Intelligent fertilization system driven by vegetable nutrient demand analysis

Abstract

The invention provides an intelligent fertilization system driven by vegetable nutrient demand analysis, and relates to the technical field of data processing.The intelligent fertilization system is characterized in that a soil area for planting target vegetables is obtained through a historical planting round determination module, and historical planting rounds are determined; the historical record acquisition module acquires fertilization record information and soil record information corresponding to each planting round; a prediction model training module performs model training according to the recorded information to generate a planting round-soil fertility change prediction model; the soil fertility matching module obtains nutrient demand data to perform soil fertility matching, and outputs a preset soil fertility threshold value; and the fertilization guidance module drives the prediction model to output a predicted planting round smaller than a preset soil fertility threshold value, and soil fertility reminding is carried out. The technical problems that the fertilization efficiency is low and the precision is insufficient due to the fact that historical planting data cannot be deeply excavated and utilized in the existing fertilization process are solved, the scientificity and the precision of the fertilization process are improved, and meanwhile the fertilization efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to an intelligent fertilization system driven by vegetable nutrient demand analysis. Background Art

[0002] In the process of vegetable planting, the fertility level in the soil directly affects the nutrient supply of vegetables, and the scientificity and accuracy of fertilization determine the growth quality and yield of crops. The existing fertilization process generally requires soil sampling and fertility testing before each planting to determine the appropriate amount of fertilizer. Frequent soil testing consumes a lot of time and human resources, increasing production costs. Secondly, the fertility and properties of the soil will change significantly after multiple cultivations. This change is not only affected by the type of crops planted before, but also closely related to the type and amount of fertilization in the past. The existing fertilization process relies on the experience of growers or is based on the results of a single soil test, while ignoring the long-term impact of historical fertilization records on soil fertility, which can easily lead to inaccurate fertilization, over-fertilization or insufficient nutrients. In addition, the existing methods lack the ability to predict future changes in soil fertility. Growers can only continuously adjust fertilization plans during crop growth, resulting in delayed fertilization decisions and affecting the healthy growth of crops. Summary of the invention

[0003] The present application provides an intelligent fertilization system driven by vegetable nutrient demand analysis, which solves the technical problem that the existing fertilization process relies on the grower's experience or soil test results, fails to deeply explore and utilize historical planting data, resulting in low fertilization efficiency and insufficient fertilization accuracy. It achieves the technical effect of improving the scientificity and accuracy of the fertilization process while improving fertilization efficiency.

[0004] In view of the above problems, the present application provides an intelligent fertilization system driven by vegetable nutrient demand analysis, the system comprising: a historical planting cycle determination module, the historical planting cycle determination module is used to obtain the soil area for planting target vegetables and determine the historical planting cycles of the soil area; a historical record collection module, the historical record collection module is used to collect the corresponding fertilization record information and soil record information under each planting cycle; a prediction model training module, the prediction model training module is used to perform model training according to the fertilization record information and the soil record information, and generate a planting cycle-soil fertility change prediction model; a soil fertility matching module, the soil fertility matching module is used to obtain the nutrient demand data of the target vegetables, perform soil fertility matching according to the nutrient demand data, and output a preset soil fertility threshold; a fertilization guidance module, the fertilization guidance module is used to drive the planting cycle-soil fertility change prediction model, output a predicted planting cycle that is less than the preset soil fertility threshold, and provide a soil fertility reminder based on the predicted planting cycle.

[0005] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0006] The historical planting cycle determination module obtains the soil area for planting target vegetables, determines the historical planting cycles of the soil area, and provides a data basis for subsequent fertilization decisions. The historical record collection module collects the corresponding fertilization record information and soil record information under each planting cycle, and provides comprehensive data support for model training. The prediction model training module performs model training according to the fertilization record information and the soil record information, and generates a planting cycle-soil fertility change prediction model that accurately predicts soil fertility, avoiding frequent soil sampling before planting and reducing operating costs. The soil fertility matching module obtains the nutrient demand data of the target vegetables, performs soil fertility matching according to the nutrient demand data, outputs a preset soil fertility threshold, and then accurately controls the amount of fertilizer used to avoid excessive or insufficient fertilization. The fertilization guidance module drives the planting cycle-soil fertility change prediction model, outputs a predicted planting cycle that is less than the preset soil fertility threshold, and performs soil fertility reminders according to the predicted planting cycle to guide the grower's fertilization operation to ensure that the soil fertility level is always within the optimal range for the growth of the target crop.

[0007] To sum up, this application collects and analyzes historical planting and fertilization data, uses prediction models to predict changes in soil fertility in advance, and performs intelligent fertilization matching and guidance based on the nutrient requirements of vegetables to ensure that crops get the right amount of nutrients. At the same time, it avoids the complex operation of frequently collecting soil fertility data in the traditional fertilization process, significantly improves the accuracy and efficiency of fertilization, and greatly improves the efficiency and scientificity of agricultural production.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A schematic diagram of the structure of an intelligent fertilization system driven by vegetable nutrient demand analysis provided in an embodiment of the present application;

[0010] Figure 2 A schematic diagram of a process for dividing soil areas according to the types of vegetables planted and obtaining multiple soil sub-areas in an intelligent fertilization system driven by vegetable nutrient demand analysis provided in an embodiment of the present application;

[0011] Figure 3A schematic diagram of the process of model training based on fertilization record information and soil record information in an intelligent fertilization system driven by vegetable nutrient demand analysis provided in an embodiment of the present application.

[0012] Explanation of the reference numerals: historical planting cycle determination module 10 , historical record collection module 20 , prediction model training module 30 , soil fertility matching module 40 , fertilization guidance module 50 . DETAILED DESCRIPTION

[0013] The embodiment of the present application solves the technical problems that the existing fertilization process relies on the grower's experience or soil test results, fails to deeply explore and utilize historical planting data, resulting in low fertilization efficiency and insufficient fertilization accuracy, by providing an intelligent fertilization system driven by vegetable nutrient demand analysis. This achieves the technical effect of improving the scientificity and accuracy of the fertilization process while improving fertilization efficiency.

[0014] like Figure 1 As shown, the embodiment of the present application provides an intelligent fertilization system driven by vegetable nutrient demand analysis, the system comprising:

[0015] The historical planting cycle determination module 10 is used to obtain the soil area for planting target vegetables and determine the historical planting cycles of the soil area.

[0016] Specifically, historical cropping cycles refer to the cycles and periods of different crops planted in a specific soil area in the past. Different cropping cycles will have a cumulative impact on soil fertility, so understanding historical cycles helps to better understand the dynamic changes of soil fertility.

[0017] The historical planting round determination module 10 obtains relevant information about the planting area where the target vegetables are planned to be planted, especially the records of planting activities carried out on this land in the past. For example, if potatoes, tomatoes and carrots are planted in turn on a plot of land, the time, crop type and duration of each round of planting need to be recorded. Agricultural management databases, GPS positioning technology, and agricultural Internet of Things devices can be used to accurately track and record the planting history of different plots of land. By obtaining this information, the historical planting conditions of the soil area can be clearly understood, which provides the necessary data basis for the subsequent analysis of soil fertility changes in the area and the prediction of future fertilization needs.

[0018] The historical record collection module 20 is used to collect the corresponding fertilization record information and soil record information in each planting cycle.

[0019] Specifically, the historical record collection module 20 collects fertilization and soil-related data in different planting cycles, including records of information such as fertilizer type, fertilizer amount, and fertilization time in each planting cycle, i.e., fertilization record information; and physical and chemical properties of the soil, such as soil pH, organic matter content, nitrogen, phosphorus, and potassium content, i.e., soil record information. For example, when planting potatoes in the first round, record how much nitrogen fertilizer is used and the organic matter content of the soil during the planting process. When planting tomatoes in the second round, record the changes in fertilization and soil. Sensors, agricultural management systems, and databases can be used to collect and manage this information. These record information provides important historical data for the prediction model training module 30, reflecting the long-term effects of different fertilization strategies on soil fertility, thereby providing an accurate reference for prediction model training.

[0020] The prediction model training module 30 is used to perform model training according to the fertilization record information and the soil record information to generate a planting cycle-soil fertility change prediction model.

[0021] Specifically, the prediction model training module 30 uses the historical data collected by the historical record collection module 20 to train a prediction model to predict future changes in soil fertility. For example, if a large amount of phosphorus fertilizer was applied when tomatoes were planted in a certain area in the past, the model will learn the impact of this fertilization behavior on the phosphorus content of the soil and predict how the phosphorus content in the soil will change after the next planting. By training a large amount of historical data, the model is automatically adjusted and optimized to predict the fertility level of the soil after a certain crop is planted in the future.

[0022] By training the model, the changing trend of soil fertility in future planting cycles can be predicted based on the existing data, so that the fertilization process can be adjusted predictively based on historical data. Frequent soil sampling is no longer required, which reduces operating costs and improves the accuracy and scientific nature of fertilization.

[0023] The soil fertility matching module 40 is used to obtain the nutrient requirement data of the target vegetables, perform soil fertility matching according to the nutrient requirement data, and output a preset soil fertility threshold.

[0024] Specifically, nutrient requirement data refers to the nutrients and quantitative data required by vegetables during their growth, such as the content of nitrogen, phosphorus, potassium and other elements. The preset soil fertility threshold is the ideal soil fertility standard set by the nutrient requirement data of the target vegetables. When it is lower than this value, it means that fertilization is needed.

[0025] The soil fertility matching module 40 first obtains the types of essential nutrients and recommended intakes of target vegetables from the agricultural database. For example, tomatoes require higher potassium and phosphorus, but relatively lower nitrogen requirements. According to the nutrient requirements of the target vegetables, the soil fertility standards required in the target vegetable planting process are matched, and the preset soil fertility threshold is output according to this standard. For example, the nitrogen, phosphorus, and potassium content in the soil are calculated.

[0026] Reasonable soil fertility thresholds can be set based on the nutrient requirements of vegetables, so that the amount of fertilizer used can be accurately controlled according to the current soil fertility conditions, ensuring that vegetables receive sufficient nutrient support during the growth process and avoiding excessive or insufficient fertilization.

[0027] The fertilization guidance module 50 is used to drive the planting cycle-soil fertility change prediction model, output a predicted planting cycle that is less than the preset soil fertility threshold, and provide soil fertility reminders based on the predicted planting cycle.

[0028] Specifically, the fertilization guidance module 50 predicts future soil fertility changes at the end of a planting cycle by driving the previously trained planting cycle-soil fertility change prediction model. If the prediction results show that the soil fertility of a certain cycle is lower than the preset soil fertility threshold, for example, the nitrogen content is too low, a corresponding fertilization suggestion is generated to remind the grower to add nitrogen fertilizer in time to ensure that the soil fertility level is always within the optimal range for the growth of the target crop. This prediction reminder mechanism avoids the lag and blindness of fertilization decisions and improves the efficiency and accuracy of fertilization.

[0029] Furthermore, the system described in the embodiment of the present application is also used to perform the following steps:

[0030] Access a digital soil management system to obtain the location of the soil area; based on the location of the soil area, determine whether the target vegetables have been continuously planted in the soil area's historical planting cycles, and if the target vegetables have been continuously planted in the soil area, obtain a global training instruction; based on the global training instruction, perform global model training on the fertilization record information and the soil record information to generate a planting cycle-soil fertility change prediction model.

[0031] Specifically, a digital soil management system is a system that uses digital technology to monitor and manage soil information, which usually includes data on the physical, chemical and biological properties of the soil, as well as planting history, fertilizer usage, etc. The system uses sensors, GPS positioning, data analysis and other tools to achieve real-time monitoring and historical analysis of soil status.

[0032] First, access the digital soil management system to obtain the location information of the soil area where the target vegetables are planted, including geographic coordinate information or spatial area range, so as to call the historical planting information and soil data of the plot. According to the historical planting records of the soil area, determine whether the target vegetable variety has been planted continuously in the soil area in the past multiple planting cycles. If it is continuous planting, it may have a cumulative effect on soil fertility and nutrient balance. For example, if a plot of land has been planted with tomatoes for three consecutive years, it is marked as continuous planting. Continuous planting of tomatoes may lead to nitrogen and potassium depletion in the soil.

[0033] If it is detected that a certain soil area has continuously planted target vegetables, a global training instruction is issued to trigger a large-scale model training of the historical fertilization and soil data of the entire soil area to generate a more comprehensive prediction model. Through the global training instruction, the model training process uses not only the data of the current planting cycle, but also the historical data of the entire soil area, especially the fertilization records and the long-term trend of soil changes, to conduct a comprehensive global model training, and finally generate a planting cycle-soil fertility change prediction model covering a larger time span and spatial range. For example, the fertilization, crop growth and soil fertility change data of the soil area in the past ten years can be used to predict the fertility change trend in the next few years.

[0034] For example, a plot of land has been continuously planted with the same vegetables (such as tomatoes) for the past five years. Through the digital soil management system, the geographical location of the plot is obtained, and its continuous planting situation is determined. Subsequently, global training is triggered, and all fertilization data and soil fertility data (such as nitrogen content, phosphorus content, and potassium content) are used to train the model to predict possible changes in soil fertility in the future. This predictive model will help growers develop more accurate fertilization plans in the next round of tomato planting, thereby avoiding nutrient imbalance problems caused by long-term planting of the same crops.

[0035] Through global training, more data dimensions can be integrated, including not only the records of a certain round of fertilization, but also the long-term changes in soil fertility and the cumulative impact of planting behaviors on the soil. The trained prediction model can guide future fertilization decisions more comprehensively and accurately.

[0036] Furthermore, the system described in the embodiment of the present application is also used to perform the following steps:

[0037] If the target vegetables are not planted continuously in the soil area, the soil area is divided according to the types of vegetables planted to obtain multiple soil sub-areas, wherein each soil sub-area corresponds to the same type of vegetables planted; based on the multiple soil sub-areas, multiple fertilization record information and multiple soil record information are obtained; based on the multiple fertilization record information and multiple soil record information, training is performed separately to output multiple planting cycle-soil fertility change prediction models; and based on the multiple planting cycle-soil fertility change prediction models, fertilization is performed on the multiple soil sub-areas in different regions.

[0038] Specifically, if the soil area where the target vegetable is planted is detected to have different types of vegetables planted in past planting cycles instead of continuously planting the same vegetables, the soil area will be divided into multiple soil sub-areas based on the types of vegetables planted in the past. For example, if a plot of land was planted with tomatoes, potatoes, and carrots at the same time in historical planting cycles, the soil area will be divided into three sub-areas, each corresponding to a specific type of vegetable.

[0039] Multiple fertilization records and soil records in these sub-areas are obtained respectively. Each soil sub-area has its own independent fertilization records and soil fertility change data. For example, the sub-area where tomatoes are planted may use more nitrogen fertilizer, while the sub-area where carrots are planted may use more potassium fertilizer. These data are used for subsequent model training.

[0040] Since each sub-region has its own unique fertilization pattern and soil changes, independent training models are needed to reflect these differences. According to the fertilization and soil data of each sub-region, training is carried out separately to generate multiple planting cycle-soil fertility change prediction models. These models can be used to predict future soil fertility changes in different sub-regions. For each different soil sub-region, based on its respective planting cycle-soil fertility change prediction model, the different soil fertility conditions of each sub-region in the future are predicted, and precise fertilizer placement is implemented in different regions. Apply an appropriate amount of nitrogen fertilizer to the tomato area, supplement phosphorus fertilizer to the potato area, and increase potassium fertilizer to the carrot area. Through this differentiated fertilization, it is ensured that the crops in each area can obtain the most suitable nutrition, improve the accuracy of the fertilization process, and optimize the nutrient distribution of the soil.

[0041] Further, such as Figure 2 As shown, the system described in the embodiment of the present application is also used to perform the following steps:

[0042] Obtain the total planting area of ​​the soil area; obtain the planting areas of multiple regions corresponding to the multiple soil sub-regions; calculate the proportional coefficients of the planting areas of the multiple regions to the total planting area, retain N soil sub-regions that are greater than or equal to a preset proportional coefficient, and output N planting cycle-soil fertility change prediction models, where N is a positive integer greater than 1.

[0043] Specifically, the total planting area of ​​the soil area is obtained, that is, the total area of ​​the entire plot that can be used for planting. Then, the planting area of ​​each soil sub-area is obtained separately. For example, a farmland has a total area of ​​5 hectares. Soil sub-area 1 (tomato) occupies 2 hectares, sub-area 2 (potato) occupies 1 hectare, and sub-area 3 (carrot) occupies 2 hectares.

[0044] Next, calculate the proportion coefficient of the planting area of ​​each region to the total area, that is, the proportion of the area of ​​each soil sub-region to the total planting area. For example, the area of ​​the tomato area is 2 hectares and the total area is 5 hectares, then the proportion coefficient is 2 / 5, that is, 40%. Set a preset proportion coefficient to retain soil sub-regions with larger areas, compare the proportion coefficient corresponding to each region with the preset proportion coefficient, and retain N soil sub-regions greater than or equal to the preset proportion coefficient, where N is a positive integer greater than 1. Model training is performed separately according to these N soil sub-regions to generate N planting cycle-soil fertility change prediction models corresponding to each soil sub-region. For example, the preset proportion coefficient is 20%. The proportion coefficient of the tomato area is 40%, the proportion coefficient of the potato area is 20%, and the proportion coefficient of the carrot area is 40%. Their area proportions are all greater than or equal to the preset 20%, then all three sub-regions will be retained.

[0045] By screening the sub-region area, resources and energy can be concentrated on those soil sub-regions that have a greater impact on the overall yield, thereby more effectively optimizing fertilization strategies and improving crop yield and quality.

[0046] Further, such as Figure 3 As shown, the prediction model training module 30 described in the embodiment of the present application is also used to perform the following steps:

[0047] Establish a Markov chain distribution between each historical planting cycle, the fertilization record information and the soil record information, wherein the fertilization record information is an incentive variable and the soil record information is a transfer variable; input the nutrient requirement data of the target vegetable, perform an initial matching evaluation on the soil record information, and obtain a soil fertility label; perform state transfer training based on the Markov chain distribution and the soil fertility label to generate a planting cycle-soil fertility change prediction model.

[0048] Specifically, the relationship between historical planting cycles and fertilization and soil record information is established through Markov chain distribution. Specifically, each planting cycle can be regarded as a "state" in the Markov chain, and each fertilization is an incentive variable that affects the state change, that is, a variable that affects the state, and the soil record information is a transfer variable that describes the change of the soil state after each fertilization behavior. The Markov chain assumes that the current state only depends on the previous state and is not affected by the previous state, so as to describe the changes in soil fertility under different planting cycles and fertilization strategies. For example, in the first round of planting, fertilization behavior (type and amount of fertilizer applied) will affect the nutrient status of the soil, and then affect the soil fertility status in the next round.

[0049] Input the nutrient requirement data of the target vegetables, such as the nitrogen, phosphorus and potassium content required by tomatoes, and perform an initial matching assessment of the soil record information under each historical planting cycle based on these data. By comparing the current soil fertility with the requirements of the target vegetables, a preliminary fertility status is obtained and a soil fertility label is generated, such as "high fertility" or "low fertility". The soil fertility label is used to indicate whether the current soil is suitable for planting the target crop.

[0050] The generated Markov chain distribution and soil fertility labels are used for state transition training. The specific training process includes: defining various possible states of the soil according to the soil fertility labels, such as high fertility state, medium fertility state, and low fertility state. These states are associated with the actual fertility level of the soil, nutrient content (such as nitrogen, phosphorus, potassium), and other chemical properties of the soil. Through the Markov chain distribution between each historical planting cycle, fertilization record information and soil record information, the fertility state transition probability is continuously adjusted until convergence. The transition probability reflects the possibility of soil fertility changes under the current state. When the state transition probability is basically stable, the model training is completed. Output the planting cycle-soil fertility change prediction model. Through this training, the generated planting cycle-soil fertility change prediction model can predict how soil fertility changes under different fertilization conditions and planting cycles, and provide accurate guidance for planting work.

[0051] Furthermore, the fertilization guidance module 50 described in the embodiment of the present application is also used to perform the following steps:

[0052] The planting cycle-soil fertility change prediction model is driven to perform prediction, and the probability that the soil fertility in each future planting cycle is less than the preset soil fertility threshold is obtained; the planting cycle whose probability reaches the expected probability is output as the predicted planting cycle.

[0053] Specifically, the fertilization guidance module 50 drives the planting cycle-soil fertility change prediction model, uses the previously trained model to simulate future planting conditions, and predicts soil fertility changes in each planting cycle. For example, the change trend of soil fertility at different time points such as the first planting round and the second planting round in the future can be predicted.

[0054] The planting cycle-soil fertility change prediction model outputs the probability that the soil fertility in each future planting cycle is lower than the preset threshold, thereby evaluating the risk of insufficient soil fertility at each time point. Set an expected probability. Only when the probability of insufficient soil fertility in a certain planting cycle reaches or exceeds this expected value will the planting cycle be further prompted or output as a potential high-risk cycle. The planting cycles with the probability reaching the expected probability are output as predicted planting cycles. These predicted planting cycles need to be fertilized to increase soil fertility to meet the nutrient requirements during the planting of target vegetables.

[0055] The fertilization guidance module 50 uses the prediction results of the planting cycle-soil fertility change prediction model to find the planting cycle targets that require fertilization adjustment, thereby guiding growers to increase fertilizer supply or adjust planting plans to achieve refined management during the planting process.

[0056] Furthermore, the historical planting round determination module 10 described in the embodiment of the present application is also used to perform the following steps:

[0057] Acquire multiple growth nodes of the target vegetable based on each planting cycle; and split and acquire fertilization record information and soil record information of each growth node; based on multiple growth nodes corresponding to each historical planting cycle, fertilization record information and soil record information of each growth node, acquire an updated Markov chain distribution.

[0058] Specifically, based on the existing historical planting cycles, the fertilization and soil data records for each key time point (i.e., growth node) in the vegetable growth process are further refined, and these refined data are integrated into the prediction model to improve the accuracy and precision of the model and achieve refined prediction of vegetable growth nodes.

[0059] Get multiple growth nodes of the target vegetable in each planting cycle. For example, a tomato may include different growth nodes such as the sowing period, germination period, flowering period and fruiting period in a planting cycle. Split each growth node and obtain the fertilization record information and soil record information corresponding to these nodes. Through the fertilization and soil data of these nodes, the soil state of each node is used as a state in the Markov chain, and the fertilization behavior of each node is used as an incentive variable to gradually transfer to the soil state of the next stage, so as to update the original Markov chain distribution. For example, soil changes after fertilization during the sowing period will affect the germination period, and fertilization during the flowering period will further change the soil fertility. The original Markov chain model only relies on the overall fertilization and soil data of each round of planting to predict fertility changes. With the addition of each growth node, the model can make more detailed predictions on the changes in soil fertility at each key time point.

[0060] For example, a tomato planting cycle includes four growth nodes: sowing period, germination period, flowering period and fruiting period. The fertilization and soil record information of each node is as follows:

[0061] Sowing period: Nitrogen fertilizer was applied, the nitrogen content in the soil increased and the pH value remained stable.

[0062] Germination period: Phosphorus fertilizer is applied and the phosphorus content in the soil increases.

[0063] Flowering period: Potassium fertilizer is applied, the potassium content of the soil increases, but the nitrogen content decreases.

[0064] Fruiting period: No fertilizer is applied and soil fertility gradually decreases.

[0065] Based on these data of each node, the soil state of each node is used as a state in the Markov chain, and the fertilization behavior of each node is used as the incentive variable to gradually transfer to the soil state of the next stage. Ultimately, through the updated Markov chain distribution, it is more accurate to predict how the soil fertility of each growth node will change in the next round of planting, and provide more precise fertilization recommendations.

[0066] Furthermore, the historical planting round determination module 10 described in the embodiment of the present application is also used to perform the following steps:

[0067] A soil fertility metabolism curve is obtained; and a step length for limiting the number of historical planting rounds is obtained according to the soil fertility metabolism curve.

[0068] Specifically, the historical planting cycle determination module 10 uses the soil fertility metabolism curve to further analyze the historical planting data and determine the historical planting cycle step length to obtain accurate and reliable model training data.

[0069] The soil fertility metabolism curve refers to the changing trend of soil fertility in different time periods. It can be obtained by describing the dynamic changes of soil fertility under the influence of various factors such as fertilization behavior, rainfall, and crop growth. For example, sensors are used to monitor the changes in the content of nutrients such as nitrogen, phosphorus, and potassium in the soil for a long time, and these data are plotted into a time series curve to show the dynamic changes in soil fertility. Through the soil fertility metabolism curve, the typical change pattern of soil fertility can be found.

[0070] Obtain soil fertility metabolic curves, and analyze the trend of soil fertility changes over time to find the key time points when soil fertility rises, stabilizes, or decreases in different planting cycles. Combined with the results of metabolic curve analysis, evaluate which historical planting cycles are most important for the current analysis, set an appropriate historical data analysis cycle or span, i.e., step length, to limit the number of planting cycles that need to be analyzed, avoid processing too much data, optimize computing resources, and analyze the most relevant planting cycles.

[0071] If the metabolic curve shows that the soil fertility changes significantly between some rounds, that is, the soil fertility changes drastically, a smaller step size can be selected to perform detailed analysis round by round. On the contrary, if the metabolic curve shows that the soil fertility changes little between some rounds, indicating that the soil fertility changes steadily, then a larger step size can be selected to skip some rounds without detailed analysis. By limiting the step size of the number of historical planting rounds, reasonably allocating computing resources, and concentrating on analyzing the data of important rounds, the accuracy of the prediction model is ensured while avoiding unnecessary computing burden.

[0072] In summary, the intelligent fertilization system driven by vegetable nutrient demand analysis provided in the embodiment of the present application has the following technical effects:

[0073] The historical planting cycle determination module 10 obtains the soil area for planting the target vegetables, determines the historical planting cycles of the soil area, and provides a data basis for subsequent fertilization decisions. The historical record collection module 20 collects the corresponding fertilization record information and soil record information under each planting cycle, and provides comprehensive data support for model training. The prediction model training module 30 performs model training according to the Markov chain based on the fertilization record information and the soil record information, and generates a planting cycle-soil fertility change prediction model that accurately predicts soil fertility, avoiding frequent soil sampling before planting and reducing operating costs. The soil fertility matching module 40 obtains the nutrient demand data of the target vegetables, performs soil fertility matching according to the nutrient demand data, outputs a preset soil fertility threshold, and then accurately controls the amount of fertilizer used to avoid excessive or insufficient fertilization. The fertilization guidance module 50 drives the planting cycle-soil fertility change prediction model, outputs a predicted planting cycle that is less than the preset soil fertility threshold, and performs soil fertility reminders according to the predicted planting cycle to guide the grower's fertilization operation to ensure that the soil fertility level is always within the optimal range for the growth of the target crop.

[0074] To sum up, this application collects and analyzes historical planting and fertilization data, uses prediction models to predict changes in soil fertility in advance, and performs intelligent fertilization matching and guidance based on the nutrient requirements of vegetables to ensure that crops get the right amount of nutrients. At the same time, it avoids the complex operation of frequently collecting soil fertility data in the traditional fertilization process, significantly improves the accuracy and efficiency of fertilization, and greatly improves the efficiency and scientificity of agricultural production.

[0075] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent fertilization system driven by vegetable nutrient demand analysis, characterized by: The system comprises: A historical planting cycle determination module, the historical planting cycle determination module is used to obtain a soil area for planting target vegetables and determine the historical planting cycles of the soil area; A historical record collection module, which is used to collect fertilization record information and soil record information corresponding to each planting cycle; A prediction model training module, the prediction model training module is used to perform model training according to the fertilization record information and the soil record information to generate a planting cycle-soil fertility change prediction model; A soil fertility matching module, which is used to obtain the nutrient requirement data of the target vegetables, perform soil fertility matching according to the nutrient requirement data, and output a preset soil fertility threshold; A fertilization guidance module is used to drive the planting cycle-soil fertility change prediction model, output a predicted planting cycle that is less than the preset soil fertility threshold, and provide soil fertility reminders based on the predicted planting cycle.

2. The intelligent fertilization system driven by vegetable nutrient demand analysis as claimed in claim 1, characterized in that: After obtaining the soil area for planting the target vegetable, the system is further used to perform the following steps: Accessing a digital soil management system to obtain the location of the soil area; According to the position of the soil area, determine whether the target vegetable has been planted continuously in the historical planting rounds of the soil area, and if the target vegetable has been planted continuously in the soil area, obtain a global training instruction; According to the global training instruction, global model training is performed on the fertilization record information and the soil record information to generate a planting cycle-soil fertility change prediction model.

3. The intelligent fertilization system driven by vegetable nutrient demand analysis as claimed in claim 2, characterized in that: The system is also configured to perform the following steps: If the target vegetables are not planted continuously in the soil area, the soil area is divided according to the types of vegetables planted to obtain a plurality of soil sub-areas, wherein each soil sub-area corresponds to the same type of vegetable planted; According to the plurality of soil sub-areas, obtaining a plurality of fertilization record information and a plurality of soil record information; Perform training respectively according to the plurality of fertilization record information and the plurality of soil record information, and output a plurality of planting cycle-soil fertility change prediction models; The multiple soil sub-regions are fertilized in different regions according to the multiple planting cycle-soil fertility change prediction models.

4. The intelligent fertilization system driven by vegetable nutrient demand analysis as claimed in claim 3, characterized in that: The system is also configured to perform the following steps: Obtaining the total planting area of ​​the soil region; Acquire multiple regional planting areas corresponding to the multiple soil sub-regions; Calculate the ratio coefficient of the planting area of ​​the multiple regions to the total planting area, retain N soil sub-regions that are greater than or equal to the preset ratio coefficient, and output N planting cycle-soil fertility change prediction models, where N is a positive integer greater than 1.

5. The intelligent fertilization system driven by vegetable nutrient demand analysis as claimed in claim 1, characterized in that: The prediction model training module is also used to perform the following steps: Establishing a Markov chain distribution between each historical planting round, the fertilization record information and the soil record information, wherein the fertilization record information is an incentive variable and the soil record information is a transfer variable; Input the nutrient requirement data of the target vegetable, perform initial matching evaluation on the soil record information, and obtain the soil fertility label; State transfer training is performed according to the Markov chain distribution and the soil fertility label to generate a planting cycle-soil fertility change prediction model.

6. The intelligent fertilization system driven by vegetable nutrient demand analysis as claimed in claim 5, characterized in that: The fertilization guidance module is also used to perform the following steps: Driving the planting cycle-soil fertility change prediction model to perform prediction, and obtaining the probability that the soil fertility in each future planting cycle is less than the preset soil fertility threshold; The planting round whose probability reaches the expected probability is output as the predicted planting round.

7. The intelligent fertilization system driven by vegetable nutrient demand analysis as claimed in claim 5, characterized in that: The historical planting round determination module is also used to perform the following steps: Obtaining multiple growth nodes of the target vegetable based on each planting cycle; And split and obtain the fertilization record information and soil record information of each growth node; Based on multiple growth nodes corresponding to each historical planting round, fertilization record information and soil record information of each growth node, an updated Markov chain distribution is obtained.

8. The intelligent fertilization system driven by vegetable nutrient demand analysis as claimed in claim 1, characterized in that: The historical planting round determination module is also used to perform the following steps: Obtain soil fertility metabolism curve; According to the soil fertility metabolism curve, a step length for limiting the number of historical planting rounds is obtained.

Images