A method and system for controlling fog droplet drift for plant protection UAV

Abstract

The present invention discloses a method and system for controlling droplet drift for plant protection UAVs, including: extracting plant protection UAV machine parameters and operation behaviors for preprocessing, determining operation parameters and climate parameters to combine and screen influencing factors, and establishing the relationship between influencing factors and droplet drift amount and drift distance; constructing a knowledge graph of influencing factors, droplet drift conditions and compensation methods, and learning to construct a compensation decision model; sensing the real-time dynamic operation behavior and environmental dynamic changes during the operation of the plant protection UAV, predicting the droplet drift conditions, and calling the compensation decision model if the preset threshold information is not met, and optimizing the droplet drift. The present invention analyzes the key factor parameters that cause droplet drift during the operation of the plant protection UAV, uses the knowledge graph to make decisions, and performs adaptive control under different operating conditions, effectively reducing the influence of natural wind and high wind on droplet descent, preventing droplet drift, and improving the accuracy of the spraying operation of the plant protection UAV.

Description

Technical Field

[0001] The present invention relates to the technical field of precision pesticide application, and more specifically, to a method and system for preventing fog droplet drift for a plant protection UAV. Background Art

[0002] Food security is the foundation for national security. In crop production, pest and disease control is a key measure to ensure good crop growth and increase grain production and income. However, there are unscientific problems in the use of pesticides in traditional agricultural production. It is necessary to accelerate the mechanization level of pesticide application equipment, improve the utilization rate of pesticides, reduce pesticide residue pollution on the basis of ensuring food security, and ensure the sustainable development of agriculture. By loading spraying equipment onto drones, quantitative and precise spraying of crops can be achieved through control systems, sensors and other equipment. The drone remote sensing technology that integrates farmland information acquisition platforms and multi-band image processing technology can effectively detect the growth of crops, accurately apply variable fertilizers, and optimize operating parameters, replacing conventional satellite remote sensing monitoring and reducing agricultural production costs. Compared with conventional agricultural machinery operations, it has the advantages of high operating efficiency and wide adaptability.

[0003] As the scope of plant protection drone spraying operations continues to expand, the drawbacks and shortcomings of plant protection drone spraying are gradually exposed, among which the problem of liquid drift caused by external wind fields and rotor wind fields is particularly obvious. When the drone flies at a high altitude, the external wind field factors have a greater impact on droplet deposition, so that the droplets cannot be effectively deposited in the target area; when the drone flies at too low an altitude, the rotor wind field factors have a greater impact on droplet deposition, and the turbulent wind field causes the crops to sway violently and cannot be effectively treated with drugs. When the flight altitude is too low, it is not safe for the plant protection drone itself, and it is easy to hit crops and crash. Therefore, how to adaptively control the anti-droplet drift according to the dynamic parameters of the environment during the spraying process of the plant protection drone is an urgent problem that needs to be solved. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method and system for preventing fog droplet drift for a plant protection UAV.

[0005] The first aspect of the present invention provides a method for controlling droplet drift for a plant protection UAV, comprising:

[0006] Extract the machine parameters and operation behaviors according to the historical operation data of the plant protection UAV for preprocessing, determine the operation parameters, combine the operation parameters with the climate parameters to screen the influencing factors, and establish the relationship between the influencing factors and the droplet drift amount and drift distance;

[0007] Generate the droplet drift status according to the droplet drift amount and drift distance, and associate it with the corresponding influencing factors, construct a knowledge graph of the influencing factors, droplet drift status and compensation methods, and learn the knowledge graph to construct a compensation decision model;

[0008] Perceive the real-time dynamic operation behavior and environmental dynamic changes of plant protection drones during operation, predict the droplet drift during spraying operations, and set threshold information based on the location characteristics of the operation area and plant characteristics;

[0009] If the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection UAV, the compensated droplet drift condition is obtained, and the knowledge graph is updated.

[0010] In this scheme, the machine parameters and operation behaviors are extracted based on the historical operation data of the plant protection UAV for preprocessing, the operation parameters are determined, and the operation parameters are combined with the climate parameters to screen the influencing factors, specifically:

[0011] Obtain historical operation data of the plant protection UAV at a preset time step, extract multi-dimensional machine parameters and operation behaviors from the historical operation data, cluster the operation behaviors of the plant protection UAV, and generate operation behavior clusters of different categories;

[0012] According to the dynamic operation behaviors of the plant protection UAV in the historical operation behaviors, different operation behavior clusters are given time series associations and logical associations, and the change rates of different machine parameters with time series when the operation behaviors change are obtained. According to the change rates, machine parameters that meet the preset standards are selected, and the operation parameters are generated after preprocessing;

[0013] The operation parameters and the climate parameters are combined to generate an initial data set, feature selection is performed in the initial data set, the mRMR algorithm is expanded using the Pearson correlation coefficient, and the correlation between each data sample in the initial data set and the correlation between the data sample and the droplet drift are calculated;

[0014] An importance ranking of data samples is generated with the goal of minimizing redundancy between data samples and maximizing correlation between data samples and droplet drift, and a preset number of influencing factors are selected based on the importance ranking.

[0015] In this scheme, the relationship between the influencing factors and the droplet drift amount and drift distance is established, specifically:

[0016] According to the influencing factors, characteristic parameters are extracted from historical operation data and climate parameters as input data, and relationship models between the influencing factors and the droplet drift amount and drift distance are respectively constructed based on the extreme learning machine;

[0017] Optimizing the key parameter combination of the extreme learning machine by using a genetic algorithm, mapping the predicted mean square error of the relationship model to a fitness function, and obtaining the optimal input layer weight and hidden layer bias of the extreme learning machine by continuously iterating individual selection, crossover, and mutation operations;

[0018] The theoretical values ​​of droplet drift amount and drift distance under different working parameters are obtained by simulating the flight state of the plant protection UAV, the droplet drift amount and drift distance are predicted using the relationship model, the prediction is verified using the theoretical values, and the relationship model is optimized based on the verification results.

[0019] In this solution, the influencing factors, droplet drift conditions and compensation methods are constructed into a knowledge graph, and the knowledge graph is learned to construct a compensation decision model, specifically:

[0020] Obtain compensation method examples for droplet drift of plant protection UAVs through a big data retrieval method, perform structured processing on the compensation method examples, evaluate the compensation effect according to the droplet drift conditions before and after compensation in the compensation method examples, and use the compensation efficiency to screen compensation method examples that meet the requirements;

[0021] The screened compensation method examples are classified according to the compensation method category, and the response relationship between the droplet drift condition and the compensation amount is fitted using the least square method in the compensation method subsets of different categories, and the optimal compensation method for the droplet drift condition is obtained according to different response relationships;

[0022] According to the influencing factors, droplet drift conditions and compensation methods in the spraying operation of plant protection UAVs, triples are represented, and a knowledge graph of droplet drift is constructed. The knowledge graph is learned and represented using a graph convolutional network to build a compensation decision model.

[0023] In this solution, the real-time dynamic operation behavior and environmental dynamic changes of the plant protection UAV during operation are perceived, the droplet drift during the spraying operation is predicted, and the threshold information is set according to the location characteristics and plant characteristics of the operation area. Specifically:

[0024] Perceive the real-time dynamic behavior and environmental dynamic changes of the plant protection UAV during the current spraying operation, extract real-time perception data using the influencing factors as indicators, call the relationship model between the influencing factors and the droplet drift amount and drift distance, and use the real-time perception data as input information;

[0025] Obtain the predicted values ​​of the droplet drift amount and drift distance of the plant protection UAV spraying work, use the predicted values ​​of the droplet drift amount and drift distance to characterize the droplet drift status of the corresponding timestamp, and set direction label information for the droplet drift status according to the ambient wind direction;

[0026] The target spraying area is gridded to obtain the operating position of the plant protection UAV, and the possible drift range is determined based on the grid of the target spraying area according to the drift status of the droplets with direction labels, and the consistency of the plant characteristics is determined within the possible drift range;

[0027] When the consistency of plant characteristics is less than the preset standard, the grid block corresponding to the mutation point of the plant characteristics is obtained, and the threshold information is set according to the shortest distance from the current grid block as the position to the grid block corresponding to the mutation point of the plant characteristics;

[0028] When the consistency of plant characteristics is not less than the preset standard, the threshold information is set according to the farthest grid block in the possible drift range.

[0029] In this scheme, if the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection drone, specifically:

[0030] When the predicted droplet drift condition during the spraying operation does not meet the preset threshold information, the compensation decision model is called to encode the entity nodes in the knowledge graph for learning representation, and the real-time perception data of the plant protection drone and the predicted droplet drift condition are located in the knowledge graph by using similarity comparison;

[0031] Obtain the target influencing factor node and the target droplet drift status node, introduce the self-attention mechanism to set the self-attention weights of the target influencing factor node and the target droplet drift status node and their neighbor nodes, and generate the initial weights by weighted aggregation based on the neighbor aggregation of graph convolution;

[0032] In the compensation decision model, the target influencing factor node and the target droplet drift condition node are used as sampling starting points for random walk to obtain the meta-path including the target influencing factor node and the target droplet drift condition node;

[0033] According to the response relationship between the droplet drift condition and the compensation amount, the weight information of different compensation method nodes is set for the compensation method node associated with the target droplet drift condition node, the initial weights and weight information of the nodes included in the meta-path are summed, and the meta-path is sorted;

[0034] The meta-path corresponding to the highest weight is obtained to extract the compensation method node corresponding to the real-time perception data as the decision output, and the response relationship between the droplet drift condition and the compensation amount in different categories of compensation method subsets is obtained to extract and output the compensation amount of the compensation method.

[0035] The second aspect of the present invention further provides a fog droplet drift prevention control system for a plant protection UAV, the system comprising: a memory, a processor, the memory comprising a fog droplet drift prevention control method program for a plant protection UAV, the fog droplet drift prevention control method program for a plant protection UAV being executed by the processor to implement the following steps:

[0036] Extract the machine parameters and operation behaviors according to the historical operation data of the plant protection UAV for preprocessing, determine the operation parameters, combine the operation parameters with the climate parameters to screen the influencing factors, and establish the relationship between the influencing factors and the droplet drift amount and drift distance;

[0037] Generate the droplet drift status according to the droplet drift amount and drift distance, and associate it with the corresponding influencing factors, construct a knowledge graph of the influencing factors, droplet drift status and compensation methods, and learn the knowledge graph to construct a compensation decision model;

[0038] Perceive the real-time dynamic operation behavior and environmental dynamic changes of plant protection drones during operation, predict the droplet drift during spraying operations, and set threshold information based on the location characteristics of the operation area and plant characteristics;

[0039] If the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection UAV, the compensated droplet drift condition is obtained, and the knowledge graph is updated.

[0040] The present invention discloses a method and system for controlling droplet drift for plant protection UAVs, including: extracting plant protection UAV machine parameters and operation behaviors for preprocessing, determining operation parameters and climate parameters to combine and screen influencing factors, and establishing the relationship between influencing factors and droplet drift amount and drift distance; constructing a knowledge graph of influencing factors, droplet drift conditions and compensation methods, and learning to construct a compensation decision model; sensing the real-time dynamic operation behavior and environmental dynamic changes during the operation of the plant protection UAV, predicting the droplet drift conditions, and calling the compensation decision model if the preset threshold information is not met, and optimizing the droplet drift. The present invention analyzes the key factor parameters that cause droplet drift during the operation of the plant protection UAV, uses the knowledge graph to make decisions, and performs adaptive control under different operating conditions, effectively reducing the influence of natural wind and high wind on droplet descent, preventing droplet drift, and improving the accuracy of the spraying operation of the plant protection UAV. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of a method for controlling fog droplet drift for a plant protection UAV according to the present invention is shown;

[0042] Figure 2 A flow chart showing the present invention for setting threshold information according to location characteristics and plant characteristics;

[0043] Figure 3 A flow chart showing the compensation optimization of the droplet drift of the plant protection UAV according to the present invention is shown;

[0044] Figure 4 A block diagram of a fog droplet drift prevention control system for a plant protection UAV of the present invention is shown. DETAILED DESCRIPTION

[0045] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0047] Figure 1 A flow chart of a method for preventing fog droplet drift for a plant protection UAV of the present invention is shown.

[0048] like Figure 1 As shown, the first aspect of the present invention provides a method for controlling fog droplet drift for a plant protection UAV, comprising:

[0049] S102, extracting machine parameters and operation behaviors according to historical operation data of the plant protection UAV for preprocessing, determining operation parameters, combining the operation parameters with climate parameters to screen influencing factors, and establishing the relationship between the influencing factors and the droplet drift amount and drift distance;

[0050] S104, generating a droplet drift status according to the droplet drift amount and drift distance, and associating it with the corresponding influencing factors, constructing a knowledge graph of the influencing factors, the droplet drift status and the compensation method, and learning the knowledge graph to construct a compensation decision model;

[0051] S106, sensing the real-time dynamic operation behavior and environmental dynamic changes of the plant protection UAV during operation, predicting the droplet drift during the spraying operation, and setting threshold information according to the location characteristics of the operation area and the characteristics of the plants;

[0052] S108, if the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection UAV, the compensated droplet drift condition is obtained, and the knowledge graph is updated.

[0053] It should be noted that the historical operation data of the plant protection UAV at a preset time step is obtained, and multidimensional machine parameters, including flight speed, flight altitude, rotor speed, spray angle, spray pressure, flow rate, etc. and operation behaviors, including UAV hovering, UAV reversing, UAV roll, etc., are extracted from the historical operation data, and the operation behaviors of the plant protection UAV are clustered to generate operation behavior clusters of different categories; according to the dynamic operation behaviors of the plant protection UAV in the historical operation behaviors, different operation behavior clusters are given time series associations and logical associations, and the change rates of different machine parameters with time series when the operation behaviors change are obtained, and the machine parameters that meet the preset standards are selected according to the change rates, and the operation parameters are generated after preprocessing; the operation parameters and climate parameters are combined to generate an initial data set, feature selection is performed in the initial data set, and the mRMR algorithm is expanded using the Pearson correlation coefficient to calculate the correlation between each data sample in the initial data set and the correlation between the data sample and the droplet drift; the expression of the expanded mRMR algorithm is: R(x i ; f) represents the data sample x i The Pearson correlation coefficient between the droplet drift f, R(x i ;x j ) represents the data sample x i With data sample x j The Pearson correlation coefficient between them, S represents the selected data samples, |S| represents the number of selected data samples, Ω s Represents a subset of data samples to be selected in the initial data set. Generate an importance ranking of data samples with the goal of minimizing redundancy between data samples and maximizing correlation between data samples and droplet drift, and select a preset number of influencing factors according to the importance ranking.

[0054] It should be noted that, according to the influencing factors, characteristic parameters are extracted from historical operation data and climate parameters as input data, and relationship models between the influencing factors and the droplet drift amount and drift distance are respectively constructed based on the extreme learning machine; since the traditional gradient descent method has low efficiency in training the weights and bias of the extreme learning machine and is prone to overfitting or underfitting, a genetic algorithm is used to optimize the key parameter combination of the extreme learning machine, which simplifies the calculation process and has better learning efficiency and generalization ability; initialize the parameters of the genetic algorithm, encode the key parameter combination, determine the maximum number of iterations, crossover probability, mutation probability, etc., map the predicted mean square error of the relationship model to the fitness function, and continuously iterate the selection, crossover and mutation operations of individuals to output the optimal individual to obtain the optimal input layer weight and hidden layer bias of the extreme learning machine; for a given training data set, determine the number of hidden layer nodes and activation function of the extreme learning machine, randomly select the input weights and bias of hidden neurons, calculate the hidden layer output matrix and calculate the output weight vector of the extreme learning machine.

[0055] The flight state simulation data of the plant protection UAV is obtained by using the big data retrieval method, the corresponding droplet motion model in the spraying of the plant protection UAV is read, and the droplet deposition and drift are numerically simulated and simulated by the CFD method. In addition, the droplet particles sprayed from the nozzle under different influencing factors are dynamically captured by a high-speed camera, and the effective feature area of ​​the droplet image is extracted, grayscale processed, and mathematical morphological operations are performed to obtain a binary image. After grid division, cross-correlation calculation is performed. According to the peak value obtained by the cross-correlation calculation, the droplet particles in the grid at the same position of the adjacent frame are selected through the obtained cross-correlation peak value to extract the particle coordinates, calculate the droplet motion speed, and obtain the droplet distribution state. The theoretical values ​​of the droplet drift amount and drift distance and the droplet distribution state under different working parameters are obtained through the flight state simulation of the plant protection UAV, and the droplet drift amount and drift distance are predicted using the relationship model, and the prediction verification is performed through the theoretical value and droplet distribution state, and the relationship model is optimized according to the verification result.

[0056] Figure 2 A flow chart of setting threshold information according to location characteristics and plant characteristics of the present invention is shown.

[0057] According to an embodiment of the present invention, the real-time dynamic operation behavior and dynamic changes of the environment during the operation of the plant protection UAV are perceived, the droplet drift during the spraying operation is predicted, and the threshold information is set according to the location characteristics and plant characteristics of the operation area, specifically:

[0058] S202, sensing the real-time dynamic behavior of the plant protection UAV and the dynamic changes of the environment during the current spraying operation, extracting real-time perception data using the influencing factors as indicators, calling a relationship model between the influencing factors and the droplet drift amount and drift distance, and using the real-time perception data as input information;

[0059] S204, obtaining predicted values ​​of the droplet drift amount and drift distance of the plant protection UAV spraying work, using the predicted values ​​of the droplet drift amount and drift distance to characterize the droplet drift status of the corresponding timestamp, and setting direction label information for the droplet drift status according to the ambient wind direction;

[0060] S206, gridding the target spraying area, obtaining the operation position of the plant protection UAV, determining a possible drift range based on the target spraying area grid according to the drift status of the droplets with direction labels, and judging the consistency of plant characteristics in the possible drift range;

[0061] S208, when the consistency of plant characteristics is less than the preset standard, the grid block corresponding to the mutation point of the plant characteristics is obtained, and the threshold information is set according to the shortest distance from the current grid block as the position to the grid block corresponding to the mutation point of the plant characteristics;

[0062] S210, when the consistency of plant characteristics is not less than a preset standard, threshold information is set according to the farthest grid block in the possible drift range.

[0063] It should be noted that the target spraying area of ​​the plant protection drone is used as the operation area. When dividing the grid blocks in the operation area, try to ensure that the plant characteristics in the area blocks are similar. In the operation area, the grid blocks are marked according to the predicted value of the droplet drift condition according to the operation position. The marked grid blocks are used to determine the possible drift range. The plant characteristics such as plant category and plant growth stage are determined in the possible drift range. The consistency of the plant characteristics in the grid blocks is obtained according to the similarity calculation. When the consistency of plant characteristics is less than the preset standard, since crops are usually planted in blocks, there are plant characteristic mutation edges between the planting blocks. The shortest distance is selected in the mutation edges to set the threshold information. When the consistency of plant characteristics is not less than the preset standard, the threshold information is set according to the farthest grid block in the possible drift range. If the drift distance is greater than the possible drift range, the operation position is re-determined and the possible drift range is re-determined.

[0064] In addition, when the consistency of plant characteristics is not less than the preset standard, the onboard visual monitoring equipment of the plant protection UAV is used in combination with the predicted droplet drift state to judge the droplet drift grid block at the current operating position, and the selected droplet drift grid block is suspended, and the weight coefficient of the suspended grid block is set according to different drift distances, and the droplet drift amounts of different suspended grid blocks are weighted by the weight coefficient; the weighted suspended grid block is visualized, and the weighted droplet drift amounts of the suspended grid blocks at different times are summed in a single grid block, and a thermal mark is generated according to the sum value. When the plant protection UAV operates to the grid block containing the thermal mark, the spraying amount of the plant protection UAV is adaptively reduced according to the thermal mark. The adaptive decision can be obtained by constructing a mapping relationship between thermal marks of different levels and reduced application amounts through a deep learning method.

[0065] Figure 3 The flowchart of the present invention for compensating and optimizing the droplet drift of the plant protection UAV is shown.

[0066] According to an embodiment of the present invention, if the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection UAV, specifically:

[0067] S302, when the predicted droplet drift condition during the spraying operation does not meet the preset threshold information, calling the compensation decision model, encoding and learning the entity nodes in the knowledge graph, and locating the real-time perception data of the plant protection drone and the predicted droplet drift condition in the knowledge graph by using similarity comparison;

[0068] S304, obtaining a target influencing factor node and a target droplet drift status node, introducing a self-attention mechanism to set the self-attention weights of the target influencing factor node and the target droplet drift status node and their neighbor nodes, and performing weighted aggregation according to the neighbor aggregation of graph convolution to generate initial weights;

[0069] S306, in the compensation decision model, taking the target influencing factor node and the target droplet drift condition node as sampling starting points to perform random walk, and obtain a meta-path including the target influencing factor node and the target droplet drift condition node;

[0070] S308, according to the response relationship between the droplet drift condition and the compensation amount, weight information of different compensation method nodes is set for the compensation method node associated with the target droplet drift condition node, initial weights and weight information of the nodes included in the meta-path are summed, and the meta-path is sorted;

[0071] S310, obtaining the meta-path corresponding to the highest weight, extracting the compensation method node corresponding to the real-time perception data as the decision output, and obtaining the response relationship between the droplet drift condition and the compensation amount in different categories of compensation method subsets to extract and output the compensation amount of the compensation method.

[0072] It should be noted that examples of compensation methods for droplet drift of plant protection UAVs, including nozzle forward compensation, flight inclination compensation, etc., are obtained through a big data retrieval method. The compensation method examples are structured, and the compensation effect is evaluated according to the droplet drift conditions before and after compensation in the compensation method examples. The compensation method examples that meet the requirements are screened using the compensation efficiency; the screened compensation method examples are classified according to the compensation method category, and the least squares method is used to fit the response relationship f between the droplet drift condition and the compensation amount in the compensation method subsets of different categories. The fitting of the least squares method is expressed as x is the droplet drift condition, y is i is the compensation amount, λ0 and λ1 are linear parameters, and n is the total number of compensation methods. The optimal compensation method for droplet drift is obtained according to different response relationships; the influencing factors, droplet drift conditions and compensation methods in the spraying operation of the plant protection UAV are represented by triples, and a knowledge graph of droplet drift is constructed. The knowledge graph is learned and represented by a graph convolutional network to construct a compensation decision model. According to the relationship between the influencing factor node, the droplet drift condition node and the compensation method node, the potential connection between the influencing factor and the compensation method, as well as the droplet drift condition and the compensation method is established through the feature propagation mechanism and the neighbor aggregation mechanism. The airborne visual monitoring equipment of the plant protection UAV is used to obtain the droplet drift condition after compensation, and the knowledge graph is updated according to the difference in the droplet drift condition.

[0073] Figure 4 A block diagram of a fog droplet drift prevention control system for a plant protection UAV of the present invention is shown.

[0074] The second aspect of the present invention further provides a fog droplet drift prevention control system 4 for a plant protection UAV, the system comprising: a memory 41, a processor 42, the memory comprising a fog droplet drift prevention control method program for a plant protection UAV, the fog droplet drift prevention control method program for a plant protection UAV being executed by the processor to implement the following steps:

[0075] Extract the machine parameters and operation behaviors according to the historical operation data of the plant protection UAV for preprocessing, determine the operation parameters, combine the operation parameters with the climate parameters to screen the influencing factors, and establish the relationship between the influencing factors and the droplet drift amount and drift distance;

[0076] Generate the droplet drift status according to the droplet drift amount and drift distance, and associate it with the corresponding influencing factors, construct a knowledge graph of the influencing factors, droplet drift status and compensation methods, and learn the knowledge graph to construct a compensation decision model;

[0077] Perceive the real-time dynamic operation behavior and environmental dynamic changes of plant protection drones during operation, predict the droplet drift during spraying operations, and set threshold information based on the location characteristics of the operation area and plant characteristics;

[0078] If the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection UAV, the compensated droplet drift condition is obtained, and the knowledge graph is updated.

[0079] The third aspect of the present invention also provides a computer-readable storage medium, which includes a program for a method for preventing droplet drift from a plant protection drone. When the program for the method for preventing droplet drift from a plant protection drone is executed by a processor, the steps of the method for preventing droplet drift from a plant protection drone as described in any one of the above items are implemented.

[0080] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0081] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0082] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0083] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0084] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0085] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for controlling fog droplet drift for a plant protection UAV, characterized in that: The following steps are involved: Extract the machine parameters and operation behaviors according to the historical operation data of the plant protection UAV for preprocessing, determine the operation parameters, combine the operation parameters with the climate parameters to screen the influencing factors, and establish the relationship between the influencing factors and the droplet drift amount and drift distance; Generate the droplet drift status according to the droplet drift amount and drift distance, and associate it with the corresponding influencing factors, construct a knowledge graph of the influencing factors, droplet drift status and compensation methods, and learn the knowledge graph to construct a compensation decision model; Perceive the real-time dynamic operation behavior and environmental dynamic changes of plant protection drones during operation, predict the droplet drift during spraying operations, and set threshold information based on the location characteristics of the operation area and plant characteristics; If the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection UAV, the compensated droplet drift condition is obtained, and the knowledge graph is updated; The influencing factors, droplet drift conditions and compensation methods are constructed into a knowledge graph, and the knowledge graph is studied to construct a compensation decision model, specifically: Obtain compensation method examples for droplet drift of plant protection UAVs through a big data retrieval method, perform structured processing on the compensation method examples, evaluate compensation effects based on the droplet drift conditions before and after compensation in the compensation method examples, and use the compensation effects to screen compensation method examples that meet the requirements; The screened compensation method examples are classified according to the compensation method category, and the response relationship between the droplet drift condition and the compensation amount is fitted using the least square method in the compensation method subsets of different categories, and the optimal compensation method for the droplet drift condition is obtained according to different response relationships; According to the influencing factors, droplet drift conditions and compensation methods in the spraying operation of plant protection UAVs, triples are represented, and a knowledge graph of droplet drift is constructed. The knowledge graph is learned and represented using a graph convolutional network to build a compensation decision model.

2. The method for controlling fog droplet drift for a plant protection UAV according to claim 1, characterized in that: According to the historical operation data of the plant protection UAV, the machine parameters and operation behaviors are extracted for preprocessing, the operation parameters are determined, and the operation parameters are combined with the climate parameters to screen the influencing factors, specifically: Obtain historical operation data of the plant protection UAV at a preset time step, extract multi-dimensional machine parameters and operation behaviors from the historical operation data, cluster the operation behaviors of the plant protection UAV, and generate operation behavior clusters of different categories; According to the dynamic operation behaviors of the plant protection UAV in the historical operation behaviors, different operation behavior clusters are given time series associations and logical associations, and the change rates of different machine parameters with time series when the operation behaviors change are obtained. According to the change rates, machine parameters that meet the preset standards are selected, and the operation parameters are generated after preprocessing; The operation parameters and the climate parameters are combined to generate an initial data set, feature selection is performed in the initial data set, the mRMR algorithm is expanded using the Pearson correlation coefficient, and the correlation between each data sample in the initial data set and the correlation between the data sample and the droplet drift are calculated; An importance ranking of data samples is generated with the goal of minimizing redundancy between data samples and maximizing correlation between data samples and droplet drift, and a preset number of influencing factors are selected based on the importance ranking.

3. The method for controlling fog droplet drift for a plant protection UAV according to claim 1, characterized in that: The relationship between the influencing factors and the droplet drift amount and drift distance is established, specifically: According to the influencing factors, characteristic parameters are extracted from historical operation data and climate parameters as input data, and relationship models between the influencing factors and the droplet drift amount and drift distance are respectively constructed based on the extreme learning machine; Optimizing the key parameter combination of the extreme learning machine by using a genetic algorithm, mapping the predicted mean square error of the relationship model to a fitness function, and obtaining the optimal input layer weight and hidden layer bias of the extreme learning machine by continuously iterating individual selection, crossover, and mutation operations; The theoretical values ​​of droplet drift amount and drift distance under different working parameters are obtained by simulating the flight state of the plant protection UAV, the droplet drift amount and drift distance are predicted using the relationship model, the prediction is verified using the theoretical values, and the relationship model is optimized based on the verification results.

4. The method for controlling fog droplet drift for a plant protection UAV according to claim 1, characterized in that: Perceive the real-time dynamic operation behavior and environmental dynamic changes of the plant protection UAV during operation, predict the droplet drift during the spraying operation, and set the threshold information according to the location characteristics and plant characteristics of the operation area, specifically: Perceive the real-time dynamic behavior and environmental dynamic changes of the plant protection UAV during the current spraying operation, extract real-time perception data using the influencing factors as indicators, call the relationship model between the influencing factors and the droplet drift amount and drift distance, and use the real-time perception data as input information; Obtain the predicted values ​​of the droplet drift amount and drift distance of the plant protection UAV spraying work, use the predicted values ​​of the droplet drift amount and drift distance to characterize the droplet drift status of the corresponding timestamp, and set direction label information for the droplet drift status according to the ambient wind direction; The target spraying area is gridded to obtain the operating position of the plant protection UAV, and the possible drift range is determined based on the grid of the target spraying area according to the drift status of the droplets with direction labels, and the consistency of the plant characteristics is determined within the possible drift range; When the consistency of plant characteristics is less than the preset standard, the grid block corresponding to the mutation point of the plant characteristics is obtained, and the threshold information is set according to the shortest distance from the current grid block as the position to the grid block corresponding to the mutation point of the plant characteristics; When the consistency of plant characteristics is not less than the preset standard, the threshold information is set according to the farthest grid block in the possible drift range.

5. The method for controlling fog droplet drift for a plant protection UAV according to claim 1, characterized in that: If the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection UAV, specifically: When the predicted droplet drift condition during the spraying operation does not meet the preset threshold information, the compensation decision model is called to encode the entity nodes in the knowledge graph for learning representation, and the real-time perception data of the plant protection drone and the predicted droplet drift condition are located in the knowledge graph by using similarity comparison; Obtain the target influencing factor node and the target droplet drift status node, introduce the self-attention mechanism to set the self-attention weights of the target influencing factor node and the target droplet drift status node and their neighbor nodes, and generate the initial weights by weighted aggregation based on the neighbor aggregation of graph convolution; In the compensation decision model, the target influencing factor node and the target droplet drift condition node are used as sampling starting points for random walk to obtain the meta-path including the target influencing factor node and the target droplet drift condition node; According to the response relationship between the droplet drift condition and the compensation amount, the weight information of different compensation method nodes is set for the compensation method node associated with the target droplet drift condition node, the initial weights and weight information of the nodes included in the meta-path are summed, and the meta-path is sorted; The meta-path corresponding to the highest weight is obtained to extract the compensation method node corresponding to the real-time perception data as the decision output, and the response relationship between the droplet drift condition and the compensation amount in different categories of compensation method subsets is obtained to extract and output the compensation amount of the compensation method.

6. A fog droplet drift prevention control system for a plant protection UAV, characterized in that: The system includes: a memory and a processor, wherein the memory includes a fog droplet drift control method program for a plant protection UAV, and when the fog droplet drift control method program for a plant protection UAV is executed by the processor, the following steps are implemented: Extract the machine parameters and operation behaviors according to the historical operation data of the plant protection UAV for preprocessing, determine the operation parameters, combine the operation parameters with the climate parameters to screen the influencing factors, and establish the relationship between the influencing factors and the droplet drift amount and drift distance; Generate the droplet drift status according to the droplet drift amount and drift distance, and associate it with the corresponding influencing factors, construct a knowledge graph of the influencing factors, droplet drift status and compensation methods, and learn the knowledge graph to construct a compensation decision model; Perceive the real-time dynamic operation behavior and environmental dynamic changes of plant protection drones during operation, predict the droplet drift during spraying operations, and set threshold information based on the location characteristics of the operation area and plant characteristics; If the droplet drift condition does not meet the preset threshold information, the compensation decision model is called to compensate and optimize the droplet drift of the plant protection UAV, the compensated droplet drift condition is obtained, and the knowledge graph is updated; The influencing factors, droplet drift conditions and compensation methods are constructed into a knowledge graph, and the knowledge graph is studied to construct a compensation decision model, specifically: Obtain compensation method examples for droplet drift of plant protection UAVs through a big data retrieval method, perform structured processing on the compensation method examples, evaluate compensation effects based on the droplet drift conditions before and after compensation in the compensation method examples, and use the compensation effects to screen compensation method examples that meet the requirements; The screened compensation method examples are classified according to the compensation method category, and the response relationship between the droplet drift condition and the compensation amount is fitted using the least square method in the compensation method subsets of different categories, and the optimal compensation method for the droplet drift condition is obtained according to different response relationships; According to the influencing factors, droplet drift conditions and compensation methods in the spraying operation of plant protection UAVs, triples are represented, and a knowledge graph of droplet drift is constructed. The knowledge graph is learned and represented using a graph convolutional network to build a compensation decision model.

7. The anti-fog drift control system for a plant protection UAV according to claim 6, characterized in that: The relationship between the influencing factors and the droplet drift amount and drift distance is established, specifically: According to the influencing factors, characteristic parameters are extracted from historical operation data and climate parameters as input data, and relationship models between the influencing factors and the droplet drift amount and drift distance are respectively constructed based on the extreme learning machine; Optimizing the key parameter combination of the extreme learning machine by using a genetic algorithm, mapping the predicted mean square error of the relationship model to a fitness function, and obtaining the optimal input layer weight and hidden layer bias of the extreme learning machine by continuously iterating individual selection, crossover, and mutation operations; The theoretical values ​​of droplet drift amount and drift distance under different working parameters are obtained by simulating the flight state of the plant protection UAV, the droplet drift amount and drift distance are predicted using the relationship model, the prediction is verified using the theoretical values, and the relationship model is optimized based on the verification results.

8. The anti-fog drift control system for a plant protection UAV according to claim 6, characterized in that: Perceive the real-time dynamic operation behavior and environmental dynamic changes of the plant protection UAV during operation, predict the droplet drift during the spraying operation, and set the threshold information according to the location characteristics and plant characteristics of the operation area, specifically: Perceive the real-time dynamic behavior and environmental dynamic changes of the plant protection UAV during the current spraying operation, extract real-time perception data using the influencing factors as indicators, call the relationship model between the influencing factors and the droplet drift amount and drift distance, and use the real-time perception data as input information; Obtain the predicted values ​​of the droplet drift amount and drift distance of the plant protection UAV spraying work, use the predicted values ​​of the droplet drift amount and drift distance to characterize the droplet drift status of the corresponding timestamp, and set direction label information for the droplet drift status according to the ambient wind direction; The target spraying area is gridded to obtain the operating position of the plant protection UAV, and the possible drift range is determined based on the grid of the target spraying area according to the drift status of the droplets with direction labels, and the consistency of the plant characteristics is determined within the possible drift range; When the consistency of plant characteristics is less than the preset standard, the grid block corresponding to the mutation point of the plant characteristics is obtained, and the threshold information is set according to the shortest distance from the current grid block as the position to the grid block corresponding to the mutation point of the plant characteristics; When the consistency of plant characteristics is not less than the preset standard, the threshold information is set according to the farthest grid block in the possible drift range.

Images