Variable pesticide application system carried by unmanned aerial vehicle and pesticide application method

Abstract

The invention belongs to the technical field of agricultural plant protection, and provides an unmanned aerial vehicle-mounted variable pesticide application system and a pesticide application method.The pesticide application system comprises an unmanned aerial vehicle, an orchard pest and disease damage distribution prescription map, a spraying system and a control system. The spraying system is mainly composed of a pesticide box, a liquid conveying pipe, a water pump and sprayers. The control system is mainly composed of a flight control module and a spraying control module. An orchard is rasterized through an orchard pest and disease damage distribution prescription map, the pesticide amount of each grid is obtained, the real-time position of the unmanned aerial vehicle is obtained in combination with a global positioning system (GPS) of the flight control module, the control system adjusts the flow of each nozzle in real time according to the pesticide amount distributed by the current grid where the unmanned aerial vehicle is located, and the variable pesticide application function is achieved. According to the pesticide application system and method, variable pesticide application of the unmanned aerial vehicle based on the orchard pest and disease damage distribution prescription map can be achieved, and the pesticide application accuracy and the pesticide utilization rate can be improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural plant protection technology, and in particular to a variable spraying system and spraying method carried by an unmanned aerial vehicle (UAV). Background Technology

[0002] Mountain orchards, as important bases for the supply of specialty agricultural products, play a vital role in ensuring fruit supply and promoting economic development in mountainous areas through their production management. However, compared with traditional plain orchards, the plant protection and pesticide application process in mountain orchards has long faced severe challenges, mainly due to their complex site environment and climatic conditions. (1) Mountain orchards are usually steep, with large slopes and fragmented plots, making it difficult for traditional large-scale ground spraying machinery to enter and carry out operations. Relying on manual labor is not only extremely inefficient and labor-intensive, but also poses safety risks such as falls from heights and pesticide poisoning, leading to problems such as untimely spraying and incomplete coverage.

[0003] (2) Due to complex factors such as canopy undulation and wind direction, the traditional constant spraying mode of UAV is prone to uneven spraying, resulting in insufficient or excessive spraying, which cannot effectively control pests and diseases. Summary of the Invention

[0004] Therefore, one objective of this invention is to provide a variable-rate drug delivery system and method carried by an unmanned aerial vehicle (UAV) to solve the problems mentioned in the background art and overcome the shortcomings of the prior art.

[0005] To achieve the above objectives, in one aspect, the present invention provides a variable-rate pesticide application system mounted on a drone, comprising: a drone, a spraying system, and a control system, wherein both the spraying system and the control system are mounted on the drone. The spraying system includes a medicine tank, an infusion pipe, a water pump, and nozzles. The medicine tank is installed below the drone via a medicine tank bracket. One end of the infusion pipe is connected to the medicine tank via the water pump, and the other end of the infusion pipe is connected to the nozzles. The water pump is a dual electromagnetic impeller pump. The nozzles are arranged in groups of two, for a total of two groups, and are respectively installed below the two rear rotors of the drone. The control system includes a flight control module and a spraying control module. The flight control module obtains the real-time position of the UAV through GPS and determines the current grid of the UAV by combining the ray method and the grid center point distance algorithm. The spraying control module constructs an optimal variable control model based on the pulse width modulation signal duty cycle and water pump pressure based on the fuzzy PID control algorithm to adjust the amount of pesticide in the nozzle in real time. Combined with the grid of the UAV, it realizes variable pesticide application based on the distribution prescription map of orchard pests and diseases. The orchard pest distribution prescription map is based on multi-angle orchard tree images collected by a high-precision unmanned aerial vehicle, a disease and pest detection model and a disease and pest grade classification model are constructed by using a convolutional neural network to identify diseases and pests of the fruit trees and classify the diseases and pests, the orchard tree images are rasterized, the disease and pest grade of each grid is evaluated based on the disease and pest grade of a single tree body, a mapping relationship between the disease and pest grade and the spraying amount is determined, and an orchard pest distribution prescription map containing grid information, amount information and position information is generated.

[0006] As preferred, the pesticide application system further comprises a data transmission system, a flow sensor module, a battery and a radiator, the data transmission system realizes communication between the unmanned aerial vehicle and a computer and a ground station, the flow sensor module acquires real-time flow at the spray head through ADC sampling and feeds back to the control system to form a negative feedback to correct the deviation at any time, the battery is installed at the rear of the fuselage of the unmanned aerial vehicle to provide power for the variable pesticide application system, and the radiator is installed at the electronic device of the unmanned aerial vehicle.

[0007] As preferred, the construction of the disease and pest detection model and the disease and pest grade classification model comprises: Multi-angle fruit disease and pest images are collected by the unmanned aerial vehicle, the positions of the diseases and pests in each image are labeled, and the original image and the image with the labeled disease and pest positions are used as a first sample training set; According to the disease degree, the diseases and pests are classified into grades, the labels are: asymptomatic, healthy, early, medium and severe, and the image with the labeled disease and pest positions and the disease grade thereof are used as a second sample training set; The first sample training set and the second sample training set are respectively expanded through image flipping, brightness adjustment, Gaussian noise addition and rain day atomization, and a disease and pest detection dataset and a disease and pest grade classification dataset are constructed; The disease and pest detection dataset and the disease and pest grade classification dataset are trained based on Mask-RCNN and ResNet50 network models to construct a disease and pest detection model and a disease and pest grade classification model.

[0008] As preferred, the optimal variable control model of the pulse width modulation signal duty cycle and the water pump pressure is constructed based on the fuzzy PID control algorithm to adjust the spray amount of the spray head in real time, which comprises: The spraying error and the error change rate are determined according to the target spraying amount and the actual spraying amount of the spray head, the fuzzy PID control algorithm is adopted, the fuzzy rules are designed according to the spraying error and the error change rate: a large control amount is output to quickly respond when the error is large, and a small control amount is output to avoid overshoot when the error is small, the output control amount is calculated through fuzzy reasoning, and the output control amount is used to adjust the pulse width modulation signal duty cycle to change the motor speed of the water pump and realize accurate control of the amount.

[0009] As preferred, the flight control module acquires the real-time position of the UAV as longitude and latitude in the world coordinate system through GPS, and converts it into local east-north coordinates through a coordinate conversion formula, and compensates for the spraying delay of the UAV spraying system according to the local east-north coordinates and based on a UAV position prediction formula.

[0010] As preferred, the coordinate conversion formula is as follows: ; wherein, is the local east coordinate, is the local north coordinate, represents the longitude in the world coordinate system, represents the latitude in the world coordinate system, represents the longitude of the coordinate conversion reference point, represents the latitude of the coordinate conversion reference point.

[0011] As preferred, the position prediction formula is as follows: ; wherein, is the predicted local east coordinate, is the current local east coordinate, is the eastward velocity component, is the predicted local north coordinate, represents the current local north coordinate, represents the northward velocity component, represents the spraying delay time.

[0012] In another aspect, the present application provides a variable pesticide application method carried by a UAV, comprising: S1: orchard information collection: high-precision UAV is used to collect orchard information to obtain high-definition pictures of the orchard trees from multiple angles; S2: orchard pest distribution prescription map generation: based on the high-definition pictures of the trees, a CNN is used to build an orchard pest detection model, which can realize the identification of tree pests and the classification of the grade of single tree pests, a chessboard segmentation method is used to grid the orchard image, the grade of each grid pest is evaluated based on the grade of single tree pests, the mapping relationship between the grade of pests and the amount of sprayed pesticides is determined by consulting the standard of pests and pesticide dosage, and an orchard pest distribution prescription map is generated, which includes grid information, pesticide amount information and position information; S3: orchard pest distribution prescription map uploading: the orchard pest distribution prescription map is uploaded to the UAV through the cloud, and the UAV receives the task; S4: body inspection and pesticide preparation: check whether the UAV rotor arm buckle is tightly clamped, detect whether the spraying system leaks, add the prepared pesticide solution according to the demand in the pesticide box, and wait for use; S5: Autonomous flight of drones: Based on the distribution prescription map of orchard pests and diseases, the drone's autonomous flight path is rationally planned at the ground station, and the flight control module controls the drone to fly autonomously and avoid obstacles according to the flight path. S6: Variable application of pesticides based on orchard pest and disease distribution prescription map: The flight control module obtains the real-time position of the drone and determines the grid where the drone is located through the ray method and the algorithm based on the distance between the grid center point. The spraying control module adjusts the amount of pesticide in real time according to the optimal variable control model of PWM signal duty cycle and water pump pressure, so as to realize the spraying operation of the drone according to the amount of pesticide corresponding to different grids. S7: Drone Return: After completing the above variable application operations, the drone returns to the takeoff point according to the planned path.

[0013] Preferably, the gridding process uses a checkerboard division method to ensure that the size of each grid matches the spray width of the drone. The spray width is set to the grid side length to ensure that the spray width covers the grid when the drone flies in from the center of the grid.

[0014] Preferably, obtaining the real-time location of the UAV and determining the grid where the UAV is located includes: obtaining the real-time location of the UAV through GPS, using the ray method and a grid center point distance algorithm to determine the current grid where the UAV is located, and combining coordinate transformation to convert the world coordinate system latitude and longitude of the UAV location to local east-north coordinates.

[0015] Preferably, the coordinate transformation formula is as follows: ; in, The coordinates are local eastward. Using local north-facing coordinates, Represents longitude in the world coordinate system. Represents latitude in the world coordinate system. Indicates the longitude of the coordinate transformation reference point. This indicates the latitude of the coordinate transformation reference point.

[0016] Preferably, in step S6, the spraying delay of the spraying system is also compensated by a position prediction formula. The position prediction formula calculates the predicted coordinates based on the current coordinates, velocity components, and delay time. The specific position prediction formula is as follows: ; in, To predict the local eastward coordinates, The current local eastward coordinates, For the eastward velocity component, To predict local northward coordinates, This indicates the current local northward coordinates. Indicates the northward velocity component. Indicates the spraying delay time.

[0017] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows: This invention employs high-precision UAV information acquisition and a convolutional neural network (CNN) to construct an orchard pest and disease detection model, enabling the identification and classification of fruit tree pests and diseases. It also combines a checkerboard segmentation method to rasterize orchard images, generating a pest and disease distribution prescription map. Simultaneously, a fuzzy PID control algorithm is used to establish an optimal variable control model for the PWM signal duty cycle and water pump pressure, allowing for real-time adjustment of the sprayer dosage. This achieves variable-rate application based on the prescription map, overcoming the uneven pesticide application caused by the constant-rate spraying of traditional plant protection UAVs. This improves application accuracy and pesticide utilization, contributing to enhanced control effectiveness and reduced costs. Furthermore, the constructed orchard pest and disease distribution prescription map includes raster information, pesticide dosage information, and location information, providing data support for subsequent orchard management and facilitating the optimization of application strategies and orchard maintenance.

[0018] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a schematic diagram of the variable drug delivery system of the present invention; Figure 2 This is a schematic diagram of the variable drug administration method of the present invention; Figure 3 This is a flowchart of a method for applying pesticides using a drone based on a prescription map of orchard pest and disease distribution, as described in an embodiment of the present invention. Figure 4 This is a schematic diagram of the distribution of pests and diseases in an orchard according to an embodiment of the present invention; Figure 5 This is a schematic diagram of a drone application operation based on a prescription map of orchard pest and disease distribution, according to an embodiment of the present invention. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0021] Example 1: like Figure 1As shown in the figure, an embodiment of the present invention provides a variable-rate pesticide application system mounted on a drone, comprising: a drone, a spraying system, and a control system, wherein both the spraying system and the control system are mounted on the drone. The spraying system includes a medicine tank, an infusion pipe, a water pump, and nozzles. The medicine tank is installed below the drone via a medicine tank bracket. One end of the infusion pipe is connected to the medicine tank via the water pump, and the other end of the infusion pipe is connected to the nozzles. The water pump is a dual electromagnetic impeller pump. The nozzles are arranged in groups of two, for a total of two groups, and are respectively installed below the two rear rotors of the drone. The control system includes a flight control module and a spraying control module. The flight control module obtains the real-time position of the UAV through GPS and determines the current grid of the UAV by combining the ray method and the grid center point distance algorithm. The spraying control module constructs an optimal variable control model based on the pulse width modulation signal duty cycle and water pump pressure based on the fuzzy PID control algorithm to adjust the amount of pesticide in the nozzle in real time. Combined with the grid of the UAV, it realizes variable pesticide application based on the distribution prescription map of orchard pests and diseases. The orchard pest and disease distribution prescription map is based on multi-angle orchard tree images collected by high-precision UAVs. A convolutional neural network is used to construct a pest and disease detection model and a pest and disease level classification model to identify and classify pest and disease levels. The orchard tree images are rasterized, and the pest and disease level of each grid is evaluated based on the pest and disease level of a single tree. The mapping relationship between pest and disease level and spraying amount is determined, and an orchard pest and disease distribution prescription map containing grid information, pesticide amount information and location information is generated.

[0022] This invention employs high-precision UAV information acquisition and a convolutional neural network (CNN) to construct an orchard pest and disease detection model, enabling the identification and classification of fruit tree pests and diseases. It also combines a checkerboard segmentation method to rasterize orchard images, generating a pest and disease distribution prescription map. Simultaneously, a fuzzy PID control algorithm is used to establish an optimal variable control model for the PWM signal duty cycle and water pump pressure, allowing for real-time adjustment of the sprayer dosage. This achieves variable-rate application based on the prescription map, overcoming the uneven pesticide application caused by the constant-rate spraying of traditional plant protection UAVs. This improves application accuracy and pesticide utilization, contributing to enhanced control effectiveness and reduced costs. Furthermore, the constructed orchard pest and disease distribution prescription map includes raster information, pesticide dosage information, and location information, providing data support for subsequent orchard management and facilitating the optimization of application strategies and orchard maintenance.

[0023] Furthermore, the application system also includes a data transmission system, a flow sensor module, a battery, and a heat sink. The data transmission system enables communication between the UAV and the computer and ground station. The flow sensor module obtains the real-time flow at the nozzle through ADC sampling and feeds it back to the control system to form negative feedback to correct deviations at all times. The battery is installed at the rear of the UAV to provide power for the variable application system. The heat sink is installed at the electronic equipment of the UAV.

[0024] Furthermore, the construction of the pest and disease detection model and the pest and disease classification model includes: Images of fruit diseases and pests were collected from multiple angles using drones. The locations of diseases and pests in each image were marked, and the original images and the images with marked disease and pest locations were used as the first sample training set. Based on the severity of the disease, pests and diseases are classified into levels and labeled as: asymptomatic, healthy, early stage, mid-stage, and severe. Images with labeled locations of pests and diseases and their corresponding disease levels are used as the second sample training set. The first and second sample training sets were expanded by image flipping, brightness adjustment, Gaussian noise addition and rain fogging respectively, to construct a pest and disease detection dataset and a pest and disease level classification dataset. The disease and pest detection model and the disease and pest classification model were trained on the Mask-RCNN and ResNet50 network models, respectively.

[0025] Furthermore, the construction of an optimal variable control model based on the fuzzy PID control algorithm, which relates the pulse width modulation signal duty cycle to the water pump pressure, for real-time adjustment of the nozzle dosage includes: The spraying error and error rate of change are determined based on the target pesticide dosage and the actual pesticide dosage sprayed by the nozzle. A fuzzy PID control algorithm is adopted, and fuzzy rules are designed based on the spraying error and error rate of change: a large control quantity is output for rapid response when the error is large, and a small control quantity is output to avoid overshoot when the error is small. The output control quantity is calculated through fuzzy inference. The output control quantity is used to adjust the duty cycle of the pulse width modulation signal to change the motor speed of the water pump, thereby achieving precise control of the pesticide dosage. (Highlighted in yellow) In one implementation, a first error threshold, a first error change rate threshold, and a first output control quantity threshold are set. When the error is large, i.e. the spraying error is greater than the first error threshold and / or the error change rate is greater than the first error change rate threshold, the output large control quantity is the control quantity greater than the first output control quantity threshold. When the error is small, i.e. the spraying error is less than the first error threshold and / or the error change rate is less than the first error change rate threshold, the output small control quantity is the control quantity less than the first output control quantity threshold.

[0026] It is understandable that the first error threshold, the first error change rate threshold, and the first output control quantity threshold can be set according to the actual situation.

[0027] Furthermore, a mapping relationship between error and / or error rate of change and output control quantity can be established, with each error and / or error rate of change corresponding to an output control quantity.

[0028] Furthermore, the flight control module obtains the real-time position of the UAV in the world coordinate system using GPS, and converts it to local east-north coordinates using a coordinate transformation formula. Based on the local east-north coordinates and the UAV position prediction formula, the spraying delay of the UAV spraying system is compensated.

[0029] Furthermore, the coordinate transformation formula is as follows: ; in, The coordinates are local eastward. Using local north-facing coordinates, Represents longitude in the world coordinate system. Represents latitude in the world coordinate system. Indicates the longitude of the coordinate transformation reference point. This indicates the latitude of the coordinate transformation reference point.

[0030] Furthermore, the location prediction formula is as follows: ; in, To predict the local eastward coordinates, The current local eastward coordinates, For the eastward velocity component, To predict local northward coordinates, This indicates the current local northward coordinates. Indicates the northward velocity component. Indicates the spraying delay time.

[0031] Example 2: A variable-rate pesticide application system carried by a drone includes a drone, an orchard pest and disease distribution prescription map, a spraying system, and a control system.

[0032] The data for the orchard pest and disease distribution prescription map mainly comes from high-precision UAV information collection, acquiring high-definition images of the orchard from multiple angles. A convolutional neural network (CNN) is used to construct an orchard pest and disease detection model to realize the identification of fruit tree pests and diseases and the classification of pest and disease levels for individual trees. The orchard images are rasterized using a chessboard segmentation method. The pest and disease levels of each grid are evaluated based on the pest and disease levels of individual trees. By consulting the standards for pests and diseases and pesticide dosage, the mapping relationship between pest and disease levels and pesticide spraying amounts is determined, and an orchard pest and disease distribution prescription map is generated.

[0033] The spraying system includes a medicine tank, an infusion pipe, a water pump, and nozzles; the medicine tank is mounted on the underside of the drone by a medicine tank bracket; one end of the infusion pipe is connected to the medicine tank via the water pump, and the other end is connected to the nozzles; the water pump is a dual electromagnetic impeller pump; the nozzles are arranged in groups of two, for a total of two groups, and are respectively installed under the two rear rotors of the drone.

[0034] The control system includes a flight control module and a spraying control module. The flight control module uses a PX4 to obtain the real-time position of the UAV via GPS and determines the current grid cell of the UAV by combining the ray casting method and a grid center point distance algorithm. The spraying control module embeds a fuzzy PID control algorithm to construct an optimal variable control model of PWM signal duty cycle and water pump pressure, adjusts the spray nozzle dosage in real time, and achieves variable application of pesticides based on the orchard pest and disease distribution prescription map by combining the grid cell of the UAV.

[0035] The variable-rate dosing system carried by the UAV also includes a data transmission system, a flow sensor module, a battery, a heat sink, and rotors. The data transmission system enables communication between the UAV and a computer and a ground station. The flow sensor module uses an ADC to sample and obtain real-time flow data, which is then fed back to the controller to form negative feedback and constantly correct deviations. The battery is installed at the rear of the UAV to provide power. The heat sink is installed on electronic devices such as the NVIDIA Jetson TX2 to ensure their normal operation.

[0036] Example 3: This invention also discloses a method for constructing the orchard pest and disease detection model and the pest and disease classification model, including the following steps: Step A1: Use a drone to collect multi-angle images of fruit diseases and pests, such as mango anthracnose. Mark the location of the disease and pests in each mango anthracnose image, and use the original mango disease and pest images and the images with marked disease and pest locations as the first sample training set; Step A2: Based on the severity of the disease, classify the mango diseases and pests in Step S1 into different levels, labeled as: asymptomatic, healthy, early stage, mid-stage, and severe. Use the images with the marked locations of mango diseases and pests and their respective disease levels as the second training set. Step A3: Expand the two sample training sets by using data augmentation techniques such as image flipping, brightness adjustment, Gaussian noise addition, and rain fogging to construct a mango anthracnose disease and pest detection dataset and a disease and pest classification dataset. Step A4: Use Mask-RCNN and ResNet50 network models to train on the two datasets respectively to build a mango disease and pest detection model and a disease and pest classification model.

[0037] Example 4: This invention also discloses an optimal variable control model for the duty cycle of the PWM signal and the water pump pressure, specifically: A fuzzy PID control algorithm is adopted. Based on the error and the rate of change of the error between the target amount of pesticide and the actual amount of pesticide, fuzzy rules are designed: when the error is large, a large control quantity is output for fast response, and when the error is small, a small control quantity is output to avoid overshoot. The output control quantity is calculated through fuzzy inference. This output control quantity is directly used to adjust the PWM duty cycle, thereby changing the motor speed of the water pump and achieving precise control of the amount of pesticide.

[0038] Example 5: On the other hand, such as Figure 2 As shown, the present invention provides a variable-rate dosing method carried by an unmanned aerial vehicle (UAV), comprising: S1: Orchard Information Collection: High-precision drones are used to collect orchard information and obtain high-definition images of orchard trees from multiple angles; S2: Orchard Pest and Disease Distribution Prescription Map Generation: Based on high-definition images of orchard trees, a convolutional neural network is used to construct a pest and disease detection model and a pest and disease level classification model to realize pest and disease identification and level classification. The high-definition images of orchard trees are rasterized, the pest and disease level of each grid is evaluated, the mapping relationship between pest and disease level and spraying amount is determined, and an orchard pest and disease distribution prescription map containing grid information, pesticide amount information and location information is generated. S3: Upload of orchard pest and disease distribution prescription map: Upload the orchard pest and disease distribution prescription map to the drone via the cloud; S4: Airframe inspection and liquid preparation: Check the tightness of the drone rotor arm latches and the water leakage of the spraying system, and add the prepared liquid to the medicine tank; S5: Autonomous flight of drones: Based on the distribution prescription map of orchard pests and diseases, the drone flight path is planned, and the flight control module controls the drone to fly autonomously and avoid obstacles according to the flight path; S6: Variable application of pesticides based on orchard pest and disease distribution prescription map: The flight control module obtains the real-time position of the drone and determines the grid where the drone is located. The spraying control module adjusts the amount of pesticide in real time according to the optimal variable control model of pulse width modulation signal duty cycle and water pump pressure to realize spraying operation according to the amount of pesticide corresponding to different grids. S7: Drone Return: After completing the spraying operation, the drone returns to its takeoff point.

[0039] Furthermore, the gridding process employs a checkerboard division method to ensure that the size of each grid matches the spray width of the drone. The spray width is set to the grid side length to ensure that the spray width covers the grid when the drone flies in from the center of the grid.

[0040] Furthermore, the step of obtaining the real-time location of the UAV and determining the grid where the UAV is located includes: obtaining the real-time location of the UAV through GPS, using the ray method and the grid center point distance algorithm to determine the current grid where the UAV is located, and combining coordinate transformation to convert the world coordinate system latitude and longitude of the UAV location to local east-north coordinates.

[0041] Furthermore, the coordinate transformation formula is as follows: ; in, The coordinates are local eastward. Using local north-facing coordinates, Represents longitude in the world coordinate system. Represents latitude in the world coordinate system. Indicates the longitude of the coordinate transformation reference point. This indicates the latitude of the coordinate transformation reference point.

[0042] Furthermore, in step S6, the spraying delay of the spraying system is compensated by a position prediction formula. The position prediction formula calculates the predicted coordinates based on the current coordinates, velocity components, and delay time. The specific position prediction formula is as follows: ; in, To predict the local eastward coordinates, The current local eastward coordinates, For the eastward velocity component, To predict local northward coordinates, This indicates the current local northward coordinates. Indicates the northward velocity component. Indicates the spraying delay time.

[0043] Example 6: like Figures 3 to 5 As shown, the drug delivery method using the variable-rate drug delivery system mounted on the aforementioned UAV is implemented through the following process: A CNN-based orchard pest and disease detection model was constructed to identify fruit tree pests and diseases and classify the pest and disease severity of individual trees. A checkerboard segmentation method was used to rasterize the orchard image. The pest and disease severity of each grid cell was evaluated based on the severity of individual trees. By consulting standards for pest and disease severity and pesticide dosage, the mapping relationship between pest and disease severity and pesticide application was determined. Furthermore, an orchard pest and disease distribution prescription map was generated by combining the location information of the orchard grid cells to achieve optimal orchard management and pesticide application strategies. The orchard pest and disease distribution prescription map includes grid information, pesticide dosage information, and location information, such as… Figure 4 As shown.

[0044] The prescription image is uploaded to the drone via the cloud, and the drone receives the task.

[0045] Before the drone takes off, it is necessary to conduct a fuselage inspection and prepare the spray solution. Check whether the drone rotor arm latches are tight, test whether the spraying system is working properly, and put the prepared spray solution into the tank in advance.

[0046] A schematic diagram of drone-based pesticide application based on orchard pest and disease distribution prescription maps is shown below. Figure 5 As shown, the numbers in parentheses represent the dosage of pesticide, and the numbers outside the parentheses represent the grid number. Based on the orchard pest and disease distribution map, the autonomous flight path of the drone is rationally planned at the ground station. The spray width is set to the grid side length to ensure the drone flies in from the center of the grid, meeting the requirement of spraying width covering the grid size. After the drone flies into the grid, it sprays according to the grid number and corresponding dosage. The flight control module controls the drone to fly autonomously and avoid obstacles according to the flight path.

[0047] The flight control module acquires the drone's real-time position via GPS and determines the grid cell where the drone is located using a ray-mapping method and a grid center point distance algorithm. The spraying control module adjusts the pesticide dosage in real-time based on an optimal variable control model that combines the PWM signal duty cycle and water pump pressure, enabling the drone to spray according to the dosage corresponding to different grid cells. The real-time drone position acquired by GPS is in latitude and longitude in the world coordinate system. Since the drone needs to acquire local E-north coordinates during flight, a coordinate transformation is required. The formula for converting latitude and longitude in the world coordinate system to local E-north coordinates is shown below: ; In the formula, A represents the local eastward coordinates, B represents the local northward coordinates, x represents the longitude in the world coordinate system, and y represents the latitude in the world coordinate system. Indicates the longitude of the coordinate transformation reference point. This indicates the latitude of the coordinate transformation reference point.

[0048] Due to communication delays and pump response time, there is a certain delay from the issuance of the control signal to the actual spraying of the liquid and its coverage of the target area. Therefore, the following formula is proposed for UAV position prediction: ; In the formula, This indicates the predicted local eastward coordinates. Indicates the current local eastward coordinates. Indicates the eastward velocity component. This indicates the predicted local northward coordinates. This indicates the current local northward coordinates. Indicates the northward velocity component. This indicates the spraying delay time. Among them, It can be adjusted according to the actual working conditions of the drone. This prediction formula can compensate for spraying system delays and improve spraying accuracy.

[0049] After completing the above-mentioned variable-rate application operations, the drone returned to its takeoff point along the planned path.

[0050] This invention utilizes the aforementioned variable application system and method mounted on a drone to establish a drone variable application system and proposes a variable application method based on orchard pest and disease distribution prescription maps, which helps to improve the accuracy of application and the utilization rate of pesticides.

[0051] This invention introduces variable-rate application into UAV technology, enabling UAVs to apply pesticides based on orchard pest and disease distribution maps. This overcomes the problem of uneven pesticide application that exists in traditional plant protection UAVs with constant-rate spraying, achieving precise application, which helps improve control effects and reduce costs.

[0052] This invention constructs a prescription map of orchard pest and disease distribution, which facilitates later orchard management.

[0053] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0054] It will be readily understood by those skilled in the art that this invention includes any combination of the inventive description and specific embodiments outlined in the foregoing specification, as well as the various parts shown in the accompanying drawings. Due to space limitations and for the sake of brevity, not all of these combinations have been described in detail. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0055] Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A variable-rate dosing system mounted on an unmanned aerial vehicle (UAV), characterized in that, include: The system includes a drone, a spraying system, and a control system, with both the spraying system and the control system mounted on the drone. The spraying system includes a medicine tank, an infusion pipe, a water pump, and nozzles. The medicine tank is installed below the drone via a medicine tank bracket. One end of the infusion pipe is connected to the medicine tank via the water pump, and the other end of the infusion pipe is connected to the nozzles. The water pump is a dual electromagnetic impeller pump. The nozzles are arranged in groups of two, for a total of two groups, and are respectively installed below the two rear rotors of the drone. The control system includes a flight control module and a spraying control module. The flight control module obtains the real-time position of the UAV through GPS and determines the current grid of the UAV by combining the ray method and the grid center point distance algorithm. The spraying control module constructs an optimal variable control model based on the pulse width modulation signal duty cycle and water pump pressure based on the fuzzy PID control algorithm to adjust the amount of pesticide in the nozzle in real time. Combined with the grid of the UAV, it realizes variable pesticide application based on the distribution prescription map of orchard pests and diseases. The orchard pest and disease distribution prescription map is based on multi-angle orchard tree images collected by high-precision UAVs. A convolutional neural network is used to construct a pest and disease detection model and a pest and disease level classification model to identify and classify pest and disease levels. The orchard tree images are rasterized, and the pest and disease level of each grid is evaluated based on the pest and disease level of a single tree. The mapping relationship between pest and disease level and spraying amount is determined, and an orchard pest and disease distribution prescription map containing grid information, pesticide amount information and location information is generated.

2. The variable-rate dosing system carried by the UAV as described in claim 1, characterized in that, It also includes a data transmission system, a flow sensor module, a battery, and a heat sink. The data transmission system enables communication between the UAV and the computer and ground station. The flow sensor module obtains the real-time flow at the nozzle through ADC sampling and feeds it back to the control system to form negative feedback to correct deviations at all times. The battery is installed at the rear of the UAV to provide power for the variable dosing system. The heat sink is installed at the electronic equipment of the UAV.

3. The variable-rate dosing system carried by the UAV as described in claim 1, characterized in that, The construction of the pest and disease detection model and the pest and disease classification model includes: Images of fruit diseases and pests were collected from multiple angles using drones. The locations of diseases and pests in each image were marked, and the original images and the images with marked disease and pest locations were used as the first sample training set. Based on the severity of the disease, pests and diseases are classified into levels and labeled as: asymptomatic, healthy, early stage, mid-stage, and severe. Images with labeled locations of pests and diseases and their corresponding disease levels are used as the second sample training set. The first and second sample training sets were expanded by image flipping, brightness adjustment, Gaussian noise addition and rain fogging respectively, to construct a pest and disease detection dataset and a pest and disease level classification dataset. Based on Mask-RCNN and ResNet50 network models, respectively, pest and disease detection datasets and pest and disease classification datasets were trained to construct pest and disease detection models and pest and disease classification models.

4. The variable-rate dosing system carried by the UAV as described in claim 1, characterized in that, The method based on fuzzy PID control algorithm, which constructs an optimal variable control model for the pulse width modulation signal duty cycle and water pump pressure to adjust the nozzle dosage in real time, includes: The spraying error and error rate of change are determined based on the target amount of pesticide and the actual amount of pesticide sprayed by the nozzle. A fuzzy PID control algorithm is adopted, and fuzzy rules are designed based on the spraying error and error rate of change: when the error is large, a large control quantity is output for fast response, and when the error is small, a small control quantity is output to avoid overshoot. The output control quantity is calculated through fuzzy inference. The output control quantity is used to adjust the duty cycle of the pulse width modulation signal to change the motor speed of the water pump, thereby adjusting the amount of pesticide sprayed by the nozzle.

5. The variable-rate dosing system carried by the UAV as described in claim 1, characterized in that, The flight control module obtains the real-time position of the UAV in the world coordinate system using GPS, converts it to local east-north coordinates using a coordinate transformation formula, and compensates for the spraying delay of the UAV spraying system based on the local east-north coordinates and the UAV position prediction formula.

6. A variable-rate application method carried by an unmanned aerial vehicle (UAV), characterized in that, include: S1: Use high-precision drones to collect orchard information and obtain high-definition images of orchard trees from multiple angles; S2: Based on high-definition images of orchard trees, a convolutional neural network is used to construct a pest and disease detection model and a pest and disease level classification model to realize pest and disease identification and level classification. The high-definition images of orchard trees are rasterized to evaluate the pest and disease level of each grid, determine the mapping relationship between pest and disease level and spraying amount, and generate an orchard pest and disease distribution prescription map containing grid information, pesticide amount information and location information. S3: Upload the orchard pest and disease distribution map to the drone via the cloud; S4: Check the tightness of the drone rotor arm latches and the water leakage of the spraying system; add the prepared medicine to the medicine tank. S5: Based on the distribution map of pests and diseases in the orchard, plan the flight path of the drone, and the flight control module controls the drone to fly autonomously and avoid obstacles according to the flight path; S6: The flight control module acquires the real-time position of the drone and determines the grid where the drone is located. The spraying control module adjusts the amount of pesticide in real time based on the optimal variable control model of pulse width modulation signal duty cycle and water pump pressure to realize spraying operation according to the amount of pesticide corresponding to different grids. S7: After completing the spraying operation, the drone returns to the takeoff point.

7. The variable-rate dosing method carried by an unmanned aerial vehicle as described in claim 6, characterized in that, The gridding process uses a checkerboard division method to ensure that the size of each grid matches the spray width of the drone. The spray width is set to the grid side length to ensure that the spray width covers the grid when the drone flies in from the center of the grid.

8. The variable-rate dosing method carried by an unmanned aerial vehicle as described in claim 6, characterized in that, The process of obtaining the real-time location of the UAV and determining the grid cell in which the UAV is located includes: obtaining the real-time location of the UAV through GPS, determining the current grid cell of the UAV using the ray method and a grid center point distance algorithm, and converting the world coordinate system latitude and longitude of the UAV's location to local east-north coordinates using coordinate transformation.

9. The variable-rate dosing method carried by an unmanned aerial vehicle as described in claim 8, characterized in that, The coordinate transformation formula is as follows: ； in, The coordinates are local eastward. Using local north-facing coordinates, Longitude in the world coordinate system Latitude in the world coordinate system The longitude of the reference point for coordinate transformation. The latitude is the reference point for coordinate transformation.

10. The variable-rate dosing method carried by an unmanned aerial vehicle as described in claim 9, characterized in that, In step S6, the spraying delay of the spraying system is also compensated by a position prediction formula. The position prediction formula calculates the predicted coordinates based on the current coordinates, velocity components, and delay time. The specific position prediction formula is as follows: ； in, To predict the local eastward coordinates, The current local eastward coordinates, For the eastward velocity component, To predict local northward coordinates, The current local northward coordinates, For the northward velocity component, This is the spraying delay time.

Images