Grain crop precise cultivation decision-making method, system, equipment and medium

Abstract

The invention belongs to the technical field of agricultural intelligent decision support and crop cultivation optimization, and particularly provides a food crop precision cultivation decision method, system and device and a medium, and the method comprises the following steps: obtaining cultivation related data of all sites in a target area; downscaling the meteorological data through a bilinear interpolation method, and merging the meteorological data with the soil data and the management measure data into a grid data set; using a random forest model and a crop growth model to determine a plurality of key variety parameters having the maximum influence on the output and the growth period importance; a random forest model is adopted to construct a mapping relation between the key variety parameters and meteorological data and soil data in the grid data set, and the key variety parameters of each grid in the target area are obtained; and utilizing the random forest model and the crop growth model to obtain the sowing date and the planting density corresponding to the target yield of each grid. According to the method, the precision breakthrough on the decision-making scale is realized, and the simulation effect precision is improved by 20-30%.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of agricultural intelligent decision support and crop cultivation optimization, and particularly relates to a precision cultivation decision method, system, device and medium for grain crops. BACKGROUND

[0002] With the continuous improvement of agricultural informatization and fine management needs, more and more researches begin to explore model-driven cultivation management optimization methods. Crop models have a strong mechanism basis and can simulate the dynamic growth process of crops under different soil, climate and management measures, and quantify yield, resource utilization and environmental effects. However, the simulation efficiency of crop models is low, and it is difficult to search and optimize large-scale, full-variable combinations on a national scale.

[0003] At the same time, artificial intelligence models, especially random forests, are widely used in crop yield, nitrogen efficiency and other indicators due to their strong robustness and high fitting ability. However, the model lacks physiological basis, and the generalization ability and interpretability of the results are limited. Some existing researches attempt to use crop model simulation data as a training set to build an AI model for rapid estimation, but most of them still remain at the regional scale or for a single target, lacking integrated recommendation capabilities across scales and targets. In addition, the number of agricultural monitoring sites is not enough and the distribution is uneven, and the spatial scale of related data is large. Therefore, the current cultivation technology recommendation is mostly based on regional experience or expert consensus, and it is difficult to achieve the purpose of high-precision, site-specific decision-making and control. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a precision cultivation decision method, system, device and medium for grain crops, which can provide high-precision decisions for different crops at a low spatial scale in a global range.

[0005] The present application provides a precision cultivation decision method for grain crops, comprising the following steps: Obtaining cultivation-related data of all stations in the target area, the cultivation-related data including meteorological data, soil data and management measure data; Downscaling the meteorological data by a bilinear interpolation method, and merging the soil data and the management measure data into a 1km x 1km resolution grid data set; Using the grid data set and a preset variety parameter set as input data, determining a number of key variety parameters with the greatest influence on yield and growth period importance by using a random forest model and a crop growth model; The random forest model is used to construct a mapping relationship between the key variety parameters and meteorological data and soil data in the grid data set, and to obtain the key variety parameters of each grid in the target area; The random forest model and the crop growth model are used to obtain the sowing date and planting density corresponding to the target yield of each grid by taking the key variety parameters, meteorological data and soil data of each grid in the target area as input data.

[0006] Further, the step of using the random forest model and the crop growth model to determine the several key variety parameters with the greatest influence on yield and growth period importance by taking the grid data set and the preset variety parameter set as input data includes the following steps: The variety parameter set is constructed, and each variety parameter is set to a random distribution within a preset range; An equal number of variety parameter sets and data sets are selected as input into the crop growth model to obtain simulated yield and simulated growth period; The random forest model is used to analyze the importance of variety parameters on simulated yield and simulated growth period, and the several key variety parameters with the greatest influence are selected.

[0007] Further, the step of using the random forest model to construct a mapping relationship between the key variety parameters and meteorological data and soil data in the grid data set, and to obtain the key variety parameters of each grid in the target area includes the following steps: The randomForest function in the random forest model is used to construct a mapping relationship between the key variety parameters and meteorological data and soil data in the grid data set; The predict function in the random forest model is used to predict the key variety parameters, and to obtain the key variety parameters of each grid in the target area.

[0008] Further, the step of using the random forest model and the crop growth model to obtain the sowing date and planting density corresponding to the target yield of each grid by taking the key variety parameters, meteorological data and soil data of each grid in the target area as input data includes the following steps: The key variety parameters, meteorological data and soil data of each grid in the target area are input into the crop growth model to obtain a simulated data set containing growth period, sowing date, planting density, daily yield and total yield; A management measure set is constructed, wherein the planting density and sowing date are set to a random distribution within a preset range, and the remaining management measures are set to the default local optimal value; The random forest model is used to construct a mapping relationship between the simulated data set and the scenario parameter set; The sowing date and planting density corresponding to the target yield of each grid are obtained by using the mapping relationship of the simulation data set and the scenario parameter set.

[0009] Further, the crop growth model is an APSIM model.

[0010] Further, the key variety parameters of corn include accumulated temperature required from crop emergence to the end of the juvenile stage, accumulated temperature required from booting to flowering, accumulated temperature required from flowering to physiological maturity, and radiation use efficiency. The key variety parameters of wheat include spring sensitivity coefficient, photoperiod sensitivity coefficient, accumulated temperature required from sowing to emergence, and accumulated temperature required from emergence to the end of the juvenile stage. The key variety parameters of rice include daily development rate in the juvenile stage, daily development rate in the photoperiod sensitive stage, daily development rate in the panicle development stage, and daily development rate in the grain filling and maturation stage.

[0011] Further, the meteorological data include daily maximum temperature, daily minimum temperature, daily precipitation, and daily radiation data. The soil data include soil bulk density, pH value, clay content, silt content, sand content, organic carbon content, and cation exchange capacity. The management measure data include latitude and longitude, year, sowing date, harvesting date, planting density, fertilizer amount, fertilizer time, irrigation amount, irrigation time, and yield.

[0012] The application also provides a precision cultivation decision-making system for grain crops, which comprises: The acquisition module is configured to acquire cultivation-related data of all stations in a target area, wherein the cultivation-related data include meteorological data, soil data, and management measure data. The preprocessing module is configured to perform downscaling processing on the meteorological data by using a bilinear interpolation method, and combine the meteorological data, the soil data, and the management measure data into a 1km✕1km resolution grid data set. The parameter determination module is configured to use the grid data set and a preset variety parameter set as input data, and determine a plurality of key variety parameters having the greatest influence on yield and growth period by using a random forest model and a crop growth model. The parameter prediction module is configured to construct a mapping relationship between the key variety parameters and the meteorological data and the soil data in the grid data set by using the random forest model, and obtain the key variety parameters of each grid in the target area. The decision-making generation module is configured to use the key variety parameters, the meteorological data, and the soil data of each grid in the target area as input data, and obtain the sowing date and planting density corresponding to the target yield of each grid by using the random forest model and the crop growth model.

[0013] The application further provides a grain crop precision cultivation decision-making device, characterized by comprising a processor and a memory, the memory is used for storing a computer program, and the processor is used for executing the computer program to realize the steps of the method.

[0014] The application further provides a computer readable and writable storage medium, characterized by storing a computer program, and the computer program is executed by a processor to realize the method.

[0015] The application has the following beneficial effects: 1. The application realizes precision breakthrough in decision scale, relies on high-resolution driving factors (i.e. variety parameters, soil data and meteorological data), and realizes 20-30% precision improvement of simulation effect compared with traditional regional uniform parameter mode.

[0016] 2. The application realizes the transition of simulation efficiency, and realizes more than 100 times efficiency improvement compared with single-point simulation efficiency of a traditional crop growth model, and reduces the simulation time to the order of hours.

[0017] 3. The application realizes reasonable optimization of management strategies, and the average yield potential of three main crops is improved by 30-80%. DETAILED DESCRIPTION

[0018] Figure 1 The figure is a flowchart of the method in the application; Figure 2 The figure is a prediction result graph of a corn crop in a crop growth model (a represents yield result, b represents growth period result, the horizontal coordinate represents a measured value, and the vertical coordinate represents a simulated value); Figure 3 The figure is a prediction result graph of a wheat crop in a crop growth model (a represents yield result, b represents growth period result, the horizontal coordinate represents a measured value, and the vertical coordinate represents a simulated value); Figure 4 The figure is a prediction result graph of a rice crop in a crop growth model (a represents yield result, b represents growth period result, the horizontal coordinate represents a measured value, and the vertical coordinate represents a simulated value); Figure 5 The figure is an effect graph of a corn crop inverting yield of a crop growth model by a random forest model (a represents training set effect, b represents test set effect, the horizontal coordinate represents crop growth model simulated yield, and the vertical coordinate represents random forest model inverted yield); Figure 6 The figure is an effect graph of a wheat crop inverting yield of a crop growth model by a random forest model (a represents training set effect, b represents test set effect, the horizontal coordinate represents crop growth model simulated yield, and the vertical coordinate represents random forest model inverted yield); Figure 7 This is a graph showing the yield of rice crop inverted by the random forest model in this invention (a represents the training set effect, b represents the test set effect, the horizontal axis represents the simulated yield of the crop growth model, and the vertical axis represents the yield inverted by the random forest model). Figure 8 This refers to the cultivation decision results corresponding to the target yield of maize crops in this invention; Figure 9 This refers to the cultivation decision results corresponding to the target yield of wheat in this invention; Figure 10 This represents the cultivation decision results corresponding to the target yield of rice in this invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0020] like Figures 1 to 10 As shown, a precision cultivation decision-making method for grain crops includes the following steps: S1. Obtain cultivation-related data for all stations within the target region from 2004 to 2023. The target region is China (data for Hong Kong, Macau, and Taiwan is currently unavailable). Cultivation-related data includes meteorological data, soil data, and management practice data. Meteorological data comes from the China Meteorological Administration's Land Surface Data Assimilation System (CLDAS-V2.0, 0.0625° resolution), which includes daily maximum temperature, daily minimum temperature, daily precipitation, and daily radiation data.

[0021] Soil data were obtained from the Sun Yat-sen University Soil Database (1km resolution, stratified data in the 0-100cm depth range), which includes soil bulk density, pH value, clay, silt, sand content, organic carbon content, and cation conversion rate.

[0022] The management data came from all monitoring stations in publicly available literature, including latitude and longitude, year, sowing date, harvest date, planting density, fertilizer application rate, fertilization time, irrigation amount, irrigation time, and yield.

[0023] S2. Since the cultivation-related data in S1 have a large spatial resolution, they all have limitations and are not conducive to high-precision prediction. Therefore, this step uses bilinear interpolation to downscale the meteorological data to obtain a 1km x 1km resolution meteorological dataset, which is then merged with the soil data and management measures data into a 1km x 1km resolution grid dataset. This achieves scale uniformity and helps to significantly improve the regional adaptability and accuracy of the model simulation in subsequent processes.

[0024] S3, Due to the large number of variety parameters of crops and the interaction between them, it is extremely challenging to identify the key parameters that have a global dominant effect on yield and growth period in a highly heterogeneous spatial environment, and the nonlinear coupling effect between parameters needs to be overcome. Therefore, this step takes the grid dataset and the preset variety parameter set as input data, uses the random forest model and the crop growth model to determine the several key variety parameters that have the greatest influence on the importance of yield and growth period. Specifically, S3 includes the following steps: S301, Construct a variety parameter set M{A i ,B i ,C i ,..N i}, each variety parameter is set to a random distribution within a preset range; A to N represent variety parameters, and i represents the sample number. According to experience, the value range of each variety parameter is preset, and the variety parameter set with a sample size of 10000 is obtained by random distribution.

[0025] S302, Select the grid dataset of 10000 grids in S2, and the variety parameter set with 10000 samples as input data, and simulate 10000 simulated yields and simulated growth periods by the crop growth model. That is, one grid corresponds to one set of variety parameters, that is, one set of simulation results. In this example, the crop growth model is the APSIM model.

[0026] S303, Based on the results of S302, use the importance function in the random forest model to analyze the importance of variety parameters on simulated yield and simulated growth period, and then sort them according to the influence and select the four key variety parameters with the greatest influence for each crop. Specifically, the key variety parameters of corn are tt_emerg_to_endjuv (accumulated temperature required from crop emergence to juvenile stage end), tt_flag_to_flower (accumulated temperature required from flag leaf to flowering stage), tt_flower_to_maturity (accumulated temperature required from flowering to physiological maturity stage), and RUE (radiation use efficiency).

[0027] The key variety parameters of wheat are vern_sens (vernalization sensitivity coefficient), photop_sens (photoperiod sensitivity coefficient), tt_emergence (accumulated temperature required from sowing to emergence), and tt_end_of_juvenile (accumulated temperature required from emergence to end of juvenile stage).

[0028] The key variety parameters of rice are DVRJ (juvenile stage daily development rate), DVRI (photoperiod sensitive stage daily development rate), DVRP (panicle development stage daily development rate), and DVRR (grain filling and maturation stage daily development rate).

[0029] The APSIM model was strictly calibrated by using the pre-collected observation data of representative site yield covering the target area, which was run based on the key variety parameters. As shown in Figures 2 to 4 , the results show that the model yield and growth period simulation effect is good, R 2 reach 0.91-0.99, which fully verifies the effectiveness of the selected key parameters and the strong adaptability of the model in the target area. For non-essential variety parameters, the optimal value is selected by default, and the APSIM model is adjusted by using the measured data to determine the optimal value of the four variety parameters.

[0030] S4, a random forest model is used to construct the mapping relationship between the key variety parameters and the meteorological data and soil data in the grid data set, and the key variety parameters of each grid in the target area are obtained. The traditional method can only determine a set of parameters for one planting area, but this step also performs a downscaling operation on the key variety parameters to obtain the key variety parameters of each grid in the country, thereby providing a basis for subsequent high-precision prediction. Specifically, S4 includes the following steps: S401, the randomForest function in the random forest model is used to construct the mapping relationship between the key variety parameters and the meteorological data and soil data in the grid data set.

[0031] S402, the predict function in the random forest model is used to predict the key variety parameters to obtain the key variety parameters of each grid in the target area.

[0032] S5, since S2-S4 performs grid processing on the related data, the data volume increases exponentially, and it is necessary to process the mapping relationship of massive grid data and multiple crops and variety parameters. Each grid has a unique combination of environmental factors (climate and soil), and there is a highly nonlinear and spatially non-stationary complex relationship between these factors and variety parameters. The predicted parameter combination not only needs to be statistically optimal, but also must meet the logical constraints of crop physiological processes (such as the sequence of accumulated temperature in the development stage, the rationality of the rate), to avoid physically unfeasible parameter values. Therefore, this step uses the key variety parameters of each grid in the target area, meteorological data and soil data as input data, and uses the random forest model and the crop growth model to obtain the sowing date and planting density corresponding to the target yield of each grid. Specifically, S5 includes the following steps: S501, input the key variety parameters, meteorological data, and soil data of each grid in the target area into the crop growth model to obtain a simulation data set containing growth period, sowing date, planting density, daily yield, and total yield; the simulation data set has the following characteristics: 1) spatial and temporal resolution: the time resolution is daily scale (statistical annual scale to facilitate AI modeling), and the spatial resolution is 1km x 1km; 2) regional division: the results are generated according to the crop planting area: corn division: Northeast Corn Belt (NEC), North China Corn Belt (NCP), Northwest Corn Belt (NWC), and Southwest Corn Belt (SWC); wheat division: North China Wheat Belt (NC), North China Wheat Belt (NCP), Southwest Wheat Belt (SWC), and Yangtze River Basin Wheat Belt (YRB); rice division: North China Rice Belt (NC), South China Rice Belt (SC), Southwest Rice Belt (SWC), and Yangtze River Basin Rice Belt (YRB).

[0033] S502, to determine the optimal planting management scheme, a management measure set Y{a j ,b j ,c j ,d j ,......,n j} is constructed, wherein the planting density and sowing date are randomly distributed within the preset range to set their values, and the remaining management measures are the default local optimal values; a to n represent management measures, and j represents the sample number. According to experience, the value range of each planting density and sowing date is preset, and the other cultivation measures are the default local high-yield field cultivation. Each division obtains a scenario parameter set with a sample size of 5000 through random distribution. This sample design ensures the representativeness of the scenario combination and can fully cover the possible value range of each parameter. Specifically, the value range of the planting density and sowing date is shown in Table 1.

[0034] Table 1 Value range of main crop planting density and sowing date

[0035] S503, a random forest model is used to construct the mapping relationship between the simulation data set and the scenario parameter set, realizing fast and accurate mapping from multi-dimensional environmental-management factors to yield indicators. As shown in Figures 5 to 7 , the training set and the test set are divided by 7:3 to ensure spatial representativeness. The random forest is set to 500 decision trees, and the maximum depth of a single tree is determined to be 15 through grid search, and the node splitting criterion adopts variance reduction. Performance optimization: ten-fold cross-validation is used to adjust the hyperparameters, and the final test set R 2The accuracy reached 0.84-0.91. While maintaining over 90% of the prediction accuracy of the mechanistic model, this model improves computational efficiency by approximately 200 times (the APSIM model requires approximately 7 seconds to simulate a single point over 20 years, while the model proxy processing time is approximately 0.035 seconds), providing a reliable intelligent tool for regional-scale agricultural production decision-making.

[0036] This step deeply integrates the physical interpretability advantage of the mechanistic model (APSIM) with the efficient computing power of machine learning, solving the dual bottlenecks of "low computational efficiency" of traditional crop models in large-scale applications and "insufficient mechanistic interpretability" of pure data-driven models.

[0037] S504. Utilizing the mapping relationship between the simulation dataset and the scenario parameter set, the sowing date and planting density corresponding to the target yield for each grid are obtained. Clearly, this step effectively captures the nonlinear correlation between multi-dimensional environmental factors and variety parameters, and achieves efficient processing of high-dimensional feature spaces and large sample sizes. During the prediction process, feature importance analysis indirectly ensures the physical rationality and regional specificity of parameter combinations, successfully establishing a 1km x 1km database of key crop variety parameters covering the target area. The core value of this database lies in completely solving the key bottleneck problem that traditional large-area uniform parameters cannot accurately reflect local environmental differences ("incompatibility"), providing crucial basic data support for subsequent high-precision, large-scale crop growth process simulation and yield prediction.

[0038] This invention also supports global optimization and spatialized decision output. The management strategy optimization engine aims to maximize yield and performs global optimization in the full variable space (sowing date × planting density). It adopts a full factorial experimental design (non-heuristic algorithm) and relies on the high-performance computing platform of China Agricultural University to achieve parallel computing of tens of thousands of scenarios, guaranteeing 100% obtaining the global optimal solution. Moreover, the solution set is not affected by random seeds and has absolute stability.

[0039] like Figures 8 to 10 As shown, a 1km raster-level decision layer is generated: using the ArcGIS spatial analysis engine, the optimized solution set is precisely matched to a 1km grid; multi-band GeoTIFF raster is output: Band 1: Optimal sowing day (Day of year); Band 2: Recommended density (plants / m²) 2 Band 3: Expected yield potential (kg / ha); Construct a WebGIS intelligent decision-making platform to support real-time extraction of parameter combinations by administrative division / geographic coordinates, directly guiding field practices.

[0040] Based on the same inventive concept, the present invention also provides a precision cultivation decision-making system for grain crops, comprising: An acquisition module is configured to acquire cultivation-related data of all stations in a target region, the cultivation-related data including meteorological data, soil data, and management measure data; A preprocessing module is configured to perform downscaling processing on the meteorological data by using a bilinear interpolation method, and combine the meteorological data, the soil data, and the management measure data into a grid data set with a resolution of 1 km x 1 km; A parameter determination module is configured to determine, by using a random forest model and a crop growth model, a plurality of key variety parameters that have the greatest influence on yield and growth period, by taking the grid data set and a preset variety parameter set as input data; A parameter prediction module is configured to construct a mapping relationship between the key variety parameters and the meteorological data and the soil data in the grid data set by using the random forest model, and obtain the key variety parameters of each grid in the target region; A decision generation module is configured to obtain, by using the random forest model and the crop growth model, sowing dates and planting densities corresponding to target yields of each grid, by taking the key variety parameters of each grid in the target region, the meteorological data, and the soil data as input data.

[0041] Based on the same inventive concept, the application further provides a precision cultivation decision device for grain crops, characterized by comprising a processor and a memory, the memory is configured to store a computer program, and the processor is configured to execute the computer program to realize the steps of the above method.

[0042] Based on the same inventive concept, the application further provides a computer readable and writable storage medium, characterized by storing a computer program, and the computer program is executed by a processor to realize the above method.

[0043] The above is only a preferred embodiment of the application, and it should be pointed out that, for those skilled in the art, without departing from the technical principles of the application, a number of improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the application.

Claims

1. A precision cultivation decision method for food crops, characterized by, comprise meteorological data, soil data and management measure data; The meteorological data is down-scaled by a bilinear interpolation method, and is combined with the soil data and the management measure data into a grid data set with a resolution of 1 km x 1 km; The grid data set and a preset variety parameter set are taken as input data, and a random forest model and a crop growth model are used to determine a plurality of key variety parameters that have the greatest influence on yield and growth period importance; A mapping relationship between the key variety parameters and meteorological data and soil data in the grid data set is constructed by using the random forest model, and the key variety parameters of each grid in the target region are obtained. The grid data set and a preset variety parameter set are taken as input data, and a random forest model and a crop growth model are used to determine a plurality of key variety parameters that have the greatest influence on yield and growth period importance, comprising the following steps: A variety parameter set is constructed, and each variety parameter is randomly distributed in a preset range to set its value; 2. The method for precision cultivation decision of grain crops according to claim 1, characterized in that, An equal number of variety parameter sets and data sets are selected to be input into a crop growth model to obtain simulated yield and simulated growth period; A random forest model is used to analyze the importance of variety parameters on simulated yield and simulated growth period, and a plurality of key variety parameters with the greatest influence are selected. The mapping relationship between the key variety parameters and meteorological data and soil data in the grid data set is constructed by using a randomForest function in the random forest model; The key variety parameters are predicted by using a predict function in the random forest model, and the key variety parameters of each grid in the target region are obtained.

3. The method for precision cultivation decision of grain crops according to claim 1, characterized in that, The key variety parameters, meteorological data and soil data of each grid in the target region are taken as input data, and a random forest model and a crop growth model are used to obtain sowing dates and planting densities corresponding to the target yield of each grid, comprising the following steps: The key variety parameters, meteorological data and soil data of each grid in the target region are input into a crop growth model to obtain a simulated data set containing growth period days, sowing dates, planting densities, daily yield and total yield; A management measure set is constructed, wherein planting density and sowing date are randomly distributed in a preset range to set their values, and the remaining management measures are set to the default local optimal values; 4. The method for precision cultivation decision of grain crops according to claim 1, characterized in that, A mapping relationship between the simulated data set and the scenario parameter set is constructed by using a random forest model; Sowing dates and planting densities corresponding to the target yield of each grid are obtained by using the mapping relationship between the simulated data set and the scenario parameter set. The crop growth model is an APSIM model.

6. The method according to claim 1, wherein ​ 5. The method for precision cultivation decision of grain crops according to claim 1, characterized in that, ​ ​ The key variety parameters of corn include accumulated temperature required from crop emergence to the end of the juvenile stage, accumulated temperature required from booting to flowering, accumulated temperature required from flowering to physiological maturity, and radiation use efficiency; The key variety parameters of wheat include spring sensitivity coefficient, photoperiod sensitivity coefficient, accumulated temperature required from sowing to emergence, and accumulated temperature required from emergence to the end of the juvenile stage; The key variety parameters of rice include daily development rate in the juvenile stage, daily development rate in the photoperiod sensitive stage, daily development rate in the panicle development stage, and daily development rate in the grain filling and maturation stage.

7. The method for precision cultivation decision of grain crops according to claim 1, characterized in that, The meteorological data includes daily maximum temperature, daily minimum temperature, daily precipitation, and daily radiation data; The soil data includes soil bulk density, pH value, clay content, silt content, sand content, organic carbon content, and cation exchange capacity; The management measure data includes latitude and longitude, year, sowing date, harvesting date, planting density, fertilizer amount, fertilizer time, irrigation amount, irrigation time, and yield.

8. A precision cultivation decision system for food crops, characterized by, The method comprises the following steps: An acquisition module is configured to acquire cultivation-related data of all stations in a target region, wherein the cultivation-related data includes meteorological data, soil data, and management measure data; A preprocessing module is configured to perform downscaling processing on the meteorological data by using a bilinear interpolation method, and combine the meteorological data, the soil data, and the management measure data into a grid data set with a resolution of 1 km x 1 km; A parameter determination module is configured to determine, by using a random forest model and a crop growth model, a plurality of key variety parameters that have the greatest influence on yield and growth period, by taking the grid data set and a preset variety parameter set as input data; A parameter prediction module is configured to construct a mapping relationship between the key variety parameters and the meteorological data and the soil data in the grid data set by using a random forest model, and obtain the key variety parameters of each grid in the target region; A decision generation module is configured to obtain sowing dates and planting densities corresponding to target yields of each grid in the target region by using a random forest model and a crop growth model, by taking the key variety parameters, the meteorological data, and the soil data of each grid in the target region as input data.

9. A precision cultivation decision device for food crops, characterized by, The computer readable storage medium stores a computer program, and the processor executes the computer program to implement the steps of the method of claim 1.

10. A computer readable and writable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the processor executes the computer program to implement the steps of the method of claim 1.

Images