Unmanned aerial vehicle autonomous spraying system and method based on crop canopy perception and prescription decision

Abstract

The invention discloses an unmanned aerial vehicle autonomous spraying system and method based on crop canopy perception and prescription decision. The system comprises an unmanned aerial vehicle, an intelligent perception module, an interaction module, a point cloud map generation module, a crop canopy semantic information module, a prescription map generation and analysis module, a control module and a variable spraying execution module. The intelligent sensing module comprises a laser radar and a positioning sub-module, the laser radar is used for collecting three-dimensional point cloud data, and the positioning sub-module is used for collecting longitude and latitude information; the point cloud map generation module is used for generating a three-dimensional point cloud map; a crop canopy semantic information module extracts cotton canopy semantic information of the three-dimensional point cloud map; the prescription map generation and analysis module is used for converting the semantic information of the cotton canopies to obtain a prescription map containing the semantic information of the cotton canopies; the variable spraying execution module is used for receiving instructions of the control module to dynamically control spraying flow. The cotton defoliation effect and the liquid medicine utilization rate can be improved, and environmental pollution is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of aerial plant protection, and in particular relates to an autonomous spraying system and method of an unmanned aerial vehicle based on crop canopy perception and prescription decision-making. Background Art

[0002] To reduce impurities in cotton fibers and improve cotton quality, defoliants are sprayed before mechanized harvesting. Drones are gaining increasing attention in this field due to their high spraying efficiency, gentleness on the cotton, and adaptability to terrain and cotton growth conditions.

[0003] Defoliant spraying by drones requires precise spray flow control. This is because when the defoliant concentration sprayed by drones is high, excessive spraying can damage crops, which in turn affects crop growth and yield, and even causes environmental pollution. When the defoliant concentration sprayed by drones is low, insufficient defoliant spraying will prevent the defoliant from effectively penetrating the crop canopy, resulting in poor defoliation effect, which in turn affects the normal management and harvesting of crops. However, at present, precise spray flow control of cotton defoliants by drones is challenging. This is because traditional drones do not incorporate characteristics such as the operating environment and crop canopy morphology into the drone's variable spraying decision-making. They lack the ability to process information within the map, resulting in the drone's inability to intelligently and accurately spray the cotton field environment, affecting the defoliation effect of the cotton. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the above-mentioned existing technologies and provide a drone autonomous spraying system based on crop canopy perception and prescription decision-making. The system can effectively extract the cotton canopy semantic information in the point cloud map of the operating area through the effective cooperation of various modules, and generate a prescription map containing the cotton canopy semantic information based on the cotton canopy semantic information, thereby effectively supporting the drone and the variable spray execution module to accurately spray the operating area, thereby effectively improving the defoliation effect of cotton and the utilization rate of the liquid medicine.

[0005] Another object of the present invention is to provide a drone autonomous spraying method based on crop canopy perception and prescription decision-making.

[0006] In order to achieve the above object, the present invention can be achieved by adopting the following technical solutions:

[0007] An autonomous drone spraying system based on crop canopy perception and prescription decision-making, the system comprising: a drone, an intelligent perception module, an interaction module, a point cloud map generation module, a crop canopy semantic information module, a prescription map generation and parsing module, a control module, and a variable spray execution module;

[0008] The drone is used to carry the intelligent sensing module and the variable spray execution module;

[0009] The intelligent perception module includes a laser radar and a positioning submodule. The laser radar is used to collect three-dimensional point cloud data of the operating area. The positioning submodule is used to collect real-time longitude and latitude data of the UAV during flight and longitude and latitude information of the UAV during flight.

[0010] The point cloud map generation module is used to process the three-dimensional point cloud data and longitude and latitude information of the operation area to generate a three-dimensional point cloud map of the operation area;

[0011] The crop canopy semantic information module is used to extract crop canopy semantic information from the three-dimensional point cloud map of the operation area to obtain cotton canopy semantic information;

[0012] The prescription map generation and parsing module is used to quantify and convert the cotton canopy semantic information to obtain cotton point cloud canopy density information, then convert the cotton point cloud canopy density information to obtain liquid spray flow information, and then construct a prescription map of the operation area based on the liquid spray flow information;

[0013] The control module is used to issue control instructions to the variable spray execution module and the drone according to the prescription map and real-time longitude and latitude data;

[0014] The interaction module is used for communication connection between the UAV, intelligent perception module, point cloud map generation module, crop canopy semantic information module, prescription map generation and analysis module, control module and variable spray execution module;

[0015] The variable spray execution module includes a controller and a spray sub-module. The controller is used to receive control instructions issued by the control module to control the spray flow of the spray sub-module.

[0016] Preferably, the working steps of the crop canopy semantic information module are as follows:

[0017] S01. Construct a pre-trained RandLA-Net network model in MATLAB software, wherein the pre-trained RandLA-Net network model uses the DALES aerial point cloud dataset as a training set;

[0018] S02. Performing transfer learning using the pre-trained RandLA-Net network model, modifying the output layer of the pre-trained RandLA-Net network model to ensure that the network's output categories are modified to a preset five-category classification, and using the RandLA-Net network model after transfer learning as a crop canopy semantic information extraction network; the five categories include vegetation, ground, road, building, and flight path;

[0019] S03. Obtain a three-dimensional point cloud map of the operating area, normalize the point cloud data in the three-dimensional point cloud map of the operating area using the normalizePointCloud function, and then divide the normalized point cloud data using the splitData function to obtain a training set and a test set;

[0020] S04, using the training set and data set obtained in step S03 to train the crop canopy semantic information extraction network to obtain a trained crop canopy semantic information extraction network;

[0021] S05. Use the trained crop canopy semantic information extraction network to extract the three-dimensional point cloud map of the operation area to obtain cotton canopy semantic information.

[0022] Preferably, the working steps of the prescription map generation and analysis module are as follows:

[0023] Step 1: First, the operation area is divided into multiple spraying operation areas of equal area according to the spraying width of the drone;

[0024] Step 2: Calculate the cotton point cloud canopy density of each spraying operation area based on the three-dimensional point cloud data and cotton canopy semantic information; the specific expression of the cotton point cloud canopy density is as follows:

[0025]

[0026] Among them D xy is the weighted plane point cloud density; N is the number of points projected in the current spraying operation area; A is the area of ​​the spraying operation area; h i is the height of point i; H max and H min The maximum and minimum heights of points within the spray operation area;

[0027] Step 3: Convert the cotton point cloud canopy density of each spraying operation area to obtain liquid spray flow information, and then construct a prescription map based on the liquid spray flow information.

[0028] Preferably, the specific representation of the liquid medicine spray flow information is as follows:

[0029]

[0030] Among them, Q is the spraying flow rate of the current spraying operation area; Q min is the lower limit of the flow rate of the spray mechanism; Q max The flow rate limit of the spray mechanism; D xy is the point cloud density of the current spraying operation area; D min and D max is the minimum and maximum value of the point cloud density in the working area.

[0031] Preferably, the working process of the point cloud map generation module is as follows:

[0032] First, the point cloud map generation module uses the SLAM mapping algorithm FAST-LIO2 to process the three-dimensional point cloud data, and then combines the processed three-dimensional point cloud data with the latitude and longitude information to generate a three-dimensional point cloud map of the operation area.

[0033] Preferably, the UAV includes a UAV body, a power mechanism, a flight control mechanism and a sensor mechanism. The power mechanism includes multiple power motors, a power distribution board and a power supply. The multiple power motors are respectively connected to the power supply through the power distribution board. The flight control mechanism includes a flight controller and a PWM control connection board. The flight controller is connected to the multiple motors through the PWM control connection board. The flight controller is communicatively connected to the positioning submodule and the control module. The sensor mechanism includes a camera, which is arranged in front of the UAV body. The camera is used to shoot the front view of the UAV. The flight controller and the camera are both connected to the power distribution board.

[0034] Preferably, the motor is a Haoying X9 model motor, the power distribution board is a six-axis power distribution board dedicated to the Yifite E610P, and the power supply is a Grignard ACE 6S16000mA model aircraft model lithium battery.

[0035] Preferably, the flight controller adopts Amovlab's ICF5 model flight controller.

[0036] Preferably, the variable spray execution module includes a controller and a spray sub-module, the controller adopts a Raspberry Pi, and the spray sub-module includes a liquid medicine tank, a multi-way relay, a PWN solenoid valve, a return solenoid valve, a flow meter, a diaphragm water pump, a spray pipe and a nozzle, the diaphragm water pump is arranged at the outlet of the liquid medicine tank, and the outlet of the diaphragm water pump is connected to the nozzle through the spray pipe, the PWN solenoid valve, the return solenoid valve and the flow meter are all arranged in the spray pipe, the Raspberry Pi is communicatively connected with the control module, the Raspberry Pi receives the control instructions issued by the control module, and then controls the multi-relay to send corresponding signals to the PWN solenoid valve and the return solenoid valve based on the control instructions to control the switching degree of the PWN solenoid valve and the return solenoid valve to achieve dynamic adjustment of the spray flow rate.

[0037] A method for autonomous spraying by a drone based on crop canopy perception and prescription decision-making, using the autonomous spraying system by a drone based on crop canopy perception and prescription decision-making, comprises the following steps:

[0038] S1. Setting flight parameters for the UAV, controlling the UAV to fly in an operation area based on the flight parameters, and obtaining three-dimensional point cloud data and longitude and latitude information of the operation area through the laser radar and positioning submodule during the flight. The point cloud map generation module generates a three-dimensional point cloud map of the operation area based on the three-dimensional point cloud data and longitude and latitude information of the operation area;

[0039] S2. extracting the three-dimensional point cloud map of the operation area using the crop canopy semantic information extraction network trained in the crop canopy semantic information module to obtain cotton canopy semantic information;

[0040] S3. Dividing the operating area into a plurality of spraying operation areas of equal area according to the spraying width of the drone; calculating the cotton point cloud canopy density of each spraying operation area based on the three-dimensional point cloud data and cotton canopy semantic information; converting the cotton point cloud canopy density of each spraying operation area to obtain liquid spray flow information, and then constructing a prescription map based on the liquid spray flow information;

[0041] S4. Control the drone and the variable spray execution module to spray each of the spraying operation areas according to the prescription map and the real-time longitude and latitude data, and complete the spraying of the operation area.

[0042] The present invention has the following advantages over the prior art:

[0043] (1) The UAV autonomous spraying system based on crop canopy perception and prescription decision-making of the present invention can effectively extract the cotton canopy semantic information in the point cloud map of the operation area through the crop canopy semantic information module, and generate a prescription map containing the cotton canopy semantic information based on the cotton canopy semantic information through the prescription map generation and analysis module, thereby effectively supporting the UAV and the variable spray execution module to accurately spray the operation area, thereby effectively improving the defoliation effect of cotton and the utilization rate of the liquid medicine, and also reducing environmental pollution and crop damage.

[0044] (2) The drone autonomous spraying method based on crop canopy perception and prescription decision-making of the present invention incorporates cotton canopy semantic information into the spray flow decision of the variable spray actuator carried by the drone. The variable spray actuator carried by the drone realizes accurate variable spraying of the working area through the prescription map containing cotton canopy semantic information, which solves the problem in the prior art that traditional drones do not incorporate characteristics such as the working environment and crop canopy morphology into the drone variable spraying decision. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a schematic diagram of the structure of the UAV autonomous spraying system based on crop canopy perception and prescription decision-making provided in Example 1 of the present invention;

[0046] Figure 2 This is a schematic diagram of generating cotton canopy semantic information provided in Example 1 of the present invention;

[0047] Figure 3 A schematic diagram of generating a prescription map provided in Example 1 of the present invention;

[0048] Figure 4 A schematic flow chart of the autonomous drone spraying method based on crop canopy perception and prescription decision-making provided in Example 2 of the present invention; DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0050] The technical solution of the present invention will be further described below with reference to the accompanying drawings and through specific implementation methods.

[0051] In the description of the present invention, it should be understood that the terms "upper", "lower", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0052] Example 1

[0053] like Figure 1-Figure 3 As shown, a UAV autonomous spraying system based on crop canopy perception and prescription decision-making includes: a UAV, an intelligent perception module, an interaction module, a point cloud map generation module, a crop canopy semantic information module, a prescription map generation and parsing module, a control module, and a variable spray execution module;

[0054] The drone is used to carry the intelligent sensing module and the variable spray execution module;

[0055] Specifically, in this embodiment, the flight altitude of the UAV is set to 10.5 m, and the flight speed is set to 2 m / s. This setting facilitates the intelligent sensing module carried by the UAV to better obtain crop canopy information.

[0056] Specifically, the UAV includes a UAV body, a power mechanism, a flight control mechanism and a sensor mechanism. The power mechanism includes multiple power motors, a power distribution board and a power supply. The multiple power motors are respectively connected to the power supply through the power distribution board. The flight control mechanism includes a flight controller and a PWM control connection board. The flight controller is connected to the multiple motors through the PWM control connection board. The flight controller is communicatively connected to the positioning submodule and the control module. The sensor mechanism includes a camera, which is arranged in front of the UAV body. The camera is used to shoot the front view of the UAV. The flight controller and the camera are both connected to the power distribution board.

[0057] Specifically, in this embodiment, there are 6 motors, and the motors are Haoying X9 model motors. This setting is because the motor integrates the motor, electric controller, and motor base into one, which can effectively reduce the loading burden of the drone. The power distribution board adopts the Yifite E610P special six-axis power distribution board, and the power supply adopts two sets of Grignard ACE 6S16000mA model aircraft lithium batteries.

[0058] Specifically, the flight controller adopts Amovlab's ICF5 model flight controller. This setting is because the ICF5 model flight controller has strong portability and can integrate the positioning submodule into the flight controller, thereby making the overall structure of the system simpler.

[0059] The intelligent perception module includes a laser radar and a positioning submodule. The laser radar is used to collect three-dimensional point cloud data of the operating area. The positioning submodule is used to collect real-time longitude and latitude data of the UAV during flight and longitude and latitude information of the UAV during flight.

[0060] Specifically, the positioning submodule adopts the Beitian BT-468A GNSS module.

[0061] The point cloud map generation module is used to process the three-dimensional point cloud data and longitude and latitude information of the operation area to generate a three-dimensional point cloud map of the operation area;

[0062] Specifically, the working process of the point cloud map generation module is as follows:

[0063] First, the point cloud map generation module uses the SLAM mapping algorithm FAST-LIO2 to process the three-dimensional point cloud data, and then combines the processed three-dimensional point cloud data with the latitude and longitude information to generate a three-dimensional point cloud map of the operation area.

[0064] The crop canopy semantic information module is used to extract crop canopy semantic information from the three-dimensional point cloud map of the operation area to obtain cotton canopy semantic information;

[0065] Specifically, the working steps of the crop canopy semantic information module are as follows:

[0066] S01. Construct a pre-trained RandLA-Net network model in MATLAB software, wherein the pre-trained RandLA-Net network model uses the DALES aerial point cloud dataset as a training set;

[0067] S02. Performing transfer learning using the pre-trained RandLA-Net network model, and modifying the output layer of the pre-trained RandLA-Net network model to ensure that the output categories of the network are modified to the preset five categories, and using the RandLA-Net network model after transfer learning as a crop canopy semantic information extraction network;

[0068] Specifically, the overall concept of the crop canopy semantic information extraction network is that the semantic segmentation effect of the point cloud data of the operation area using the pre-trained RandLA-Net network model is not ideal. In addition, the aerial points of the DALES dataset are classified into eight categories, while we actually need five categories. Therefore, it is necessary to improve the RandLA-Net network model to adapt to the five-category effect that we actually need the model output; specifically, the five categories include vegetation, ground, road, building, and flight path;

[0069] Specifically, the construction process of the crop canopy semantic information extraction network is as follows:

[0070] A pretrained RandLA-Net network model was constructed in MATLAB software. The pretrained RandLA-Net network model used the DALES aerial point cloud dataset as a training set. Since the input layer of the pretrained RandLA-Net network model had already learned rich features in the DALES dataset, transfer learning was used to retain the structure of the input layer. The number of output neurons in the fully connected layer of the pretrained RandLA-Net network model was then changed to 5. The original output layer of the Softmax layer was then replaced to adapt it to the five-category category output. A new classification layer was then used to replace the original classification layer. The new classification layer performed classification according to the category labels and output the classification results. The improved model was used as a crop canopy semantic information extraction network.

[0071] Specifically, to accelerate model training and prevent overfitting, and avoid retraining the encoder (i.e., the low-level feature extraction portion of the network), we can freeze some layers of the network and set the learning rate of these frozen layers to 0. This prevents these layers from being updated during transfer learning. These layers have already learned rich features on the DALES dataset, and only the new classification head needs to be trained. Therefore, during transfer learning, we do not need to retrain the weights of these layers and freeze all the encoder layers.

[0072] S03. Obtain a three-dimensional point cloud map of the operating area, normalize the point cloud data in the three-dimensional point cloud map of the operating area using the normalizePointCloud function, and then divide the normalized point cloud data using the splitData function to obtain a training set and a test set;

[0073] S04, using the training set and data set obtained in step S03 to train the crop canopy semantic information extraction network to obtain a trained crop canopy semantic information extraction network;

[0074] Specifically, during the training process of the crop canopy semantic information extraction network, the trainNetwork function was used to limit the maximum training rounds of the model to 20 times to avoid overfitting, and then the frequency of validation data was set to be evaluated every 30 steps.

[0075] S05. Use the trained crop canopy semantic information extraction network to extract the three-dimensional point cloud map of the operation area to obtain cotton canopy semantic information.

[0076] Specifically, such as Figure 2 As shown, the specific process of step S05 is as follows:

[0077] First, in MATLAB software, the three-dimensional point cloud map of the work area is preprocessed, and the preprocessed three-dimensional point cloud map is input into the trained crop canopy semantic information extraction network. The trained crop canopy semantic information extraction network performs semantic segmentation extraction based on five categories on the input point cloud, and generates a five-category segmentation result map based on the three-dimensional point cloud map. In order to further extract point cloud data containing only cotton vegetation points, the other four category labels (flight route, road, building, ground) in the five categories are specified to remove the other four category point cloud data in the five category segmentation result map, thereby retaining the cotton vegetation point cloud data. Finally, in order to more clearly display the vegetation segmentation effect, the cotton point cloud layer color (originally green) is modified to "parula" color, so as to better display the segmentation effect of the vegetation points. Finally, a point cloud map containing only cotton vegetation is obtained, and the cotton vegetation point cloud map is used as cotton canopy semantic information.

[0078] The prescription map generation and parsing module is used to quantify and convert the cotton canopy semantic information to obtain cotton point cloud canopy density information, then convert the cotton point cloud canopy density information to obtain liquid spray flow information, and then construct a prescription map of the operation area based on the liquid spray flow information;

[0079] Specifically, the working steps of the prescription map generation and analysis module are as follows:

[0080] Step 1: First, the operation area is divided into multiple spraying operation areas of equal area according to the spraying width of the drone;

[0081] Step 2: Calculating the cotton point cloud canopy density of each spraying operation area based on the three-dimensional point cloud data and cotton canopy semantic information;

[0082] Specifically, in this example, when calculating the cotton point cloud canopy density, the weighted point cloud density formula is used to characterize the actual canopy coverage within the plot due to the consideration of the spatial distribution of the point cloud and the influence of the terrain. This setting can improve the accuracy of the point cloud density in characterizing the actual coverage of the crop canopy by considering the height distribution characteristics of the canopy points. The introduction of the height parameter makes higher vegetation points have a greater influence in the density calculation, avoiding the overestimation of low-altitude points or miscellaneous points.

[0083] The specific expression of the cotton point cloud canopy density is as follows:

[0084]

[0085] Among them D xyis the weighted plane point cloud density; N is the number of points projected in the current spraying operation area; A is the area of ​​the spraying operation area; h i is the height of point i; H max and H min The maximum and minimum heights of points within the spray operation area;

[0086] Step 3: Convert the cotton point cloud canopy density of each spraying operation area to obtain liquid spray flow information, and then construct a prescription map based on the liquid spray flow information.

[0087] Based on the cotton point cloud canopy density, through linear interpolation, the flow rate increases with the increase of point cloud density, ensuring that more liquid is obtained in the dense canopy coverage area. The specific representation of the liquid spray flow rate information is as follows:

[0088]

[0089] Among them, Q is the spraying flow rate of the current spraying operation area; Q min is the lower limit of the flow rate of the spray mechanism; Q max The flow rate limit of the spray mechanism; D xy is the point cloud density of the current spraying operation area; D min and D max is the minimum and maximum value of the point cloud density in the working area.

[0090] The control module is used to issue control instructions to the variable spray execution module and the drone according to the prescription map and real-time longitude and latitude data;

[0091] The interaction module is used for communication connection between the UAV, intelligent perception module, point cloud map generation module, crop canopy semantic information module, prescription map generation and analysis module, control module and variable spray execution module;

[0092] The variable spray execution module includes a controller and a spray sub-module. The controller is used to receive control instructions issued by the control module to control the spray flow of the spray sub-module.

[0093] Specifically, the variable spray execution module includes a controller and a spray sub-module. The controller adopts Raspberry Pi. The spray sub-module includes a liquid medicine tank, a multi-way relay, a PWN solenoid valve, a return solenoid valve, a flow meter, a diaphragm water pump, a spray pipe and a nozzle. The diaphragm water pump is arranged at the outlet of the liquid medicine tank, and the outlet of the diaphragm water pump is connected to the nozzle through the spray pipe. The PWN solenoid valve, return solenoid valve and flow meter are all arranged in the spray pipe. The Raspberry Pi is communicatively connected with the control module. The Raspberry Pi receives the control instructions issued by the control module, and then controls the multi-relay to send corresponding signals to the PWN solenoid valve and the return solenoid valve based on the control instructions to control the switching degree of the PWN solenoid valve and the return solenoid valve to achieve dynamic adjustment of the spray flow rate.

[0094] The drone autonomous spraying system based on crop canopy perception and prescription decision-making is also provided with a touch display module, which includes a display screen and a touch screen. The display screen is used to display the real-time camera perspective of the drone and the real-time data received by the control module, and the touch screen is used to adjust the operating instructions of the control module.

[0095] Example 2

[0096] like Figure 4 As shown, a drone autonomous spraying method based on crop canopy perception and prescription decision-making adopts the drone autonomous spraying system based on crop canopy perception and prescription decision-making as described in Example 1, including the following steps:

[0097] S1. Setting flight parameters for the UAV, controlling the UAV to fly in an operation area based on the flight parameters, and obtaining three-dimensional point cloud data and longitude and latitude information of the operation area through the laser radar and positioning submodule during the flight. The point cloud map generation module generates a three-dimensional point cloud map of the operation area based on the three-dimensional point cloud data and longitude and latitude information of the operation area;

[0098] S2. extracting the three-dimensional point cloud map of the operation area using the crop canopy semantic information extraction network trained in the crop canopy semantic information module to obtain cotton canopy semantic information;

[0099] S3. Dividing the operating area into a plurality of spraying operation areas of equal area according to the spraying width of the drone; calculating the cotton point cloud canopy density of each spraying operation area based on the three-dimensional point cloud data and cotton canopy semantic information; converting the cotton point cloud canopy density of each spraying operation area to obtain liquid spray flow information, and then constructing a prescription map based on the liquid spray flow information;

[0100] S4. Control the drone and the variable spray execution module to spray each of the spraying operation areas according to the prescription map and the real-time longitude and latitude data, and complete the spraying of the operation area.

[0101] In summary, the drone autonomous spraying system based on crop canopy perception and prescription decision-making of the present invention can effectively extract cotton canopy semantic information from the point cloud map of the operating area through the crop canopy semantic information module, and generate a prescription map containing cotton canopy semantic information based on the cotton canopy semantic information through the prescription map generation and analysis module, thereby effectively supporting the drone and the variable spray execution module to accurately spray the operating area, thereby effectively improving the defoliation effect of cotton and the utilization rate of the liquid medicine, and also reducing environmental pollution and crop damage. In addition, the present invention also provides a drone autonomous spraying method based on crop canopy perception and prescription decision-making. This method incorporates cotton canopy semantic information into the spray flow decision of the variable spray actuator carried by the drone. The variable spray actuator carried by the drone realizes accurate variable spraying of the operating area through the prescription map containing cotton canopy semantic information, which solves the problem in the prior art that traditional drones do not incorporate characteristics such as the operating environment and crop canopy morphology into the drone variable spraying decision.

[0102] In the description of the present invention, it should be noted that, unless otherwise expressly specified or agreed upon, the terms "disposed," "installed," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0103] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A drone autonomous spraying system based on crop canopy perception and prescription decision-making, characterized by: include: UAV, intelligent perception module, interaction module, point cloud map generation module, crop canopy semantic information module, prescription map generation and parsing module, control module and variable spray execution module; The drone is used to carry the intelligent sensing module and the variable spray execution module; The intelligent perception module includes a laser radar and a positioning submodule. The laser radar is used to collect three-dimensional point cloud data of the operating area. The positioning submodule is used to collect real-time longitude and latitude data of the UAV during flight and longitude and latitude information of the UAV during flight. The point cloud map generation module is used to process the three-dimensional point cloud data and longitude and latitude information of the operation area to generate a three-dimensional point cloud map of the operation area; The crop canopy semantic information module is used to extract crop canopy semantic information from the three-dimensional point cloud map of the operation area to obtain cotton canopy semantic information; The prescription map generation and parsing module is used to quantify and convert the cotton canopy semantic information to obtain cotton point cloud canopy density information, then convert the cotton point cloud canopy density information to obtain liquid spray flow information, and then construct a prescription map of the operation area based on the liquid spray flow information; The control module is used to issue control instructions to the variable spray execution module and the drone according to the prescription map and real-time longitude and latitude data; The interaction module is used for communication connection between the UAV, intelligent perception module, point cloud map generation module, crop canopy semantic information module, prescription map generation and analysis module, control module and variable spray execution module; The variable spray execution module includes a controller and a spray sub-module. The controller is used to receive control instructions issued by the control module to control the spray flow of the spray sub-module.

2. The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to claim 1 is characterized in that: The working steps of the crop canopy semantic information module are as follows: S01. Construct a pre-trained RandLA-Net network model in MATLAB software, wherein the pre-trained RandLA-Net network model uses the DALES aerial point cloud dataset as a training set; S02. Performing transfer learning using the pre-trained RandLA-Net network model, modifying the output layer of the pre-trained RandLA-Net network model to ensure that the network's output categories are modified to a preset five-category classification, and using the RandLA-Net network model after transfer learning as a crop canopy semantic information extraction network; the five categories include vegetation, ground, road, building, and flight path; S03. Obtain a three-dimensional point cloud map of the operating area, normalize the point cloud data in the three-dimensional point cloud map of the operating area using the normalizePointCloud function, and then divide the normalized point cloud data using the splitData function to obtain a training set and a test set; S04, using the training set and data set obtained in step S03 to train the crop canopy semantic information extraction network to obtain a trained crop canopy semantic information extraction network; S05. Use the trained crop canopy semantic information extraction network to extract the three-dimensional point cloud map of the operation area to obtain cotton canopy semantic information.

3. The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to claim 1 is characterized in that: The working steps of the prescription map generation and analysis module are as follows: Step 1: First, the operation area is divided into multiple spraying operation areas of equal area according to the spraying width of the drone; Step 2: Calculate the cotton point cloud canopy density of each spraying operation area based on the three-dimensional point cloud data and cotton canopy semantic information; the specific expression of the cotton point cloud canopy density is as follows: Among them D xy is the weighted plane point cloud density; N is the number of points projected in the current spraying operation area; A is the area of ​​the spraying operation area; h i is the height of point i; H max and H min The maximum and minimum heights of points within the spray operation area; Step 3: Convert the cotton point cloud canopy density of each spraying operation area to obtain liquid spray flow information, and then construct a prescription map based on the liquid spray flow information.

4. The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to claim 3 is characterized in that: The specific representation of the liquid spray flow information is as follows: Among them, Q is the spraying flow rate of the current spraying operation area; Q min is the lower limit of the flow rate of the spray mechanism; Q max The flow rate limit of the spray mechanism; D xy is the point cloud density of the current spraying operation area; D min and D max is the minimum and maximum value of the point cloud density in the working area.

5. The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to claim 1 is characterized in that: The working process of the point cloud map generation module is as follows: First, the point cloud map generation module uses the SLAM mapping algorithm FAST-LIO2 to process the three-dimensional point cloud data, and then combines the processed three-dimensional point cloud data with the latitude and longitude information to generate a three-dimensional point cloud map of the operation area.

6. The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to claim 1 is characterized in that: The UAV includes a UAV body, a power mechanism, a flight control mechanism and a sensor mechanism. The power mechanism includes multiple power motors, a power distribution board and a power supply. The multiple power motors are respectively connected to the power supply through the power distribution board. The flight control mechanism includes a flight controller and a PWM control connection board. The flight controller is connected to the multiple motors through the PWM control connection board. The flight controller is communicatively connected to the positioning submodule and the control module. The sensor mechanism includes a camera, which is arranged in front of the UAV body. The camera is used to shoot the front view of the UAV. The flight controller and the camera are both connected to the power distribution board.

7. The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to claim 6 is characterized in that: The motor adopts the Haoying X9 model motor, the power distribution board adopts the Yifite E610P special six-axis power distribution board, and the power supply adopts the Grignard ACE 6S16000mA model aircraft model lithium battery.

8. The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to claim 6 is characterized in that: The flight controller adopts Amovlab's ICF5 model flight controller.

9. The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to claim 1 is characterized in that: The variable spray execution module includes a controller and a spray sub-module. The controller adopts Raspberry Pi. The spray sub-module includes a liquid medicine tank, a multi-way relay, a PWN solenoid valve, a return solenoid valve, a flow meter, a diaphragm water pump, a spray pipe and a nozzle. The diaphragm water pump is arranged at the outlet of the liquid medicine tank, and the outlet of the diaphragm water pump is connected to the nozzle through the spray pipe. The PWN solenoid valve, return solenoid valve and flow meter are all arranged in the spray pipe. The Raspberry Pi is communicatively connected with the control module. The Raspberry Pi receives the control instructions issued by the control module, and then controls the multi-relay to send corresponding signals to the PWN solenoid valve and the return solenoid valve based on the control instructions to control the switching degree of the PWN solenoid valve and the return solenoid valve to achieve dynamic adjustment of the spray flow rate.

10. A drone autonomous spraying method based on crop canopy perception and prescription decision-making, characterized in that: The UAV autonomous spraying system based on crop canopy perception and prescription decision-making according to any one of claims 1 to 9 comprises the following steps: S1. Setting flight parameters for the UAV, controlling the UAV to fly in an operation area based on the flight parameters, and obtaining three-dimensional point cloud data and longitude and latitude information of the operation area through the laser radar and positioning submodule during the flight. The point cloud map generation module generates a three-dimensional point cloud map of the operation area based on the three-dimensional point cloud data and longitude and latitude information of the operation area; S2. extracting the three-dimensional point cloud map of the operation area using the crop canopy semantic information extraction network trained in the crop canopy semantic information module to obtain cotton canopy semantic information; S3. Dividing the operating area into a plurality of spraying operation areas of equal area according to the spraying width of the drone; calculating the cotton point cloud canopy density of each spraying operation area based on the three-dimensional point cloud data and cotton canopy semantic information; converting the cotton point cloud canopy density of each spraying operation area to obtain liquid spray flow information, and then constructing a prescription map based on the liquid spray flow information; S4. Control the drone and the variable spray execution module to spray each of the spraying operation areas according to the prescription map and the real-time longitude and latitude data, and complete the spraying of the operation area.

Images