Original papersDevelopment of a LiDAR-guided section-based tree canopy density measurement system for precision spray applications
Introduction
Conventional agriculture relies heavily on a high-level use of plant protection products, commonly known as pesticides. Pesticides play a critical role in increasing crop quality and productivity. Oerke et al. (2012) suggested that failure to use plant protection products against insects, diseases, pests, and weeds could result in up to 65% of crop yield losses. On the contrary, pesticide misuse represents a serious concern about their adverse impact on the non-targets, including humans, environment, and ecosystems (Alavanja et al., 1996, Deveau, 2009). Additionally, pesticide use caused about $8.2 billion in annual environmental and economic losses in the United States (Pimentel & Burgess, 2014). To address this concern, the reduction of plant protection products is very important and crucial when considering agricultural sustainability and profitability.
Current development of innovative management strategies has shown a significant reduction of pesticides and improves efficacy and safety by adopting the modern breakthrough in electronics (Ampatzidis et al., 2018). Precision spraying is one of the modern crops management strategies that assist management decisions (e.g. spraying) according to estimated variability in the field, aiming to reduce agricultural inputs. Precision spraying strategies have been utilized by researchers in the recent few decades for site-specific managements including weeds (Hunter et al., 2020), diseases (Yang, 2020), and pests (Zhong et al., 2018). The core concept of precision spraying is to adjust the spray volume by controlling nozzle flow rate.
Adjustment of the spray deposits according to the tree canopy characteristics offers the chance of decreasing pesticide use and environmental contamination (Nan et al., 2019). The tree canopy foliage plays an important role in determining the amount of spray volume required in an individual tree. However, tree canopies are not uniform in terms of density and volume. The variability of canopy foliage density is shown in Fig. 1. Identifying tree canopy foliage density can characterize the tree structure and determine the appropriate spray volume for precise pesticide applications (Chen et al., 2012, Hu and Whitty, 2019, Wei and Salyani, 2005). It also helps to adjust the pesticide application rate, spray flow rate, and air supply volume to better manage the orchards while spraying (Gil et al., 2007, Jeon and Zhu, 2012, Llorens et al., 2010, Shen et al., 2017).
A range of techniques used for measuring tree canopy foliage density characteristics has included visible-range camera sensor (Asaei et al., 2019), ultrasonic sensor (Gil et al., 2007), spectral sensor, infrared sensor (He et al., 2011), and laser sensor (Chen et al., 2012, Liu and Zhu, 2016). Despite the considerable efforts reflected in characterizing tree canopies, challenges still exist to accurately implement the developed strategies in real-time field conditions, due to uncontrollable weather conditions and system limitations. The performance of the precision management system is significantly reduced in the field conditions due to illumination variations, wind speed and direction, and system vibrations when a camera-based sensing system is used (Asaei et al., 2019). Ultrasonic based sensing systems provide inconsistent data due to the large angle of divergence of ultrasonic waves and uncontrollable environmental conditions in fields (Zhang et al., 2018). Similarly, studies have shown the difficulty of recording accurate data using spectral and infrared sensors due to high sensitivity to the outdoor field illumination and weather conditions (Zhang et al., 2018). Conversely, laser-based sensing techniques are not affected by the outdoor field weather conditions and provide more accurate detection results (Liu & Zhu, 2016).
LiDAR (light detection and ranging) sensing is an active laser scanner-based remote sensing technique applied widely for tree canopy characterizations (Brandtberg et al., 2003, Holmgren and Persson, 2004, Hosoi and Omasa, 2006, Omasa et al., 2007). The LiDAR sensor emits an electromagnetic signal that can bounce off of the vegetation canopy enabling a view of the exterior structure and three-dimensional information of the tree. Calculation of tree canopy foliage density characteristics using a LiDAR sensor have been reported (Auat Cheein et al., 2015, Berk et al., 2020, Chakraborty et al., 2019, Hu and Whitty, 2019). Auat Cheein et al. (2015) estimated three-dimensional structure of orchard trees; in particular, real-time measurement of canopy volume and shape using a LiDAR sensor and computational geometry analysis. Results reported that the accuracy was decreased up to 30%. Underwood et al. (2016) measured the tee canopy volume using terrestrial LiDAR scanner and achieved coefficient of determination (R2 = 0.77) for establishing the relationship between canopy volume and yield. Chakraborty et al. (2019) used a mobile 3D LiDAR mapping system to measure canopy volume for apple trees and grapevines. They reported correlation values of 0.81 and 0.51 between manual and automatic measurements using Convex hull and Voxel grid methods, respectively. However, the Voxel grid method is computationally intense and is affected by the voxel size. The Convex hull method showed inferior performance compared to the Voxel grid method. Hu and Whitty (2019) evaluated a tree canopy density mapping system for a trellis-structured apple orchard where all points generated from an individual tree were included. However, a trellis-structured apple orchard may have many points produced by the wire-plane and also from the main tree trunk that need to be removed before canopy density calculation. Berk et al. (2020) established a relationship by conducting laboratory experiments for measuring tree leaf area, but low accuracies were reported. Among the studies surveyed, most researchers have tried to measure tree canopy density based on the volume of individual trees considering all points; however, the density of the whole tree cannot precisely guide the sprayer unit because the precision sprayer may have multiple nozzles on each side which need to be controlled separately. Section-based canopy density measurement leads to assessment of foliage density by dividing the tree into sections (e.g., bottom, middle, and top, etc.). The computed density information of the canopy sections can separately guide/control the corresponding nozzle facing each section. Since the precision spraying system requires nozzle flow rates to be continuously controlled during orchard spraying, the spray decision input, i.e., section-based tree canopy density information needs to be measured automatically.
The primary goal of this study was to develop an automated section-based tree canopy density measurement system for precise pesticide spraying using a LiDAR sensor. The specific objectives were to: (i) establish a relationship between the point cloud and canopy foliage without considering trellis-wires, support poles, and tree trunk (ii) predict the number of leaves in each section with the density measurement algorithm (iii) measure the tree canopy volume and generate the canopy density map for providing guideline information for variable-rate spraying.
In this study, we performed three major tasks based on LiDAR scanned data for tree canopy density and volume measurements: point cloud data acquisition, tree canopy points segmentation, and canopy density and volume measurements. Data were acquired through LiDAR integrated sensing system. A sample consensus algorithm was used to remove the ground points. Unnecessary points from tree trunks, trellis wires, and support poles were removed using a processing algorithm aimed to segment only canopy points. The canopy density and volume were measured, and canopy density map was generated. The canopy density map generated in this work provides a graphical view of tree leaves distributions in different sections, which can be used later for spraying operation in the orchards.
Section snippets
Test orchards
Two orchard sites with GoldRush (site 1: 39°56′15.8″N, 77°15′21.0″W) and Fuji (site 2: 39°56′19.1″N, 77°15′17.5″W) apple varieties, located at Penn State Fruit Research and Extension Center (FREC), Biglerville, PA, USA, were used (Fig. 2). Both orchards use a trellis system to support trees, including three tiers of trellis wires and support poles. For the GoldRush block, the trees were trained as a tall spindle structure. The trees were planted in 2009 with an inter-row spacing of 6.10 m and
Canopy identification in sections
The point cloud data of five consequent trees in each orchard site were processed using the developed algorithms. Fig. 12, Fig. 13 show the tree canopies in the two test sites with and without trellis wire, poles, and tree trunks. These points were marked in different colors to represent four sections of tree canopy divided by the trellis wires.
Fig. 12, Fig. 13(a) show the point cloud data with points removed for the ground and neighboring rows. The trellis wires, poles, and tree trunks were
Conclusions
Apple tree canopy density was assessed using a ground-based LiDAR guided sensing system. Point cloud data were acquired from two orchard sites with GoldRush and Fuji apple varieties. A processing algorithm was scripted in the MATLAB® programming environment. Experiments were conducted with T-trellis structured orchards; therefore, tree trunk, trellis wires, and support poles were extracted to separate only the canopy points from the acquired data. Apple leaves were counted manually by a
CRediT authorship contribution statement
Md Sultan Mahmud: Conceptualization, Investigation, Methodology, Validation, Writing - original draft. Azlan Zahid: Investigation, Methodology, Validation. Long He: Conceptualization, Supervision, Writing - review & editing, Funding acquisition. Daeun Choi: Supervision, Writing - review & editing. Grzegorz Krawczyk: Supervision, Writing - review & editing. Heping Zhu: Supervision, Writing - review & editing. Paul Heinemann: Supervision, Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
This study was supported in part by United States Department of Agriculture (USDA)’s National Institute of Food and Agriculture (NIFA) Federal Appropriations under Project PEN04653 and Accession No. 1016510, a USDA NIFA Crop Protection and Pest Management Program (CPPM) competitive grant (Award No. 2019-70006-30440), and a Northeast Sustainable Agriculture Research and Education (SARE) Graduate Student Grant GNE20-234-34268. The authors would like to give special thanks to Xiaohu Jiang
References (36)
- et al.
Site-specific orchard sprayer equipped with machine vision for chemical usage management
Comput. Electron. Agric.
(2019) - et al.
Real-time approaches for characterization of fully and partially scanned canopies in groves
Comput. Electron. Agric.
(2015) - et al.
Digital evaluation of leaf area of an individual tree canopy in the apple orchard using the lidar measurement system
Comput. Electron. Agric.
(2020) - et al.
Detection and analysis of individual leaf-off tree crowns in small footprint, high sampling density lidar data from the eastern deciduous forest in North America
Remote Sens. Environ.
(2003) - et al.
Evaluation of mobile 3D light detection and ranging based canopy mapping system for tree fruit crops
Comput. Electron. Agric.
(2019) - et al.
Variable rate application of plant protection products in vineyard using ultrasonic sensors
Crop Prot.
(2007) - et al.
Identifying species of individual trees using airborne laser scanner
Remote Sens. Environ.
(2004) - et al.
An evaluation of an apple canopy density mapping system for a variable-rate sprayer
IFAC-PapersOnLine
(2019) - et al.
Variable rate dosing in precision viticulture: use of electronic devices to improve application efficiency
Crop Prot.
(2010) - et al.
Estimating leaf area density of Osmanthus trees using ultrasonic sensing
Biosyst. Eng.
(2019)
Sensitivity of tree volume measurement to trajectory errors from a terrestrial lidar scanner
Agric. For. Meteorol.
Mapping almond orchard canopy volume, flowers, fruit and yield using lidar and vision sensors
Comput. Electron. Agric.
Reduction of fungal disease spread in cultivar mixtures: Impact of canopy architecture on rain-splash dispersal and on crop microclimate
Agric. For. Meteorol.
Remote sensing and precision agriculture technologies for crop disease detection and management with a practical application example
Engineering
Semantic segmentation of sparse 3D point cloud based on geometrical features for trellis-structured apple orchard
Biosyst. Eng.
The agricultural health study
Environ. Health Perspect.
Voice-controlled and wireless solid set canopy delivery (VCW-SSCD) system for mist-cooling
Sustainability (Switzerland)
Development of a variable-rate sprayer with laser scanning sensor to synchronize spray outputs to tree structures
Trans. ASABE
Cited by (41)
Force and motion behaviour of crop leaves during spraying
2023, Biosystems EngineeringVision-based normalized canopy area estimation for variable nitrogen application in apple orchards
2023, Smart Agricultural TechnologyEffects of laser scanner quality and tractor speed to characterise apple tree canopies
2023, Smart Agricultural TechnologyUnmanned aerial vehicle based tree canopy characteristics measurement for precision spray applications
2023, Smart Agricultural TechnologyCitation Excerpt :An alpha shape algorithm (Dylan [26]) was used with three alpha values (0.35, 0.40, and 0.45) to calculate individual tree volume (Fig. 2). The canopy point clouds preprocessing and processing steps were described in detail by Mahmud et al. [9], while it did not measure any ground-based canopy coverage due to the complexity and impracticality compared to aerial measurement. Images were acquired using a rotary-wing unmanned aerial vehicle (UAV) DJI Matrice 200 (DJI Technology Inc., Shenzhen, China) with a high-resolution RGB Zenmuse X5S camera (DJI Technology Inc., Shenzhen, China) on June 21, 2021 (Fig. 3).
Real-time detection of street tree crowns using mobile laser scanning based on pointwise classification
2023, Biosystems Engineering