A crop data processing method and apparatus

Abstract

The application provides a crop data processing method and device, which can obtain a first crop characteristic value; the first crop characteristic value is a crop characteristic value of a target grid in a target period; the first crop characteristic value is input into a trained crop yield regression model to obtain a first crop yield initial value output by the crop yield regression model; the first crop yield initial value is a crop yield initial value of the target grid in the target period, and the first crop yield initial value is generated by regression analysis of the crop yield regression model on the first crop characteristic value; a first crop yield final value is determined based on the first crop yield initial value; wherein the first crop yield final value is a crop yield final value of the target grid in the target period. The application can effectively determine the crop yield value of the grid, obtain the grid crop yield data, and effectively enhance the reliability of the generation of the grid crop yield data.

Description

Technical Field

[0001] This invention relates to the field of crop data processing technology, and in particular to a crop data processing method and apparatus. Background Technology

[0002] Against the backdrop of global change, extreme weather events such as high temperatures, droughts, and floods occur frequently, severely impacting global food production and threatening world food security.

[0003] Rasterized crop yield data plays an important role in analyzing the relationship between regional and even global food production and climate change.

[0004] However, existing methods for generating rasterized crop yield data have low reliability. Summary of the Invention

[0005] This invention provides a crop data processing method and apparatus to address the shortcomings of existing rasterized crop yield data generation methods in terms of low reliability. It effectively determines the crop yield value of the grid, obtains gridded crop yield data, and significantly enhances the reliability of gridded crop yield data generation.

[0006] This invention provides a crop data processing method, comprising:

[0007] Obtain the first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid within the target time period;

[0008] The first crop feature value is input into the trained crop yield regression model to obtain the first initial crop yield value output by the crop yield regression model; wherein, the first initial crop yield value is the initial crop yield value of the target grid in the target time period, and the first initial crop yield value is generated by the crop yield regression model performing regression analysis on the first crop feature value;

[0009] Based on the initial value of the first crop yield, the final value of the first crop yield is determined; wherein, the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period.

[0010] Optionally, the target grid is a grid within the target region; determining the final value of the first crop yield based on the initial value of the first crop yield includes:

[0011] Obtain a second initial value for crop yield, which includes the initial values ​​of crop yield for grids in the target area other than the target grid during the target time period;

[0012] Calculate the sum of the initial yield of the second crop and the initial yield of the first crop;

[0013] Calculate the ratio of the initial value of the first crop yield to the sum value;

[0014] Determine the total statistical value of crop yield in the target area during the target time period;

[0015] The product of the total statistical value of crop yield and the ratio is determined as the final value of the first crop yield.

[0016] Optionally, the target grid is a grid within the target region; before inputting the first crop feature value into the trained crop yield regression model, the crop data processing method further includes:

[0017] Obtain multiple sample sets related to the target region;

[0018] A combined sample set is generated based on the multiple sample sets;

[0019] The crop yield regression model is trained using the combined sample set.

[0020] Optionally, obtaining multiple sample sets related to the target region includes:

[0021] Obtain a first sample set, a second sample set, a third sample set, and a fourth sample set; wherein, the first sample set is a sample set of regions at the same latitude as the target region within the target time period, the second sample set is a sample set of adjacent regions of the target region within the first time period, the third sample set is a sample set of the target region within the second time period, and the fourth sample set is a sample set of the target region within the target time period, the grid samples in the first, second, and third sample sets include crop characteristic values ​​and crop yield values, and the grid samples in the fourth sample set include crop characteristic values.

[0022] Optionally, generating a combined sample set based on the plurality of sample sets includes:

[0023] Using a transfer learning algorithm, target grid samples with a matching degree of not less than a preset matching degree threshold are selected from the first sample set and the second sample set.

[0024] The target grid sample is merged with the third sample set to generate the combined sample set;

[0025] The step of training the crop yield regression model using the combined sample set includes:

[0026] Using the combined sample set, multiple models to be evaluated are trained for regression analysis to determine crop yield values;

[0027] Evaluate the performance of each of the models to be evaluated;

[0028] The model to be evaluated with the best model performance is determined as the crop yield regression model.

[0029] Optionally, the target grid is a grid within the target area, and the first crop feature value includes the final harvested area value; the process of determining the final harvested area value includes the following steps:

[0030] Obtain a first initial value of harvested area, which includes the initial value of harvested area of ​​all grids in the target area during the target time period;

[0031] Calculate the sum of the initial values ​​of the first harvested area;

[0032] Obtain the ratio of the initial value of the second harvested area to the sum; wherein, the initial value of the second harvested area is the initial value of the harvested area of ​​the target grid within the target time period;

[0033] Determine the total harvested area of ​​the target region during the target time period;

[0034] The product of the ratio and the total statistical value of the harvested area is determined as the final value of the harvested area.

[0035] Optionally, obtaining the initial value of the first harvested area includes:

[0036] The harvested area of ​​the first grid in the third time period is obtained. Based on the inverse distance weighting method and the harvested area, an initial value of the harvested area of ​​the first grid in the target time period is generated. The first grid includes all grids in the target area.

[0037] Optionally, the target grid is a grid within the target region; after determining the final value of the first crop yield based on the initial value of the first crop yield, the crop data processing method further includes:

[0038] Obtain the final value of the second crop yield; wherein, the final value of the second crop yield includes the final value of the crop yield of the grid cells in the target area other than the target grid cell during the target time period;

[0039] According to the grid distribution location, the final values ​​of the first crop yield and the second crop yield are spliced ​​together to generate the first overall crop yield data; the first overall crop yield data is the overall crop yield data of the target area within the target time period.

[0040] Optionally, after concatenating the final values ​​of the first and second crop yields according to the grid distribution to generate the overall crop yield data of the target area within the target time period, the crop data processing method further includes:

[0041] Obtain second overall crop yield data; wherein, the second overall crop yield data includes the overall crop yield data of the first region during the target time period;

[0042] According to the regional distribution location, the first overall crop yield data and the second overall crop yield data are spliced ​​together to generate overall crop yield data for the first region and the target region.

[0043] The present invention also provides a crop data processing apparatus, comprising: a first obtaining unit, an input unit, a second obtaining unit, and a first determining unit; wherein:

[0044] The first obtaining unit is used to obtain a first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid in the target time period;

[0045] The input unit is used to input the first crop feature value into the trained crop yield regression model;

[0046] The second obtaining unit is used to obtain a first initial value of crop yield output by the crop yield regression model; wherein, the first initial value of crop yield is the initial value of crop yield of the target grid in the target time period, and the first initial value of crop yield is generated by the crop yield regression model performing regression analysis on the first crop feature value;

[0047] The first determining unit is configured to determine the final value of the first crop yield based on the initial value of the first crop yield; wherein the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period.

[0048] The crop data processing method and apparatus provided by this invention can obtain a first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid within a target time period; the first crop feature value is input into a trained crop yield regression model to obtain an initial value of the first crop yield output by the crop yield regression model; wherein, the initial value of the first crop yield is the initial value of the crop yield of the target grid within the target time period, and the initial value of the first crop yield is generated by the crop yield regression model through regression analysis of the first crop feature value; based on the initial value of the first crop yield, the final value of the first crop yield is determined; wherein, the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period. This invention can utilize a crop yield regression model to perform regression analysis on the first crop feature value to obtain the initial value of the first crop yield, and correct the initial value of the first crop yield to obtain the final value of the first crop yield, which can effectively realize the determination of the crop yield value of the grid, obtain grid crop yield data, and effectively enhance the reliability of grid crop yield data generation. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0050] Figure 1 This is one of the flowcharts of a crop data processing method provided in an embodiment of the present invention;

[0051] Figure 2 This is a second schematic flowchart of a crop data processing method provided in an embodiment of the present invention;

[0052] Figure 3 This is the third flowchart of a crop data processing method provided in an embodiment of the present invention;

[0053] Figure 4 This is a schematic diagram of the structure of a crop data processing device provided in an embodiment of the present invention;

[0054] Figure 5 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0056] The following is combined with Figures 1-3 The crop data processing method of the present invention is described.

[0057] like Figure 1 As shown in the figure, the present invention proposes a first crop data processing method, which may include the following steps:

[0058] S101. Obtain the first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid in the target time period;

[0059] It should be noted that a grid is one of the multiple grids obtained after rasterizing a certain area.

[0060] The target grid can be a grid within a certain region. It should be noted that this region can be an administrative unit, such as a country or province.

[0061] The target period can be a specific time period, such as a particular year or several consecutive years.

[0062] The crop characteristic values ​​can include crop-related indicators that are related to crop yield and have a causal relationship with variables, such as crop harvested area, agricultural meteorological indicators, and agricultural condition indicators. Specifically, the first crop characteristic value is the crop characteristic value of the target grid within the target time period.

[0063] Specifically, this invention can obtain the first crop characteristic value from a data source of crop data specifically for a particular region. This data source may include the CropWatch database, the Spatial Production Allocation Model (SPAM) dataset, the Global Agro-Ecological Zones (GAEZ) dataset, and statistical data from the Food and Agriculture Organization of the United Nations (FAO).

[0064] The CropWatch database can include crop-related indicators for global regions at different times, such as potential cumulative biomass, cumulative precipitation, average temperature, photosynthetically active radiation, arable land planting ratio, optimal vegetation status index, leaf area index, biomass, multiple cropping index, irrigation type, and agroecological zones.

[0065] Specifically, the present invention can obtain crop-related index data from the CropWatch database, perform data preprocessing to achieve a spatial resolution of 10 kilometers, and then extract the first crop feature value from the preprocessed data.

[0066] S102. Input the first crop feature value into the trained crop yield regression model;

[0067] Among them, the crop yield regression model is a model that estimates or predicts the dependent variable by performing regression analysis on the independent variable with a given value.

[0068] Specifically, crop yield regression models can be used to analyze the crop yield of a grid cell within a certain region during a certain period based on the crop characteristic values ​​of that grid cell.

[0069] It should be noted that there is a certain correlation between the crop yield of a grid and the crop characteristic value of that grid, and this correlation is consistent within local homogeneous regions. Therefore, this invention can use a large amount of grid data within local homogeneous regions as samples, and through regression analysis, uncover the association between crop yield and characteristic values, thus obtaining a correlation model between crop yield and characteristic values, i.e., a crop yield regression model.

[0070] Specifically, this invention can collect crop characteristic values ​​and crop yield values ​​from grids in certain areas over certain time periods as samples. These samples are then used to train a model with regression analysis capabilities, enhancing the model's regression analysis ability. When the model's regression analysis ability meets the requirements, the model is determined as the trained crop yield regression model. It is understood that the crop characteristic values ​​and crop yield values ​​in the samples are all known data.

[0071] Specifically, after obtaining the first crop feature value, the present invention can input the first crop feature value into the trained crop yield regression model.

[0072] S103. Obtain the initial value of the first crop yield output by the crop yield regression model; wherein, the initial value of the first crop yield is the initial value of the crop yield of the target grid in the target time period, and the initial value of the first crop yield is generated by the crop yield regression model through regression analysis of the first crop characteristic value;

[0073] The initial crop yield value is the crop yield value output by the crop yield regression model after performing regression analysis on the input crop feature values.

[0074] Specifically, the initial value of the first crop yield is the initial value of the crop yield output by the crop yield regression model through regression analysis of the first crop characteristic value.

[0075] Specifically, after inputting the first crop characteristic value into the crop yield regression model, the present invention can obtain the initial value of the first crop yield obtained and output by the crop yield regression model through regression analysis of the first crop characteristic value.

[0076] S104. Based on the initial value of the first crop yield, determine the final value of the first crop yield; wherein, the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period.

[0077] Specifically, after obtaining the initial value of the first crop yield, the present invention can correct the initial value of the first crop yield to determine the corresponding final value of the first crop yield.

[0078] Optionally, in other crop data processing methods proposed in this embodiment of the invention, the target grid is a grid within the target region; in this case, step S104 may include:

[0079] Obtain the initial value of the second crop yield, which includes the initial value of the crop yield of the grid cells in the target area, excluding the target grid cells, during the target time period;

[0080] Calculate the sum of the initial yield of the second crop and the initial yield of the first crop;

[0081] Calculate the ratio of the initial value of the first crop yield to the sum of the yields;

[0082] Determine the total statistical value of crop yield in the target area during the target time period;

[0083] The product of the total crop yield statistics and the ratio is determined as the final value of the first crop yield.

[0084] Specifically, the initial value of the second crop yield can include the initial values ​​of the crop yield of all grids in the target area, excluding the target grid, during the target time period.

[0085] It is understood that the present invention can determine the initial crop yield of other grids in the target area according to the process of determining the initial crop yield of the target grid in the target time period, i.e., the first initial crop yield.

[0086] Specifically, the present invention can determine the total crop yield statistics of a target region within a target time period from the target statistical data.

[0087] Among them, the target statistical data can be statistical data that includes the total crop yield of different regions at different times, such as FAO statistical data.

[0088] Specifically, the present invention can obtain the initial crop yield values ​​of all grids in the target area, sum them up, and calculate the corresponding sum value; then, the present invention can determine the ratio of the initial crop yield value to the sum value; then, the present invention can multiply the ratio by the total statistical value of crop yield, and determine the product as the final value of the first crop yield.

[0089] It should be noted that the present invention is achieved through... Figure 1 Steps S101 to S104 in the process can use a crop yield regression model to perform regression analysis on the characteristic value of the first crop, obtain the initial value of the first crop yield, and correct the initial value of the first crop yield to obtain the final value of the first crop yield. This can effectively determine the crop yield value within the grid, obtain crop yield data within the grid, and effectively enhance the reliability of the generation of crop yield data within the grid.

[0090] The crop data processing method proposed in this invention can obtain a first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid within a target time period; the first crop feature value is input into a trained crop yield regression model to obtain an initial first crop yield value output by the crop yield regression model; wherein, the initial first crop yield value is the initial crop yield value of the target grid within the target time period, and the initial first crop yield value is generated by the crop yield regression model through regression analysis of the first crop feature value; based on the initial first crop yield value, the final first crop yield value is determined; wherein, the final first crop yield value is the final crop yield value of the target grid within the target time period. This invention can utilize a crop yield regression model to perform regression analysis on the first crop feature value to obtain the initial first crop yield value, and then correct the initial first crop yield value to obtain the final first crop yield value, which can effectively realize the determination of the crop yield value of the grid, obtain grid crop yield data, and effectively enhance the reliability of grid crop yield data generation.

[0091] based on Figure 1 This invention proposes a second crop data processing method. In this method, the target grid is a grid within the target area; in this case, before step S102, the method may further include S1011, S1012, and S1013, wherein:

[0092] S1011. Obtain multiple sample sets related to the target region;

[0093] It should be noted that this invention can collect a sample set and use it to train a crop yield regression model. Since the crop yield regression model needs to determine the initial crop yield value for the target grid within the target region, this invention can collect the sample set from regions related to the target region to reduce the difference between the sample set and the target grid. This allows the trained crop yield regression model to have better performance when determining the initial crop yield value within the target grid.

[0094] Specifically, this invention can collect multiple sample sets for training a crop yield regression model based on a target region. It should be noted that the sample set may include one or more grid samples, and each grid sample may include crop feature values ​​and crop yield values ​​within the grid.

[0095] Specifically, this invention can collect samples from a grid within a target area;

[0096] Specifically, the present invention can collect samples from grids of other regions related to the target region, such as regions at the same latitude and adjacent regions.

[0097] S1012. Generate a combined sample set based on multiple sample sets;

[0098] The combined sample set is a sample set generated by combining samples from multiple sample sets.

[0099] Specifically, after obtaining multiple sample sets, the present invention can obtain a combined sample set based on these multiple sample sets.

[0100] Specifically, after obtaining multiple sample sets, the present invention can select samples with small differences from the target grid from the multiple sample sets and combine the selected samples into a combined sample set.

[0101] S1013. Using a combined sample set, a crop yield regression model is trained.

[0102] Specifically, after obtaining a combined sample set, the present invention can use the grid samples in the combined sample set to train the initial model to be trained, thereby training a crop yield regression model whose performance meets the requirements.

[0103] Optionally, in other crop data processing methods proposed in this invention, step S1011 may include:

[0104] Obtain a first sample set, a second sample set, a third sample set, and a fourth sample set; wherein, the first sample set is the sample set of the same latitude region of the target region within the target time period, the second sample set is the sample set of the adjacent regions of the target region within the first time period, the third sample set is the sample set of the target region within the second time period, and the fourth sample set is the sample set of the target region within the target time period. The grid samples in the first, second, and third sample sets include crop characteristic values ​​and crop yield values, and the grid samples in the fourth sample set include crop characteristic values.

[0105] Specifically, the first sample set can be a sample set collected from the crop data of the grid in the same latitude region of the target area within the target time period. Each grid sample includes the corresponding crop feature value and crop yield value.

[0106] Specifically, the second sample set can be a sample set collected from the crop data of the grid of the adjacent area of ​​the target area in the first time period, and each grid sample includes the corresponding crop feature value and crop yield value;

[0107] The first time period can be a different time period than the target time period. For example, when the target time period is a certain year, the first time period can be a different year than the target time period.

[0108] Specifically, the third sample set can be a sample set collected from the crop data of the grid in the target area during the second time period, and each grid sample includes the corresponding crop feature value and crop yield value;

[0109] The second time period can be a different time period than the target time period. For example, when the target time period is a certain year, the second time period can be a different year than the target time period.

[0110] Specifically, the fourth sample set can be a sample set collected from the crop data of the grid in the target area within the target time period. Each grid sample only includes crop feature values ​​and does not include crop yield values.

[0111] Optionally, step S1012 may include:

[0112] Using the transfer learning algorithm, target grid samples with a matching degree of not less than a preset matching degree threshold are selected from the first and second sample sets.

[0113] The target grid sample is merged with the third sample set to generate a combined sample set;

[0114] It should be noted that the third sample set is a sample set collected from the grid of the target region. The grid samples in the third sample set have a certain degree of local homogeneity with the grid data in the target region. This invention can use the third sample set as a whole as the training sample set for the crop yield regression model. The first and second sample sets, however, are not sample sets collected from the grid of the target region. This invention can select grid samples with low differences from the grid data of the target region from the first and second sample sets as training samples for the crop yield regression model.

[0115] Specifically, the present invention can utilize transfer learning algorithms to filter grid samples from the first and second sample sets to select grid samples with a high degree of matching with grid samples in the fourth sample set.

[0116] The target grid sample is the grid sample selected from the grid samples in the first and second sample sets that has a matching degree with the grid samples in the fourth sample set that is not lower than a preset matching degree threshold.

[0117] The preset matching threshold can be determined by staff based on the actual situation, and this invention does not limit it.

[0118] Specifically, the present invention can merge the selected target grid samples with a third sample set to obtain a combined sample set.

[0119] Optionally, step S1013 may include:

[0120] By using a combined sample set, multiple models to be evaluated are trained for regression analysis to determine crop yield values;

[0121] Evaluate the performance of each model to be evaluated separately;

[0122] The model with the best performance was selected as the crop yield regression model.

[0123] Specifically, this invention can prepare multiple basic models with regression analysis capabilities, i.e., models to be evaluated, such as Catboost, XGBoost, random forest, and support vector machine models. After obtaining a combined sample set, each basic model can be trained using the combined sample set. During the training process, K-fold cross-validation can be used to adaptively adjust the optimal parameters of the model and adaptively evaluate the model's performance. Finally, based on the performance evaluation results, the model with the best performance and the best accuracy can be selected as the crop yield regression model.

[0124] It should be noted that, through the above steps S1011, S1012 and S1013, the present invention obtains multiple sample sets, generates a combined sample set based on the multiple sample sets, and uses the combined sample set to train a crop yield regression model, thereby ensuring the sufficiency and effectiveness of the sample data and enhancing the reliability and accuracy of the crop yield regression model.

[0125] The crop data processing method proposed in this invention can obtain multiple sample sets, generate a combined sample set based on the multiple sample sets, and use the combined sample set to train a crop yield regression model, ensuring the sufficiency and effectiveness of the sample data and enhancing the reliability and accuracy of the crop yield regression model.

[0126] based on Figure 1 This invention proposes a third method for crop data processing. In this method, the target grid is a grid within the target region, and the first crop feature value includes the final harvested area value; the process of determining the final harvested area value may include the following steps:

[0127] S201. Obtain the initial value of the first harvested area, which includes the initial value of the harvested area of ​​all grids in the target area during the target time period.

[0128] Specifically, the initial value of the first harvest area may include the initial value of the harvest area of ​​all grids in the target area during the target time period.

[0129] Optionally, step S201 may include:

[0130] Obtain the harvested area of ​​the first grid in the third time period. Based on the inverse distance weighting method and the harvested area, generate the initial value of the harvested area of ​​the first grid in the target time period. The first grid includes all grids in the target area.

[0131] The third time period may include one or more time periods that are different from the target time period.

[0132] Specifically, for any grid in the first grid, the present invention can first obtain the harvested area of ​​the grid in the third time period, and then use the inverse distance weighting method to calculate the harvested area of ​​the grid in the third time period to calculate the initial value of the harvested area of ​​the grid in the target time period.

[0133] To better illustrate the process of determining the initial value of the harvested area, the present invention provides the following Example 1.

[0134] Example 1: When it is necessary to determine the initial value of the harvested area of ​​a target grid in a target year, this invention can first obtain the harvested area A of the target grid in four years (2000, 2005, 2010, and 2015), which are different from the target year. 2000 A 2005 A2010 and A 2015 This refers to the harvested area of ​​the target grid during the third time period. Subsequently, this invention can use the inverse distance weighting method to estimate the initial value of the harvested area of ​​the target grid in the target year, as shown in the following formula:

[0135]

[0136] It is understood that the present invention can take each grid of the target area as the processing object and determine the initial value of the harvest area of ​​each grid of the target area within the target time period.

[0137] S202. Calculate the sum of the initial values ​​of the first harvested area;

[0138] Specifically, the sum of the initial values ​​of the first harvested area is the sum of the initial values ​​of the harvested area of ​​all grids in the target area during the target time period.

[0139] S203. Obtain the ratio of the initial value of the second harvested area to the sum; wherein, the initial value of the second harvested area is the initial value of the harvested area of ​​the target grid within the target time period;

[0140] S204. Determine the total harvested area of ​​the target region within the target time period;

[0141] Specifically, the present invention can determine the total harvested area of ​​the target region within the target time period from the target statistical data.

[0142] Among them, the target statistical data can be statistical data that includes the total harvested area of ​​different regions at different times, such as FAO statistical data.

[0143] S205. The product of the ratio and the total statistical value of the harvested area is determined as the final value of the harvested area.

[0144] It should be noted that, through the above steps S201, S202, S203, S204 and S205, the present invention can effectively determine the final value of the harvested area of ​​the target grid within the target time period, effectively ensure the determination of the first crop characteristic value, thereby effectively determining the crop yield of the target grid and enhancing the reliability of crop yield determination.

[0145] The crop data processing method proposed in this invention can effectively determine the final harvest area of ​​the target grid within the target time period, effectively ensure the determination of the first crop characteristic value, and thus effectively determine the crop yield of the target grid, enhancing the reliability of crop yield determination.

[0146] based on Figure 1This invention proposes a fourth method for processing crop data. In this method, the target grid is a grid within the target region; after step S104, the method may further include:

[0147] Obtain the final value of the second crop yield; wherein, the final value of the second crop yield includes the final value of the crop yield of the grid cells in the target area other than the target grid cell during the target time period;

[0148] According to the grid distribution location, the final values ​​of the first crop yield and the second crop yield are spliced ​​together to generate the first overall crop yield data; the first overall crop yield data is the overall crop yield data of the target area within the target time period.

[0149] It is understood that the present invention can determine the final crop yield of other grids within the target time period by following the process of determining the final crop yield of the target grid within the target time period.

[0150] Specifically, after determining the final values ​​of the first and second crop yields, the present invention can stitch together the final values ​​of the first and second crop yields according to the distribution position of each grid in the target area to generate the first overall crop yield data, that is, the rasterized crop yield data of the target area in the target time period.

[0151] The crop data processing method proposed in this invention can effectively generate rasterized crop yield data for a target area, further enhancing the reliability of crop yield data generation.

[0152] Based on the fourth crop data processing method described above, this invention proposes a fifth crop data processing method. In this fifth method, after concatenating the final values ​​of the first and second crop yields according to the grid distribution to generate the first overall crop yield data, the method may further include:

[0153] Obtain second overall crop yield data; wherein, the second overall crop yield data includes the overall crop yield data of the first region within the target time period;

[0154] Based on the regional distribution, the first overall crop yield data and the second overall crop yield data are spliced ​​together to generate overall crop yield data for the first region and the target region.

[0155] Specifically, the present invention can determine the overall crop yield data of other regions by following the process of determining the overall crop yield data of the target region.

[0156] Specifically, after determining the first overall crop yield data and the second overall crop yield data, the present invention can splice the first overall crop yield data and the second overall crop yield data according to the distribution relationship of each region to generate overall crop yield data of the first region and the target region.

[0157] It should be noted that when the target region and the first region are various countries around the world, the overall crop yield data of the first region and the target region generated by this invention is the global rasterized crop yield data.

[0158] The crop data processing method proposed in this invention can effectively generate rasterized crop yield data over a large area, further enhancing the reliability of crop yield data generation.

[0159] Based on the aforementioned crop data processing methods, this invention proposes a sixth crop data processing method. In this method, the invention comprehensively utilizes image processing, transfer learning, regression analysis, cross-validation, and other methods. Based on the global agricultural weather and crop condition observation index dataset monitored by CropWatch and other publicly available yield datasets, it can rasterize crop yield data at the administrative unit scale (i.e., national unit), enabling the production of high-precision, high-spatial-resolution rasterized crop yield data for major crop types at the global or regional scale.

[0160] It should be noted that CropWatch divides the world into 110 agro-ecological zones. This invention can treat each agro-ecological zone as a region, process crop data for each agro-ecological zone separately to obtain rasterized crop yield data for each agro-ecological zone, and finally stitch together the rasterized crop yield data of the 110 agro-ecological zones to obtain global rasterized crop yield data for the target year.

[0161] like Figure 2As shown, this invention can use the CropWatch database, SPAM dataset, GAEZ dataset, and FAO statistical data as data sources to extract the required basic crop data. The CropWatch data spans from 2000 to 2022, and is updated at the end of each quarter. The SPAM dataset can include crop harvested area and yield data for multiple years, with a rasterized GeoTIFF grid data type and global coverage. After preprocessing, the spatial resolution of the data can be 10 kilometers. The GAEZ dataset contains data on the planted area, harvested area, and yield of major crops worldwide, including crop harvested area and yield data for multiple years, with a rasterized GeoTIFF grid data type and global coverage. After preprocessing, the spatial resolution of the data is 10 kilometers. The FAO statistical data can be based on administrative units (countries or regions), including yield and harvested area data for major crop types, with a data table data type and covering major countries and regions worldwide.

[0162] like Figure 2 As shown, after starting crop data processing, the present invention can specify the year (target time period) for which data needs to be generated and the agricultural ecological zone (target area) to be processed, and process each agricultural ecological zone one by one.

[0163] Specifically, this invention can determine the rasterized crop yield data of a target area within a target time period by estimating the harvested area, dynamically selecting training samples, adaptively training the regression model, and calculating the grid yield.

[0164] Specifically, the present invention can use data from the CropWatch database from 2000 to 2022, data from the SPAM dataset from 2000, 2005, and 2010, data from the GAEZ dataset from 2015, and data from the FAO statistics from 2000 to 2021 as data sources to extract the required basic crop data.

[0165] Specifically, in the process of estimating the harvested area, this invention can obtain the crop harvested area of ​​different years in the region from the basic data, and then use the inverse distance weighted interpolation method to estimate the harvested area, thereby obtaining the estimated value of the crop harvested area of ​​the region in that year, i.e., the initial value of the crop harvested area, and obtain crop harvested area data at the administrative region scale. The initial value of the crop harvested area is then globally corrected to obtain the corrected crop harvested area of ​​the region in that year. The crop harvested area of ​​the region in that year is used as a component of the crop feature value to generate a sample set of the region in that year.

[0166] In this invention, during the dynamic screening of training samples, usable samples can be selected and extracted from the data source, including sample sets of the same year in the same latitude region, sample sets of different years in adjacent regions, and sample sets of different years in the same region. Based on the sample set of the region and the sample set of the same year, transfer learning is performed on the sample set of the same year in the same latitude region and the sample set of different years in adjacent regions to select target grid samples. The target grid samples are then combined with the sample set of the region and the sample set of the same year to generate a combined sample set.

[0167] In this invention, the adaptive training regression model can use Catboost model, XGBoost model, random forest model and support vector machine model as the base model, and use a combined sample set to train each base model regression model. During the training process, K-fold cross-validation is used to select the optimal model and determine the crop yield regression model.

[0168] In this invention, the calculation of grid yield involves obtaining total crop yield data at the administrative region level from a data source. A crop yield regression model is used to determine the initial crop yield value for each grid cell in that region for that year. Based on the total crop yield data at the administrative region level, the initial crop yield values ​​are globally corrected to obtain the crop yield for each grid cell in that region for that year. Finally, crop yield data from all regions for that year are merged to obtain global crop yield grid data in GeoTiff format for that year. This concludes the crop data processing flow.

[0169] like Figure 3 As shown, the present invention can sequentially read basic data, specify the target year T, and read the list of all administrative units, and then process each administrative unit as follows:

[0170] The following processes were performed on each agricultural ecological zone within the administrative unit: estimated harvest area, dynamic screening of training samples, adaptive training of regression models, and calculation of crop yield.

[0171] Subsequently, this invention can synthesize global crop yield data for the target year T.

[0172] The crop data processing method proposed in this invention can effectively generate rasterized global crop yield data and enhance the reliability of crop yield data generation.

[0173] The crop data processing apparatus provided by the present invention is described below. The crop data processing described below can be referred to in correspondence with the crop data processing method described above.

[0174] like Figure 4 As shown, the present invention proposes a crop data processing device, comprising: a first obtaining unit 401, an input unit 402, a second obtaining unit 403, and a first determining unit 404; wherein:

[0175] The first obtaining unit 401 is used to obtain a first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid in the target time period;

[0176] Input unit 402 is used to input the first crop feature value into the trained crop yield regression model;

[0177] The second obtaining unit 403 is used to obtain the first initial value of crop yield output by the crop yield regression model; wherein, the first initial value of crop yield is the initial value of crop yield of the target grid in the target time period, and the first initial value of crop yield is generated by the crop yield regression model performing regression analysis on the first crop characteristic value;

[0178] The first determining unit 404 is used to determine the final value of the first crop yield based on the initial value of the first crop yield; wherein the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period.

[0179] It should be noted that the specific processing procedures of the first obtaining unit 401, the input unit 402, the second obtaining unit 403, and the first determining unit 404, and their beneficial effects, can be found in the present invention. Figure 1 The relevant explanations of steps S101, S102, S103 and S104 will not be repeated here.

[0180] Optionally, the target grid is a grid in the target region; the first determining unit 404 includes: a third obtaining unit, a first calculation unit, a second calculation unit, a second determining unit, and a third determining unit;

[0181] The third obtaining unit is used to obtain the initial value of the second crop yield, which includes the initial value of the crop yield of the grid cells in the target area excluding the target grid cells during the target time period.

[0182] The first calculation unit is used to calculate the sum of the initial value of the second crop yield and the initial value of the first crop yield;

[0183] The second calculation unit is used to calculate the ratio of the initial value of the first crop yield to the sum value;

[0184] The second determining unit is used to determine the total statistical value of crop yield in the target area during the target time period;

[0185] The third determining unit is used to determine the final value of the first crop yield by multiplying the total statistical value of crop yield by the ratio.

[0186] Optionally, the target grid is a grid within the target area; the crop data processing device further includes: a fourth acquisition unit, a first generation unit, and a first training unit;

[0187] The fourth obtaining unit is used to obtain multiple sample sets related to the target region before inputting the first crop feature value into the trained crop yield regression model;

[0188] The first generation unit is used to generate a combined sample set based on multiple sample sets;

[0189] The first training unit is used to train a crop yield regression model using a combined sample set.

[0190] Optionally, the fourth obtaining unit is used to obtain a first sample set, a second sample set, a third sample set, and a fourth sample set before inputting the first crop feature value into the trained crop yield regression model; wherein, the first sample set is a sample set of the same latitude region of the target region within the target time period, the second sample set is a sample set of the adjacent regions of the target region within the first time period, the third sample set is a sample set of the target region within the second time period, and the fourth sample set is a sample set of the target region within the target time period, the grid samples in the first, second, and third sample sets include crop feature values ​​and crop yield values, and the grid samples in the fourth sample set include crop feature values.

[0191] Optionally, the first generation unit includes: a filtering unit and a merging unit; the first training unit includes: a second training unit, an evaluation unit and a fourth determination unit;

[0192] The filtering unit is used to use a transfer learning algorithm to filter out target grid samples from the first and second sample sets that have a matching degree of not less than a preset matching degree threshold with the fourth sample set.

[0193] The merging unit is used to merge the target grid sample with the third sample set to generate a combined sample set;

[0194] The second training unit is used to train multiple models to be evaluated for regression analysis of crop yield values ​​using a combined sample set.

[0195] Evaluation unit, used to evaluate the performance of each model to be evaluated;

[0196] The fourth determination unit is used to determine the crop yield regression model as the model with the best performance.

[0197] Optionally, the target grid is a grid within the target area, and the first crop feature value includes the final harvested area value; the process for determining the final harvested area value is set as follows:

[0198] Obtain the initial value of the first harvested area, which includes the initial value of the harvested area of ​​all grids in the target area during the target time period;

[0199] Calculate the sum of the initial values ​​of the first harvested area;

[0200] Obtain the ratio of the initial value of the second harvested area to the sum value; wherein, the initial value of the second harvested area is the initial value of the harvested area of ​​the target grid within the target time period;

[0201] Determine the total harvested area of ​​the target region within the target time period;

[0202] The final value of the harvested area is determined by multiplying the ratio by the total statistical value of the harvested area.

[0203] Optionally, the initial value for obtaining the first harvested area is set as follows:

[0204] Obtain the harvested area of ​​the first grid in the third time period. Based on the inverse distance weighting method and the harvested area, generate the initial value of the harvested area of ​​the first grid in the target time period. The first grid includes all grids in the target area.

[0205] Optionally, the target grid is a grid within the target area; the crop data processing device also includes: a fifth acquisition unit and a first stitching unit;

[0206] The fifth obtaining unit is used to obtain the second crop yield final value after determining the first crop yield final value based on the first crop yield initial value; wherein, the second crop yield final value includes the crop yield final value of the grid cells in the target area excluding the target grid cell during the target time period;

[0207] The first splicing unit is used to splice the final values ​​of the first crop yield and the second crop yield according to the grid distribution position to generate the first overall crop yield data; the first overall crop yield data is the overall crop yield data of the target area within the target time period.

[0208] Optionally, the crop data processing device may also include: a sixth acquisition unit and a second splicing unit;

[0209] The sixth obtaining unit is used to splice the final values ​​of the first and second crop yields according to the grid distribution to generate the overall crop yield data of the target area within the target time period, and then obtain the second overall crop yield data; wherein, the second overall crop yield data includes the overall crop yield data of the first area within the target time period;

[0210] The second splicing unit is used to splice the first overall crop yield data and the second overall crop yield data according to the regional distribution location, so as to generate overall crop yield data for the first region and the target region.

[0211] The crop data processing device proposed in this invention can obtain a first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid within a target time period; the first crop feature value is input into a trained crop yield regression model to obtain an initial value of the first crop yield output by the crop yield regression model; wherein, the initial value of the first crop yield is the initial value of the crop yield of the target grid within the target time period, and the initial value of the first crop yield is generated by the crop yield regression model through regression analysis of the first crop feature value; based on the initial value of the first crop yield, the final value of the first crop yield is determined; wherein, the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period. This invention can utilize a crop yield regression model to perform regression analysis on the first crop feature value to obtain the initial value of the first crop yield, and then correct the initial value of the first crop yield to obtain the final value of the first crop yield, which can effectively realize the determination of the crop yield value of the grid, obtain grid crop yield data, and enhance the reliability of grid crop yield data generation.

[0212] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a crop data processing method, which includes:

[0213] Obtain the first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid within the target time period;

[0214] The first crop feature value is input into the trained crop yield regression model to obtain the initial value of the first crop yield output by the crop yield regression model; wherein, the initial value of the first crop yield is the initial value of the crop yield of the target grid in the target time period, and the initial value of the first crop yield is generated by the crop yield regression model through regression analysis of the first crop feature value;

[0215] Based on the initial value of the first crop yield, the final value of the first crop yield is determined; wherein, the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period.

[0216] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0217] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to perform the crop data processing method provided by the above methods, the method comprising:

[0218] Obtain the first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid within the target time period;

[0219] The first crop feature value is input into the trained crop yield regression model to obtain the initial value of the first crop yield output by the crop yield regression model; wherein, the initial value of the first crop yield is the initial value of the crop yield of the target grid in the target time period, and the initial value of the first crop yield is generated by the crop yield regression model through regression analysis of the first crop feature value;

[0220] Based on the initial value of the first crop yield, the final value of the first crop yield is determined; wherein, the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period.

[0221] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the crop data processing methods provided by the methods described above, the method comprising:

[0222] Obtain the first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid within the target time period;

[0223] The first crop feature value is input into the trained crop yield regression model to obtain the initial value of the first crop yield output by the crop yield regression model; wherein, the initial value of the first crop yield is the initial value of the crop yield of the target grid in the target time period, and the initial value of the first crop yield is generated by the crop yield regression model through regression analysis of the first crop feature value;

[0224] Based on the initial value of the first crop yield, the final value of the first crop yield is determined; wherein, the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period.

[0225] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0226] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0227] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A crop data processing method, characterized in that, include: Obtain the first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid within the target time period; The first crop feature value is input into the trained crop yield regression model to obtain the first initial crop yield value output by the crop yield regression model; wherein, the first initial crop yield value is the initial crop yield value of the target grid in the target time period, and the first initial crop yield value is generated by the crop yield regression model performing regression analysis on the first crop feature value; Based on the initial value of the first crop yield, the final value of the first crop yield is determined; wherein, the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period; The target grid is a grid within a target area; determining the final value of the first crop yield based on the initial value of the first crop yield includes: obtaining a second initial value of the crop yield, the second initial value of the crop yield including the initial crop yield values ​​of the grids in the target area other than the target grid within the target time period; calculating the sum of the second initial value of the crop yield and the first initial value of the crop yield; calculating the ratio of the first initial value of the crop yield to the sum; determining the total statistical value of the crop yield in the target area within the target time period; and determining the final value of the first crop yield as the product of the total statistical value of the crop yield and the ratio. Before inputting the first crop feature value into the trained crop yield regression model, the crop data processing method further includes: obtaining multiple sample sets related to the target region; generating a combined sample set based on the multiple sample sets; and training the crop yield regression model using the combined sample set. The step of obtaining multiple sample sets related to the target area includes: obtaining a first sample set, a second sample set, a third sample set, and a fourth sample set; wherein, the first sample set is a sample set of regions at the same latitude as the target area within the target time period, the second sample set is a sample set of adjacent regions of the target area within the first time period, the third sample set is a sample set of the target area within the second time period, and the fourth sample set is a sample set of the target area within the target time period; the grid samples in the first, second, and third sample sets include crop characteristic values ​​and crop yield values, and the grid samples in the fourth sample set include crop characteristic values; The step of generating a combined sample set based on the multiple sample sets includes: using a transfer learning algorithm to select target grid samples from the first and second sample sets whose matching degree with the fourth sample set is not lower than a preset matching degree threshold; merging the target grid samples with the third sample set to generate the combined sample set; and using the combined sample set to train the crop yield regression model includes: using the combined sample set to train multiple models to be evaluated for regression analysis to obtain crop yield values; evaluating the model performance of each model to be evaluated; and determining the model to be evaluated with the best model performance as the crop yield regression model.

2. The crop data processing method according to claim 1, characterized in that, The target grid is a grid within the target area, and the first crop characteristic value includes the final harvested area value; the process of determining the final harvested area value includes the following steps: Obtain a first initial value of harvested area, which includes the initial value of harvested area of ​​all grids in the target area during the target time period; Calculate the sum of the initial values ​​of the first harvested area; Obtain the ratio of the initial value of the second harvested area to the sum; wherein, the initial value of the second harvested area is the initial value of the harvested area of ​​the target grid within the target time period; Determine the total harvested area of ​​the target region during the target time period; The product of the ratio and the total statistical value of the harvested area is determined as the final value of the harvested area.

3. The crop data processing method according to claim 2, characterized in that, Obtaining the initial value of the first harvested area includes: The harvested area of ​​the first grid in the third time period is obtained. Based on the inverse distance weighting method and the harvested area, an initial value of the harvested area of ​​the first grid in the target time period is generated. The first grid includes all grids in the target area.

4. The crop data processing method according to claim 1, characterized in that, After determining the final value of the first crop yield based on the initial value of the first crop yield, the crop data processing method further includes: Obtain the final value of the second crop yield; wherein, the final value of the second crop yield includes the final value of the crop yield of the grid cells in the target area other than the target grid cell during the target time period; According to the grid distribution location, the final values ​​of the first crop yield and the second crop yield are spliced ​​together to generate the first overall crop yield data; the first overall crop yield data is the overall crop yield data of the target area within the target time period.

5. The crop data processing method according to claim 4, characterized in that, After concatenating the final values ​​of the first and second crop yields according to the grid distribution to generate the overall crop yield data for the target area within the target time period, the crop data processing method further includes: Obtain second overall crop yield data; wherein, the second overall crop yield data includes the overall crop yield data of the first region during the target time period; According to the regional distribution location, the first overall crop yield data and the second overall crop yield data are spliced ​​together to generate overall crop yield data for the first region and the target region.

6. A crop data processing device, characterized in that, include: The system comprises a first obtaining unit, an input unit, a second obtaining unit, and a first determining unit; wherein: The first obtaining unit is used to obtain a first crop feature value; wherein, the first crop feature value is the crop feature value of the target grid in the target time period; The input unit is used to input the first crop feature value into the trained crop yield regression model; The second obtaining unit is used to obtain a first initial value of crop yield output by the crop yield regression model; wherein, the first initial value of crop yield is the initial value of crop yield of the target grid in the target time period, and the first initial value of crop yield is generated by the crop yield regression model performing regression analysis on the first crop feature value; The first determining unit is configured to determine the final value of the first crop yield based on the initial value of the first crop yield; wherein the final value of the first crop yield is the final value of the crop yield of the target grid within the target time period; The target grid is a grid within the target region; the first determining unit includes: a third obtaining unit, a first calculation unit, a second calculation unit, a second determining unit, and a third determining unit; The third obtaining unit is used to obtain the initial value of the second crop yield, which includes the initial value of the crop yield of the grid cells in the target area excluding the target grid cells during the target time period. The first calculation unit is used to calculate the sum of the initial value of the second crop yield and the initial value of the first crop yield; The second calculation unit is used to calculate the ratio of the initial value of the first crop yield to the sum value; The second determining unit is used to determine the total statistical value of crop yield in the target area during the target time period; The third determining unit is used to determine the final value of the first crop yield by multiplying the total statistical value of crop yield by the ratio. The crop data processing device also includes: a fourth acquisition unit, a first generation unit, and a first training unit; The fourth obtaining unit is used to obtain multiple sample sets related to the target region before inputting the first crop feature value into the trained crop yield regression model; The first generation unit is used to generate a combined sample set based on multiple sample sets; The first training unit is used to train a crop yield regression model using a combined sample set; The fourth obtaining unit is used to obtain a first sample set, a second sample set, a third sample set, and a fourth sample set before inputting the first crop feature value into the trained crop yield regression model; wherein, the first sample set is a sample set of the same latitude region of the target region within the target time period, the second sample set is a sample set of the adjacent regions of the target region within the first time period, the third sample set is a sample set of the target region within the second time period, and the fourth sample set is a sample set of the target region within the target time period, the grid samples in the first, second, and third sample sets include crop feature values ​​and crop yield values, and the grid samples in the fourth sample set include crop feature values; The first generation unit includes a screening unit and a merging unit; the first training unit includes a second training unit, an evaluation unit, and a fourth determination unit. The filtering unit is used to use a transfer learning algorithm to filter out target grid samples from the first and second sample sets that have a matching degree of not less than a preset matching degree threshold with the fourth sample set. The merging unit is used to merge the target grid sample with the third sample set to generate a combined sample set; The second training unit is used to train multiple models to be evaluated for regression analysis of crop yield values ​​using a combined sample set. Evaluation unit, used to evaluate the performance of each model to be evaluated; The fourth determination unit is used to determine the crop yield regression model as the model with the best performance.

Images