A smart spraying machine and a method for precise target targeting and optimal pesticide dosage control.

Abstract

This invention relates to the field of pesticide spraying, and discloses an intelligent spraying machine and a method for precise target aiming and optimal pesticide dosage control. The machine includes a motion mechanism, a control mechanism, a spraying mechanism, and a navigation module for path planning and real-time obstacle avoidance. During its operation, the machine uses an onboard depth camera to acquire information on the canopy status of fruit trees and data on the location and extent of pests and diseases, thereby building a canopy model. Simultaneously, the image information of the fruit trees is transmitted back to the control mechanism for analysis to determine the target canopy. Based on the height of the target canopy center relative to the ground, the height of the spray nozzles and their corresponding rotation angles are calculated to achieve precise target aiming. Then, an improved BP neural network PID control algorithm is used to control the amount of pesticide sprayed by the nozzles, achieving variable-rate application to reduce pesticide waste.

Description

Technical Field

[0001] This invention relates to the field of pesticide spraying, specifically to an intelligent spraying machine and a method for precise target targeting and optimal pesticide dosage control. Background Technology

[0002] China's fruit industry has become the third largest agricultural planting industry in the country, and its orchard area and total fruit output have consistently ranked first in the world.

[0003] In real life, pesticides are generally applied by broad spraying because manual spraying is inefficient. Broad spraying, lacking targeted application to pests and diseases, results in a large discrepancy between the amount of pesticides sprayed and the actual utilization rate. It also leads to problems such as high pollution and high labor costs.

[0004] Therefore, there is an urgent need to invent a new type of precision spraying machine and method to improve the utilization rate of pesticides in order to address these problems. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent spraying machine and a method for precise target targeting and optimal pesticide dosage control, thereby solving the above-mentioned technical problems.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] An intelligent spraying machine includes: a control mechanism, a motion mechanism, a spraying mechanism, and a navigation module; the motion mechanism serves as the carrier of the entire robot, and the control mechanism, spraying mechanism, and navigation module are mounted on the motion mechanism.

[0008] The motion mechanism includes: a machine body, tracks, a DC motor, and a reducer; the tracks are installed at the bottom of the machine body, the DC motor is connected to the tracks and is used to drive the tracks to move, so as to realize the forward and backward movement of the machine, and the reducer is connected to the DC motor.

[0009] The DC motor is connected to the track. After being powered on, the DC motor rotates, driving the track to move. The speed is reduced by a reducer while the output torque is increased, thus enabling the machine to move forward and backward.

[0010] Furthermore, the spraying mechanism includes: a battery, a depth camera, a liquid storage tank, a stirring rod, a nozzle, a water pump, a solenoid valve, a servo motor, a guide rail, and a hydraulic cylinder;

[0011] The battery is located on the upper part of the body, a lidar is located on the upper part of the battery, a display is located on the upper part of the lidar, and the battery has an external square charging port that is electrically connected to the control mechanism.

[0012] The depth camera is located on the top of the display and is used to collect data on the state of the fruit tree canopy and the area and extent of fruit tree diseases and pests. It is connected to the control mechanism via wireless signal.

[0013] The liquid storage tank is located on one side of the depth camera. The liquid storage tank is connected to the nozzle through a liquid infusion hose. The liquid storage tank supplies liquid to the nozzle through a water pump and a liquid infusion hose.

[0014] The guide rail is fixedly installed on the rear side of the machine body and is symmetrical about the left and right along the axis of the machine body. The nozzle is slidably installed on the guide rail, which can realize the up and down movement of the nozzle.

[0015] The hydraulic cylinder is fixedly installed at the rear of the machine body, and a servo motor is fixedly connected to the top of the hydraulic cylinder. The output shaft end of the servo motor is fixedly connected to the nozzle. The servo motor rotates forward and backward by receiving rotation commands issued by the control mechanism, thereby rotating the nozzle.

[0016] The solenoid valve is connected to the infusion tubing, and the control mechanism controls the opening and closing of the solenoid valve to control the output liquid flow rate.

[0017] The stirring rod is located inside the storage tank and is used to stir the liquid inside the storage tank.

[0018] When powered on, the electric stirring rod can stir liquid media.

[0019] Furthermore, the navigation module includes a lidar, which is used to rotate and scan the surrounding environment to achieve path planning and real-time obstacle avoidance.

[0020] The control mechanism includes: a controller;

[0021] The controller is installed on the upper part of the machine body and is used to collect and process tree canopy status information and pest and disease information acquired by the depth camera. It is connected to the solenoid valve, servo motor, hydraulic cylinder and display through control circuit, and connected to the lidar through wireless signal to control the movement status and spraying operation of the entire machine.

[0022] A method for precise target targeting and optimal drug dosage control includes:

[0023] S1: Based on real-time data collection of fruit tree canopy status information, fruit tree disease and pest occurrence areas and severity data using a depth camera, build a fruit tree canopy model and output canopy images;

[0024] S2: The controller receives the tree canopy image and processes it according to preset rules to determine the target tree canopy;

[0025] S3: Obtain the center of the tree canopy and the height of the tree canopy center above the ground based on the determined target tree canopy. Determine the extension distance of the hydraulic cylinder based on the height of the tree canopy center above the ground. After the controller controls the hydraulic cylinder to execute, the position of the servo motor and the nozzle is adjusted.

[0026] S4: Calculate the rotation angle of the servo motor based on its height relative to the ground after adjustment, and control the nozzle to rotate and align with the target tree canopy;

[0027] S5: Open the solenoid valve to control the water pump to supply liquid to the nozzle through the infusion hose. During the process of the nozzle spraying the target tree canopy, the amount of spraying is controlled in real time through the BP neural network incremental PID control strategy.

[0028] Furthermore, the working process of S2 is as follows:

[0029] The tree canopy images were segmented using the Mask R-CNN algorithm, and redundant non-target tree canopies were filtered out. The area of ​​each tree canopy was calculated according to the following formula, and the tree canopy with the largest area was the target tree canopy.

[0030]

[0031] S max =(S1,S2,S3…) max ;

[0032] Where S is the area of ​​the two-dimensional fruit tree canopy, S i It is the area of ​​the i-th sector, r i Let dθ be the radius of the i-th sector, dθ be the central angle of the sector, and S be the radius of the sector. max It refers to the target canopy area.

[0033] Furthermore, the process of determining the extension and retraction distance of the hydraulic cylinder based on the height of the tree canopy center above the ground is as follows:

[0034] Obtain the height of the tree canopy center above the ground The extension / retraction distance h of the hydraulic cylinder is calculated using the following formula. x :

[0035]

[0036] Where h1 is the reference height of the hydraulic cylinder and h2 is the limit height of the hydraulic cylinder;

[0037] like When this happens, the hydraulic cylinder is controlled to not extend or retract;

[0038] like When this happens, the hydraulic cylinder is controlled to extend to...

[0039] like Then the hydraulic cylinder is controlled to extend to h2.

[0040] Furthermore, the process of obtaining the servo motor rotation angle is as follows:

[0041] Obtain the adjusted servo height relative to the ground

[0042] Obtain the horizontal straight-line distance X between the servo motor and the center of the tree canopy;

[0043] Through the formula:

[0044]

[0045] Calculate the rotation angle F of the servo motor θ ,

[0046] Furthermore, the process of controlling the amount of pesticide sprayed includes:

[0047] Through the formula:

[0048] Δu(k)=u(k)-u(k-1);

[0049] Δu(k)=K p [e(k)-e(k-1)]+K i e(k)+K d [e(k)-2e(k-1)+e(k-2)];

[0050] Where Δu(k) is the increment of the control liquid output at time k, K p K is the PID proportional control coefficient. i K represents the integral control coefficient of the PID controller. d Here, u(k) represents the PID differential control coefficient, k is the sampling number (k = 0, 1, 2, ...), u(k) is the actual liquid output quantity in the kth sampling, and e(k) is the deviation of the liquid output quantity in the kth sampling.

[0051] Furthermore, before S1, a spraying path plan is developed based on the current distribution of pests and diseases in the orchard;

[0052] The specific process of spray path planning includes:

[0053] The orchard is divided into multiple zones according to a preset area. A drone equipped with a high-definition camera is used to inspect the fruit trees in each zone according to a preset flight path, and real-time images of the fruit trees in each zone are collected.

[0054] The regions with pests and diseases are marked, and then the fruit tree images of the regions with pests and diseases are input into a pre-trained neural network model for processing, and the location of the fruit trees with pests and diseases and the type of pests and diseases are output.

[0055] Get the number of fruit trees with pests and diseases in a partition with pests and diseases within a given time period;

[0056] Through the formula:

[0057]

[0058] Calculate and obtain the j-th outlier value ρ of the fruit trees in the current partition. j ;

[0059] Where, m jc Let m be the number of fruit trees affected by pests in the j-th instance. jb M represents the number of fruit trees with disease in the j-th instance. j Let be the total number of fruit trees in the current partition, α and β be weighting coefficients, R be the total number of drone patrols within a time period, and μ be the stability coefficient. The conversion factor;

[0060] ρ i Compare with the preset abnormal threshold ρ0;

[0061] If ρ i If ≥ρ0, then the current partition is determined to be in an abnormal state;

[0062] If ρ i If <ρ0, then the current partition is considered to be in a normal state;

[0063] Zones in abnormal condition will be included in the spraying plan, and the abnormal value ρ of fruit trees will be used to determine the spraying method. j Sort the numbers 1, 2, ..., N in descending order, and use the order of the partitions corresponding to the numbers as the spraying paths.

[0064] Furthermore, the process of obtaining the stability coefficient is as follows:

[0065] Obtain the curve of the number of diseased fruit trees over time within a work cycle. jc (t) and the curves showing the change in the number of fruit trees affected by pests over time. jb (t);

[0066] Based on historical data, obtain the standard variation curve of the number of diseased fruit trees over time within a work cycle. jc0 (t) and the standard variation curve of the number of fruit trees infested with pests over time m jc0 (t);

[0067] Through the formula:

[0068]

[0069]

[0070] Calculate and obtain the stability coefficient μ;

[0071] Among them, δ1 and δ2 are proportionality coefficients, which are selected and determined based on historical data; τ0 is the preset threshold of τ; and γ is the conversion coefficient, which is selected and determined based on historical data and experimental data.

[0072] The beneficial effects of this invention are:

[0073] (1) The depth camera can capture the shape of the fruit tree canopy and the location of the fruit tree during the journey. The Mask R-CNN network is used to identify and segment the tree canopy in the image. Since the background of a single tree canopy is complex, there is more than one tree canopy in the collected image. The model may identify multiple tree canopies. The algorithm filters out the redundant non-target tree canopies and calculates the area of ​​each tree canopy. The tree canopy with the largest area is the target tree canopy.

[0074] (2) After determining the shape of the fruit tree canopy, the distance between the servo motor and the ground and the distance between the center of the canopy and the ground are used to calculate the corresponding rotation angle of the servo motor after the extension and retraction of the hydraulic cylinder. The controller issues a control command to rotate the servo motor and controls the nozzle to aim at the center of the fruit tree canopy to achieve precise spraying.

[0075] (3) The improved neural network is used to tune the PID control variable application algorithm. By leveraging the approximation characteristics of the PID function, the parameters are continuously corrected to reduce the deviation of liquid output, thereby achieving retuning of the control parameters. The parameters are then fed back to the controller to control the spraying pressure of the pesticide, achieving precise quantitative spraying. At the same time, the algorithm only calculates the control output increments of the control system in the most recent three times, so the calculation speed is greatly improved, and the influence of cumulative error is avoided, thus improving the control effect.

[0076] (4) The self-learning ability of the BP neural network is used to adjust the specific weights of the parameters so that the PID control parameters output by the neural network can be adjusted accordingly to the changes in the control system. In addition, a calculation method is proposed to address the difficulty in determining the number of hidden layer neurons in the three-layer structure of the BP neural network, which is beneficial to balance the neural network's ability to acquire effective information and its generalization ability. Attached Figure Description

[0077] The invention will now be further described with reference to the accompanying drawings.

[0078] Figure 1 This is a three-dimensional structural diagram of the present invention;

[0079] Figure 2 This is a rear view of the present invention;

[0080] Figure 3 This is a logic diagram of the method of the present invention;

[0081] Figure 4This is a schematic diagram of the tree canopy of the present invention;

[0082] Figure 5 This is a schematic diagram of the height control of the hydraulic cylinder in this invention;

[0083] Figure 6 This is a schematic diagram of the neural network topology in this invention.

[0084] Attached diagram descriptions: 1. Motion mechanism; 5. Track; 6. DC motor; 7. Reducer; 8. Battery;

[0085] 2. Control mechanism; 12. Controller;

[0086] 3. Spraying mechanism; 10. Depth camera; 11. Display; 13. Storage tank; 14. Guide rail; 15. Hydraulic cylinder; 16. Nozzle; 17. Solenoid valve; 18. Servo motor; 19. Square charging port; 20. Stirring rod; 22. Infusion hose;

[0087] 4. Navigation module; 9. LiDAR. Detailed Implementation

[0088] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0089] Please see Figures 1-6 As shown, the present invention is an intelligent spraying machine, comprising: a control mechanism 2, a motion mechanism 1, a spraying mechanism 3, and a navigation module 4; the motion mechanism 1 serves as the carrier of the entire robot, and the control mechanism 2, the spraying mechanism 3, and the navigation module 4 are mounted on the motion mechanism 1;

[0090] The motion mechanism 1 includes: a body, a track 5, a DC motor 6, and a reducer 7; the track 5 is installed at the bottom of the body, the DC motor 6 is connected to the track 5 and is used to drive the track 5 to move, so as to realize the forward and backward movement of the machine, and the reducer 7 is connected to the DC motor 6.

[0091] Furthermore, the spraying mechanism 3 includes: a battery 8, a depth camera 10, a liquid storage tank 13, a stirring rod 20, a nozzle 16, a water pump, a solenoid valve 17, a servo motor 18, a guide rail 14, and a hydraulic cylinder 15.

[0092] The battery 8 is located on the upper part of the body, and a lidar 9 is located on the upper part of the battery 8. A display 11 is located on the upper part of the lidar 9. The battery 8 has an external square charging port 19 and is electrically connected to the control mechanism 2.

[0093] The depth camera 10 is located on the upper part of the display 11 and is used to collect data on the status of the fruit tree canopy and the area and extent of fruit tree diseases and pests. It is connected to the control mechanism 2 via wireless signal.

[0094] The liquid storage tank 13 is located on one side of the depth camera 10. The liquid storage tank 13 is connected to the nozzle 16 through the infusion hose 22. The liquid storage tank 13 supplies liquid to the nozzle 16 through the water pump and the infusion hose 22.

[0095] The guide rail 14 is fixedly installed on the rear side of the machine body and is symmetrical about the left and right along the axis of the machine body. The nozzle 16 is slidably installed on the guide rail 14, which enables the nozzle 16 to move up and down.

[0096] The hydraulic cylinder 15 is fixedly installed at the tail of the machine body. The top of the hydraulic cylinder 15 is fixedly connected to the servo motor 18. The output shaft end of the servo motor 18 is fixedly connected to the nozzle 16. The servo motor 18 rotates forward and backward by receiving the rotation command issued by the control mechanism 2, thereby realizing the rotation of the nozzle 16.

[0097] The solenoid valve 17 is connected to the infusion tubing 22, and the control mechanism 2 controls the opening and closing of the solenoid valve 17 to control the output liquid flow rate.

[0098] The stirring rod 20 is located inside the liquid storage tank 13 and is used to stir the liquid inside the liquid storage tank 13.

[0099] Furthermore, the navigation module 4 includes a lidar 9, which is used to rotate and scan the surrounding environment to achieve path planning and real-time obstacle avoidance.

[0100] The control mechanism 2 includes: a controller 12;

[0101] The controller 12 is installed on the upper part of the machine body and is used to collect and process the tree canopy status information and pest and disease information acquired by the depth camera 10. It is connected to the solenoid valve 17, servo motor 18, hydraulic cylinder 15 and display 11 through control lines, and connected to the lidar 9 through wireless signals to control the movement status and spraying operation of the entire machine.

[0102] In this embodiment, the machine uses LiDAR 9 to rotate and scan the surrounding environment to achieve path planning and real-time obstacle avoidance. During the process, it uses depth camera 10 and controller 12 to obtain information about the target tree canopy. After determining the shape of the fruit tree canopy, it controls the hydraulic cylinder 15 to extend and retract according to the instructions issued by controller 12. After the extension and retraction of hydraulic cylinder 15 is completed, the distance between the servo motor 18 and the ground and the distance between the center of the canopy and the ground are used to calculate the corresponding rotation angle of the servo motor. Controller 12 issues a control command to rotate the servo motor 18, controlling the nozzle 16 to aim at the center of the fruit tree canopy to achieve the purpose of precise spraying.

[0103] A method for precise target targeting and optimal drug dosage control includes:

[0104] S1: Based on the real-time acquisition of fruit tree canopy status information, fruit tree disease and pest occurrence areas and severity data using a depth camera 10, build a fruit tree canopy model and output canopy images;

[0105] S2: The controller 12 receives the tree canopy image and processes it according to preset rules to determine the target tree canopy;

[0106] S3: Obtain the center of the tree canopy and the height of the tree canopy center above the ground based on the determined target tree canopy. Determine the extension distance of the hydraulic cylinder 15 based on the height of the tree canopy center above the ground. After the controller 12 controls the hydraulic cylinder 15 to execute, the position of the servo motor 18 and the nozzle 16 is adjusted.

[0107] S4: Based on the adjusted height of the servo motor 18 relative to the ground, calculate the rotation angle of the servo motor 18, and control the nozzle 16 to rotate and align with the target tree canopy;

[0108] S5: Open the solenoid valve 17 to control the water pump to supply liquid to the nozzle 16 through the infusion hose 22. During the process of the nozzle 16 spraying the target tree canopy, the amount of spraying is controlled in real time through the incremental BP neural network PID control strategy.

[0109] Furthermore, the working process of S2 is as follows:

[0110] The tree canopy images were segmented using the Mask R-CNN algorithm, and redundant non-target tree canopies were filtered out. The area of ​​each tree canopy was calculated according to the following formula, and the tree canopy with the largest area was the target tree canopy.

[0111]

[0112] S max =(S1,S2,S3…) max ;

[0113] Where S is the area of ​​the two-dimensional fruit tree canopy, S i It is the area of ​​the i-th sector, r iLet dθ be the radius of the i-th sector, dθ be the central angle of the sector, and S be the radius of the sector. max It refers to the target canopy area.

[0114] This invention utilizes a depth camera 10 to capture the shape of the fruit tree canopy and the location of the fruit tree during the process of movement. The Mask R-CNN network is used to identify and segment the tree canopy in the image. Since the background of a single tree canopy is complex, there is more than one tree canopy in the acquired image. The model may identify multiple tree canopies. The algorithm filters out the redundant non-target tree canopies and calculates the area of ​​each tree canopy. The tree canopy with the largest area is the target tree canopy.

[0115] Furthermore, the process of determining the extension and retraction distance of the hydraulic cylinder 15 based on the height of the tree canopy center above the ground is as follows:

[0116] Obtain the height of the tree canopy center above the ground The extension distance h of hydraulic cylinder 15 is calculated using the following formula. x :

[0117]

[0118] Where h1 is the reference height of hydraulic cylinder 15, and h2 is the limit height of hydraulic cylinder 15;

[0119] like When this occurs, hydraulic cylinder 15 is controlled to not extend or retract;

[0120] like When this happens, the hydraulic cylinder 15 is controlled to extend to...

[0121] like Then control the hydraulic cylinder 15 to extend to h2.

[0122] Furthermore, the process of obtaining the rotation angle of servo motor 18 is as follows:

[0123] Obtain the adjusted height of servo 18 relative to the ground.

[0124] Obtain the horizontal straight-line distance X between servo motor 18 and the center of the tree canopy;

[0125] Through the formula:

[0126]

[0127] The rotation angle F of servo motor 18 was calculated. θ ,

[0128] After determining the shape of the fruit tree canopy, the present invention calculates the corresponding rotation angle of the servo motor 18 by using the distance between the servo motor 18 and the ground and the distance between the center of the canopy and the ground after the extension and retraction of the hydraulic cylinder 15. The controller 12 issues a control command to rotate the servo motor 18, and controls the nozzle 16 to be aligned with the center of the fruit tree canopy, so as to achieve precise spraying.

[0129] Furthermore, the process of controlling the amount of pesticide sprayed includes:

[0130] Through the formula:

[0131] Δu(k)=u(k)-u(k-1);

[0132] Δu(k)=K p [e(k)-e(k-1)]+K i e(k)+K d [e(k)-2e(k-1)+e(k-2)];

[0133] Where Δu(k) is the increment of the control liquid output at time k, K p K is the PID proportional control coefficient. i K represents the integral control coefficient of the PID controller. d Here, u(k) represents the PID differential control coefficient, k is the sampling number (k = 0, 1, 2, ...), u(k) is the actual liquid output quantity in the kth sampling, and e(k) is the deviation of the liquid output quantity in the kth sampling.

[0134] The BP neural network incremental PID control consists of an incremental PID controller and a BP neural network. The PID controller is responsible for regulating the liquid output, while the K... p K i and K d The system is controlled by a backpropagation (BP) neural network, which adjusts three parameters based on the system's operating status. The system employs a three-layer BP neural network, and the method for calculating the number of hidden layer nodes has been optimized and improved. The improved calculation formula is as follows:

[0135]

[0136] In the formula N hid N represents the number of hidden layer neurons. in Number of neurons in the input layer, N out This represents the number of neurons in the output layer.

[0137] An improved BP neural network-tuned pesticide application algorithm using PID control variables is developed. Leveraging the approximation properties of the PID function, parameters are continuously corrected to reduce liquid output deviation, achieving retuning of control parameters. These parameters are then fed back to the controller to control the spraying pressure of the pesticide, enabling precise quantitative spraying. Furthermore, the algorithm only calculates the increments of the control output from the control system's three most recent operations, significantly improving computational speed and avoiding the influence of accumulated errors, thus enhancing control effectiveness. The self-learning capability of the BP neural network is utilized to adjust the specific weights of the parameters, allowing the PID control parameters output by the neural network to adapt to changes in the control system. A calculation method is proposed to address the difficulty in determining the number of hidden neurons in the three-layer structure of the BP neural network, balancing the neural network's ability to acquire effective information with its generalization capabilities.

[0138] Furthermore, before S1, a spraying path plan is developed based on the current distribution of pests and diseases in the orchard;

[0139] The specific process of spray path planning includes:

[0140] The orchard is divided into multiple zones according to a preset area. A drone equipped with a high-definition camera is used to inspect the fruit trees in each zone according to a preset flight path, and real-time images of the fruit trees in each zone are collected.

[0141] The regions with pests and diseases are marked, and then the fruit tree images of the regions with pests and diseases are input into a pre-trained neural network model for processing, and the location of the fruit trees with pests and diseases and the type of pests and diseases are output.

[0142] Get the number of fruit trees with pests and diseases in a partition with pests and diseases within a given time period;

[0143] Through the formula:

[0144]

[0145] Calculate and obtain the j-th outlier value ρ of the fruit trees in the current partition. j ;

[0146] Where, m jc Let m be the number of fruit trees affected by pests in the j-th instance. jb M represents the number of fruit trees with disease in the j-th instance. j Let be the total number of fruit trees in the current partition, α and β be weighting coefficients, R be the total number of drone patrols within a time period, and μ be the stability coefficient. The conversion factor;

[0147] ρ i Compare with the preset abnormal threshold ρ0;

[0148] If ρ i If ≥ρ0, then the current partition is determined to be in an abnormal state;

[0149] If ρ i If <ρ0, then the current partition is considered to be in a normal state;

[0150] Zones in abnormal condition will be included in the spraying plan, and the abnormal value ρ of fruit trees will be used to determine the spraying method. j Sort the numbers 1, 2, ..., N in descending order, and use the order of the partitions corresponding to the numbers as the spraying paths.

[0151] In this embodiment, the orchard is divided into multiple zones according to a preset range. A drone equipped with a high-definition camera inspects the fruit trees in each zone according to a preset flight path, collecting real-time images of the fruit trees in each zone. Then, the number of fruit trees with pests and diseases in each zone within a given time period is obtained; using the formula: Calculate and obtain the j-th outlier value ρ of the fruit trees in the current partition. j , will ρ i Compare with the preset abnormal threshold ρ0; if ρ i If ≥ρ0, then the current partition is determined to be in an abnormal state; if ρ i If the value is less than ρ0, the current partition is considered to be in a normal state; partitions in an abnormal state are included in the spraying plan, and the abnormal value ρ of the fruit trees is used to determine the spraying status. j Arrange the pesticides in descending order as 1, 2, ..., N, and use the corresponding partition order as the spraying path. Through the above process, timely feedback can be provided on the development trend of pests and diseases in each partition, thereby planning the spraying path according to the severity of the deterioration trend, thus improving the effectiveness of spraying and enhancing the safety of fruit trees.

[0152] Furthermore, the process of obtaining the stability coefficient is as follows:

[0153] Obtain the curve of the number of diseased fruit trees over time within a work cycle. jc (t) and the curves showing the change in the number of fruit trees affected by pests over time. jb (t);

[0154] Based on historical data, obtain the standard variation curve of the number of diseased fruit trees over time within a work cycle. jc0 (t) and the standard variation curve of the number of fruit trees infested with pests over time m jb0 (t);

[0155] Through the formula:

[0156]

[0157]

[0158] Calculate and obtain the stability coefficient μ;

[0159] Among them, δ1 and δ2 are proportionality coefficients, which are selected and determined based on historical data; τ0 is the preset threshold of τ; and γ is the conversion coefficient, which is selected and determined based on historical data and experimental data.

[0160] In this embodiment, the curve m of the number of diseased fruit trees over time within a working cycle is first obtained. jc (t) and the curves showing the change in the number of fruit trees affected by pests over time. jb (t); then, based on historical data, obtain the standard variation curve m of the number of diseased fruit trees over time within a work cycle. jc0 (t) and the standard variation curve of the number of fruit trees infested with pests over time m jb0 (t);

[0161] Through the formula: The stability coefficient μ is calculated. The above formula allows for objective adjustment of the stability coefficient based on the difference between the actual and standard trends of disease and pest changes within a zone. This dynamic change in the stability coefficient influences the abnormal values ​​of fruit trees, ensuring the accuracy of these values ​​and preventing misjudgments of zone status that could lead to errors in spraying path planning, resulting in fruit trees that should have been sprayed in time not being sprayed and thus deteriorating or even dying.

[0162] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for precise target targeting and optimal dosage control, applicable to intelligent spraying machines, characterized in that, include: S1: Based on the real-time acquisition of fruit tree canopy status information, fruit tree disease and pest occurrence area and degree data by depth camera (10), build a fruit tree canopy model and output canopy images; S2: The controller (12) receives the tree canopy image and processes the tree canopy image according to preset rules to determine the target tree canopy; S3: Obtain the center of the tree canopy and the height of the tree canopy center relative to the ground based on the determined target tree canopy. Determine the extension distance of the hydraulic cylinder (15) based on the height of the tree canopy center relative to the ground. After the controller (12) controls the hydraulic cylinder (15) to execute, the position of the servo motor (18) and the nozzle (16) is adjusted. S4: Based on the height of the adjusted servo motor (18) relative to the ground, calculate the rotation angle of the servo motor (18) and control the nozzle (16) to rotate and align with the target tree canopy; S5: Open the solenoid valve (17) and control the water pump to supply liquid to the nozzle (16) through the infusion hose (22). During the process of the nozzle (16) spraying the target tree canopy, the amount of spraying is controlled in real time by the BP neural network incremental PID control strategy. Before S1, a spraying path plan is developed based on the current distribution of pests and diseases in the orchard; The specific process of spray path planning includes: The orchard is divided into multiple zones according to a preset area. A drone equipped with a high-definition camera is used to inspect the fruit trees in each zone according to a preset flight path, and real-time images of the fruit trees in each zone are collected. The regions with pests and diseases are marked, and then the fruit tree images of the regions with pests and diseases are input into a pre-trained neural network model for processing, and the location of the fruit trees with pests and diseases and the type of pests and diseases are output. Get the number of fruit trees with pests and diseases in a partition with pests and diseases within a given time period; Through the formula: Calculate and obtain the j-th outlier value of the fruit tree in the current partition. ; in, Let be the number of fruit trees affected by pests in the j-th instance. Let be the number of fruit trees that exhibit disease in the j-th instance. This represents the total number of fruit trees in the current partition. , These are the weighting coefficients. The total number of drone patrols within a given time period. For stability coefficient, The conversion factor; Will Compared with the preset abnormal threshold Compare; like ≥ If so, the current partition is determined to be in an abnormal state; like If so, then the current partition is determined to be in a normal state; Zones exhibiting abnormal conditions will be included in the spraying plan, and abnormal values ​​of fruit trees will be used to identify these zones. Sort the numbers 1, 2, ..., N in descending order, and use the order of the partitions corresponding to the numbers as the spraying paths; The process of obtaining the stability coefficient is as follows: Obtain the curve showing the change in the number of diseased fruit trees over time within a work cycle. (t) and the curves showing the change in the number of fruit trees affected by pests over time. (t); Based on historical data, obtain a standard curve showing the change in the number of diseased fruit trees over time within a work cycle. (t) and standard variation curves of the number of fruit trees infested with pests over time. (t); Through the formula: Calculate and obtain the stability coefficient ; in, , The proportionality coefficient is selected and determined based on historical data; for The preset threshold; The conversion factor is selected and determined based on historical and experimental data.

2. The method for precise target targeting and optimal drug dosage control according to claim 1, characterized in that, The working process of S2 is as follows: The tree canopy images were segmented using the Mask R-CNN algorithm, and redundant non-target tree canopies were filtered out. The area of ​​each tree canopy was calculated according to the following formula, and the tree canopy with the largest area was the target tree canopy. ； ； in, It is the area of ​​the two-dimensional fruit tree canopy. It is the area of ​​the i-th sector. It is the first i The radius of the sector, The central angle of the sector is... It refers to the target canopy area.

3. The method for precise target targeting and optimal drug dosage control according to claim 1, characterized in that, The process of determining the extension distance of the hydraulic cylinder (15) based on the height of the tree canopy center above the ground is as follows: Obtain the height of the tree canopy center above the ground The extension distance of the hydraulic cylinder (15) is calculated using the following formula. : in, The reference height for hydraulic cylinder (15) The limit height of the hydraulic cylinder (15); like When this occurs, the hydraulic cylinder (15) is controlled to not extend or retract; like When the hydraulic cylinder (15) is extended, it is controlled to extend to... ; like Then control the hydraulic cylinder (15) to extend to .

4. The method for precise target targeting and optimal drug dosage control according to claim 1, characterized in that, The process of obtaining the rotation angle of the servo motor (18) is as follows: Obtain the height of the adjusted servo (18) relative to the ground. ; Obtain the horizontal straight-line distance X between the servo motor (18) and the center of the tree canopy; Through the formula: The rotation angle of the servo motor (18) was calculated. , .

5. The method for precise target targeting and optimal drug dosage control according to claim 1, characterized in that, The process of controlling the amount of pesticide sprayed includes: Through the formula: ； ； in, for Control the liquid output increment at all times. The proportional control coefficient for PID control. For PID integral control coefficients, These are the PID derivative control coefficients. For sampling sequence number, =0, 1, 2, ..., For the first The actual liquid output volume was sampled once. For the first Deviation in liquid output volume during secondary sampling.

Images