Crop pest space-time prediction method and device based on meteorological soil factors

Abstract

The invention relates to a meteorological soil factor-based crop pest spatio-temporal prediction method and device. The method comprises the following steps of: acquiring daily pest information and meteorological soil data in a full life cycle of crops; performing data checking and preprocessing on the meteorological soil data to generate research area meteorological soil raster data with consistent spatial resolution; and training an insect pest prediction model by using the training set, and taking the meteorological soil raster data as input data of the insect pest prediction model to realize insect pest quantity prediction. The crop insect pest time-space prediction method and the crop insect pest time-space prediction device provided by the invention have the beneficial effects of effectively solving the problems of excessive dependence on expert experience and complexity in relation between prediction factors and insect pests, surpassing the limitation of previous small-range prediction, and realizing accurate prediction of crop insect pest time sequence and spatial distribution.

Description

Technical Field

[0001] The present invention relates to the field of agricultural information technology, and more specifically, to a method and device for spatiotemporal prediction of crop pests based on meteorological and soil factors. Background Art

[0002] Traditional crop pest prediction methods mostly rely on manual experience judgment and mathematical statistics prediction, which is not only time-consuming and laborious, but also the prediction results are often limited by fixed experience and patterns, making it difficult to capture the nonlinear relationship between prediction factors and pests. With the advancement of information technology, machine learning models have the ability to reveal the complex relationship between pests and prediction factors due to their powerful data integration capabilities, data processing capabilities and prediction accuracy, and can continuously optimize the prediction model through adaptive learning, providing timely and accurate pest warnings for agricultural production, showing great potential in the field of crop pest prediction.

[0003] However, existing crop pest prediction methods and devices based on machine learning still face the following two challenges: (a) crop pest prediction and early warning mainly focus on the model construction during the pest outbreak period, and lack pest prediction and early warning for the entire growth process of crops; (b) they only focus on the prediction of pests in time series, and do not comprehensively study the spatial distribution and time series changes of pests. Summary of the invention

[0004] The purpose of the present invention is to address the deficiencies of the prior art and to propose a method and device for spatiotemporal prediction of crop pests based on meteorological and soil factors.

[0005] In the first aspect, a method for spatiotemporal prediction of crop pests based on meteorological and soil factors is provided, comprising:

[0006] S1. Set up a sex trapping and reporting system in the study area to obtain daily pest information and first meteorological soil data throughout the life cycle of crops;

[0007] S2. Obtain the second meteorological soil data of crops in the whole life cycle within the study area;

[0008] S3, verifying and preprocessing the second meteorological soil data obtained in S2 to generate meteorological soil raster data of the study area with consistent spatial resolution;

[0009] S4. Divide the data obtained by the neutral trapping and reporting system in S1 into a training set and a validation set, use the training set to train the pest prediction model, and use the meteorological soil grid data as input data of the pest prediction model to achieve pest quantity prediction.

[0010] Preferably, in S1, the pest information includes pest types, quantities and pest images; the first meteorological soil data is composed of a number of meteorological soil factors, and the meteorological soil factors include average temperature, rainfall, wind speed, wind direction, atmospheric pressure, dew point temperature, soil temperature and soil moisture.

[0011] Preferably, in S2, meteorological soil data are obtained through a meteorological station in the study area, and / or, second meteorological soil data are obtained through a reanalysis data set provided by the European Centre for Medium-Range Weather Forecasts.

[0012] Preferably, in S3, the data verification includes: checking whether there are abnormal values ​​and missing values ​​in the data, and deleting the abnormal values ​​and missing values;

[0013] The pre-processing comprises:

[0014] For the meteorological soil data obtained from the meteorological station in the study area, the inverse distance weighted interpolation method is used to interpolate the point data into raster data. The spatial resolution of the raster data is determined according to the actual distribution of the meteorological station sites.

[0015] The second meteorological soil data obtained through the reanalysis dataset provided by the European Centre for Medium-Range Weather Forecasts are resampled using the bilinear interpolation method to generate raster data with the required spatial resolution.

[0016] Preferably, in S4, the training set takes the date of the entire life cycle of the crop as a time series, meteorological and soil factors as independent variables, and the number of pests as a dependent variable; and the pest prediction model is an XGBoost model.

[0017] Preferably, in S4, the step of training the pest prediction model using the training set includes:

[0018] Determine the input parameters of the XGBoost model and their parameter ranges; the input parameters include learning rate, maximum number of iterations, maximum depth of the tree, minimum child node weight and regularization term;

[0019] Using a grid search technique, systematically traverse all possible combinations of the input parameters within their value ranges;

[0020] For each parameter combination, leave-one-out cross validation was performed;

[0021] The mean absolute error (MAE) of several validations is used as an indicator to evaluate the performance of the parameter combination;

[0022] The MAE values ​​of all parameter combinations are compared, and the parameter combination and value corresponding to the minimum MAE are selected as the optimal parameter combination of the XGBoost model, and the XGBoost optimal model is determined based on the optimal parameter combination of the XGBoost model.

[0023] As a preference, it also includes:

[0024] S5. Compare the validation set with the predicted pest quantity value, and use evaluation indicators to evaluate the performance of the pest prediction model; the evaluation indicators include: mean absolute error, root mean square error and Pearson correlation coefficient.

[0025] In a second aspect, a device for spatiotemporal prediction of crop pests based on meteorological and soil factors is provided, characterized in that it is used to execute any method described in the first aspect, including:

[0026] The first acquisition module is used to deploy a sex-attracting detection and reporting system in the study area to obtain daily pest information and the first meteorological soil data throughout the life cycle of crops;

[0027] The second acquisition module is used to obtain the second meteorological soil data in the whole life cycle of crops within the study area;

[0028] A generation module, used for performing data verification and preprocessing on the second meteorological soil data acquired by the second acquisition module, and generating meteorological soil raster data of the study area with consistent spatial resolution;

[0029] The prediction module is used to divide the data obtained by the neutral attractant system of the first acquisition module into a training set and a verification set, use the training set to train the pest prediction model, and use the meteorological soil grid data as input data of the pest prediction model to realize the prediction of pest quantity.

[0030] According to a third aspect, a computer storage medium is provided, wherein a computer program is stored in the computer storage medium; when the computer program is executed on a computer, the computer executes any method described in the first aspect.

[0031] The beneficial effects of the present invention are as follows: the present invention collects crop planting information, pest information and meteorological soil data in the study area. The meteorological soil data is preprocessed so that the data of different meteorological soil factors have the consistency and accuracy of spatial characteristics. On this basis, the present invention adopts the XGBoost model to predict the daily spatial distribution of the number of pests, and selects the best parameter combination of model parameters through grid search and leave-one-out cross-validation method, which effectively solves the problem of parameter tuning of the XGBoost model on complex data sets. In addition, the present invention combines the spatial distribution data of crop planting to perform spatial masking on the prediction results, improves the accuracy and practicality of the prediction results, and provides a scientific basis for crop pest management. Compared with traditional pest prediction methods, the spatiotemporal prediction method and device for crop pests provided by the present invention effectively solve the problems of over-reliance on expert experience and the complexity of the relationship between prediction factors and pests, while surpassing the limitations of previous small-scale predictions, and realizing accurate prediction of the time series and spatial distribution of crop pests. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a technical flow chart of a method and device for spatiotemporal prediction of crop pests based on meteorological soil factors;

[0033] Figure 2 Schematic diagram of the study area and experimental field in the embodiment of the present invention;

[0034] Figure 3 A schematic diagram of the deployment of a sexual inducement detection system and information collection in the research area in an embodiment of the present invention;

[0035] Figure 4 This is a schematic diagram of data collection and processing in the study area in an embodiment of the present invention;

[0036] Figure 5 This is the meteorological soil factor preprocessing diagram of the study area on April 8, 2023 in the embodiment of the present invention

[0037] Figure 6 This is a spatial prediction distribution map of the cumulative number of Chilo suppressalis in the study area in the embodiment of the present invention;

[0038] Figure 7 It is a scatter plot of the prediction results of Jiulong Wangdao in the embodiment of the present invention. DETAILED DESCRIPTION

[0039] The present invention is further described below in conjunction with embodiments. The description of the following embodiments is only used to help understand the present invention. It should be noted that for ordinary persons in the art, without departing from the principle of the present invention, the present invention can also be modified in some ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

[0040] Embodiment 1:

[0041] Rice is the main crop in Ningbo, Zhejiang. According to statistics, in 2023, the early rice area in Ningbo will be 227,000 mu, with a yield of 423.8 kg per unit area; the single-season late rice area will be 665,000 mu, with a yield of 539.1 kg per unit area; the continuous late rice area will be 207,200 mu, with a yield of 448.36 kg per unit area. As one of the main pests of rice, the Chilo suppressalis continuously bores the stems and feeds from the tillering stage to the milky stage of rice, which will cause a 3% to 5% reduction in rice production, and in severe cases it can reach 18.5%, affecting the stability of rice production and the quality of rice. The spatiotemporal prediction method of crop pests can provide timely and accurate prevention and control guidance for farmers and agricultural management departments by analyzing the occurrence and spatial distribution of pests, and reduce the damage of pests to crops. In this regard, a spatiotemporal prediction method of crop pests based on meteorological soil factors was applied to Ningbo, Zhejiang Province. This method uses time series meteorological soil factors with spatial characteristics, combines grid search and leave-one-out cross-validation technology to optimize XGBoost model parameters, and obtains the optimal spatiotemporal prediction model for pests. Finally, the prediction results are masked in combination with crop planting spatial distribution data to achieve accurate prediction of crop pest time series and spatial distribution.

[0042] Specifically, Example 1 of the present application provides a method for spatiotemporal prediction of crop pests based on meteorological and soil factors, including:

[0043] S1. Set up a sex-attracting detection and reporting system in the study area to obtain daily pest information and the first meteorological soil data throughout the life cycle of crops.

[0044] like Figure 2 As shown, rice planting and harvesting dates were obtained through field observations, and sex-induced reporting systems were set up in three experimental fields: Jiulong Wangdao in Zhenhai District, Ningbo City, Modern Seed Incubation Base in Yinzhou District, and Mailang Farm in Fenghua District.

[0045] like Figure 3 As shown, the sex-attracting detection and reporting system consists of two parts: sex-attracting equipment and meteorological monitoring equipment, which are used to obtain pest information and first meteorological soil data respectively. The sex-attracting equipment uses sex attractants for trapping, and is set to automatically shoot and identify at 10:30 a.m. every day, so as to collect the number and image of the Chilo suppressalis; the meteorological monitoring equipment collects meteorological soil data every hour, and forms 24-hour average daily data, and the study uses daily meteorological soil data. These meteorological soil data are composed of several meteorological soil factors, including average temperature, rainfall, wind speed, wind direction, atmospheric pressure, dew point temperature, soil temperature and soil moisture. It should be noted that the first meteorological soil data is used for the training of the pest prediction model (please refer to S4).

[0046] S2, such as Figure 4As shown, the second meteorological soil data of the entire life cycle of crops in the study area are obtained.

[0047] For example, the data of each meteorological station in Ningbo City, including average temperature, rainfall, wind speed and wind direction, are obtained through the Ningbo Meteorological Station; in addition, the reanalysis data set (ERA5) is obtained through the European Center for Medium-Range Weather Forecasts (ECMWF), including atmospheric pressure, dew point temperature, soil temperature and soil moisture. It should be noted that the second meteorological soil data is used for spatial prediction (please refer to S3).

[0048] S3, such as Figure 5 As shown, the second meteorological soil data obtained by S2 is verified and preprocessed to generate meteorological soil raster data of the study area with consistent spatial resolution.

[0049] The data verification refers to checking whether there are outliers and missing values ​​in the data, and deleting the outliers and missing values.

[0050] In addition, the preprocessing of the second meteorological soil data obtained from the meteorological station in the study area includes: using the inverse distance weighted interpolation method to interpolate the point data into raster data, and the spatial resolution of the raster data is determined according to the actual distribution of the meteorological station sites. The spatial resolution of the raster data is determined to be 5km based on the distribution of the meteorological stations in Ningbo, and the calculation formula is:

[0051]

[0052] In the above formula, Z(P) is the value of the point to be interpolated, Z(P i ) is the known sample point P i The value of d(P,P i ) is the interpolation point P and the known sample point P i The distance between them. k is the exponential parameter of the distance, and N is the number of known points involved in the interpolation.

[0053] The preprocessing of the second meteorological soil data obtained through the reanalysis dataset (ERA5) provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) includes: resampling using the bilinear interpolation method to generate raster data of the required spatial resolution, and the calculation formula is:

[0054]

[0055] In the above formula, f(x1,y1), f(x2,y1), f(x1,y2), f(x2,y2) are the function values ​​of four known neighboring points, (x1,y1), (x2,y1), (x1,y2), (x2,y2) are the coordinates of the known points, and (x,y) are the coordinates of the target point, that is, the point you want to estimate.

[0056] S4. Divide the data obtained by the neutral trapping and reporting system in S1 into a training set and a validation set, use the training set to train the pest prediction model, and use the meteorological soil grid data as input data of the pest prediction model to achieve pest quantity prediction.

[0057] Specifically, the data collected by the sex-induced detection and reporting system from two experimental fields, the Modern Seed Industry Incubation Base in Yinzhou District, Ningbo City and Mailang Farm in Fenghua District, were used as training sets, and the data from Jiulongwang Rice Experimental Field in Zhenhai District, Ningbo City were used as validation sets. The training set uses the date of the entire life cycle of crops as a time series, meteorological and soil factors as independent variables, and the cumulative number of rice stem borers as a dependent variable. Specifically, the training set is mainly used to train the pest prediction model to find the parameters of the optimal prediction model; the pest quantity prediction uses the meteorological and soil raster data as the input data for pest prediction, and uses the determined optimal XGBoost model to generate the pest quantity prediction results; the validation set is used to evaluate the effect of the prediction results.

[0058] In S4, the step of training the pest prediction model using the training set includes:

[0059] Determine the input parameters of the XGBoost model and their parameter ranges; the input parameters include learning rate, maximum number of iterations, maximum depth of the tree, minimum child node weight and regularization term.

[0060] A grid search technique is used to systematically traverse all possible combinations of the input parameters within their value ranges.

[0061] For each parameter combination, leave-one-out cross validation is performed. Leave-one-out cross validation is to use each sample in the data set as a validation set in turn, and the remaining samples as training sets, and repeat this process until each sample is validated once to evaluate the performance of the model on each sample.

[0062] The mean absolute error (MAE) of several (eg, 5) validations is used as an indicator for evaluating the performance of the parameter combination.

[0063] The MAE values ​​of all parameter combinations are compared, and the parameter combination and value corresponding to the minimum MAE are selected as the optimal parameter combination of the XGBoost model, and the XGBoost optimal model is determined based on the optimal parameter combination of the XGBoost model.

[0064] Among them, the smaller the MAE value, the better the prediction performance of the model. The calculation formula of MAE is:

[0065]

[0066] Where n is the number of samples, y i is the actual observation value of the ith is the i-th predicted value, and MAE measures the mean absolute error between the predicted value and the actual value.

[0067] Embodiment 2:

[0068] Based on Example 1, Example 2 of the present application provides a more specific method for spatiotemporal prediction of crop pests based on meteorological and soil factors. On the basis of existing machine learning algorithms, it fully considers the dynamic changes of meteorological and soil factors and the response of pests to these changes, and constructs a method and device for spatiotemporal prediction of crop pests based on meteorological and soil factors, so as to achieve accurate prediction of the number of pests in time and space during the entire life cycle of crops in a large area, thereby improving the predictability and planning of crop pest control and enhancing the sustainability and safety of agricultural production.

[0069] Specifically, Figure 1 As shown, the method includes:

[0070] S1. Set up a sex-attracting detection and reporting system in the study area to obtain daily pest information and the first meteorological soil data throughout the life cycle of crops.

[0071] S2. Obtain the second meteorological soil data during the entire life cycle of crops in the study area.

[0072] S3. Verify and preprocess the second meteorological soil data obtained in S2 to generate meteorological soil raster data of the study area with consistent spatial resolution.

[0073] S4. Divide the data obtained by the neutral attractant reporting system in S1 into a training set and a validation set, use the training set to train the pest prediction model, and use the first meteorological soil grid data as input data of the pest prediction model to achieve pest quantity prediction.

[0074] like Figure 6 As shown in the figure, the optimal XGBoost model is used to predict the daily spatial distribution of the number of insect pests, and the spatial distribution data of crop planting in the study area are combined for masking to obtain the daily spatial prediction results of the number of crop pests.

[0075] S5. Compare the validation set with the predicted pest quantity value, and use evaluation indicators to evaluate the performance of the pest prediction model; the evaluation indicators include: mean absolute error, root mean square error and Pearson correlation coefficient.

[0076] like Figure 7 As shown in the figure, the validation set is compared with the predicted pest quantity value, and the model performance is evaluated using evaluation indicators to achieve quantitative analysis of the model performance.

[0077] It should be noted that the parts in this embodiment that are the same or similar to those in Embodiment 1 can be referenced to each other and will not be described in detail in this application.

[0078] Embodiment 3:

[0079] On the basis of Example 1, Example 3 of the present application provides a spatiotemporal prediction device for crop pests based on meteorological and soil factors, which is used to execute any of the methods described in the first aspect, including:

[0080] The first acquisition module is used to deploy a sex-attracting detection and reporting system in the study area to obtain daily pest information and the first meteorological soil data throughout the life cycle of crops;

[0081] The second acquisition module is used to obtain the second meteorological soil data in the whole life cycle of crops within the study area;

[0082] A generation module, used for performing data verification and preprocessing on the second meteorological soil data acquired by the second acquisition module, and generating meteorological soil raster data of the study area with consistent spatial resolution;

[0083] The prediction module is used to divide the data obtained by the neutral attractant system of the first acquisition module into a training set and a verification set, use the training set to train the pest prediction model, and use the meteorological soil grid data as input data of the pest prediction model to realize the prediction of pest quantity.

[0084] Specifically, the system provided in this embodiment is a system corresponding to the method provided in Embodiment 1. Therefore, the parts in this embodiment that are the same or similar to those in Embodiment 1 can be referenced to each other and will not be repeated in this application.

Claims

1. A method for spatiotemporal prediction of crop pests based on meteorological and soil factors, characterized in that: include: S1. Set up a sex trapping and reporting system in the study area to obtain daily pest information and first meteorological soil data throughout the life cycle of crops; S2. Obtain the second meteorological soil data of crops in the whole life cycle within the study area; S3, verifying and preprocessing the second meteorological soil data obtained in S2 to generate meteorological soil raster data of the study area with consistent spatial resolution; S4. Divide the data obtained by the neutral trapping and reporting system in S1 into a training set and a validation set, use the training set to train the pest prediction model, and use the meteorological soil grid data as input data of the pest prediction model to achieve pest quantity prediction.

2. The method for spatiotemporal prediction of crop pests based on meteorological and soil factors according to claim 1, characterized in that: In S1, the pest information includes pest types, quantities and pest images; the first meteorological soil data is composed of a number of meteorological soil factors, and the meteorological soil factors include average temperature, rainfall, wind speed, wind direction, atmospheric pressure, dew point temperature, soil temperature and soil moisture.

3. The method for spatiotemporal prediction of crop pests based on meteorological and soil factors according to claim 2, characterized in that: In S2, the second meteorological soil data are obtained through the meteorological station where the study area is located, and / or the meteorological soil data are obtained through the reanalysis dataset provided by the European Centre for Medium-Range Weather Forecasts.

4. The method for spatiotemporal prediction of crop pests based on meteorological and soil factors according to claim 3, characterized in that: In S3, the data verification includes: checking whether there are abnormal values ​​and missing values ​​in the data, and deleting the abnormal values ​​and missing values; The pre-processing comprises: For the second meteorological soil data obtained from the meteorological station in the study area, the point data is interpolated into raster data using the inverse distance weighted interpolation method. The spatial resolution of the raster data is determined according to the actual distribution of the meteorological station sites. The meteorological soil data obtained through the reanalysis dataset provided by the European Centre for Medium-Range Weather Forecasts are resampled using the bilinear interpolation method to generate raster data with the required spatial resolution.

5. The method for spatiotemporal prediction of crop pests based on meteorological and soil factors according to claim 4, characterized in that: In S4, the training set takes the date of the entire life cycle of crops as a time series, meteorological and soil factors as independent variables, and the number of pests as a dependent variable; the pest prediction model is an XGBoost model.

6. The method for spatiotemporal prediction of crop pests based on meteorological and soil factors according to claim 5, characterized in that: In S4, the step of training the pest prediction model using the training set includes: Determine the input parameters of the XGBoost model and their parameter ranges; the input parameters include learning rate, maximum number of iterations, maximum depth of the tree, minimum child node weight and regularization term; Using a grid search technique, systematically traverse all possible combinations of the input parameters within their value ranges; For each parameter combination, leave-one-out cross validation was performed; The mean absolute error (MAE) of several validations is used as an indicator to evaluate the performance of the parameter combination; The MAE values ​​of all parameter combinations are compared, and the parameter combination and value corresponding to the minimum MAE are selected as the optimal parameter combination of the XGBoost model, and the XGBoost optimal model is determined based on the optimal parameter combination of the XGBoost model.

7. The method for spatiotemporal prediction of crop pests based on meteorological and soil factors according to claim 6, characterized in that: Also includes: S5. Compare the validation set with the predicted pest quantity value and use the evaluation index to evaluate the performance of the pest prediction model; The evaluation indicators include: mean absolute error, root mean square error and Pearson correlation coefficient.

8. A device for spatiotemporal prediction of crop pests based on meteorological and soil factors, characterized in that: Used to perform the method according to any one of claims 1 to 7, comprising: The first acquisition module is used to deploy a sex-attracting detection and reporting system in the study area to obtain daily pest information and the first meteorological soil data throughout the life cycle of crops; The second acquisition module is used to obtain the second meteorological soil data in the whole life cycle of crops within the study area; A generation module, used for performing data verification and preprocessing on the second meteorological soil data acquired by the second acquisition module, and generating meteorological soil raster data of the study area with consistent spatial resolution; The prediction module is used to divide the data obtained by the neutral attractant system of the first acquisition module into a training set and a verification set, use the training set to train the pest prediction model, and use the meteorological soil grid data as input data of the pest prediction model to realize the prediction of pest quantity.

9. A computer storage medium, characterized in that The computer storage medium stores a computer program; when the computer program is executed on a computer, the computer executes any one of the methods described in claims 1 to 7.

Images