Crop yield prediction method based on whole-course mechanized operation data
By collecting and analyzing data from the entire mechanized operation process, a yield prediction model based on the characteristic data of each operation stage was established, which solved the problem of inaccurate crop yield prediction in existing technologies and achieved more accurate yield prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-23
- Publication Date
- 2026-03-27
AI Technical Summary
Current crop yield prediction technologies rely on human experience, resulting in insufficient accuracy and failing to meet the needs of precision agriculture.
By collecting data from the entire mechanized operation process, extracting characteristic data from each operation stage, and combining this data with historical data to input into the yield prediction model, a yield prediction model based on the operation stages is established, taking into account the dynamic changes in crop growth.
It enables more accurate crop yield forecasting, improves forecast accuracy, and adapts to the dynamic changes in the crop growth cycle.
Smart Images
Figure CN121745385A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of agricultural intelligence, in particular to a crop yield prediction method based on full mechanization operation data. BACKGROUND
[0002] With the continuous development of agricultural production mode, precision agriculture has gradually become an important development direction of modern agriculture. Yield prediction is one of the core applications of precision agriculture, which provides a scientific basis for field management and production planning through early estimation of crop yield, and the combination of the two can greatly improve the efficiency of agricultural production and resource utilization.
[0003] However, the current crop yield prediction in the art relies on artificial experience, and is estimated by soil quality, crop species and meteorological factors, etc. This prediction method has the problem of insufficient accuracy and cannot meet the needs of precision agriculture.
[0004] Therefore, how to realize more accurate crop yield prediction is a problem to be solved by those skilled in the art. SUMMARY
[0005] The purpose of the present application is to solve the problem of inaccurate prediction in the current technology by simply relying on experience and initial conditions for crop yield prediction. Therefore, the present application provides a crop yield prediction method based on full mechanization operation data to consider the dynamic changes of crop growth and realize more accurate yield prediction.
[0006] In order to solve the above technical problems, the present application provides a crop yield prediction method based on full mechanization operation data, comprising:
[0007] Collecting operation data in the current operation link and extracting current link feature data therefrom; the operation link is a link for performing operation actions according to the growth cycle of crops;
[0008] The current link feature data and the historical link feature data extracted from the historical operation link are input into a yield prediction model to predict the yield of crops; wherein the yield prediction model is a model established according to the relationship between the feature data in different operation links and the final crop yield.
[0009] Preferably, the operation link includes a tillage link, a seeding link, a fertilization link, an irrigation link and a plant protection link;
[0010] The operation data of the tillage link is the total energy consumption of mechanical operation; the operation data of the seeding link is the total seeding amount of mechanical operation; the operation data of the fertilization link is the total fertilization amount of mechanical operation; the operation data of the irrigation link is the total irrigation amount of mechanical operation; and the operation data of the plant protection link is the total pesticide spraying amount of mechanical operation.
[0011] Preferably, the calculation formula of the tillage link characteristic data is: ;
[0012] wherein, is the tillage load coefficient, is the energy consumption per unit area, , is the total energy consumption of mechanical operation, is the actual operation coverage area, is the reference energy consumption per unit area;
[0013] The calculation formula of the sowing link characteristic data is: ;
[0014] wherein, is the sowing difference coefficient, is the sowing amount per unit area, wherein is the total sowing amount of mechanical operation, is the recommended sowing amount per unit area;
[0015] The calculation formula of the fertilization link characteristic data is: ;
[0016] wherein, is the fertilization deviation coefficient, is the fertilization intensity per unit area, wherein is the total fertilization amount of mechanical operation, is the recommended fertilization amount per unit area;
[0017] The calculation formula of the irrigation link characteristic data is: ;
[0018] wherein, is the water difference coefficient, is the water supply intensity per unit area, wherein is the total irrigation amount of mechanical operation, is the crop water requirement per unit area;
[0019] The calculation formula of the plant protection link characteristic data is: ;
[0020] wherein, is the prevention and control intensity coefficient, is the plant protection effective equivalent intensity per unit area, , is the total pesticide spraying amount of mechanical operation, is the effective ingredient concentration of pesticide liquid, The recommended drug intensity per unit area is taken as a reference.
[0021] Preferably, the establishment of the yield prediction model comprises: extracting historical feature data of different operation links according to historical operation data in the training data set;
[0022] The corresponding historical measured yield is taken as a supervision signal, and the random forest model is trained according to all the historical feature data collected in different operation links to obtain the yield prediction model.
[0023] Preferably, the method further comprises:
[0024] The yield prediction model is called, and historical feature data extracted from each historical operation link in the test data set is used for yield prediction;
[0025] The prediction error is confirmed according to the deviation between the predicted yield and the measured yield, and the influence weight of the feature data extracted from different operation links on yield prediction is confirmed;
[0026] The yield prediction model is modified based on the influence weight.
[0027] Preferably, the method further comprises:
[0028] The operation field is divided into grids according to a fixed side length ratio;
[0029] Operation data is collected based on the divided grids.
[0030] Preferably, before the step of extracting current link feature data, the method further comprises: screening out abnormal data in the operation data.
[0031] The present application provides a crop yield prediction method based on whole-process mechanized operation data. Compared with the current technology which simply relies on experience and initial conditions for crop yield prediction, the present application has the problem of inaccurate prediction. The present application collects operation data when the crop performs operation actions in different growth periods, and extracts feature data therefrom. The feature data extracted from previous operation links is input into a yield prediction model for yield prediction, thereby realizing dynamic yield prediction according to different operation links and improving prediction accuracy. By using the present technical solution, the dynamic changes of crop growth are considered, different types of operation data are collected in different operation links, and feature data capable of representing crop yield are extracted therefrom for yield prediction. The present technical solution not only enables yield prediction in different growth periods of crops, but also comprehensively predicts yield in combination with data collected in previous operation links, thereby improving prediction accuracy. BRIEF DESCRIPTION OF DRAWINGS
[0032] In order to more clearly illustrate the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings described in the following are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.
[0033] Figure 1 A flow chart of a crop yield prediction method based on whole-process mechanized operation data provided by the embodiments of the present application. DETAILED DESCRIPTION
[0034] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without any creative effort are within the protection scope of the present application.
[0035] The core of the present application is to provide a crop yield prediction method based on whole-process mechanized operation data, so as to realize more accurate yield prediction by considering the dynamic changes of crop growth.
[0036] In order to make the person skilled in the art better understand the present application, the present application will be further described in the following with reference to the drawings and specific embodiments.
[0037] Figure 1 A flow chart of a crop yield prediction method based on whole-process mechanized operation data provided by the embodiments of the present application is shown in Figure 1 The method comprises the following steps.
[0038] S10: collecting operation data in a current operation link and extracting current link feature data therefrom; the operation link is a link in which operation actions are performed according to a crop growth cycle;
[0039] S11: inputting the current link feature data and historical link feature data extracted from historical operation links into a yield prediction model to perform crop yield prediction; wherein the yield prediction model is a model established according to the relationship between feature data in different operation links and final crop yield.
[0040] The crop yield prediction method based on full-mechanized operation data provided in the present application is mainly applied to full-mechanized operation of agricultural production, and is used for realizing prediction of crop yield. The execution subject of the method can be a crop yield prediction device based on full-mechanized operation data, which realizes implementation of the crop yield prediction method based on full-mechanized operation data. In specific implementation, the device can specifically include a memory and a processor, wherein the memory is used for storing a computer program, and the processor is used for realizing steps of the crop yield prediction method based on full-mechanized operation data provided in the present application when executing the computer program. In some embodiments, the crop yield prediction device based on full-mechanized operation data can further include a display, a touch screen and other human-computer interaction devices. In specific implementation, the crop yield prediction device based on full-mechanized operation data provided in the present application can specifically include but is not limited to a smart phone, a tablet computer, a notebook computer or a desktop computer and the like.
[0041] Of course, it can be understood that if the method in the embodiments of the present application is realized in the form of a software function unit and is sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, all or part of the technical solutions of the present application can be embodied in the form of a software product. The computer software product is stored in a storage medium, and executes all or part of the steps of the method described in each embodiment of the present application.
[0042] The prediction of crop yield in the present application relies on data collected in different operation links. It can be understood that crop planting to harvesting is a long-term process, which needs to go through multiple operation links such as fertilization and watering. The operation data of each link reflects the growth state of the crop to a certain extent, so the crop yield can be predicted through the operation data.
[0043] For the collection of operation data, the operation data can be collected by sensors and other devices installed on the machine during mechanical operation, or obtained by manual calculation. In step S10, the operation data of the link is collected when the machine performs the operation action, and the operation link reflects the influence of the mechanical operation on the crop growth and the final yield. The application divides the operation link into tillage link, seeding link, fertilization link, irrigation link and plant protection link. In order to realize the representation of operation data on yield, different operation data is collected for different links. The operation data of tillage link is total energy consumption of mechanical operation; the operation data of seeding link is total seeding amount of mechanical operation; the operation data of fertilization link is total fertilization amount of mechanical operation; the operation data of irrigation link is total irrigation amount of mechanical operation; and the operation data of plant protection link is total pesticide spraying amount of mechanical operation. It can be understood that such data can be collected by sensors on the machine. It should be noted that when operating a whole field, due to geological, soil and other factors, different regions have different demands for water and fertilizer, and the final crop yield is also different. If the data of the whole field is directly used for yield prediction, the prediction will be inaccurate. Therefore, in the specific implementation, the operation field can be divided into grids according to a fixed side length ratio, and the operation data is collected based on the divided grids. Finally, the predicted yields of each grid are summed to obtain the predicted yield of the whole field, thereby improving the accuracy of the prediction. When collecting operation data, the operation data of different operation links of each grid is collected.
[0044] After collecting the operation data, the characteristic data representing the influence of the data on the yield needs to be extracted from the operation data. The application provides a specific extraction method. The calculation formula of the characteristic data of the tillage link is: ; wherein, is the tillage load coefficient, is the unit area energy consumption, , is the total energy consumption of mechanical operation, is the actual operation coverage area, is the reference unit area energy consumption, which can be obtained by looking up or calibrating the soil type, tillage depth and operation speed working condition. The calculation formula of the characteristic data of the seeding link is: ; wherein, is the seeding difference coefficient, is the unit area seeding amount, , wherein is the total seeding amount of mechanical operation, is the reference unit area recommended seeding amount, which can be set according to the agronomic technical regulations, crop planting standards or expert experience. The calculation formula of the characteristic data of the fertilization link is: ; wherein, is the fertilization deviation coefficient, is the fertilization intensity per unit area, wherein is the total fertilization amount of mechanical operation, is the recommended fertilization amount per unit area of the reference, which can be given by an agronomic guide, a crop fertilizer requirement model, or expert experience; the calculation formula of the irrigation link characteristic data is: ; wherein, is the water difference coefficient, is the water supply intensity per unit area, wherein is the total irrigation amount of mechanical operation, is the crop water requirement per unit area of the reference, which can be obtained by a crop model, calculation of meteorological evapotranspiration, or agronomic recommended value; the calculation formula of the plant protection link characteristic data is: ; wherein, is the prevention and control intensity coefficient, is the plant protection effective equivalent intensity per unit area, , is the total pesticide spraying amount of mechanical operation, is the effective ingredient concentration of pesticide liquid, is the recommended pesticide use intensity per unit area of the reference, which can be obtained by consulting a plant protection technical manual or according to expert experience.
[0045] In the application to the extraction of characteristic data of each grid after division, for example, for grid , the calculation formula of the tillage link characteristic data is: ; wherein, is the tillage load coefficient of grid , is the unit area energy consumption of grid , , is the total energy consumption of the tillage operation mechanical operation in grid , is the actual operation coverage area of grid , wherein is the effective operation width (when the width is approximately constant) of the tillage operation mechanical operation in grid , is the effective operation track length of the tillage operation mechanical operation in grid . The calculation formula of the sowing link characteristic data is: ; wherein, is the sowing difference coefficient of grid , is the unit area sowing amount of grid , wherein is the total energy consumption of the sowing operation mechanical operation in grid ,The total seeding amount during mechanized seeding operations; the formula for calculating characteristic data of the fertilization process is: ;in, For grid Fertilization deviation coefficient For grid Fertilizer application intensity per unit area ,in For in the grid The total fertilizer application rate during mechanical fertilization operations; the calculation formula for irrigation characteristic data is: ;in, For grid Moisture difference coefficient For grid Water supply intensity per unit area ,in For in the grid The total irrigation volume during mechanical irrigation operations; the calculation formula for characteristic data of the plant protection process is: ;in, For grid Prevention and control intensity coefficient For grid Effective equivalent intensity of plant protection per unit area , For in the grid Total amount of pesticides sprayed during the operation of plant protection machinery.
[0046] In practice, to ensure data validity and avoid wasting computing power, abnormal data in the task data can be filtered out before extracting the feature data of the current stage. This abnormal data mainly includes abrupt changes in the task data and values that do not conform to reality.
[0047] After extracting the feature data, in step S11, the current stage feature data, combined with historical stage feature data extracted from historical operation stages, is input into the yield prediction model to predict crop yield. The yield prediction model is established based on the relationship between feature data in different operation stages and the final crop yield. It is understood that different yield prediction models can be set for different grids, and the same yield prediction model can be used when the relationship between yield and feature data in each grid is nearly equal.
[0048] In the present application, the yield prediction model is pre-trained according to the relevant data of historical crop harvesting records, the data in the records are divided into a training data set and a test data set, the training data set is used to build the model, and the test data set is used to correct the model. Specifically, the building of the yield prediction model includes: extracting historical feature data of different operation links according to historical operation data in the training data set; taking the corresponding historical measured yield as a supervision signal, training the random forest model according to all the historical feature data collected in different operation links to obtain the yield prediction model. Then, the yield prediction model is called, and the historical feature data extracted from each historical operation link in the test data set is used for yield prediction; the prediction error is confirmed according to the deviation between the predicted yield and the measured yield, and the influence weight of the feature data extracted from different operation links on yield prediction is confirmed; the yield prediction model is corrected based on the influence weight.
[0049] In step S11, all the extracted feature data is input into the yield prediction model to obtain the predicted yield, which can not only realize yield prediction in different operation links, but also input the data extracted from the current operation link and the data extracted from the previous operation link into the model to improve the accuracy of yield prediction. In specific implementation, the feature data of each link can be normalized and processed, and different weights can be set for the data of each link in the model to accurately reflect the influence of different links on yield prediction.
[0050] The crop yield prediction method based on whole-process mechanized operation data provided in the present application can solve the problem of inaccurate prediction in the current technology which simply relies on experience and initial conditions for crop yield prediction. The present application collects operation data when performing operation actions in different growth periods of crops, extracts feature data therefrom, inputs the feature data extracted from previous operation links into a yield prediction model for yield prediction, realizes dynamic yield prediction according to different operation links, and improves the prediction accuracy. By using the technical solution, the dynamic changes of crop growth are considered, different types of operation data are collected in different operation links, and feature data capable of representing crop yield are extracted therefrom for yield prediction. The present application not only can realize yield prediction in different growth periods of crops, but also can comprehensively predict yield by combining the data collected in previous operation links, thereby improving the prediction accuracy.
[0051] The above provides a detailed description of the crop yield prediction method based on data from fully mechanized operations provided by this invention. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. It should be noted that those skilled in the art can make various improvements and modifications to this invention without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this invention.
Claims
1. A method for predicting crop yield based on data from fully mechanized operations, characterized in that, include: Collect operational data from the current operational stage and extract feature data for that stage. The aforementioned operational steps are those that involve actions performed according to the crop growth cycle; The current stage feature data, combined with historical stage feature data extracted from historical operation stages, is input into the yield prediction model to predict crop yield; wherein, the yield prediction model is a model established based on the relationship between feature data in different operation stages and the final crop yield.
2. The crop yield prediction method based on data from fully mechanized operations according to claim 1, characterized in that, The operational steps include: tillage, sowing, fertilization, irrigation, and plant protection. The operational data for the tillage stage is the total energy consumption of mechanical operations; the operational data for the sowing stage is the total sowing amount of mechanical operations; the operational data for the fertilization stage is the total fertilization amount of mechanical operations; the operational data for the irrigation stage is the total irrigation amount of mechanical operations; and the operational data for the plant protection stage is the total spraying amount of mechanical operations.
3. The crop yield prediction method based on data from fully mechanized operations according to claim 2, characterized in that, The formula for calculating characteristic data in the tillage stage is: ; in, This is the tillage load coefficient. Energy consumption per unit area , This represents the total energy consumption of mechanized operations. This refers to the actual area covered by the operation. Energy consumption per unit area as a baseline; The formula for calculating the characteristic data of the sowing stage is: ; in, This is the sowing difference coefficient. This refers to the seeding rate per unit area. ,in This represents the total seeding amount for mechanized operations. Recommended seeding rate per unit area; The formula for calculating characteristic data of the fertilization process is: ; in, This is the fertilization deviation coefficient. Fertilizer application intensity per unit area ,in This represents the total amount of fertilizer applied by mechanical operations. Recommended fertilizer application rate per unit area; The formula for calculating characteristic data of irrigation is as follows: ; in, This is the moisture difference coefficient. The intensity of water supply per unit area. ,in This represents the total irrigation volume achieved through mechanized operations. Water requirement per unit area of crop; The formula for calculating characteristic data in the plant protection process is as follows: ; in, To prevent and control the intensity coefficient, The effective equivalent intensity of plant protection per unit area, , This represents the total amount of pesticide sprayed by the mechanized operation. This refers to the concentration of the effective components in the drug solution. Recommended drug intensity per unit area.
4. The crop yield prediction method based on data from fully mechanized operations according to any one of claims 1 to 3, characterized in that, The production prediction model is built by extracting historical feature data of different operational stages based on historical operational data in the training dataset. The corresponding historical measured output is used as a supervision signal, and the random forest model is trained based on all historical feature data collected from different operation stages to obtain the output prediction model.
5. The crop yield prediction method based on data from fully mechanized operations according to claim 4, characterized in that, Also includes: The production prediction model is invoked to predict production based on historical feature data extracted from each historical operation stage in the test dataset. The prediction error is determined by the deviation between the predicted output and the actual output, and the influence weight of the feature data extracted from different operation stages on the output prediction is also determined. The output prediction model is modified based on the influence weights.
6. The crop yield prediction method based on data from fully mechanized operations according to claim 1, characterized in that, Also includes: The work area is divided into grids according to a fixed side length ratio; Data collection is performed based on the divided grid.
7. The crop yield prediction method based on data from fully mechanized operations according to claim 6, characterized in that, Before the step of extracting the feature data of the current stage, the method also includes: filtering out abnormal data in the operation data.