ^("a,* "){ }^{\text {a,* }}, Lei Zhang ^(b){ }^{\mathrm{b}}, Yubin Lan ^(c){ }^{\mathrm{c}}, Xiaoshuai Chen ^(a){ }^{\mathrm{a}} 罗少勇 ^("a "){ }^{\text {a }} ,温盛 ^("a,* "){ }^{\text {a,* }} ,张磊 ^(b){ }^{\mathrm{b}} ,兰玉彬 ^(c){ }^{\mathrm{c}} ,陈小帅 ^(a){ }^{\mathrm{a}}^("a "){ }^{\text {a }} National Center for International Collaboration Research on Precision Agriculture Aviation Pesticides Spraying Technology, College of Engineering, South China Agricultural University, Guangzhou 510642, China ^("a "){ }^{\text {a }} 华南农业大学工程学院,国家精准农业航空施药技术国际联合研究中心,广州 510642,中国^("b "){ }^{\text {b }} College of Agriculture, South China Agricultural University, Guangzhou 510642, China 华南农业大学农学院,中国广州 510642^("c "){ }^{\text {c }} National Center for International Collaboration Research on Precision Agriculture Aviation Pesticides Spraying Technology, College of Electronic Engineering, South China Agricultural University, Guangzhou 510642, China 华南农业大学电子工程学院,国家精准农业航空施药技术国际联合研究中心,中国广州 510642
ARTICLE INFO 文章信息
Keywords: 关键词:
Variable-rate spray system 变量速率喷雾系统
Light detection and ranging sensor 激光雷达传感器
Canopy volume 冠层体积
Prescription map 处方图
Precision spraying 精准喷洒
Abstract 摘要
Implementation of precision agriculture aerial applications using unmanned aerial vehicles (UAVs) represents a pivotal technological aspect in precision agriculture. However, current UAV applications are still far from realizing true precision variable spraying. The use of UAVs to analyze and model crop growth, pest and disease conditions, and other pertinent information to enable targeted precision spraying poses a formidable challenge. Herein, a UAV variable-rate spray system with adaptive flow control based on crop canopy morphology was developed using a UAV equipped with a light detection and ranging (LiDAR) sensor to support precision spraying of cotton. Point cloud data captured by the LiDAR sensor initially undergo processing using the simple morphological filter (SMRF) algorithm and feature extraction function. Subsequently, cotton canopy volumes measured by the alpha shape algorithm, convex hull by slices shape algorithm, and voxel-based method were compared with manual measurements with coefficient of determination R^(2)R^{2} values of 0.89,0.780.89,0.78, and 0.82 , respectively. The results of the validation experiments denote that the cotton canopy volume obtained using the alpha shape algorithm is in good agreement with that obtained using manual measurements; thus, this algorithm is employed for volume calculations in this design. Models were constructed to relate UAV flight parameters, cotton canopy volume, and spraying parameters to the duty cycle of variable control signals, which enabled the UAV to adjust the flow rate in real time to match the cotton canopy structure. In addition, a polygonal prescription map decoding algorithm based on the cotton canopy morphology was introduced. This algorithm is used for real-time decoding of prescription maps and extraction of flow, latitude, and longitude information. Finally, field experiments were conducted, and the results indicate that the UAV variable-rate spray system designed in this study exhibits biological efficacy (effective application thresholds provided by the pesticide manufacturer) comparable to that of the constant-rate spray system at a lower application rate. Furthermore, it exhibited superior spray deposition efficiency and coverage compared with the constant-rate spray system. Notably, the use of the variable-rate spray system in cotton crops resulted in a 43.37%43.37 \% reduction in spray volume while maintaining high application quality, providing an opportunity for pesticide and labor cost savings. 利用无人机(UAV)实施精准农业航空作业是精准农业中的一项关键技术。然而,当前无人机应用距离实现真正的精准变量喷洒仍有较大差距。通过无人机分析并建模作物生长、病虫害状况等相关信息以实现针对性精准喷洒,是一项极具挑战性的任务。为此,本研究开发了一种基于作物冠层形态自适应流量控制的无人机变量喷洒系统,该系统搭载激光雷达(LiDAR)传感器,用于支持棉花的精准喷洒。LiDAR 传感器获取的点云数据首先通过简单形态学滤波(SMRF)算法和特征提取功能进行处理。随后,采用 alpha 形状算法、切片凸包形状算法及基于体素的方法测量的棉花冠层体积与人工测量结果进行了对比,其决定系数 R^(2)R^{2} 值分别为 0.89,0.780.89,0.78 和 0.82。 验证实验结果表明,采用 alpha 形状算法获取的棉花冠层体积与人工测量结果高度吻合,因此本设计选用该算法进行体积计算。研究构建了无人机飞行参数、棉花冠层体积及喷施参数与变量控制信号占空比之间的关联模型,使无人机能够实时调节流量以适应棉花冠层结构。此外,还引入了基于棉花冠层形态的多边形处方图解码算法,该算法用于实时解码处方图并提取流量、纬度及经度信息。最终田间试验表明,本研究设计的无人机变量喷施系统在较低施药量下,其生物效能(农药制造商提供的有效施用阈值)与恒量喷施系统相当,且在雾滴沉积效率和覆盖均匀性方面优于恒量喷施系统。 值得注意的是,在棉花作物中采用变量喷雾系统实现了喷雾量 43.37%43.37 \% 的减少,同时保持了较高的施药质量,为农药和劳动力成本节约提供了机会。
1. Introduction 1. 引言
Cotton stands as a principal economic crop and is recognized for its labor-intensive and time-consuming nature. Beyond the initial sowing stage, subsequent production phases, including pest and disease control, topping, and harvesting, require substantial labor input (Wang et al., 2022). Nevertheless, the pace of urbanization acceleration has resulted in a scarcity of affordable labor for cotton cultivation, and the swift 棉花作为主要经济作物,以其劳动密集和耗时特性而著称。除初始播种阶段外,后续生产环节如病虫害防治、打顶及收获等均需大量劳动力投入(Wang 等,2022)。然而,城市化进程的加快导致棉花种植面临廉价劳动力短缺问题,且迅速...
surge in labor costs has consequently triggered a continual increase in the overall expenses associated with cotton farming (Feng et al., 2022). 劳动力成本的激增进而引发了棉花种植相关总费用的持续上涨(Feng 等,2022)。
With the rapid advancements in cotton generation mechanization technology, the adoption of mechanized harvesting is a crucial avenue for mitigating labor intensity and enhancing cotton cultivation efficiency (Qiao, 2023). Cotton defoliation and ripening technology constitutes a vital prerequisite for achieving mechanical harvesting of cotton, representing a critical juncture in this process (Wang and 随着棉花生产机械化技术的快速发展,采用机械化采收是减轻劳动强度、提升棉花种植效率的关键途径(Qiao,2023)。棉花脱叶催熟技术是实现棉花机械采收的重要前提,也是该过程中的关键环节(Wang 与
Memon, 2020). The application of agricultural unmanned aerial vehicle (UAV) spraying presents remarkable advantages, including costeffectiveness, water conservation, high efficiency, and negligible impact on cotton plants. In particular, this method is well suited to the intensive cotton planting patterns prevalent in China (Yu et al., 2022). Consequently, it has evolved into a crucial instrument for the application of defoliant catalysts in China (Patil et al., 2023). Memon,2020)。农业无人机喷施技术的应用展现出显著优势,包括成本效益高、节水、高效且对棉花植株影响微乎其微。尤其适合中国普遍采用的高密度棉花种植模式(Yu 等,2022)。因此,该技术已发展成为中国脱叶剂喷施的关键工具(Patil 等,2023)。
However, suitable key technologies for intelligent and precise spraying control using UAVs that can promote effective defoliation and ripening of cotton are still considered to be the main bottleneck that limits the quality of cotton defoliation (He et al., 2017). Although defoliation and ripening spraying of cotton using agricultural UAVs are a gradually emerging technology in recent years, current research in this aspect is till mainly focused on the feasibility and experimental validation of the relevant operational parameters. 然而,适用于无人机智能精准喷施以促进棉花有效脱叶与催熟的关键技术,仍被视为制约棉花脱叶质量的主要瓶颈(He 等,2017)。尽管近年来利用农业无人机进行棉花脱叶与催熟喷施是一项逐渐兴起的技术,当前该领域的研究仍主要集中在相关作业参数的可行性与实验验证上。
In two field trials, Cavalaris et al. (2022) evaluated the efficacy of UAV spraying compared with conventional ground application for cotton. The outcomes indicated optimal results for low-altitude UAV operations at 2 m , demonstrating the superiority of all UAV applications over traditional field sprayers. Meng et al. (2019) conducted a comparative trial to evaluate the efficacy of UAV and ground-based mechanical spraying for cotton defoliant and ripening agents. After experimental verification, the best defoliation effect was shown for a single-rotor agricultural UAV (3WQF120-12, Quanfeng, China) when the spray volume was controlled at 22.5L//22.5 \mathrm{~L} / ha and the flight speed was 4 m//s\mathrm{m} / \mathrm{s}. Additionally, the defoliant catalyst sprayed by the agricultural UAV did not significantly affect cotton yield and fiber quality components. 在两项田间试验中,Cavalaris 等人(2022)评估了无人机喷洒与传统地面施药在棉花上的效果对比。结果表明,2 米低空飞行的无人机作业效果最佳,所有无人机应用均优于传统田间喷雾机。Meng 等人(2019)开展了一项对比试验,评估无人机与地面机械喷洒对棉花脱叶剂和催熟剂的效果。经实验验证,当喷雾量控制在 22.5L//22.5 \mathrm{~L} / 公顷且飞行速度为 4 m//s\mathrm{m} / \mathrm{s} 时,单旋翼农用无人机(3WQF120-12,中国全丰)展现出最佳脱叶效果。此外,农用无人机喷洒的脱叶催化剂对棉花产量和纤维品质成分无显著影响。
In a field experiment conducted within the northwest cotton production region of China, Chen et al. (2021) explored two test sites. The main effects and interactions between the operational parameters, including spray volume, droplet size Dv50, and flight altitude, were analyzed and discussed. The findings indicated that flight altitude, spray volume, and droplet size played pivotal roles in influencing spray penetration. Lowering the flight altitude of the UAV improved the distribution of droplets at the base of the vegetation canopy. Xiao et al. (2019) evaluated the impact of incorporating aerial spray additives on cotton defoliation, droplet deposition, and boll opening using a quadrotor agricultural UAV for the application of a cotton defoliant and ripening agent. The experimental findings indicated that the incorporation of aerial spray additives substantially enhanced the droplet deposition of the defoliant maturing agent. In the experimental area, the defoliation rate of cotton reached 80.31%80.31 \% and the boll opening rate was 90.61%1590.61 \% 15 days after the UAV sprayed the defoliant and ripening agent. 在中国西北棉花产区的田间实验中,Chen 等人(2021 年)选取了两处试验点,重点分析并探讨了作业参数(包括喷施量、雾滴粒径 Dv50 及飞行高度)的主效应及其交互作用。研究发现,飞行高度、喷施量与雾滴粒径对雾滴穿透性具有关键影响:降低无人机飞行高度能显著改善雾滴在植被冠层基部的分布状况。Xiao 等(2019 年)通过四旋翼农用无人机喷施棉花脱叶催熟剂,评估了航空喷雾助剂对棉花脱叶率、雾滴沉积及吐絮效果的影响。实验结果表明,添加航空喷雾助剂使脱叶催熟剂的雾滴沉积量显著提升,试验区棉花脱叶率在无人机喷药 80.31%80.31 \% 天后达到标准,吐絮率为 90.61%1590.61 \% 15 。
Agricultural UAVs have become increasingly prevalent in the realm of pest and disease control as well as in defoliant catalyst spraying operations in cotton. However, challenges remain, including issues associated with an inadequate defoliation rate. After the second spraying, the average defoliation rate achieved by the defoliant catalyst sprayed by agricultural UAVs was approximately 85%85 \%, with a flocculating rate of around 90%90 \% (Kong et al., 2020). However, this falls short of the preferred defoliation rate of 90%90 \% and the flocculating rate exceeding 95 %\%, as required by machine-picked cotton (Lu et al., 2021). This phenomenon substantially contributes to the lower defoliation rate and higher impurity content observed in machine-picked cotton in Northwest China, which is attributed to the application of agricultural UAVs for spraying. Such practices have implications for the fiber quality of cotton (Wang et al., 2019). 农业无人机在病虫害防治及棉花脱叶剂喷洒作业中日益普及,但仍面临脱叶率不足等问题。二次喷洒后,农业无人机施用的脱叶催化剂平均脱叶率约为 85%85 \% ,絮凝率约 90%90 \% (Kong 等,2020)。然而,这未达到机采棉要求的理想脱叶率 90%90 \% 和絮凝率 95 %\% 以上标准(Lu 等,2021)。该现象显著导致西北地区机采棉脱叶率偏低、含杂率偏高,究其原因与农业无人机喷洒作业直接相关,此类操作对棉花纤维品质具有重要影响(Wang 等,2019)。
The primary factor contributing to the low defoliation rate in cotton resulting from UAV spraying is the absence of appropriate key technologies for intelligent and precise control of UAV spraying (Maddikunta et al., 2021). The defoliation efficacy of cotton defoliants is intricately associated with various operational conditions, including crop canopy, flight parameters, spray flow rate, and particle size (Chen et al., 2021). Presently, UAV onboard precision spraying control models that match decision making based on cotton canopy characteristics and operation parameters, as well as integrating geolocation information, 无人机喷施导致棉花脱叶率低的主要原因是缺乏智能精准控制无人机喷施的关键技术(Maddikunta 等,2021)。棉花脱叶剂的脱叶效果与多种作业条件密切相关,包括作物冠层、飞行参数、喷雾流量和雾滴粒径(Chen 等,2021)。目前,能够根据棉花冠层特征与作业参数匹配决策并结合地理定位信息的无人机机载精准喷施控制模型仍较为匮乏,
are scarce. Moreover, precise spraying tailored to different cotton varieties and growth conditions is currently unattainable, which results in suboptimal defoliation effects. 且针对不同棉花品种及生长状况的精准喷施尚无法实现,从而导致脱叶效果欠佳。
This study investigates UAV airborne intelligent spraying control technology, incorporating flight, spraying, and cotton canopy parameters. It further explores UAV intelligent spraying mechanisms, which involve acquiring cotton canopy information and making decisions on spraying operation parameters. The primary contributions of this study are as follows: 本研究探讨了融合飞行、喷施及棉花冠层参数的无人机机载智能喷施控制技术,并进一步研究了涉及棉花冠层信息获取与喷施作业参数决策的无人机智能喷施机制。本研究的主要贡献如下:
An information acquisition platform for agricultural UAVs was constructed to acquire high-precision 3D point cloud data of a cotton canopy. Subsequently, 3D light detection and ranging (LiDAR) sensor-scanned point cloud data were preprocessed, and a simple morphological filter (SMRF) algorithm, along with a feature extraction function, was used to categorize the point cloud feature points and efficiently extract the parameters of the cotton vegetation point cloud. 构建了农业无人机信息采集平台,用于获取棉花冠层高精度三维点云数据。随后对三维激光雷达(LiDAR)传感器扫描的点云数据进行预处理,采用简单形态学滤波(SMRF)算法结合特征提取功能,对点云特征点进行分类,并高效提取棉花植被点云参数。
A methodology for creating a prescription map for UAV spraying based on the volume density of the cotton canopy was proposed. The alpha shape algorithm was used to compute the cotton canopy volume, and a mapping correlation between the cotton canopy volume and the UAV spray application volume was established. Incorporating canopy morphology into spray volume decisions, a prescription map based on cotton canopy volume density was created. This map was designed for agricultural plant protection UAVs to execute precise spraying operations. 提出了一种基于棉花冠层体积密度生成无人机喷施处方图的方法。利用 alpha shape 算法计算棉花冠层体积,建立冠层体积与无人机喷施量的映射关系,结合冠层形态学特征制定喷施决策,最终生成基于棉花冠层体积密度的处方图,供农业植保无人机执行精准喷施作业。
A multivariate UAV spraying system was designed based on cotton canopy volume. Models depicting the associations between UAV flight parameters, cotton canopy volume, spraying parameters, and the duty cycle (DC) of variable control signals were constructed. A polygonal prescription map decoding algorithm was devised for realtime decoding of the prescription map and extraction of position details and flow information. Real-time variable spraying was performed using UAV location information acquired in real time, and subsequent field experiments were conducted. 基于棉花冠层体积设计了一套多变量无人机施药系统,构建了无人机飞行参数、棉花冠层体积、施药参数与变量控制信号占空比(DC)之间的关联模型。开发了多边形处方图解码算法,用于实时解码处方图并提取位置细节与流量信息。利用实时获取的无人机位置信息实施变量施药,并开展了后续田间试验。
2. Materials and methods 2. 材料与方法
2.1. Experimental site 2.1. 试验地点
The field experiment was conducted in September 2023 at the Cotton Research Institute of Jiangxi Province, China. As shown in Fig. 1a, the experimental plot was at latitude 29.7112563^(@)N29.7112563^{\circ} \mathrm{N} and longitude 115.8433393^(@)E115.8433393^{\circ} \mathrm{E}. The panoramic perspective of the experimental plot captured by DJI Mavic3 (Shenzhen, China) is shown in Fig. 1b. The sowing of cotton was conducted on June 2, 2023, using a fully mechanized method. The selected variety was Zhongshengmian 10, and the row spacing was 18 xx76cm18 \times 76 \mathrm{~cm}. 田间试验于 2023 年 9 月在中国江西省棉花研究所进行。如图 1a 所示,试验田位于纬度 29.7112563^(@)N29.7112563^{\circ} \mathrm{N} 和经度 115.8433393^(@)E115.8433393^{\circ} \mathrm{E} 处。图 1b 展示了由大疆 Mavic3(中国深圳)拍摄的试验田全景视角。棉花于 2023 年 6 月 2 日采用全程机械化方式播种,选用品种为中棉所 10 号,行距为 18 xx76cm18 \times 76 \mathrm{~cm} 。
To study the structural characteristics of the cotton canopy, we focused on the cotton bolling stage of growth, when the cotton canopy has been shaped, which is an important time for cotton field management. The selected experimental area for this investigation covered 36 xx36 \times 18 m with a total of 22 columns. The entire experimental field was subdivided into 18 plots measuring 6xx6m6 \times 6 \mathrm{~m} (Fig. 1b), which were used to calculate various vegetation indices and canopy attributes and to conduct subsequent analyses in the field experiment. The cotton point cloud canopy height model was used with a grid size of 0.05 m per pixel, ensuring clear observation of cotton height and sparseness, as presented in Fig. 1c. Due to its current processing power, the onboard processor is unable to cope with the real-time processing of large-scale laser point clouds. The UAV conducted a two-phase flight experiment. The first phase involved generating the point cloud, performing data processing, and extracting cotton canopy volume parameters. The second phase was dedicated to spraying, including constant- and variable-rate spraying experiments. 为研究棉花冠层结构特征,我们聚焦于棉花结铃期这一冠层形态已定型的生长阶段,此时期是棉田管理的关键时段。本实验选取的研究区域覆盖 36 xx36 \times 18 米,共 22 列。整个试验田被细分为 18 个 6xx6m6 \times 6 \mathrm{~m} 大小的地块(图 1b),用于计算各类植被指数与冠层属性,并开展田间试验的后续分析。棉花点云冠层高度模型采用每像素 0.05 米的网格尺寸,确保能清晰观测棉花高度与疏密状况,如图 1c 所示。受限于当前处理能力,机载处理器无法实时处理大规模激光点云数据。无人机分两阶段开展飞行实验:第一阶段生成点云数据并进行处理,提取棉花冠层体积参数;第二阶段专用于喷施作业,包括恒量与变量喷施实验。