ANN-GWO-based plant protection unmanned aerial vehicle spray drift amount prediction method and related equipment

Abstract

The invention discloses a plant protection unmanned aerial vehicle spray drift amount prediction method based on ANN-GWO and related equipment, and the method comprises the steps: collecting test parameter data and drift amount data of a plurality of candidate feature parameters in a test stage, carrying out the correlation analysis through a random forest algorithm, determining at least one target feature parameter for the prediction of the spray drift amount, and carrying out the prediction of the spray drift amount through the at least one target feature parameter. Through cooperative use of multiple candidate characteristic parameters and the random forest algorithm, the correlation and reliability of the selected target characteristic parameters on spray drift amount prediction are ensured, and the data amount of post-model training and spray drift amount prediction is reduced at the same time. An ANN-GWO artificial neural network model is used, global search and local development capabilities of a GWO optimization algorithm are effectively utilized, the problem that convergence is easy to occur when part of target feature parameters are used is solved, weight uniformity of multiple target feature parameters is effectively taken into consideration, and the robustness of the target feature parameters is improved. And the reliability of predicting the spray drift amount of different target downwind distances is improved.

Description

Technical Field

[0001] This application generally relates to the technical field of plant protection unmanned aerial vehicles, and specifically relates to a method and related equipment for predicting the spray drift amount of a plant protection unmanned aerial vehicle based on ANN-GWO. Background Art

[0002] During the process of crop planting, in order to effectively manage farmland pests, diseases and weeds, growers usually spray chemical agents such as insecticides, fungicides and herbicides for pest, disease and weed control to ensure that the yield and quality of crops are not affected. In recent years, plant protection unmanned aerial vehicles have developed rapidly in China due to their characteristics of not being restricted by terrain and crop growth, low purchase and maintenance costs, simple operation, strong mobility, etc., and are suitable for plant protection spray operations in plots where it is difficult for humans and ground machinery to enter, expanding the application scenarios of forestry and agricultural aviation plant protection machinery and application technology. However, compared with ground plant protection machinery, plant protection unmanned aerial vehicles have low liquid application volume, fine atomized droplets, high flight altitude and fast operation speed, and are prone to pesticide drift under the influence of natural wind, resulting in phytotoxicity to non-target crops and organisms and environmental pollution. In plant protection operations, the unmanned aerial vehicle is usually controlled to fly at a fixed distance from the edge of the planting area according to its effective spray width. At the same time, the actual measurement of spray drift amount often requires a lot of time, manpower and material resources. Therefore, how to accurately predict the spray drift amount of a plant protection unmanned aerial vehicle at different downwind distances under different characteristic parameters (flight parameters, application parameters, meteorological parameters) has important practical application value for pesticide application risk assessment, formulation of safety specifications for aerial spraying operations, improvement of pesticide utilization rate and reduction of pesticide use. Summary of the Invention

[0003] In view of the above-mentioned defects or deficiencies in the prior art, it is desirable to provide a method and device for predicting the spray drift amount of a plant protection unmanned aerial vehicle based on ANN-GWO, which can use the ANN-GWO artificial neural network model to predict the spray drift amount of a plant protection unmanned aerial vehicle at a target downwind distance.

[0004] In a first aspect, an embodiment of the present application provides a method for predicting the spray drift amount of a plant protection unmanned aerial vehicle based on ANN-GWO. In the test stage, historical test data for testing the spray drift amount of the plant protection unmanned aerial vehicle is obtained. The historical test data at least includes test parameter data and drift amount data of multiple candidate characteristic parameters under different test conditions. The candidate characteristic parameters at least include flight altitude, flight speed, droplet size, average ambient wind speed, average temperature, average humidity, nozzle rotation speed and downwind distance;

[0005] Preprocess the historical test data to obtain preprocessed historical test data;

[0006] Using the random forest algorithm, perform a correlation analysis based on the preprocessed historical test data to determine at least one target feature parameter for predicting the spray drift amount, and obtain the preprocessed test parameter data corresponding to the target feature parameter to construct a training set and a test set; the target feature parameter at least includes the downwind distance;

[0007] Construct an initial ANN-GWO artificial neural network model, and use the training set and the test set to train the initial ANN-GWO artificial neural network model to obtain a target prediction model for predicting the spray drift amount;

[0008] Use the target prediction model to predict the spray drift amount of the plant protection UAV at a target downwind distance.

[0009] In some embodiments, in the test stage, obtaining the historical test data of the spray drift amount test of the plant protection UAV includes:

[0010] Arrange 9 ground fog droplet collection devices with a diameter of 15 cm at 3 m, 5 m, 10 m, 15 m, 20 m, 30 m, and 50 m downwind and at the edge of the spray width of the plant protection UAV;

[0011] In the test stage, use a simulated liquid medicine with a concentration of 0.1% of ABF fluorescent tracer and OP-10 non-ionic surfactant to perform the spray operation of the plant protection UAV, and use the ground fog droplet collection devices to collect the fog droplets drifting on the ground;

[0012] Elute the fog droplet collection devices with deionized water to obtain the fluorescence values corresponding to each fog droplet collection device, and determine the drift amount data corresponding to the current test through the fluorescence values.

[0013] In some embodiments, preprocessing the historical test data to obtain preprocessed historical test data includes:

[0014] Delete the missing values in the historical test data; and

[0015] Perform normalization and standardization processing on the historical test data after missing value processing;

[0016] Among them, the following formula is used for normalization processing:

[0017]

[0018] Among them, x i is the original historical test data to be normalized, is the historical test data after normalization, and its value range is [0, 1], x maxis the maximum value in the original historical test data;

[0019] The following formula is used for standardization:

[0020]

[0021] where μ is the mean of the normalized historical test data, σ is the standard deviation of the normalized historical test data, and x standardized is the normalized historical test data. After standardization, the historical test data conforms to the standard normal distribution with a mean of 0 and a standard deviation of 1.

[0022] In some embodiments, the correlation analysis is performed on the preprocessed historical test data using the random forest algorithm to determine at least one target feature parameter for predicting the spray drift amount, including:

[0023] Obtain the difference in Gini index before and after the decision tree branches for each candidate feature parameter;

[0024] Use the difference in Gini index of each candidate feature parameter to determine the feature contribution amount of each candidate feature parameter;

[0025] Use the ratio of the feature contribution amount of each candidate feature parameter to the total sum of the feature contribution amounts of all candidate feature parameters to determine the contribution value of the candidate feature parameter;

[0026] Take at least one candidate feature parameter whose contribution value meets the preset condition as the target feature parameter.

[0027] In some embodiments, the following formula is used to calculate the difference in Gini index:

[0028] VIM jm = GI m - GI l - GI r

[0029] where VIM jm represents the change value of the Gini index of candidate feature parameter j at node m, GI m represents the Gini index before branching, and GI l and GI r are the Gini indices of the two new nodes generated after node m branches;

[0030] In addition, the following formula is used to calculate the Gini index:

[0031]

[0032] where p k represents the weight of the k-th category.

[0033] In some embodiments, constructing an initial ANN-GWO artificial neural network model, and training the initial ANN-GWO artificial neural network model by using the training set and the test set to obtain a target prediction model for predicting spray drift amount, including:

[0034] Constructing an initial ANN artificial neural network model and GWO algorithm parameters;

[0035] Taking the structural features of the initial ANN artificial neural network model as prey and calculating the position vectors of the wolf pack;

[0036] Calculating the fitness corresponding to the current position vectors of the wolf pack based on the training set;

[0037] Calculating the information of the three gray wolves with the optimal fitness, updating their position vectors and the position vectors of the wolf pack, and performing iterative training accordingly;

[0038] When the termination condition is satisfied, assigning the position vectors of the wolf pack in the current iteration round to the ANN artificial neural network model to obtain the target prediction model.

[0039] In some embodiments, it further includes:

[0040] Calculating the fitness corresponding to the current position vectors of the wolf pack by using the improved root mean square error.

[0041] In a second aspect, an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the method described in the embodiment of the present application is implemented.

[0042] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the method described in the embodiment of the present application is implemented.

[0043] In a fourth aspect, an embodiment of the present application provides a computer program product, including a computer program, characterized in that when the computer program is executed by a processor, the method described in the embodiment of the present application is implemented.

[0044] The method for predicting the spray drift amount of a plant protection UAV based on ANN-GWO provided by the embodiment of the present application collects the test parameter data of multiple candidate feature parameters and the drift amount data during the test phase, and uses the random forest algorithm for correlation analysis to determine at least one target feature parameter for predicting the spray drift amount. By effectively using the combination of multiple candidate feature parameters and the random forest algorithm, it ensures the relevance and reliability of the selected target feature parameter for predicting the spray drift amount, while reducing the data volume for subsequent model training and spray drift amount prediction, improving the prediction accuracy of the target prediction model while increasing the training calculation cost and reducing the cost. Moreover, by using the ANN-GWO artificial neural network model, it effectively utilizes the global search and local development capabilities of the GWO optimization algorithm, solves the problem of difficult convergence that is prone to occur when using some target feature parameters, and effectively takes into account the weight uniformity of multiple target feature parameters, improving the reliability of predicting the spray drift amount at different target downwind distances.

[0045] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objects, and advantages of the present application will become more apparent:

[0047] Figure 1 FIG. shows a schematic flow chart of a method for predicting the spray drift amount of a plant protection UAV based on ANN-GWO provided by an embodiment of the present application;

[0048] Figure 2 FIG. shows a schematic flow chart of a method for predicting the spray drift amount of a plant protection UAV based on ANN-GWO provided by another embodiment of the present application;

[0049] Figure 3 FIG. shows a schematic structural diagram of a computer system of an electronic device or a server suitable for implementing the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] The present application will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related invention, rather than limiting the invention. In addition, it should be noted that for the sake of description, only the parts related to the invention are shown in the drawings.

[0051] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and embodiments.

[0052] To further illustrate the technical solution provided by the embodiments of the present application, the following will be described in detail with reference to the accompanying drawings and specific implementation manners. Although the embodiments of the present application provide the method operation instruction steps as shown in the following embodiments or drawings, more or fewer operation instruction steps may be included in the method based on routine or non-creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided by the embodiments of the present application. When the method is actually processed or executed by the device, it can be executed in the method order shown in the embodiments or drawings or executed in parallel.

[0053] Please refer to Figure 1 , Figure 1 , which shows a schematic flowchart of a method for predicting the spray drift amount of a plant protection UAV based on ANN-GWO provided by an embodiment of the present application. As Figure 1 shown, the method includes:

[0054] Step 101, in the test stage, obtain the historical test data of the spray drift amount test of the plant protection UAV.

[0055] Among them, the historical test data at least includes the test parameter data and drift amount data of multiple candidate feature parameters under different test conditions. The candidate feature parameters at least include flight altitude, flight speed, droplet size, average ambient wind speed, average temperature, average humidity, nozzle rotation speed, and downwind distance.

[0056] That is to say, in the test stage, the plant protection UAV can be controlled to perform tests multiple times at different flight altitudes and different flight speeds, and the corresponding average ambient wind speed, average temperature, and average humidity during each flight test process can be obtained. The droplet size and nozzle rotation speed of the spray of the plant protection UAV can be the same or different multiple times. Preferably, different nozzle rotation speeds and droplet sizes are selected during each test flight.

[0057] After each test flight ends, calculate the drift amount data corresponding to the test flight. Specifically, before the test flight, 9 ground droplet collectors with a diameter of 15 cm are respectively arranged at 3 m, 5 m, 10 m, 15 m, 20 m, 30 m, and 50 m in the downwind direction and at the edge of the spray width of the plant protection UAV. Among them, the opening of the ground droplet collector is not less than its diameter. Preferably, the ground droplet collector is a plastic petri dish.

[0058] In the test stage, a simulated liquid medicine with a concentration of 10% of both the configured ABF fluorescent tracer and OP-10 non-ionic surfactant is used for the spraying operation of the plant protection unmanned aerial vehicle, and a ground droplet collection device is used to collect the droplets drifting on the ground. After each test flight, the droplet collection device is eluted with deionized water to obtain the fluorescence values corresponding to each droplet collection device, and the drift amount data corresponding to the current test is determined through the fluorescence values.

[0059] Preferably, the drift amount data is the average value of the drift amount data corresponding to 9 ground droplet collection devices at this distance position.

[0060] Then, multiple sets of historical test data are generated based on the test parameter data of the candidate feature parameters and the drift amount data corresponding to each test flight. Preferably, a set of test data includes the test parameter data of 7 candidate feature parameters and a test downwind distance and its corresponding drift amount data. In other words, one test flight can obtain 7 sets of historical test data corresponding one by one to 7 downwind distances.

[0061] Step 102, preprocess the historical test data to obtain the preprocessed historical test data.

[0062] It should be noted that preprocessing the historical test data is beneficial to reducing the interference of incorrect data and improving the reliability and accuracy of the data.

[0063] In the embodiment of the present application, preprocessing the historical test data includes deleting the missing values in the historical test data. Specifically, a set of historical test data with missing values is deleted. Preferably, a set of historical test data with missing values in the test parameter data corresponding to the candidate feature parameters is deleted.

[0064] Furthermore, the historical test data after missing value processing in the embodiment of the present application is subjected to normalization and standardization processing.

[0065] It should be noted that normalizing the historical test data can scale the test parameter data corresponding to multiple candidate feature parameters to a unified measurement range, such as [0, 1], thereby avoiding abnormal weight effects in subsequent analysis due to large differences between the test parameter data. Also, standardizing the historical test data can make the processed data conform to the standard normal distribution, further improving the reliability and accuracy of the data in subsequent use.

[0066] Specifically, the following formula can be used for normalization processing:

[0067]

[0068] where x iis the original historical test data to be normalized, is the historical test data after normalization, and its value range is [0,1], x max is the maximum value in the original historical test data;

[0069] The following formula is used for standardization:

[0070]

[0071] where μ is the mean of the historical test data after normalization, σ is the standard deviation of the historical test data after normalization, and x standardized is the historical test data after standardization. After standardization, the historical test data conforms to the standard normal distribution, with a mean of 0 and a standard deviation of 1.

[0072] It should also be noted that when standardizing the historical test data, the test parameter data obtained after normalizing each candidate feature parameter can be standardized separately to improve the feature expression ability of each candidate feature parameter itself.

[0073] Step 103: Use the random forest algorithm to perform correlation analysis based on the preprocessed historical test data, determine at least one target feature parameter for predicting the spray drift amount, and obtain the preprocessed test parameter data corresponding to the target feature parameter, and construct a training set and a test set.

[0074] Among them, the target feature parameter at least includes the downwind distance.

[0075] It should be noted that the random forest algorithm is an algorithm for judging the importance of features in a decision tree. Specifically, in the random forest algorithm, the Gini index is used to calculate the importance of each feature. Among them, the Gini index is an index used to measure the impurity or uncertainty of data.

[0076] Specifically, the following formula is used to calculate the Gini index:

[0077]

[0078] where p k represents the weight of the kth category.

[0079] Further, the random forest algorithm is used to perform correlation analysis based on the preprocessed historical test data to determine at least one target feature parameter for predicting the spray drift amount, including: obtaining the difference in Gini index before and after the decision tree branches for each candidate feature parameter, using the difference in Gini index of each candidate feature parameter to determine the feature contribution amount of each candidate feature parameter, using the ratio of the feature contribution amount of each candidate feature parameter to the total sum of the feature contribution amounts of all candidate feature parameters to determine the contribution value of the candidate feature parameter, and taking at least one candidate feature parameter whose contribution value meets the preset conditions as the target feature parameter.

[0080] Exemplarily, for the candidate feature parameter j, the change in Gini index at node m can be obtained by calculating the difference between the Gini index before branching of the candidate feature parameter j at this node and the Gini index after branching. Specifically, it can be expressed by the following formula:

[0081] VIM jm =GI m -GI l -GI r

[0082] Among them, VIM jm represents the change value of the Gini index of the candidate feature parameter j at node m, GI m represents the Gini index before branching, GI l and GI r are the Gini indices of the two new nodes generated after branching of node m.

[0083] Therefore, the feature contribution amount of the candidate feature parameter j in the decision tree i is:

[0084] VIM ij =∑ m∈M VIM jm

[0085] If there are n decision trees in the random forest, the feature contribution amount of the candidate feature parameter j is:

[0086]

[0087] The value obtained by normalizing the contribution amount of the candidate feature parameter j is the contribution value of the candidate feature parameter j:

[0088]

[0089] Among them, VIM′ j represents the contribution amount of the normalized feature j, represents the sum of the differences in Gini indices of all features.

[0090] It should be understood that the candidate feature parameter with a larger contribution value is considered to have a greater impact on the decision. Therefore, by calculating the Gini index of each candidate feature parameter, the importance degree of multiple candidate feature parameters in the historical test data can be analyzed, and then at least one candidate feature parameter with a higher importance degree is used as the target feature parameter.

[0091] Optionally, the N candidate feature parameters with the highest contribution values are used as the target feature parameters, or at least one candidate feature parameter with a contribution value ratio sum greater than 50% is used as the target parameter.

[0092] Preferably, when the downwind distance of the candidate feature parameter meets the selection strategy of the target feature parameter, the downwind distance is directly used as the target feature parameter; when the downwind distance of the candidate feature parameter does not meet the selection strategy of the target feature parameter, the downwind distance is simulated as the target feature parameter.

[0093] Furthermore, the preprocessed test parameter data corresponding to the target feature parameter is obtained, and a training set and a test set are constructed. Preferably, the preprocessed test parameter data corresponding to the target feature parameter can be divided into a training set and a test set according to a ratio of 8:2.

[0094] Step 104, construct an initial ANN-GWO artificial neural network model, and use the training set and the test set to train the initial ANN-GWO artificial neural network model to obtain a target prediction model for drift amount prediction.

[0095] It should be noted that GWO (Grey Wolf Optimizer) takes the weights and biases of the neural network model as optimization variables, and conducts global search and local development by simulating the hunting behavior of the grey wolf group, so as to find the optimal structure combination.

[0096] In the embodiment of the present application, different parameter combinations of the ANN artificial neural network model are randomly initialized in the exploration space through the GWO algorithm to train the ANN artificial neural network model. By iteratively updating these solutions, the optimal solution is gradually approximated to determine the best ANN artificial neural network model structure for predicting the drift amount.

[0097] Specifically, as Figure 2 shown, it includes the following steps:

[0098] Step 201, construct an initial ANN artificial neural network model and GWO algorithm parameters.

[0099] It should be noted that the initial ANN artificial neural network model needs to be constructed according to the number of target feature parameters. The GWO algorithm parameters include the number of grey wolves and the maximum number of iterations, where the number of grey wolves is greater than or equal to 3.

[0100] Step 202: Take the structural features of the initial ANN artificial neural network model as prey and calculate the position vectors of the wolf pack.

[0101] Before optimization, all weights and biases of the initial ANN artificial neural network model need to be flattened into a one-dimensional vector and mapped to GWO:

[0102] D = (input_dim × hidden_dim) + hidden_dim + (hidden_dim × output_dim) + output_dim

[0103] Where input_dim is the number of input target feature parameters, hidden_dim is the number of neurons in each layer, and output_dim is the number of output results.

[0104] Preferably, randomly generate the position vectors of the initial wolf pack. Among them, the position vector of each gray wolf in the wolf pack corresponds to a set of parameters of the ANN artificial neural network model.

[0105] Step 203: Calculate the fitness corresponding to the current wolf pack position vector based on the training set.

[0106] Specifically, calculate the fitness corresponding to the position vector of each gray wolf in the wolf pack.

[0107] It should be noted that the fitness is used to evaluate the difference between the predicted value generated by the ANN artificial neural network model corresponding to the current wolf pack position vector based on the training set and the actual value corresponding to the training set. Among them, the fitness can adopt the coefficient of determination R 2 , root mean square error RMSE, mean absolute error MAE, etc.

[0108] In the embodiments of the present application, since during subsequent model prediction, prediction needs to be performed according to the target downwind distance, that is, attention is paid to the input of the downwind distance feature parameter. Therefore, during model training, it is necessary to at least achieve attention to the downwind distance and take into account other target feature parameters at the same time.

[0109] Preferably, in the embodiments of the present application, the improved root mean square error is used to calculate the fitness corresponding to the current wolf pack position vector.

[0110] Specifically, the present application adopts a multi-objective optimization strategy, that is, a balanced solution is selected through the Pareto front to simultaneously optimize the model performance and feature weight distribution. The fitness function is:

[0111]

[0112] Where n is the number of training times, ω i is the feature weight, Oi is the corresponding drift amount data in the training set, P i is the prediction result according to the test parameter data in the training set.

[0113] It should be noted that the optimization objective of the fitness function is to minimize the model error (root mean square error RMSE), and at the same time maximize the uniformity of the feature weights, so as to ensure that when the ANN artificial neural network model pays sufficient attention to the downwind distance of the target feature parameter, it can effectively take into account the influence of other target feature parameters on the spray drift amount, so as to ensure that when predicting the spray drift amount at the target downwind distance later, it can rely on the uniform weights of other target feature parameters to predict the drift amount at the target downwind distance.

[0114] Step 204: Calculate the information of the three gray wolves with the optimal fitness, and update their position vectors and the position vectors of the wolf pack, so as to perform iterative training.

[0115] Step 205: When the termination condition is met, assign the position vector of the wolf pack in the current iteration round to the ANN artificial neural network model to obtain the target prediction model.

[0116] It should be noted that in the embodiment of the present application, the preset maximum number of iterations of GWO is used as the training termination condition. Then, assign the position vector of a gray wolf with the optimal fitness to the ANN artificial neural network model to obtain the target prediction model. Among them, the target prediction model has multiple input ends, which correspond one by one to the number of target feature parameters.

[0117] Step 105: Use the target prediction model to predict the drift amount of the plant protection UAV at the target downwind distance.

[0118] Specifically, after obtaining the target prediction model, based on the type of at least one target feature parameter determined by the random forest algorithm, determine the parameter data to be predicted during the flight to be predicted, such as one or more of flight altitude, flight speed, droplet size, average ambient wind speed, average temperature, average humidity, and nozzle rotation speed. Then select at least one target downwind distance, and input the parameter data to be predicted and the target downwind distance into the target prediction model, and the predicted drift amount data at the target downwind distance under the condition of the parameter data to be predicted can be obtained.

[0119] It should be understood that the parameter data to be predicted can be determined according to the flight environment to be flown. For example, according to the predicted average ambient wind speed, average temperature, and average humidity during the actual flight period to be flown, select the flight speed within the third-level wind (wind speed 3.3 m / s) of the highest wind speed at which the plant protection UAV can perform spraying operations and the flight altitude at which it can fly reliably, etc. The present application does not make specific limitations on this process.

[0120] Therefore, the ANN-GWO-based prediction method for the spray drift amount of plant protection UAVs provided by the embodiments of the present application collects the test parameter data and drift amount data of multiple candidate feature parameters during the test phase, and uses the random forest algorithm for correlation analysis to determine at least one target feature parameter for predicting the spray drift amount. By effectively using the combination of multiple candidate feature parameters and the random forest algorithm, it ensures the relevance and reliability of the selected target feature parameters for spray drift amount prediction, while reducing the amount of invalid data in subsequent model training and spray drift amount prediction. It improves the prediction accuracy of the target prediction model while increasing the training calculation cost and reducing the cost. Moreover, by using the ANN-GWO artificial neural network model, it effectively utilizes the global search and local development capabilities of the GWO optimization algorithm, solves the problem of difficult convergence that is prone to occur when using some target feature parameters, and effectively takes into account the weight uniformity of multiple target feature parameters, improving the reliability of predicting the spray drift amount at different target downwind distances.

[0121] It should be noted that although the operations of the method of the present invention are described in a specific order in the drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result.

[0122] Reference is now made to Figure 3 , Figure 3 which shows a schematic structural diagram of a computer system suitable for use in an electronic device or server implementing the embodiments of the present application.

[0123] As Figure 3 shown, the computer system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage section 308 into the random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation instructions of the system are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other through a bus 304. The input / output (I / O) interface 305 is also connected to the bus 304.

[0124] The following components are connected to the I / O interface 305; an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as required. A removable medium 311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is installed on the drive 310 as required so that a computer program read therefrom is installed into the storage section 308 as required.

[0125] Specifically, according to an embodiment of the present application, the process described above with reference to the flowchart Figure 2 can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product that includes a computer program carried on a computer-readable medium, and the computer program includes program codes for performing the method shown in the flowchart. In such an embodiment, the computer program includes program codes for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 309, and / or installed from the removable medium 311. When the computer program is executed by a central processing unit (CPU) 301, the above functions defined in the system of the present application are executed.

[0126] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of a computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, and this computer-readable medium can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0127] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operation instructions of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram can represent a module, a program segment, or a part of code, and the foregoing module, program segment, or part of code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two connected blocks can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for executing specified functions or operation instructions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0128] As another aspect, the present application also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments, or may exist separately without being assembled into the electronic device. The above computer-readable storage medium stores one or more programs, and when the above programs are executed by one or more processors, they are used to implement the ANN-GWO-based prediction method for the spray drift amount of plant protection UAVs described in the present application.

[0129] The above description is only the preferred embodiment of the present application and the explanation of the applied technical principles. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the foregoing disclosure concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the present application.

Claims

1. A method for predicting the spray drift amount of a plant protection unmanned aircraft based on ANN-GWO, characterized in that, Including: In the test stage, obtain historical test data of the spray drift amount test of the plant protection UAV. The historical test data at least includes test parameter data and drift amount data of multiple candidate characteristic parameters under different test conditions. The candidate characteristic parameters at least include flight altitude, flight speed, droplet size, average ambient wind speed, average temperature, average humidity, nozzle rotation speed, and downwind distance; Preprocess the historical test data to obtain preprocessed historical test data; Use the random forest algorithm to perform correlation analysis based on the preprocessed historical test data, determine at least one target characteristic parameter for spray drift amount prediction, and obtain the preprocessed test parameter data corresponding to the target characteristic parameter, and construct a training set and a test set; the target characteristic parameter at least includes the downwind distance; Construct an initial ANN-GWO artificial neural network model, and use the training set and the test set to train the initial ANN-GWO artificial neural network model to obtain a target prediction model for spray drift amount prediction; Use the target prediction model to predict the spray drift amount of the plant protection UAV at the target downwind distance.

2. The method for predicting the spray drift amount of a plant protection UAV based on ANN-GWO according to claim 1, characterized in that, The step of, in the test stage, obtaining historical test data of the spray drift amount test of the plant protection UAV includes: Arrange 9 ground droplet collection devices with a diameter of 15 cm at 3 m, 5 m, 10 m, 15 m, 20 m, 30 m, and 50 m downwind and at the edge of the spray width of the plant protection UAV respectively; In the test stage, use a simulated liquid medicine with a concentration of 0.1% of ABF fluorescent tracer and OP-10 non-ionic surfactant to perform spray operation of the plant protection UAV, and use the ground droplet collection devices to collect the droplets drifting on the ground; Elute the droplet collection devices with deionized water to obtain the fluorescence values corresponding to each droplet collection device, and determine the drift amount data corresponding to the current test through the fluorescence values.

3. The ANN-GWO-based prediction method for the spray drift amount of a plant protection UAV according to claim 1, wherein The step of preprocessing the historical test data to obtain preprocessed historical test data includes: Delete the missing values in the historical test data; and Perform normalization and standardization processing on the historical test data after missing value processing; Among them, the following formula is used for normalization processing: where x i is the original historical test data to be normalized, is the historical test data after normalization, whose value range is [0, 1], and x max is the maximum value in the original historical test data; The following formula is used for standardization processing: where μ is the mean of the normalized historical test data, σ is the standard deviation of the normalized historical test data, and x standardized is the normalized historical test data. After normalization, the historical test data follows a standard normal distribution with a mean of 0 and a standard deviation of 1.

4. The method for predicting the spray drift amount of a plant protection UAV based on ANN-GWO according to claim 1, characterized in that The step of using the random forest algorithm to perform correlation analysis based on the preprocessed historical test data to determine at least one target characteristic parameter for spray drift amount prediction includes: Obtain the difference in Gini index before and after the decision tree branches for each candidate characteristic parameter; Use the difference in Gini index of each candidate characteristic parameter to determine the characteristic contribution amount of each candidate characteristic parameter; Use the ratio of the characteristic contribution amount of each candidate characteristic parameter to the total characteristic contribution amount of all candidate characteristic parameters to determine the contribution value of the candidate characteristic parameter; Use at least one candidate characteristic parameter whose contribution value meets the preset conditions as the target characteristic parameter.

5. The ANN-GWO-based prediction method for the spray drift amount of plant protection UAV according to claim 4, wherein, Calculate the difference in Gini index using the following formula: VIM jm = GI m -GI l -GI r Among them, VIM jm represents the change value of the Gini index of the candidate feature parameter j on the node m, GI m represents the Gini index before branching, GI l and GI r are the Gini indices of the two new nodes generated after the node m branches; And, calculate the Gini index using the following formula: Among them, p k represents the weight of the k-th category.

6. The method for predicting the spray drift amount of a plant protection UAV based on ANN-GWO according to claim 1, wherein Construct the initial ANN-GWO artificial neural network model, and use the training set and the test set to train the initial ANN-GWO artificial neural network model to obtain a target prediction model for predicting the spray drift amount, including: Construct the initial ANN artificial neural network model and GWO algorithm parameters; Take the structural features of the initial ANN artificial neural network model as the prey, and calculate the position vectors of the wolf pack; Calculate the fitness corresponding to the current wolf pack position vectors based on the training set; Calculate the information of the three gray wolves with the best fitness, update their position vectors and the position vectors of the wolf pack, and perform iterative training accordingly; When the termination condition is met, assign the position vectors of the wolf pack in the current iteration round to the ANN artificial neural network model to obtain the target prediction model.

7. The method for predicting the spray drift amount of a plant protection UAV based on ANN-GWO according to claim 6, wherein It further includes: Calculate the fitness corresponding to the current wolf pack position vectors using the improved root mean square error.

8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the ANN-GWO-based method for predicting the spray drift amount of a plant protection UAV according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the ANN-GWO-based method for predicting the spray drift amount of a plant protection UAV according to any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the ANN-GWO-based method for predicting the spray drift amount of a plant protection UAV according to any one of claims 1-7.

Images