Water and fertilizer integrated irrigation method and system based on plant nutrient dynamic tracking

Abstract

This application provides a water and fertilizer integrated irrigation method and system based on dynamic tracking of plant nutrients, relating to the field of alternating irrigation technology. The method includes: acquiring plant nutrient characterization parameters at multiple spatial locations within a target field, and constructing a nutrient spatial distribution model based on these parameters; calculating the nutrient deficit at each spatial location based on the nutrient spatial distribution model and a preset target nutrient baseline value, and generating a nutrient deficit distribution map; dividing the target field into several irrigation management zones with different nutrient requirements based on the nutrient deficit distribution map; generating variable fertilization instructions based on the nutrient deficit characteristics of each irrigation management zone, and controlling a three-level water and fertilizer integrated irrigation system according to these instructions to perform differentiated irrigation for the several irrigation management zones. This solves the technical problem of nutrient spatial heterogeneity caused by the expansion of field area in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of alternating irrigation technology, and in particular to an integrated water and fertilizer irrigation method and system based on dynamic tracking of plant nutrients. Background Technology

[0002] Integrated water and fertilizer technology combines irrigation and fertilization, using an irrigation system to precisely deliver fertilizer to the roots of crops, achieving coordinated water and fertilizer supply. This modern agricultural technology can effectively improve water and fertilizer utilization efficiency and reduce agricultural non-point source pollution.

[0003] However, with the expansion of individual field areas, the traditional uniform fertilization and irrigation methods have ignored the spatial heterogeneity of crop nutrient requirements within the field, resulting in excessive fertilization in some areas, causing fertilizer waste and environmental pollution, while insufficient fertilization in other areas affects crop yield.

[0004] Meanwhile, existing variable fertilization technologies are mostly based on soil nutrient sampling or remote sensing image inversion, lacking real-time tracking of the dynamic changes in plant nutrients. Furthermore, the division of irrigation areas is often disconnected from the spatial distribution characteristics of nutrients, making it difficult to achieve precise variable fertilization. Summary of the Invention

[0005] This invention addresses the technical problems in existing technologies, such as significantly increased spatial heterogeneity of nutrients in fields due to the expansion of individual plots, the inability of traditional uniform fertilization to match local differences in nutrient requirements, the lack of a real-time response mechanism for dynamic nutrient demand of plants in fertilization decisions, and the insufficient fine-tuning ability of existing fertigation systems to control fertilizer concentration, resulting in low fertilizer utilization efficiency, uneven growth between regions, and untapped crop yield potential. It provides a fertigation method and system based on dynamic tracking of plant nutrients.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides a water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients, comprising: Obtain plant nutrient characterization parameters at multiple spatial locations within the target field, and construct a nutrient spatial distribution model based on the plant nutrient characterization parameters; Based on the nutrient spatial distribution model and the preset target nutrient baseline value, the nutrient deficit at each spatial location is calculated, and a nutrient deficit distribution map is generated based on the nutrient deficit. Based on the nutrient deficit distribution map, the target field is divided into several irrigation management areas with different nutrient requirements. Initial variable fertilization instructions are generated based on the nutrient deficit characteristics of each irrigation management area, and the initial variable fertilization instructions are dynamically optimized to generate an optimized set of refined irrigation instructions. The intelligent irrigation system, laid in the field ditch network, is controlled according to the refined irrigation instruction set to perform differentiated irrigation for several irrigation management areas. The field ditch network is pre-constructed by the ridging and furrowing seeder used during sowing. The intelligent irrigation system includes a fertilizer applicator, a main conveying pipeline, a flexible conveying pipeline laid in the field ditch network, and intelligent switches with unique identification IDs installed at the openings of the flexible conveying pipelines.

[0007] Secondly, the present invention provides an integrated water and fertilizer irrigation system based on dynamic tracking of plant nutrients, comprising: The nutrient spatial division module is used to obtain plant nutrient characterization parameters at multiple spatial locations within the target field, and to construct a nutrient spatial distribution model based on the plant nutrient characterization parameters. The nutrient deficit calculation module is used to calculate the nutrient deficit at each spatial location based on the nutrient spatial distribution model and the preset target nutrient benchmark value, and to generate a nutrient deficit distribution map based on the nutrient deficit. The irrigation management area division module is used to divide the target field into several irrigation management areas with different fertilizer requirements based on the nutrient deficit distribution map. The irrigation instruction generation module is used to generate initial variable fertilization instructions based on the nutrient deficit characteristics of each irrigation management area, and to dynamically optimize the initial variable fertilization instructions to generate an optimized refined irrigation instruction set. The differentiated irrigation module is used to control the intelligent irrigation system laid in the field ditch network according to the refined irrigation instruction set, and to perform zoned differentiated irrigation for several irrigation management areas. The field ditch network is pre-constructed by the ridging and furrowing seeder used during sowing. The intelligent irrigation system includes a fertilizer applicator, a main conveying pipeline, a flexible conveying pipeline laid in the field ditch network, and an intelligent switch with a unique identification ID installed at the opening of the flexible conveying pipeline.

[0008] The beneficial effects of this invention are: Compared to existing technologies, this application first acquires plant nutrient characterization parameters at multiple spatial locations within the target field and constructs a nutrient spatial distribution model, transforming discrete sampling data into continuous spatial distribution information, thus achieving a comprehensive understanding of the nutrient status of the entire field. Second, based on the nutrient spatial distribution model and preset target nutrient baseline values, it calculates the nutrient deficit at each spatial location and generates a nutrient deficit distribution map, transforming abstract nutrient differences into visualized quantitative indicators, providing a scientific basis for precise decision-making. Third, based on the nutrient deficit distribution map, the target field is divided into several irrigation management zones with different fertilizer requirements, transforming the continuously changing nutrient spatial distribution into a continuous spatial distribution. The differences between irrigation areas are transformed into clearly defined and internally consistent operational units, solving the problem that traditional uniform management cannot respond to local differences in nutrient demand. Then, based on the nutrient deficit characteristics of each irrigation management area, initial variable fertilization instructions are generated and dynamically optimized by combining real-time monitoring data and nutrient leakage risk prediction to generate a refined irrigation instruction set. Finally, the intelligent irrigation system laid in the field ditch network is controlled according to the refined irrigation instruction set, and the intelligent switches with unique identification IDs are used to perform zoned differentiated irrigation for each irrigation management area, realizing intelligent control of the entire process from nutrient perception, deficit quantification, area segmentation, instruction optimization to precise execution.

[0009] Through the above technical solutions, this application constructs a complete closed loop based on the dynamic tracking of plant nutrients and the linkage of intelligent irrigation equipment. It effectively solves the technical problems of traditional uniform irrigation and fertilization ignoring spatial heterogeneity, coarse division of irrigation areas, and static commands being unable to adapt to dynamic environments. It improves water and fertilizer utilization efficiency and crop yield consistency, reduces environmental pollution caused by nutrient leakage, and provides reliable technical support for precision agriculture. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the process of the integrated water and fertilizer irrigation method based on dynamic tracking of plant nutrients provided by the present invention; Figure 2 This is a schematic diagram of the integrated water and fertilizer irrigation system based on dynamic tracking of plant nutrients provided by the present invention. Figure 3 , Figure 4 , Figure 5 Three views of the precision strip direct seeding machine provided by the present invention; Figure 6 , Figure 7 A schematic diagram of the field furrow network formed after sowing by the precision strip direct seeding machine provided by the present invention.

[0011] In the attached diagram, the components represented by each number are as follows: 1. Fertilizer and water ditch shaping plow, 2. Seeding pipe, 3. Water storage ditch shaping plow, 4. Ridge leveling board, 5. Central axis, 11. Nutrient space division module, 12. Nutrient deficit calculation module, 13. Irrigation management area division module, 14. Irrigation instruction generation module, 15. Differentiated irrigation module. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0014] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0015] Example 1, as Figure 1 As shown, this embodiment of the invention provides a water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients, including: S10: Obtain plant nutrient characterization parameters at multiple spatial locations within the target field, and construct a nutrient spatial distribution model based on the plant nutrient characterization parameters.

[0016] Because soil fertility, water distribution, and plant growth vary in different locations within the target field, adopting a uniform fertilization and irrigation model can easily lead to localized nutrient deficiencies or over-fertilization.

[0017] Therefore, this application first needs to obtain plant nutrient characterization parameters at multiple spatial locations within the target field, and then construct a nutrient spatial distribution model based on the plant nutrient characterization parameters, thereby realizing the quantitative expression of spatial differences in plant nutrients.

[0018] Specifically, step S10 in the method includes: The target field is divided into multiple sampling grids according to a preset sampling density, and the center point of each sampling grid is determined as the in-situ detection point; Multispectral image data covering the target field is collected by a multispectral imaging device carried by a drone, and combined with field measurement data collected at each in-situ detection point, the plant nutrient characterization parameters corresponding to each in-situ detection point are determined. The plant nutrient characterization parameters include at least the normalized vegetation index and the relative chlorophyll content. The spatial coordinate data of each in-situ detection point is obtained, and the plant nutrient characterization parameters and corresponding spatial coordinate data of the multiple sampling grids are used as input. A spatial interpolation algorithm is used to interpolate and extrapolate the unsampled area to generate a nutrient spatial distribution model that covers the entire target field and has a continuous spatial distribution.

[0019] In this embodiment, the target field is first divided into multiple sampling grids according to a preset sampling density, and the center point of each sampling grid is determined as an in-situ detection point. Here, sampling density refers to the number of sampling points deployed per unit area, which can be predetermined by those skilled in the art based on the size of the target field, the complexity of the terrain, and the required accuracy. A sampling grid is a set of rectangular or square units obtained by regularly dividing the target field. An in-situ detection point refers to the actual spatial location where data is collected, and the center point of each sampling grid is determined as an in-situ detection point.

[0020] For example, for a paddy field with an area of ​​100 mu, if the preset sampling density is 4 sampling points per mu, the target field is divided into 400 sampling grids of equal size. The size of each grid is approximately 12.9 meters × 12.9 meters (estimated at approximately 667 square meters per mu). The center point of each grid is an in-situ detection point. Finally, a total of 400 in-situ detection points are evenly distributed in the target field.

[0021] Secondly, multispectral imagery data covering the target field is collected using a drone equipped with a multispectral imaging device. This data is then combined with field measurement data collected at each in-situ monitoring point to determine the plant nutrient characterization parameters corresponding to each monitoring point. These parameters include at least the Normalized Difference Vegetation Index (NDVI) and relative chlorophyll content. Specifically, a drone is used as a remote sensing platform, equipped with a multispectral imaging device, to cruise and scan the target field along a pre-set flight path based on its geographical location characteristics, acquiring multispectral imagery data covering the entire field. Simultaneously, technicians enter the field with measuring equipment to collect plant samples or perform in-situ measurements at each monitoring point, obtaining field measurement data. By combining multispectral imagery data with field measurement data, the accuracy and reliability of the plant nutrient characterization parameters can be improved.

[0022] Among them, the Normalized Difference Vegetation Index (NDVI) is a vegetation index calculated by red light band reflectance and near-infrared band reflectance. It is a commonly used remote sensing indicator that reflects the vegetation growth status and nitrogen nutrition level. The relative chlorophyll content refers to the relative value obtained by measuring leaves using a portable chlorophyll meter. It is usually expressed as SPAD value. This parameter is highly correlated with the nitrogen concentration in leaves.

[0023] Finally, the spatial coordinate data of each in-situ detection point is acquired. Using the plant nutrient characterization parameters and corresponding spatial coordinate data from multiple sampling grids as input, a spatial interpolation algorithm is employed to interpolate and extrapolate the unsampled areas, generating a nutrient spatial distribution model covering the entire target field with continuous spatial distribution. The spatial coordinate data refers to the geographic coordinates of each in-situ detection point, which can be obtained through a drone positioning system or a handheld GPS device with centimeter-level accuracy. The spatial interpolation algorithm is a mathematical method that extrapolates the values ​​of unknown areas based on the values ​​of known sampling points. Optionally, spatial interpolation algorithms such as Kriging interpolation, inverse distance weighted interpolation, and spline function interpolation can be selected.

[0024] For example, using the plant nutrient characterization parameters and their corresponding spatial coordinates from 400 in-situ detection points as input, the spatial interpolation algorithm analyzes the variogram between known points based on the principle of spatial correlation to calculate the parameter values ​​at any unsampled location within the target field. After interpolation, a nutrient spatial distribution model covering the entire target field with continuous spatial distribution is finally generated. This nutrient spatial distribution model can be represented as a two-dimensional raster dataset that completely corresponds to the geographic space of the target field, with each raster cell having a corresponding nutrient characterization parameter value.

[0025] It should be noted that the above spatial interpolation algorithm is a prior art known in the field. The implementation details not described in this application, such as the fitting method of the variogram, the selection strategy of the search neighborhood, and the cross-validation of the interpolation accuracy, can be reasonably selected and configured by those skilled in the art based on the actual application scenario, field characteristics and accuracy requirements, by consulting relevant technical literature or referring to existing software implementation schemes, without the need for creative labor.

[0026] Furthermore, the phrase "collecting multispectral image data covering the target field using a multispectral imaging device mounted on a drone, and combining it with field measurement data collected at each in-situ detection point to determine the plant nutrient characterization parameters corresponding to each in-situ detection point" includes: The drone is controlled to cruise and scan the target field along a preset route, and multispectral image data covering the entire target field is collected by the multispectral imaging device. The multispectral image data includes at least red band reflectance and near-infrared band reflectance. The acquired multispectral image data is geometrically corrected and radiometrically calibrated to generate standardized multispectral images with geographic coordinate information. Based on the spatial coordinates of each in-situ detection point, the red band reflectance and near-infrared band reflectance at the corresponding locations are extracted from the standardized multispectral image. The normalized vegetation index is calculated based on the red light band reflectance and the near-infrared band reflectance. Plant samples were collected at each in-situ detection point, and the relative chlorophyll content at each in-situ detection point was obtained using a portable chlorophyll meter.

[0027] In this embodiment, the UAV is first controlled to cruise and scan the target field along a preset flight path, and multispectral image data covering the entire target field is collected using a multispectral imaging device. The multispectral image data includes at least red-band reflectance and near-infrared-band reflectance. Specifically, before takeoff, a preset flight path is planned based on the boundary coordinates of the target field, ensuring that the preset flight path completely covers the target field with a certain overlap rate to guarantee image stitching quality. After takeoff, the UAV automatically flies along the preset flight path, and the onboard multispectral imaging device continuously collects multispectral image data during flight.

[0028] Among them, multispectral imaging equipment can simultaneously acquire images of multiple specific bands. For vegetation monitoring, the red band (center wavelength of about 660nm) and the near-infrared band (center wavelength of about 790nm) are key bands for calculating the normalized vegetation index. Therefore, multispectral image data must include reflectance data of at least these two bands.

[0029] Secondly, the acquired multispectral image data undergoes geometric correction and radiometric calibration to generate standardized multispectral images with geographic coordinate information. Geometric correction eliminates geometric distortions caused by factors such as UAV attitude changes, terrain undulations, and lens distortion, ensuring that each pixel in the image accurately corresponds to its actual location in real geographic space. Radiometric calibration converts the raw digital values ​​of the multispectral image data into surface reflectance, eliminating radiometric errors caused by factors such as lighting conditions, atmospheric effects, and differences in sensor response, thus ensuring the comparability of multispectral image data acquired at different times and locations.

[0030] For example, geometric correction can be achieved through ground control points or a positioning and orientation system mounted on the UAV. The positioning and orientation system is a high-precision positioning and attitude measurement device integrated on the UAV, comprising a Global Navigation Satellite System (GNSS) receiver and an Inertial Measurement Unit (INS). The GNSS receiver acquires the precise spatial coordinates of the UAV at the moment of exposure, while the INS simultaneously records the UAV's pitch, roll, and yaw angles. By fusing this data with multispectral imagery, the precise geographic coordinates of each pixel can be calculated, thus achieving high-precision registration between multispectral imagery and real geographic space.

[0031] For example, radiometric calibration can be achieved using standard whiteboard data acquired during data collection or pre-calibrated parameters in a laboratory. Reference data can be obtained by photographing a standard whiteboard with known reflectivity before or after drone takeoff; alternatively, rigorous radiometric calibration of the multispectral imaging equipment can be performed in a laboratory environment to obtain conversion factors that convert the raw digital values ​​into reflectivity. Applying these reference data or conversion factors to the multispectral image data allows the raw digital value of each pixel to be converted into a physically meaningful surface reflectivity value.

[0032] Thus, after two steps of geometric correction and radiometric calibration, a standardized multispectral image with geographic coordinate information is generated. This standardized multispectral image has both accurate geographic location information and true surface reflectance values, providing a high-quality data foundation for subsequent band extraction and vegetation index calculation.

[0033] Next, based on the spatial coordinates of each in-situ detection point, the red band reflectance and near-infrared band reflectance at the corresponding location are extracted from the standardized multispectral image. Specifically, the latitude and longitude coordinates of each in-situ detection point are input into the geographic information system, the corresponding pixel location is located on the standardized multispectral image, and the red band reflectance and near-infrared band reflectance of that pixel are read.

[0034] Optionally, for situations requiring high positioning accuracy, those skilled in the art can use methods such as bilinear interpolation or cubic convolution to interpolate and calculate the reflectance value corresponding to the precise location of the detection point from the values ​​of multiple surrounding pixels.

[0035] Furthermore, the Normalized Difference Vegetation Index (NDVI) is calculated based on the red and near-infrared reflectance. The formula for NDVI is: NDVI = (Near-infrared reflectance - Red reflectance) / (Near-infrared reflectance + Red reflectance). The NDVI ranges from -1 to 1. For vegetated areas, the NDVI is typically positive, and a higher value indicates more abundant vegetation and more sufficient nitrogen nutrition.

[0036] For example, if the near-infrared reflectance of a certain in-situ detection point is 0.45 and the red light reflectance is 0.15, then the normalized vegetation index (NVI) can be calculated using the above formula: Normalized Vegetation Index = (0.45 - 0.15) / (0.45 + 0.15) = 0.50. The NVI can be used as one of the plant nutrient characterization parameters for this in-situ detection point.

[0037] Finally, plant samples were collected at each in-situ detection point, and the relative chlorophyll content at each point was obtained using a portable chlorophyll meter. Specifically, technicians randomly selected several representative plants at each in-situ detection point and used the portable chlorophyll meter to measure the middle of the leaves. The portable chlorophyll meter calculates a relative value, commonly known as the SPAD value, by measuring the transmittance of the leaves in both red and near-infrared bands. The SPAD value is highly correlated with the chlorophyll content of the leaves, and consequently highly correlated with the nitrogen concentration in the leaves, serving as a direct indicator of the plant's nitrogen nutrition status. Optionally, the SPAD-502 model portable chlorophyll meter can be selected.

[0038] For example, the SPAD values ​​of five leaves were measured at each in-situ detection point using a portable chlorophyll meter, and the average value was taken as the relative chlorophyll content of that in-situ detection point. For instance, if the five SPAD values ​​at a certain in-situ detection point were 41.2, 42.5, 40.8, 41.9, and 42.1, the average value was 41.7, which was then taken as the relative chlorophyll content of that in-situ detection point.

[0039] In summary, compared to existing technologies, this application obtains plant nutrient characterization parameters at multiple spatial locations within the target field and constructs a nutrient spatial distribution model based on these parameters. This allows for precise acquisition of the spatial distribution of plant nutrients within the target field, enabling dynamic tracking of nutrient spatial distribution and providing accurate nutrient data support for subsequent differentiated irrigation and fertilization.

[0040] S20: Based on the nutrient spatial distribution model and the preset target nutrient baseline value, calculate the nutrient deficit at each spatial location, and generate a nutrient deficit distribution map based on the nutrient deficit.

[0041] In the process of integrated water and fertilizer irrigation based on dynamic tracking of plant nutrients, accurately quantifying the gap between plant nutrient supply and demand in different spatial locations within the target field is a key prerequisite for achieving differentiated water and fertilizer supply by region and improving the accuracy of irrigation and fertilization.

[0042] However, existing technologies struggle to accurately quantify the nutrient supply and demand gap among plants at different spatial locations within the target field, and cannot intuitively present the spatial distribution characteristics of nutrient deficits, resulting in a lack of precise data support for subsequent water and fertilizer regulation. Meanwhile, the nutrient spatial distribution model obtained in the aforementioned steps can comprehensively reflect the continuous spatial distribution of plant nutrients within the target field, providing a reliable foundation for quantifying the nutrient supply and demand gap.

[0043] Therefore, this application calculates the nutrient deficit at each spatial location based on the nutrient spatial distribution model and the preset target nutrient benchmark value, and generates a nutrient deficit distribution map based on the nutrient deficit, so as to solve the shortcomings of the prior art.

[0044] Specifically, step S20 in the method includes: Based on the variety of crops planted in the target field, their current growth stage, and the preset target yield, obtain the corresponding preset target nutrient baseline values; Traverse each spatial location in the nutrient spatial distribution model, perform a difference calculation between the nutrient characterization parameter value at each spatial location and the preset target nutrient benchmark value to obtain the corresponding nutrient deficit. A positive nutrient deficit indicates a nutrient surplus state, while a negative nutrient deficit indicates a nutrient deficit state. The nutrient deficit at all spatial locations is mapped according to their respective spatial coordinates to generate a two-dimensional raster dataset that completely corresponds to the geographic space of the target field. This two-dimensional raster dataset is used as a nutrient deficit distribution map.

[0045] In this embodiment, the corresponding preset target nutrient baseline value is first obtained based on the variety of crop planted in the target field, its current growth stage, and the preset target yield. The preset target nutrient baseline value refers to the standard nutrient level that the plant should achieve for a specific crop variety and at a specific growth stage in order to achieve the preset target yield.

[0046] Specifically, the first step is to determine the crop variety to be planted in the target field, such as Ningxiangjing 9 rice; then, determine the current growth stage, such as the tillering stage, jointing stage, or booting stage; finally, determine the preset target yield, such as 750 kg / mu. Based on these three factors, and through multi-year variety trial data, field calibration trials, or historical experience data, obtain the target normalized vegetation index and target relative chlorophyll content that the variety should have to achieve the target yield at this growth stage, and use these as the preset target nutrient baseline values.

[0047] For example, for the Ningxiangjing 9 rice variety, in order to achieve a target yield of 750 kg / mu during the peak tillering stage, the preset target nutrient baseline value can be set as a target normalized vegetation index of 0.75 and a target relative chlorophyll content of 42 SPAD.

[0048] Secondly, the model iterates through each spatial location in the nutrient spatial distribution model, calculating the difference between the nutrient representation parameter value at each location and the preset target nutrient baseline value to obtain the corresponding nutrient deficit. A positive nutrient deficit indicates a nutrient surplus, while a negative nutrient deficit indicates a nutrient deficit. Specifically, the nutrient spatial distribution model is treated as a two-dimensional raster dataset, with each raster cell having its corresponding nutrient representation parameter value. The model is then iterated raster-by-raster, and for each raster cell, its nutrient deficit is calculated.

[0049] The formula for calculating nutrient deficit is: Nutrient deficit = Nutrient characterization parameter value - Preset target nutrient baseline value. When nutrient deficit > 0, it means that the measured nutrient level of the grid cell is higher than the preset target nutrient baseline value, and it is in a nutrient surplus state, that is, the nutrient is excessive at this location; when nutrient deficit < 0, it means that the measured nutrient level is lower than the preset target nutrient baseline value, and it is in a nutrient deficit state, that is, the nutrient needs to be supplemented at this location; when nutrient deficit = 0, it means that the target level has been reached.

[0050] For example, if the normalized vegetation index (NVI) of a certain grid cell is 0.70, and the target NVI in the preset target nutrient baseline is 0.75, then the nutrient deficit of this grid cell is -0.05, indicating a nutrient deficit state of 5%. Conversely, if the NVI of another grid cell is 0.80, then its nutrient deficit is +0.05, indicating a nutrient surplus state of 5%.

[0051] Finally, the nutrient deficit values ​​at all spatial locations are mapped according to their corresponding spatial coordinates to generate a two-dimensional raster dataset that perfectly corresponds to the geographic space of the target field. This two-dimensional raster dataset serves as the nutrient deficit distribution map. Specifically, the nutrient deficit value of each raster cell is reassigned according to its original spatial coordinates to generate a new two-dimensional raster dataset. This new dataset has the exact same spatial extent, raster size, and geographic coordinate reference as the original nutrient spatial distribution model, but the value of each raster changes from the original nutrient representation parameter value to the nutrient deficit value. This newly generated two-dimensional raster dataset serves as the nutrient deficit distribution map.

[0052] The nutrient deficit distribution map visually displays the nutrient surplus and deficit status of various areas within the target field, serving as a direct basis for subsequent irrigation management area division and variable fertilization instruction generation. Furthermore, negative values ​​on the nutrient deficit distribution map indicate areas requiring focused nutrient supplementation, positive values ​​indicate areas requiring controlled fertilization, and values ​​near zero represent areas with adequate nutrient levels.

[0053] In summary, compared to existing technologies, this application calculates the nutrient deficit at each spatial location based on the aforementioned nutrient spatial distribution model and a preset target nutrient baseline value, and generates a nutrient deficit distribution map based on the nutrient deficit. This allows for precise quantification of the nutrient supply and demand gap for plants at each spatial location within the target field, intuitively presenting the spatial distribution pattern of nutrient deficit and surplus, and providing accurate and intuitive core data support for subsequent target field zoning and differentiated water and fertilizer supply.

[0054] S30: Based on the nutrient deficit distribution map, the target field is divided into several irrigation management areas with different nutrient requirements.

[0055] Traditional agricultural management often treats large fields as a homogeneous whole and applies fertilizer in a one-size-fits-all manner, ignoring the spatial differences in nutrients within the field caused by soil, topography, or previous management. This results in lodging due to over-fertilization in high-yield areas and reduced yields due to nutrient deficiency in low-yield areas.

[0056] To address the aforementioned issues, this application, based on the nutrient deficit distribution map, divides the target field into several irrigation management zones with different nutrient requirements. This transforms the abstract nutrient deficit distribution map into several irrigation management zones with consistent internal characteristics but significant differences among them. This provides a clear geospatial basis for generating targeted variable fertilization instructions, ensuring that differentiated irrigation by zone can be truly implemented.

[0057] Specifically, step S30 in the method includes: The nutrient deficit value of each grid cell in the nutrient deficit distribution map is used as the topographic elevation value of the grid cell to construct a nutrient deficit topographic map. Identify all local minima in the nutrient deficit topographic map and designate each local minima as a seed point, where each seed point represents the center location of a potential irrigation management area. Simultaneously simulate the flooding process of gradually rising water level starting from each seed point. During the rise of water level, the catchment area corresponding to each seed point gradually expands to the surrounding area. When the boundaries of the catchment areas corresponding to any two adjacent seed points are about to touch during the rise in water level, calculate the difference in nutrient deficit between the two adjacent seed points and compare the difference in nutrient deficit with a preset irrigation management difference threshold. If the difference in nutrient deficit is less than or equal to the irrigation management difference threshold, then the water collection areas corresponding to the two adjacent seed points are merged into one irrigation management area, and the water level rise process continues to be simulated. If the difference in nutrient deficit is greater than the irrigation management difference threshold, a watershed boundary line is generated at the contact position of the two adjacent water catchment areas, and the water catchment areas corresponding to the two adjacent seed points are determined as two independent irrigation management areas. Repeat the above process until the catchment areas corresponding to all seed points have been expanded and the watershed boundaries between all adjacent catchment areas have been generated. Based on all the generated watershed boundaries, the target field is divided into several spatially continuous irrigation management areas with relatively consistent internal nutrient deficit characteristics.

[0058] In this embodiment, the nutrient deficit value of each grid cell in the nutrient deficit distribution map is first used as the topographic elevation value of the grid cell to construct a nutrient deficit topographic map. This nutrient deficit topographic map is a virtual three-dimensional terrain model that transforms the two-dimensional nutrient deficit distribution map into a three-dimensional visualization model with an elevation dimension. In the nutrient deficit topographic map, the nutrient deficit value of each grid cell is used as the topographic elevation value of that point. Areas with more negative nutrient deficit (more severe nutrient deficit) correspond to lower terrain, while areas with more positive nutrient deficit (more nutrient surplus) correspond to higher terrain.

[0059] For example, the nutrient deficit distribution map can be input into image processing software (optionally, geographic information system software or remote sensing image processing software with raster data processing and 3D visualization functions, such as ArcGIS or GlobalMapper), and the nutrient deficit value of each raster can be used as the Z value (i.e., the terrain elevation value) of that raster, while the original X and Y coordinates remain unchanged, thereby constructing a virtual nutrient deficit topographic map.

[0060] Secondly, all local minima in the nutrient deficit topographic map are identified, and each local minima is designated as a seed point. A local minima is a point on the topographic surface of the nutrient deficit map whose elevation value is lower than the elevation values ​​of all surrounding adjacent grid cells. In the nutrient deficit topographic map, local minima correspond to the areas with the most severe nutrient deficit, i.e., the areas most in need of nutrient replenishment, and are therefore selected as the center points of potential irrigation management areas. Each seed point represents the central location of a potential irrigation management area and also serves as the starting point for the simulated water level rise process.

[0061] For example, a neighborhood comparison method can be used to identify local minima: for each grid cell in a nutrient deficit topographic map, its topographic elevation value is compared with the topographic elevation values ​​of all grid cells in its surrounding 8 or 4 neighborhoods. If the topographic elevation value of the grid cell is strictly less than the topographic elevation values ​​of all neighboring grid cells, it is marked as a local minima and determined as a seed point.

[0062] For example, in the nutrient deficit topographic map, 15 local minimum points were identified. The nutrient deficit values ​​corresponding to these points were -0.32, -0.28, and -0.25, respectively, which were all significantly lower than the surrounding areas. Therefore, these 15 points were determined as 15 seed points.

[0063] Secondly, starting from each seed point, a synchronous simulation of the gradual rise in water level during inundation is performed. As the water level rises, the catchment area corresponding to each seed point gradually expands outwards. Specifically, using a simulated inundation method, the nutrient deficit topographic map is inverted, and the simulated water level rises gradually from the lowest point where each seed point is located. As the water level rises, the slightly higher-elevation grid cells around the seed point are gradually submerged. These submerged grid cells, together with the seed point, constitute the catchment area corresponding to that seed point. Furthermore, multiple seed points can start this process simultaneously, with their respective catchment areas expanding outwards synchronously.

[0064] Furthermore, when the boundaries of the catchment areas corresponding to any two adjacent seed points are about to touch during the water level rise, the difference in nutrient deficit between the two adjacent seed points is calculated, and this difference is compared with a preset irrigation management difference threshold. The nutrient deficit difference refers to the absolute value of the difference between the nutrient deficit values ​​corresponding to the two adjacent seed points, reflecting the difference in the degree of nutrient deficit between the two potential area centers. The preset irrigation management difference threshold is a predetermined critical value used to determine whether the nutrient difference between two adjacent areas is large enough to require separate management.

[0065] Specifically, during the water level rise, the proximity of the boundaries of different catchment areas is monitored in real time. When it is detected that the boundaries of two adjacent catchment areas are about to touch, the water level rise is paused, the nutrient deficit values ​​of the seed points corresponding to these two areas are extracted, and the absolute value of the difference between the two is calculated to obtain the nutrient deficit difference value. Then, the nutrient deficit difference value is compared with the preset irrigation management difference threshold.

[0066] Furthermore, if the difference in nutrient deficit is less than or equal to the irrigation management difference threshold, the catchment areas corresponding to the two adjacent seed points are merged into one irrigation management area, and the water level rise process continues to be simulated. Specifically, if the calculated difference in nutrient deficit is less than or equal to the irrigation management difference threshold, it indicates that the nutrient deficit levels of the two potential area centers are not significantly different, and their nutrient requirements are similar. They can be merged for management to improve operational efficiency. In this case, the two catchment areas are merged into a larger catchment area, their boundary is removed, and the water level rise process continues.

[0067] For example, the nutrient deficit of seed point A is -0.28 and the nutrient deficit of seed point B is -0.25, with a difference of 0.03. If the preset irrigation management difference threshold is 0.05, since 0.03≤0.05, the water collection areas corresponding to seed point A and seed point B are merged.

[0068] Furthermore, if the difference in nutrient deficit exceeds the irrigation management difference threshold, a watershed boundary line is generated at the contact point of the two adjacent catchment areas, defining the catchment areas corresponding to the two adjacent seed points as two independent irrigation management areas. Specifically, if the calculated difference in nutrient deficit exceeds the irrigation management difference threshold, it indicates a significant difference in the degree of nutrient deficit between the centers of the two potential areas, suggesting different nutrient requirements, necessitating separate management. In this case, a watershed boundary line is generated at the point where the two catchment areas are about to contact, serving as a permanent dividing line. These two catchment areas are thus defined as two independent irrigation management areas, and they will not merge again even if the water level continues to rise.

[0069] For example, the nutrient deficit of seed point C is -0.32 and the nutrient deficit of seed point D is -0.18, with a difference of 0.14. If the preset irrigation management difference threshold is 0.05, since 0.14 > 0.05, a watershed boundary line is generated at the contact position of the water catchment areas of seed point C and seed point D, thus defining seed point C and seed point D as two independent irrigation management areas.

[0070] Furthermore, the above process is repeated until all catchment areas corresponding to seed points have expanded and watershed boundaries between all adjacent catchment areas have been generated. Specifically, after completing one merge or segmentation judgment, the water level continues to rise, and the catchment areas continue to expand. Whenever a new adjacent catchment area boundary is about to touch, the above judgment process is repeated. This process iterates continuously until the water level rises to the highest point (i.e., covering all grid cells), all catchment areas corresponding to seed points have expanded, and watershed boundaries have been generated at all locations that need to be segmented.

[0071] Finally, based on all generated watershed boundaries, the target field is divided into several spatially continuous irrigation management zones with relatively consistent internal nutrient deficit characteristics. Specifically, the final generated watershed boundaries divide the entire nutrient deficit topographic map into multiple independent, spatially continuous catchment basins. These catchment basins are then mapped back to the original geographic space of the target field, with each catchment basin constituting an irrigation management zone. Because the segmentation process is based on nutrient deficit characteristics, the nutrient deficit characteristics within each irrigation management zone are relatively consistent, while significant differences exist between different zones.

[0072] For example, after the above process, the initially identified 15 seed points are merged and divided to form 8 irrigation management areas. Among them, 3 irrigation management areas are larger areas formed by merging multiple seed points, and the other 5 irrigation management areas are smaller areas that remain independent.

[0073] Thus, because the entire segmentation process is strictly based on the spatial distribution characteristics of nutrient deficit and incorporates an irrigation management difference threshold as the criterion for merging and segmentation, the nutrient deficit characteristics at each location point within each irrigation management area are relatively consistent, while significant differences in nutrient deficit characteristics exist between different irrigation management areas. These segmented irrigation management areas provide clear operational units for generating targeted variable fertilization instructions. Each irrigation management area, due to its relatively consistent nutrient deficit characteristics, can adopt a unified fertilization strategy; while different areas, due to significant differences in nutrient deficit characteristics, will implement differentiated fertilization strategies, thereby achieving precise matching and on-demand supply of nutrients across the entire target field.

[0074] Furthermore, the process for determining the "preset irrigation management difference threshold" includes: Based on the nutrient deficit values ​​of all grid cells in the nutrient deficit topographic map, the ratio of the standard deviation to the mean of the nutrient deficit values ​​of all grid cells is calculated, and the ratio is used as the global topographic variation coefficient. Obtain the regulation capability parameters of the integrated water and fertilizer irrigation system, wherein the regulation capability parameters include at least the minimum concentration adjustment absolute accuracy of the simple fertilizer applicator and the minimum compensation absolute accuracy of the passive flow regulation device. Obtain a preset baseline difference threshold and a target fertilizer application benchmark value corresponding to a preset target nutrient benchmark value, wherein the target fertilizer application benchmark value refers to the standard fertilizer application amount required to achieve the preset target nutrient benchmark value; Divide the minimum concentration adjustment absolute accuracy and the minimum compensation absolute accuracy by the target fertilization rate benchmark value to obtain the dimensionless concentration adjustment relative accuracy and the dimensionless compensation relative accuracy. The basic difference threshold, the global terrain variation coefficient, the concentration adjustment relative accuracy, and the compensation relative accuracy are input into the preset threshold calculation formula. After weighted summation, the irrigation management difference threshold used in this segmentation process is obtained.

[0075] In this embodiment, firstly, based on the nutrient deficit values ​​of all grid cells in the nutrient deficit topographic map, the ratio of the standard deviation to the mean of the nutrient deficit values ​​of all grid cells is calculated, and this ratio is used as the global topographic variation coefficient. The formula for calculating the global topographic variation coefficient is: Global Topographic Variation Coefficient = Standard Deviation of Nutrient Deficit Values ​​of All Grid Cells / Mean of Nutrient Deficit Values ​​of All Grid Cells. The global topographic variation coefficient is a statistical indicator reflecting the degree of spatial variation in nutrient distribution across the entire target field. The standard deviation measures the dispersion of the data, while the mean reflects the overall level of the data. The ratio of the two eliminates the influence of dimensions, making the degree of variation in different fields and at different times comparable.

[0076] Specifically, firstly, the mean and standard deviation of nutrient deficit values ​​for all grid cells in the nutrient deficit topographic map are calculated. Then, the ratio of the standard deviation to the mean of all grid cell nutrient deficit values ​​is calculated to obtain the global topographic variation coefficient. The larger the global topographic variation coefficient, the more significant the spatial difference in nutrients within the field, which may require more refined regional division. Conversely, the smaller the global topographic variation coefficient, the more uniform the nutrient distribution within the field, which may allow for a more coarse division.

[0077] Secondly, obtain the regulation capability parameters of the integrated water and fertilizer irrigation system. These parameters should include at least the minimum concentration adjustment absolute accuracy of the simple fertilizer applicator and the minimum compensation absolute accuracy of the passive flow regulator. The minimum concentration adjustment absolute accuracy of the simple fertilizer applicator refers to the minimum concentration change achievable by the applicator, reflecting the precision of the main concentration regulation stage (e.g., ±0.5 kg / mu). The minimum compensation absolute accuracy of the passive flow regulator refers to the minimum additional fertilizer absorption achievable by the passive flow regulator, reflecting the precision of the unit precision compensation stage (e.g., ±0.2 kg / mu). These parameters are provided by the equipment manufacturers of the simple fertilizer applicator and the passive flow regulator, or obtained through on-site calibration tests.

[0078] Next, a preset baseline difference threshold is obtained, along with a target fertilization rate benchmark value corresponding to the preset target nutrient baseline value. The target fertilization rate benchmark value refers to the standard fertilization amount required to achieve the preset target nutrient baseline value. The baseline difference threshold is an empirical value reflecting the minimum degree of nutrient difference required to determine whether two regions need separate management from an agronomic perspective. For example, the baseline difference threshold can be preset based on factors such as the nutrient sensitivity of the crop variety, local climate conditions, and historical planting experience.

[0079] For example, for conventional rice varieties, a basic difference threshold of 0.10 can be set, meaning that when the difference in nutrient deficit between adjacent areas reaches more than 10%, it is considered necessary to manage them separately from an agronomic perspective.

[0080] It should be noted that the basic difference threshold is not fixed and can be dynamically adjusted by those skilled in the art according to the actual application scenario. For example, for nutrient-sensitive varieties such as high-quality rice, it can be set to 0.05 to achieve more refined zoning. For fields with large variations in soil fertility, the threshold can be appropriately increased to reduce the number of areas and improve operational efficiency.

[0081] Specifically, the target nutrient baseline value is an agronomic indicator reflecting the nutritional status of plants, and achieving this indicator requires a corresponding actual fertilization rate. Optionally, a conversion relationship between nutrient characterization parameter values ​​and actual fertilization rates can be established through field calibration experiments, thereby obtaining a target fertilization rate baseline value corresponding to the preset target nutrient baseline value. The target fertilization rate baseline value has the same dimensions as the control capacity parameter, such as kg / acre.

[0082] For example, for the Ningxiangjing 9 rice variety, based on years of field trial data, it was determined that during the peak tillering period, to achieve the target level of 0.75 normalized vegetation index, 15 kg / mu of nitrogen fertilizer needs to be applied under standard conditions. Therefore, the benchmark value for the target fertilizer application rate is 15 kg / mu.

[0083] Furthermore, the minimum concentration adjustment absolute accuracy and the minimum compensation absolute accuracy are divided by the target fertilization rate benchmark value to obtain the dimensionless concentration adjustment relative accuracy and the dimensionless compensation relative accuracy. Specifically, by introducing the target fertilization rate benchmark value with the same dimensions as the denominator, the dimensional absolute accuracy is transformed into dimensionless relative accuracy, allowing it to be mathematically calculated with the same dimensionless basic difference threshold and global terrain variation coefficient. The calculation formulas for the dimensionless concentration adjustment relative accuracy and the dimensionless compensation relative accuracy are: Concentration adjustment relative accuracy = Minimum concentration adjustment absolute accuracy / Target fertilization rate benchmark value; Compensation relative accuracy = Minimum compensation absolute accuracy / Target fertilization rate benchmark value.

[0084] For example, if the minimum concentration adjustment absolute accuracy is 0.5 kg / mu and the target fertilizer application rate is 15 kg / mu, then the concentration adjustment relative accuracy = 0.5 / 15 ≈ 0.033, indicating that the simple fertilizer applicator can achieve a relative adjustment accuracy of approximately 3.3%. If the minimum compensation absolute accuracy is 0.2 kg / mu, then the compensation relative accuracy = 0.2 / 15 ≈ 0.013, indicating that the passive flow rate adjustment device can achieve a relative compensation accuracy of approximately 1.3%.

[0085] Finally, the basic difference threshold, global topographic variation coefficient, concentration regulation relative accuracy, and compensation relative accuracy are input into the preset threshold calculation formula. After weighted summation, the irrigation management difference threshold used in this segmentation process is obtained. For example, the preset threshold calculation formula can be in the form of a product, which can comprehensively reflect the influence of various factors on the regional division: Irrigation management difference threshold = basic difference threshold × (1 + α × global topographic variation coefficient) × (1 + β × P), where P is the maximum value of concentration regulation relative accuracy and compensation relative accuracy, i.e., P = max(concentration regulation relative accuracy, compensation relative accuracy), and α and β are preset weighting coefficients, which are used to adjust the degree of influence of spatial variation characteristics and hardware control capabilities on the final threshold, respectively.

[0086] The weighting coefficients α and β can be dynamically set according to the actual application scenario. For example, for fields with complex terrain and large nutrient variability, the value of α can be appropriately increased to amplify the influence of spatial variability factors; for fields with low precision fertilization equipment, the value of β can be appropriately increased to emphasize the limitations of hardware capabilities. The specific values ​​of α and β can be determined through field trials. For example, different combinations of α and β can be set in preliminary trials, the effects of different zones and yield performance can be compared, and the optimal parameters can be selected.

[0087] Optionally, for conventional rice cultivation, α=0.5 and β=0.3 can be set. Those skilled in the art can reasonably adjust the weighting coefficients based on factors such as crop type, field characteristics, and management objectives through limited experiments or experience, without requiring creative effort.

[0088] Thus, the irrigation management difference threshold calculated by the preset threshold calculation formula takes into account both the spatial variation characteristics of the field itself (i.e., the global terrain variation coefficient) and the control capability of the hardware system (i.e., the relative accuracy of concentration adjustment and the relative accuracy of compensation). It is a dynamic and adaptive threshold that can generate reasonable regional division standards for different fields and different real-world conditions.

[0089] In summary, compared to existing technologies, this application, based on the aforementioned nutrient deficit distribution map, divides the target field into several irrigation management zones with different nutrient requirements. Thus, by transforming the continuously distributed nutrient deficit distribution map into irrigation management zones with clearly defined boundaries and consistent internal characteristics, it provides precise spatial execution units for subsequent implementation of differentiated irrigation by zone, ensuring a high degree of matching between fertilization decisions and the actual spatial distribution of nutrient requirements in the field.

[0090] S40: Generate initial variable fertilization instructions based on the nutrient deficit characteristics of each irrigation management area, and dynamically optimize the initial variable fertilization instructions to generate an optimized refined irrigation instruction set.

[0091] The irrigation management zones, segmented by the nutrient deficit distribution map, exhibit relatively consistent levels of nutrient deficit within their respective areas, but significant differences exist between zones. Generating fertilization instructions directly based on the total cumulative deficit in each zone fails to consider actual soil moisture changes, nutrient absorption rates, and the risk of leakage between adjacent zones during irrigation, potentially leading to fertilizer waste or secondary pollution.

[0092] To address the aforementioned issues, this application generates initial variable fertilization instructions based on the nutrient deficit characteristics of each irrigation management area, and dynamically optimizes these initial variable fertilization instructions to generate an optimized set of refined irrigation instructions.

[0093] Specifically, step S40 in the method includes: For each segmented irrigation management area, the nutrient deficit at all spatial locations within the irrigation management area is statistically analyzed, and the total cumulative nutrient deficit and area parameters of the irrigation management area are calculated. Based on the total cumulative nutrient deficit in the region, combined with the growth stage of the crops planted in the irrigation management area and the fertilizer requirement coefficient per unit area of ​​the growth stage, the total amount of topdressing required in the irrigation management area is calculated, and an initial variable fertilization instruction containing the target fertilization concentration and fertilization duration is generated. Acquire real-time monitoring data for each irrigation unit, wherein the real-time monitoring data includes at least the rate of change of soil moisture, soil solution conductivity, and plant nutrient absorption rate; The real-time monitoring data is input into a pre-trained nutrient leakage prediction model, and the nutrient leakage risk coefficient between adjacent irrigation units is output. When the nutrient leakage risk coefficient exceeds the preset risk threshold, with the goal of minimizing nutrient loss and maximizing fertilizer utilization, a multi-objective optimization algorithm is used to optimize and adjust the initial variable fertilization instruction, generating a refined irrigation instruction that includes the optimized start time, optimized start duration and optimized target fertilization concentration for each irrigation unit. The refined irrigation instructions for all irrigation management areas within the target field are integrated to form a refined irrigation operation instruction set covering the target field.

[0094] In this embodiment, firstly, for each segmented irrigation management area, the nutrient deficit at all spatial locations within the irrigation management area is statistically analyzed, and the total cumulative nutrient deficit and area parameters of the irrigation management area are calculated. Specifically, all spatial locations within each irrigation management area are traversed, and the nutrient deficit at each spatial location (negative values ​​can be taken as absolute values) is summed to obtain the total cumulative nutrient deficit of the area, representing the total nutrients that need to be supplemented in the area. At the same time, the area parameters (e.g., mu) of the area are statistically analyzed to calculate the amount of topdressing per unit area.

[0095] For example, if an irrigation management area has an area of ​​10 mu and the total cumulative nutrient deficit in the area is 50 kg, it means that the area needs to be supplemented with 50 kg of nutrients, and the fertilizer requirement per unit area is 5 kg / mu.

[0096] Secondly, based on the total cumulative nutrient deficit in the region, combined with the growth stage of the crops planted within the irrigation management area and the nutrient requirement coefficient per unit area during that growth stage, the total amount of topdressing required for the irrigation management area is calculated, generating an initial variable fertilization instruction that includes the target fertilization concentration and fertilization duration. The nutrient requirement coefficient per unit area refers to the proportion of nutrients that the crop needs to supplement per unit area to achieve the target yield during that growth stage. Its value can be determined by consulting agricultural technical manuals or field trial data based on the crop type and growth stage. For example, rice has a high nitrogen requirement during the tillering stage, and the nutrient requirement coefficient per unit area can be taken as 0.8–1.2; while the requirement is lower during the maturity stage, and the coefficient can be taken as 0.2–0.4. The nutrient requirement coefficient per unit area is used to adjust the total amount of topdressing to match the fertilization plan with the needs of the crop's growth stage. The target fertilization concentration refers to the mass concentration of nutrients in the fertilizer solution (e.g., kg / m³). 3 The fertilization duration refers to the duration (e.g., minutes) during which each irrigation unit is fertilized. Together, these two constitute the initial variable, the fertilization instruction.

[0097] Specifically, the calculation first multiplies the total cumulative nutrient deficit in the region by the fertilizer requirement coefficient per unit area to obtain the total amount of topdressing required for the region. Then, based on the total amount of topdressing, the design flow rate of the irrigation unit, and the area of ​​the region, the target fertilizer concentration and fertilization duration are calculated to generate the initial variable fertilization instruction.

[0098] For example, suppose an irrigation management area has an area of ​​10 mu (approximately 1.65 acres), a cumulative nutrient deficit of 50 kg, and the rice planted in this area is in the tillering stage, with a nutrient requirement coefficient of 1.0 per unit area. Then, the total amount of topdressing fertilizer would be 50 kg × 1.0 = 50 kg. If the designed irrigation flow rate for this area is 5 cubic meters per hour, and the planned fertilization duration is 2 hours, then the total irrigation water volume is 10 cubic meters. Therefore, the target fertilizer concentration = total topdressing fertilizer / total irrigation water volume = 50 kg / 10 cubic meters = 5 kg / cubic meter. The resulting initial variable fertilization instruction is: target fertilizer concentration 5 kg / cubic meter, fertilization duration 2 hours.

[0099] Secondly, real-time monitoring data for each irrigation unit should be obtained. This data should include at least the rate of change of soil moisture, soil solution conductivity, and the rate of nutrient absorption by plants. Specifically, the rate of change of soil moisture reflects the soil's water retention capacity; a rapid change indicates that water is easily lost, increasing the risk of nutrient leakage. Soil solution conductivity reflects the concentration of soluble salts (i.e., nutrients) in the soil; excessive conductivity indicates nutrient excess, exceeding the soil's colloid retention capacity, and is prone to downward or lateral leakage. The rate of nutrient absorption by plants reflects the efficiency of nutrient absorption by crops; a slow absorption rate indicates that the current fertilizer concentration may be too high or the crop's demand may be reduced, requiring an appropriate reduction in fertilizer concentration or an extension of fertilization time to avoid nutrient accumulation and leakage.

[0100] For example, the rate of change of soil moisture content is monitored in real time by a soil moisture sensor (optionally a time domain reflectometer, TDR or frequency domain reflectometer, FDR) installed in the field, the conductivity of soil solution is continuously acquired by a conductivity probe (i.e., EC sensor) buried in the root layer, and the nutrient absorption rate of the plant is estimated by a stem flow meter or plant physiological sensor.

[0101] Furthermore, real-time monitoring data is input into a pre-trained nutrient leakage prediction model, which outputs a nutrient leakage risk coefficient between adjacent irrigation units. Specifically, the nutrient leakage prediction model is used to predict the risk of nutrient leakage between adjacent irrigation units during irrigation, avoiding fertilizer waste and environmental pollution caused by nutrient leakage. The nutrient leakage risk coefficient is an indicator that quantifies leakage risk, ranging from 0 to 1. The closer the nutrient leakage risk coefficient is to 1, the higher the nutrient leakage risk; the closer it is to 0, the lower the leakage risk.

[0102] Furthermore, when the nutrient leakage risk coefficient exceeds a preset risk threshold, a multi-objective optimization algorithm is used to optimize and adjust the initial variable fertilization command, aiming to minimize nutrient loss and maximize fertilizer utilization. This generates a refined irrigation command that includes the optimized start time, optimized start duration, and optimized target fertilizer concentration for each irrigation unit. Specifically, when the leakage risk is high, the initial variable fertilization command needs to be optimized to avoid nutrient leakage while ensuring maximum fertilizer utilization. The preset risk threshold is used to determine whether command optimization is necessary. Its value can be determined based on the risk coefficient distribution of leakage events in historical irrigation data. For example, it can be the 75th percentile of the risk coefficients of all historical leakage events, or set to 0.7 based on soil type and crop tolerance.

[0103] For example, a particle swarm optimization algorithm can be used to optimize and adjust the initial variable fertilization command. The parameter configuration can be as follows: number of particles: 30, each particle representing a possible combination of refined command parameters; number of iterations: 100, the algorithm outputs the optimal solution after 100 iterations; inertia weight: 0.7, controlling the degree to which particles maintain their original direction of motion; individual learning factor: 1.5, the acceleration coefficient for particles to move towards their historical best position; social learning factor: 1.5, the acceleration coefficient for particles to move towards the global best position; objective function: ,in, Estimated nutrient leakage loss (unit: kg). The objective function, denoted as fertilizer utilization rate (value 0-1), balances minimizing nutrient loss with maximizing fertilizer utilization. During optimization, each particle's position vector contains three decision variables: the activation timing of each irrigation unit (offset relative to the baseline time, in minutes), activation duration (minutes), and target fertilizer concentration (kg / m³). 3 Particle swarm optimization (PSO) iteratively updates particle velocity and position, ultimately outputting the combination of decision variables that minimizes the objective function value; this is the refined irrigation instruction.

[0104] Finally, the refined irrigation instructions for all irrigation management areas within the target field are integrated to form a refined irrigation operation instruction set covering the target field. Specifically, the refined instructions for all irrigation units within each irrigation management area are categorized and integrated according to spatial location and smart switch ID, forming a complete instruction set that includes the activation timing, activation duration, and target fertilizer concentration for each smart switch, providing a basis for the precise control of the subsequent intelligent irrigation system.

[0105] Furthermore, the process of constructing the nutrient leakage prediction model includes: Based on the field ditch network layout pre-constructed in the target field by the ridging and ditching seeder, as well as the soil texture parameters and groundwater level data of each irrigation unit, a digital twin space that completely corresponds to the geographic space of the target field is constructed. The digital twin space includes the spatial coordinates, adjacency relationships, hydraulic connectivity paths, and soil hydraulic parameters of each irrigation unit. In the digital twin space, a spatiotemporal correlation network describing the nutrient transport patterns between irrigation units is constructed, with each irrigation unit as a node and the water transport coefficient and nutrient diffusion coefficient between adjacent irrigation units as edge weights. Based on deep learning algorithms and the spatiotemporal correlation network, an initial nutrient leakage prediction model is constructed. Historical real-time monitoring data were collected from historical irrigation data to form a training sample set; Obtain the nutrient leakage event data actually observed within the time window corresponding to the training sample set, quantify and classify the nutrient leakage event data according to the degree of leakage, obtain discrete nutrient leakage risk level labels, and form a supervision label set corresponding to the training sample set; The training sample set and the supervision label set are input into the initial nutrient leakage prediction model for supervised training until the verification convergence, thus obtaining the pre-trained nutrient leakage prediction model.

[0106] In this embodiment, a digital twin space is first constructed based on the pre-built field ditch network layout within the target field using a ridging, ditching, and seeding machine, as well as the soil texture parameters and groundwater level data of each irrigation unit. This digital twin space completely corresponds to the geographic space of the target field. The digital twin space includes the spatial coordinates, adjacency relationships, hydraulic connectivity paths, and soil hydraulic parameters of each irrigation unit. This digital twin space is a virtual replica of the target field, accurately simulating its geographic layout, soil conditions, and hydraulic characteristics, providing a virtual simulation environment for nutrient leakage prediction. Specifically, during construction, the field ditch network layout, soil texture parameters, and groundwater level data are input into a geographic information system or general computing platform to generate a 1:1 digital twin space corresponding to the target field. This digital twin space can be dynamically updated based on the actually collected data.

[0107] It should be noted that geographic information systems or general computing platforms are publicly available, such as open-source or commercial software like ArcGIS and QGIS, as well as spatial modeling toolkits in scientific computing environments like MATLAB and Python. Those skilled in the art can choose a suitable platform to construct a digital twin space based on their actual needs, without the need to develop additional specialized software.

[0108] Secondly, in the digital twin space, a spatiotemporal relational network describing the nutrient transport patterns between irrigation units is constructed, with each irrigation unit as a node and the water transport coefficient and nutrient diffusion coefficient between adjacent irrigation units as edge weights. Here, nodes represent irrigation units, and edge weights represent the intensity of nutrient transport between adjacent irrigation units. Larger water transport and nutrient diffusion coefficients indicate more frequent and intense nutrient transport between adjacent units, and a higher risk of leakage. Specifically, in the construction process, each pair of adjacent irrigation units is first determined based on the hydraulic connectivity paths in the digital twin space; then, Darcy's law is used to calculate the water transport coefficient. Where k is the soil saturated hydraulic conductivity, Δh is the hydraulic head difference between adjacent cells, and L is the distance between cell centers; then, Fick's first law is used to estimate the nutrient diffusion coefficient. ,in, The effective diffusion coefficient is given by ΔC, where ΔC is the nutrient concentration difference and L is the distance between the center of the unit cell. You can consult empirical values ​​for the corresponding soil texture in a soil physics handbook based on the soil type and moisture conditions, such as sandy soil. ≈1.5×10 -9 m 2 / s, loam ≈0.8×10 -9 m 2 / s, clay ≈0.2×10 -9 m 2 The nutrient concentration can be estimated based on actual field monitoring of changes in nutrient concentration. Using the calculated water transport coefficient and nutrient diffusion coefficient as edge weights, a spatiotemporal correlation network characterizing nutrient transport patterns can be constructed. This network clearly presents the nutrient transport paths and intensities between irrigation units, providing a topological foundation for subsequent leakage prediction models.

[0109] Next, based on deep learning algorithms and spatiotemporal correlation networks, an initial nutrient leakage prediction model is constructed. Specifically, the initial nutrient leakage prediction model takes the real-time monitoring data (soil moisture change rate, soil solution conductivity, and plant nutrient absorption rate) of each irrigation unit in the spatiotemporal correlation network as input and the edge weights (water transport coefficient and nutrient diffusion coefficient) between adjacent units as input, and outputs the nutrient leakage risk coefficient (a continuous value from 0 to 1) between adjacent irrigation units.

[0110] For example, the process of constructing the initial nutrient leakage prediction model can be referred to as follows: The initial nutrient leakage prediction model is constructed based on a graph attention network, mainly consisting of an input layer, a graph attention layer, a temporal processing layer, and an output layer. Specifically, the node feature dimension of the input layer is 3 (corresponding to soil moisture change rate, soil solution conductivity, and plant nutrient absorption rate, respectively), and the edge feature dimension is 2 (corresponding to water transport coefficient and nutrient diffusion coefficient, respectively). The graph attention layer consists of two layers, with output feature dimensions of 64 and 32 for each layer, and 4 attention heads, using the LeakyReLU activation function. The temporal processing layer uses a long short-term memory network to process time series data, with a hidden layer dimension of 64 and 2 layers. The output layer is a fully connected layer that maps the aggregated features to a single output value, which is then activated by the Sigmoid function to obtain the nutrient leakage risk coefficient (range 0~1).

[0111] Furthermore, historical real-time monitoring data is collected from historical irrigation data to form a training sample set. Specifically, real-time monitoring data for each irrigation unit within a continuous time window (e.g., one sampling point every 30 minutes) is extracted from historical irrigation records of different soil types, crop varieties, and irrigation schemes. This data includes at least the rate of change of soil moisture, soil solution conductivity, and plant nutrient absorption rate. These data are then organized according to time sequence and spatial location, outliers are removed, and training samples are formed. All training samples constitute the training sample set.

[0112] Furthermore, nutrient leakage event data actually observed within the time window corresponding to the training sample set is obtained. This nutrient leakage event data is then quantified and graded according to the degree of leakage, resulting in discrete nutrient leakage risk level labels, forming a supervisory label set corresponding to the training sample set. Specifically, nutrient leakage event data refers to leakage conditions actually observed during historical irrigation processes through leakage monitoring devices (optionally, leakage pans or groundwater level sensors), including the time, location, and amount of leakage. To convert leakage events into supervisory signals, the degree of leakage needs to be quantified and graded according to the amount of leakage. For example, three thresholds can be set: when the leakage amount is less than A kg / acre, it is marked as "no leakage" (level 0); when the leakage amount is between A and B, it is marked as "slight leakage" (level 1); when it is between B and C, it is marked as "moderate leakage" (level 2); and when it is greater than C, it is marked as "severe leakage" (level 3). Different levels correspond to discrete risk level labels, such as 0, 1, 2, and 3. The actual leakage level within the corresponding time window for each training sample is used as a supervision label, which is matched one-to-one with the input features to form a supervision label set, which is used for the supervised training of the model in the future.

[0113] Finally, the training sample set and the supervision label set are input into the initial nutrient leakage prediction model for supervised training until validation convergence, resulting in a pre-trained nutrient leakage prediction model. For example, the nutrient leakage prediction model can be trained using the following technical path: 1. Data preparation: The training sample set and the supervision label set are randomly divided into a training set, a validation set, and a test set in a ratio of 7:1.5:1.5. The training set is used for model parameter updates, the validation set is used for hyperparameter tuning and convergence judgment, and the test set is used for final model performance evaluation.

[0114] 2. Model Training: Historical real-time monitoring data from the training set (i.e., soil moisture change rate, soil solution conductivity, and plant nutrient absorption rate for each irrigation unit) are used as input features, and the corresponding nutrient leakage risk level labels are used as supervision labels. The cross-entropy loss function is used to calculate the error between the model prediction and the true label. The optimizer is Adam, with an initial learning rate of 0.001, a batch size of 32, and a maximum of 200 training epochs. After each training epoch, the loss value and classification accuracy are calculated on the validation set. When the validation set loss no longer decreases for 20 consecutive epochs (optionally, the decrease is less than 0.0001), the model is considered to have converged, and training is stopped.

[0115] 3. Model saving: Save the model parameters with the minimum loss on the validation set as a pre-trained nutrient leakage prediction model. Finally, evaluate the classification accuracy of the model on the test set to ensure that it meets the requirements of practical applications.

[0116] In summary, compared to existing technologies, this application generates initial variable fertilization instructions based on the nutrient deficit characteristics of each irrigation management area, and dynamically optimizes these initial variable fertilization instructions to generate an optimized, refined irrigation instruction set. Thus, by combining real-time monitoring data and nutrient leakage risk prediction, and employing a multi-objective optimization algorithm to dynamically adjust fertilization instructions, nutrient leakage loss is reduced, fertilizer utilization efficiency is improved, and precise, safe, and efficient variable irrigation fertilization is achieved.

[0117] S50: Control the intelligent irrigation system laid in the field ditch network according to the refined irrigation instruction set, and perform zoned differentiated irrigation for several irrigation management areas.

[0118] After generating precise irrigation instructions, these instructions need to be translated into actual control actions to drive the field irrigation equipment. Simultaneously, since the field irrigation network is pre-constructed during sowing, and the flexible delivery pipes are equipped with smart switches with unique IDs at their openings, precise control of the timing and duration of each switch's activation allows for differentiated fertilization and irrigation in various irrigation management areas. This ensures the optimized instructions are effectively implemented, achieving the goal of supplying water and fertilizer on demand.

[0119] To address the aforementioned issues, this application controls an intelligent irrigation system laid in a field ditch network based on the refined irrigation instruction set, and performs differentiated irrigation for several irrigation management areas.

[0120] Specifically, step S50 in the method includes: According to the refined irrigation instructions, the fertilizer applicator is controlled to adjust the fertilizer concentration in the main delivery pipeline to the optimized target fertilizer concentration, and the adjusted fertilizer is continuously delivered to the main delivery pipeline. According to the optimized activation timing in the refined irrigation instruction, an activation signal is sent to the corresponding smart switch. The smart switch is installed at the opening of the flexible delivery pipe in the field ditch network pre-constructed during rice sowing. Each smart switch has a unique identification ID and corresponds to one irrigation unit. Fertilizer and water flow out from the corresponding openings of the flexible delivery pipes via the activated smart switch, naturally diffuse along the field ditch network, and infiltrate into the root zone of the corresponding irrigation unit. After the smart switch has been on for the optimized duration, a shutdown signal is sent to the smart switch to stop the supply of fertilizer and water to the irrigation unit until the differentiated irrigation of the target irrigation management area is completed.

[0121] In this embodiment, the fertilizer applicator is first controlled according to the refined irrigation instructions to adjust the fertilizer concentration in the main delivery pipeline to the optimized target fertilizer concentration, and then continuously delivers the adjusted fertilizer solution to the main delivery pipeline. Specifically, the fertilizer applicator is the core equipment for adjusting the fertilizer concentration. Based on the optimized target fertilizer concentration specified in the refined irrigation instructions, the fertilizer injection flow rate of the fertilizer applicator and the water flow rate of the main water pipeline are controlled to precisely adjust the fertilizer concentration in the main pipeline to the optimized target fertilizer concentration. After adjustment, the fertilizer applicator continuously delivers the prepared fertilizer solution to the main delivery pipeline, preparing for subsequent diversion.

[0122] For example, if the target fertilizer concentration for an irrigation management area is 0.5%, the fertilizer applicator will mix water-soluble fertilizer with water in proportion to stabilize the fertilizer-water concentration in the main pipeline at 0.5%.

[0123] Secondly, based on the optimized activation timing specified in the refined irrigation instructions, an activation signal is sent to the corresponding smart switch. This smart switch is installed at the opening of the flexible delivery pipes laid in the field furrow network pre-constructed during rice sowing. Each smart switch has a unique identification ID, corresponding to one irrigation unit. Specifically, the smart switch is the actuator for implementing zoned differentiated irrigation; each switch corresponds to one irrigation unit and has a unique ID. According to the optimized activation timing specified in the refined irrigation instructions, an activation signal is sent to the smart switch with the corresponding ID via wireless or wired communication. Because the switch ID is unique, it ensures that the instructions are accurately delivered to the target irrigation unit, avoiding misoperation.

[0124] For example, if the start time of an irrigation unit in the refined instruction is 8:00, then the start instruction will be sent to the smart switch corresponding to the unit precisely at 8:00.

[0125] Secondly, the fertilized water flows out from the corresponding opening in the flexible delivery pipe via a smart switch, naturally diffusing and infiltrating along the field ditch network to the root zone of the corresponding irrigation unit. Specifically, the flexible delivery pipe is laid in a pre-constructed field ditch network, with its openings corresponding to the root zone of the irrigation unit. When the smart switch is activated, the fertilized water in the main pipe enters the flexible pipe and flows out from the opening into the field ditch network. The fertilized water diffuses naturally through the ditch network using capillary action and gravity, evenly infiltrating the soil of the irrigation unit and ultimately reaching the crop root layer for absorption and utilization. In this way, this delivery method avoids surface runoff and deep seepage, improving water and fertilizer utilization efficiency.

[0126] like Figure 3 , 4 As shown in Figure 5, the field ditch network is constructed synchronously by the precision strip direct seeder during sowing: the precision strip direct seeder drives each working component through the central shaft (5), the fertilizer and water ditch opening and shaping plow (1) first opens fertilizer and water ditches in the field; the seeding pipe (2) sows the seeds into the seedbed; the water storage ditch opening and shaping plow (3) opens water storage ditches on the side or the other side of the fertilizer and water ditches to store irrigation water, thereby enhancing the water retention capacity of the field; the ridge leveling board (4) then levels and backfills the soil turned up by the ditch opening, forming a regular ridge structure. Through the coordinated operation of the above components, a field ditch network composed of fertilizer and water ditches and water storage ditches is constructed synchronously during the sowing process. This field ditch network provides a physical basis for the laying of flexible conveying pipelines and the installation of intelligent switches, enabling the irrigation system to achieve precise diffusion and infiltration of fertilizer and water using the natural ditch network.

[0127] like Figure 6 , 7 The diagram shows a field furrow network formed after sowing by a precision strip direct seeder.

[0128] Finally, after the smart switch has been on for the optimized duration, a shutdown signal is sent to the smart switch to stop the supply of fertilizer and water to that irrigation unit, until the differentiated irrigation for the target irrigation management area is completed. Specifically, the on-time of each smart switch is monitored in real time, and when the optimized on-time specified in the refined irrigation instruction is reached, a shutdown signal is immediately sent to that switch to stop the supply of fertilizer and water. By sequentially controlling the on and off of the smart switches of each irrigation unit, differentiated irrigation operations for all irrigation management areas of the entire target field are completed.

[0129] For example, if the optimized operating time of an irrigation unit is 30 minutes, it will automatically turn off after 30 minutes of operation to ensure that the unit receives a precise supply of water and fertilizer.

[0130] In summary, compared to existing technologies, this application controls an intelligent irrigation system laid in a field ditch network based on the aforementioned refined irrigation instruction set, performing differentiated irrigation for several irrigation management areas. Thus, through intelligent switches with unique IDs and a pre-constructed field ditch network, independent and precise control of each irrigation unit is achieved, ensuring the optimized refined instructions are implemented effectively, improving the targeting and utilization rate of water and fertilizer supply, and reducing waste and environmental pollution.

[0131] In summary, the embodiments of this application have at least the following technical effects: Compared to existing technologies, this application first obtains plant nutrient characterization parameters at multiple spatial locations within the target field and constructs a nutrient spatial distribution model, enabling dynamic tracking of plant nutrient spatial distribution and providing precise data support for subsequent differentiated irrigation by region. Secondly, based on the nutrient spatial distribution model and preset target nutrient baseline values, it calculates the nutrient deficit at each spatial location and generates a nutrient deficit distribution map, accurately quantifying the gap between plant nutrient supply and demand and intuitively presenting the spatial distribution pattern of nutrient deficit and surplus, providing a core basis for differentiated water and fertilizer supply by region. Thirdly, based on the nutrient deficit distribution map, the target field is divided into several irrigation management areas with different fertilizer requirements, transforming continuously distributed nutrient information into spatially defined execution units with consistent internal characteristics, ensuring a high degree of matching between fertilization decisions and the actual spatial distribution of fertilizer requirements in the field. Furthermore, initial variable fertilization instructions are generated based on the nutrient deficit characteristics of each irrigation management area. These instructions are then dynamically optimized using real-time monitoring data and nutrient leakage risk prediction to generate an optimized, refined irrigation instruction set. This reduces nutrient leakage loss, improves fertilizer utilization efficiency, and achieves precise, safe, and efficient variable irrigation and fertilization. Finally, the intelligent irrigation system laid in the field ditch network is controlled according to the refined irrigation instruction set. Intelligent switches with unique identifiers (IDs) are used to perform differentiated irrigation for several irrigation management areas, ensuring the optimized refined instructions are implemented effectively. This improves the targeting and utilization rate of water and fertilizer supply, and reduces waste and environmental pollution.

[0132] Through the above technical solutions, this application constructs a complete closed loop based on the dynamic tracking of plant nutrients and the linkage of intelligent irrigation equipment. It effectively solves the technical problems of traditional uniform irrigation and fertilization ignoring spatial heterogeneity, coarse division of irrigation areas, and static instructions being unable to adapt to dynamic environments. It significantly improves water and fertilizer utilization efficiency and crop yield consistency, reduces environmental pollution caused by nutrient leakage, and provides reliable technical support for precision agriculture.

[0133] Example 2, as Figure 2 As shown, based on the same inventive concept as the fertigation method based on dynamic tracking of plant nutrients provided in Embodiment 1, this embodiment of the invention also provides a fertigation system based on dynamic tracking of plant nutrients, comprising: The nutrient spatial division module 11 is used to obtain plant nutrient characterization parameters at multiple spatial locations within the target field, and to construct a nutrient spatial distribution model based on the plant nutrient characterization parameters. The nutrient deficit calculation module 12 is used to calculate the nutrient deficit at each spatial location based on the nutrient spatial distribution model and the preset target nutrient benchmark value, and to generate a nutrient deficit distribution map based on the nutrient deficit. The irrigation management area division module 13 is used to divide the target field into several irrigation management areas with different fertilizer requirements based on the nutrient deficit distribution map. The irrigation instruction generation module 14 is used to generate initial variable fertilization instructions based on the nutrient deficit characteristics of each irrigation management area, and to dynamically optimize the initial variable fertilization instructions to generate an optimized refined irrigation instruction set. The differentiated irrigation module 15 is used to control the intelligent irrigation system laid in the field ditch network according to the refined irrigation instruction set, and to perform zoned differentiated irrigation for several irrigation management areas. The field ditch network is pre-constructed by the ridging and furrowing seeder used during sowing. The intelligent irrigation system includes a fertilizer applicator, a main conveying pipeline, a flexible conveying pipeline laid in the field ditch network, and an intelligent switch with a unique identification ID installed at the opening of the flexible conveying pipeline.

[0134] The nutrient space division module 11 is specifically used for: The target field is divided into multiple sampling grids according to a preset sampling density, and the center point of each sampling grid is determined as the in-situ detection point; Multispectral image data covering the target field is collected by a multispectral imaging device carried by a drone, and combined with field measurement data collected at each in-situ detection point, the plant nutrient characterization parameters corresponding to each in-situ detection point are determined. The plant nutrient characterization parameters include at least the normalized vegetation index and the relative chlorophyll content. The spatial coordinate data of each in-situ detection point is obtained, and the plant nutrient characterization parameters and corresponding spatial coordinate data of the multiple sampling grids are used as input. A spatial interpolation algorithm is used to interpolate and extrapolate the unsampled area to generate a nutrient spatial distribution model that covers the entire target field and has a continuous spatial distribution.

[0135] Furthermore, the phrase "collecting multispectral image data covering the target field using a multispectral imaging device mounted on a drone, and combining it with field measurement data collected at each in-situ detection point to determine the plant nutrient characterization parameters corresponding to each in-situ detection point" includes: The drone is controlled to cruise and scan the target field along a preset route, and multispectral image data covering the entire target field is collected by the multispectral imaging device. The multispectral image data includes at least red band reflectance and near-infrared band reflectance. The acquired multispectral image data is geometrically corrected and radiometrically calibrated to generate standardized multispectral images with geographic coordinate information. Based on the spatial coordinates of each in-situ detection point, the red band reflectance and near-infrared band reflectance at the corresponding locations are extracted from the standardized multispectral image. The normalized vegetation index is calculated based on the red light band reflectance and the near-infrared band reflectance. Plant samples were collected at each in-situ detection point, and the relative chlorophyll content at each in-situ detection point was obtained using a portable chlorophyll meter.

[0136] The nutrient deficit calculation module 12 is specifically used for: Based on the variety of crops planted in the target field, their current growth stage, and the preset target yield, obtain the corresponding preset target nutrient baseline values; Traverse each spatial location in the nutrient spatial distribution model, perform a difference calculation between the nutrient characterization parameter value at each spatial location and the preset target nutrient benchmark value to obtain the corresponding nutrient deficit. A positive nutrient deficit indicates a nutrient surplus state, while a negative nutrient deficit indicates a nutrient deficit state. The nutrient deficit at all spatial locations is mapped according to their respective spatial coordinates to generate a two-dimensional raster dataset that completely corresponds to the geographic space of the target field. This two-dimensional raster dataset is used as a nutrient deficit distribution map.

[0137] The irrigation management area division module 13 is specifically used for: The nutrient deficit value of each grid cell in the nutrient deficit distribution map is used as the topographic elevation value of the grid cell to construct a nutrient deficit topographic map. Identify all local minima in the nutrient deficit topographic map and designate each local minima as a seed point, where each seed point represents the center location of a potential irrigation management area. Simultaneously simulate the flooding process of gradually rising water level starting from each seed point. During the rise of water level, the catchment area corresponding to each seed point gradually expands to the surrounding area. When the boundaries of the catchment areas corresponding to any two adjacent seed points are about to touch during the rise in water level, calculate the difference in nutrient deficit between the two adjacent seed points and compare the difference in nutrient deficit with a preset irrigation management difference threshold. If the difference in nutrient deficit is less than or equal to the irrigation management difference threshold, then the water collection areas corresponding to the two adjacent seed points are merged into one irrigation management area, and the water level rise process continues to be simulated. If the difference in nutrient deficit is greater than the irrigation management difference threshold, a watershed boundary line is generated at the contact position of the two adjacent water catchment areas, and the water catchment areas corresponding to the two adjacent seed points are determined as two independent irrigation management areas. Repeat the above process until the catchment areas corresponding to all seed points have been expanded and the watershed boundaries between all adjacent catchment areas have been generated. Based on all the generated watershed boundaries, the target field is divided into several spatially continuous irrigation management areas with relatively consistent internal nutrient deficit characteristics.

[0138] Specifically, the process for determining the "preset irrigation management difference threshold" includes: Based on the nutrient deficit values ​​of all grid cells in the nutrient deficit topographic map, the ratio of the standard deviation to the mean of the nutrient deficit values ​​of all grid cells is calculated, and the ratio is used as the global topographic variation coefficient. Obtain the regulation capability parameters of the integrated water and fertilizer irrigation system, wherein the regulation capability parameters include at least the minimum concentration adjustment absolute accuracy of the simple fertilizer applicator and the minimum compensation absolute accuracy of the passive flow regulation device. Obtain a preset baseline difference threshold and a target fertilizer application benchmark value corresponding to a preset target nutrient benchmark value, wherein the target fertilizer application benchmark value refers to the standard fertilizer application amount required to achieve the preset target nutrient benchmark value; Divide the minimum concentration adjustment absolute accuracy and the minimum compensation absolute accuracy by the target fertilization rate benchmark value to obtain the dimensionless concentration adjustment relative accuracy and the dimensionless compensation relative accuracy. The basic difference threshold, the global terrain variation coefficient, the concentration adjustment relative accuracy, and the compensation relative accuracy are input into the preset threshold calculation formula. After weighted summation, the irrigation management difference threshold used in this segmentation process is obtained.

[0139] The irrigation instruction generation module 14 is specifically used for: For each segmented irrigation management area, the nutrient deficit at all spatial locations within the irrigation management area is statistically analyzed, and the total cumulative nutrient deficit and area parameters of the irrigation management area are calculated. Based on the total cumulative nutrient deficit in the region, combined with the growth stage of the crops planted in the irrigation management area and the fertilizer requirement coefficient per unit area of ​​the growth stage, the total amount of topdressing required in the irrigation management area is calculated, and an initial variable fertilization instruction containing the target fertilization concentration and fertilization duration is generated. Acquire real-time monitoring data for each irrigation unit, wherein the real-time monitoring data includes at least the rate of change of soil moisture, soil solution conductivity, and plant nutrient absorption rate; The real-time monitoring data is input into a pre-trained nutrient leakage prediction model, and the nutrient leakage risk coefficient between adjacent irrigation units is output. When the nutrient leakage risk coefficient exceeds the preset risk threshold, with the goal of minimizing nutrient loss and maximizing fertilizer utilization, a multi-objective optimization algorithm is used to optimize and adjust the initial variable fertilization instruction, generating a refined irrigation instruction that includes the optimized start time, optimized start duration and optimized target fertilization concentration for each irrigation unit. The refined irrigation instructions for all irrigation management areas within the target field are integrated to form a refined irrigation operation instruction set covering the target field.

[0140] Specifically, the process of constructing the nutrient leakage prediction model includes: Based on the field ditch network layout pre-constructed in the target field by the ridging and ditching seeder, as well as the soil texture parameters and groundwater level data of each irrigation unit, a digital twin space that completely corresponds to the geographic space of the target field is constructed. The digital twin space includes the spatial coordinates, adjacency relationships, hydraulic connectivity paths, and soil hydraulic parameters of each irrigation unit. In the digital twin space, a spatiotemporal correlation network describing the nutrient transport patterns between irrigation units is constructed, with each irrigation unit as a node and the water transport coefficient and nutrient diffusion coefficient between adjacent irrigation units as edge weights. Based on deep learning algorithms and the spatiotemporal correlation network, an initial nutrient leakage prediction model is constructed. Historical real-time monitoring data were collected from historical irrigation data to form a training sample set; Obtain the nutrient leakage event data actually observed within the time window corresponding to the training sample set, quantify and classify the nutrient leakage event data according to the degree of leakage, obtain discrete nutrient leakage risk level labels, and form a supervision label set corresponding to the training sample set; The training sample set and the supervision label set are input into the initial nutrient leakage prediction model for supervised training until the verification convergence, thus obtaining the pre-trained nutrient leakage prediction model.

[0141] Specifically, the differentiated irrigation module 15 is used for: According to the refined irrigation instructions, the fertilizer applicator is controlled to adjust the fertilizer concentration in the main delivery pipeline to the optimized target fertilizer concentration, and the adjusted fertilizer is continuously delivered to the main delivery pipeline. According to the optimized activation timing in the refined irrigation instruction, an activation signal is sent to the corresponding smart switch. The smart switch is installed at the opening of the flexible delivery pipe in the field ditch network pre-constructed during rice sowing. Each smart switch has a unique identification ID and corresponds to one irrigation unit. Fertilizer and water flow out from the corresponding openings of the flexible delivery pipes via the activated smart switch, naturally diffuse along the field ditch network, and infiltrate into the root zone of the corresponding irrigation unit. After the smart switch has been on for the optimized duration, a shutdown signal is sent to the smart switch to stop the supply of fertilizer and water to the irrigation unit until the differentiated irrigation of the target irrigation management area is completed.

[0142] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0143] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0144] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0145] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0146] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0147] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0148] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients, characterized in that, The method includes: Obtain plant nutrient characterization parameters at multiple spatial locations within the target field, and construct a nutrient spatial distribution model based on the plant nutrient characterization parameters; Based on the nutrient spatial distribution model and the preset target nutrient baseline value, the nutrient deficit at each spatial location is calculated, and a nutrient deficit distribution map is generated based on the nutrient deficit. Based on the nutrient deficit distribution map, the target field is divided into several irrigation management areas with different nutrient requirements. Initial variable fertilization instructions are generated based on the nutrient deficit characteristics of each irrigation management area, and the initial variable fertilization instructions are dynamically optimized to generate an optimized set of refined irrigation instructions. The intelligent irrigation system, laid in the field ditch network, is controlled according to the refined irrigation instruction set to perform differentiated irrigation for several irrigation management areas. The field ditch network is pre-constructed by the ridging and furrowing seeder used during sowing. The intelligent irrigation system includes a fertilizer applicator, a main conveying pipeline, a flexible conveying pipeline laid in the field ditch network, and intelligent switches with unique identification IDs installed at the openings of the flexible conveying pipelines.

2. The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients according to claim 1, characterized in that, Obtain plant nutrient characterization parameters at multiple spatial locations within the target field, and construct a nutrient spatial distribution model based on the plant nutrient characterization parameters, including: The target field is divided into multiple sampling grids according to a preset sampling density, and the center point of each sampling grid is determined as the in-situ detection point; Multispectral image data covering the target field is collected by a multispectral imaging device carried by a drone, and combined with field measurement data collected at each in-situ detection point, the plant nutrient characterization parameters corresponding to each in-situ detection point are determined. The plant nutrient characterization parameters include at least the normalized vegetation index and the relative chlorophyll content. The spatial coordinate data of each in-situ detection point is obtained, and the plant nutrient characterization parameters and corresponding spatial coordinate data of the multiple sampling grids are used as input. A spatial interpolation algorithm is used to interpolate and extrapolate the unsampled area to generate a nutrient spatial distribution model that covers the entire target field and has a continuous spatial distribution.

3. The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients according to claim 2, characterized in that, Multispectral image data covering the target field was collected using a multispectral imaging device mounted on a drone. Combined with field measurement data collected at each in-situ detection point, plant nutrient characterization parameters corresponding to each in-situ detection point were determined, including: The drone is controlled to cruise and scan the target field along a preset route, and multispectral image data covering the entire target field is collected by the multispectral imaging device. The multispectral image data includes at least red band reflectance and near-infrared band reflectance. The acquired multispectral image data is geometrically corrected and radiometrically calibrated to generate standardized multispectral images with geographic coordinate information. Based on the spatial coordinates of each in-situ detection point, the red band reflectance and near-infrared band reflectance at the corresponding locations are extracted from the standardized multispectral image. The normalized vegetation index is calculated based on the red light band reflectance and the near-infrared band reflectance. Plant samples were collected at each in-situ detection point, and the relative chlorophyll content at each in-situ detection point was obtained using a portable chlorophyll meter.

4. The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients according to claim 1, characterized in that, Based on the nutrient spatial distribution model and the preset target nutrient baseline value, the nutrient deficit at each spatial location is calculated, and a nutrient deficit distribution map is generated based on the nutrient deficit, including: Based on the variety of crops planted in the target field, their current growth stage, and the preset target yield, obtain the corresponding preset target nutrient baseline values; Traverse each spatial location in the nutrient spatial distribution model, perform a difference calculation between the nutrient characterization parameter value at each spatial location and the preset target nutrient benchmark value to obtain the corresponding nutrient deficit. A positive nutrient deficit indicates a nutrient surplus state, while a negative nutrient deficit indicates a nutrient deficit state. The nutrient deficit at all spatial locations is mapped according to their respective spatial coordinates to generate a two-dimensional raster dataset that completely corresponds to the geographic space of the target field. This two-dimensional raster dataset is used as a nutrient deficit distribution map.

5. The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients according to claim 1, characterized in that, Based on the aforementioned nutrient deficit distribution map, the target field is divided into several irrigation management zones with different nutrient requirements, including: The nutrient deficit value of each grid cell in the nutrient deficit distribution map is used as the topographic elevation value of the grid cell to construct a nutrient deficit topographic map. Identify all local minima in the nutrient deficit topographic map and designate each local minima as a seed point, where each seed point represents the center location of a potential irrigation management area. Simultaneously simulate the flooding process of gradually rising water level starting from each seed point. During the rise of water level, the catchment area corresponding to each seed point gradually expands to the surrounding area. When the boundaries of the catchment areas corresponding to any two adjacent seed points are about to touch during the rise in water level, calculate the difference in nutrient deficit between the two adjacent seed points and compare the difference in nutrient deficit with a preset irrigation management difference threshold. If the difference in nutrient deficit is less than or equal to the irrigation management difference threshold, then the water collection areas corresponding to the two adjacent seed points are merged into one irrigation management area, and the water level rise process continues to be simulated. If the difference in nutrient deficit is greater than the irrigation management difference threshold, a watershed boundary line is generated at the contact position of the two adjacent water catchment areas, and the water catchment areas corresponding to the two adjacent seed points are determined as two independent irrigation management areas. Repeat the above process until the catchment areas corresponding to all seed points have been expanded and the watershed boundaries between all adjacent catchment areas have been generated. Based on all the generated watershed boundaries, the target field is divided into several spatially continuous irrigation management areas with relatively consistent internal nutrient deficit characteristics.

6. The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients according to claim 5, characterized in that, The process of determining the preset irrigation management difference threshold includes: Based on the nutrient deficit values ​​of all grid cells in the nutrient deficit topographic map, the ratio of the standard deviation to the mean of the nutrient deficit values ​​of all grid cells is calculated, and the ratio is used as the global topographic variation coefficient. Obtain the regulation capability parameters of the integrated water and fertilizer irrigation system, wherein the regulation capability parameters include at least the minimum concentration adjustment absolute accuracy of the simple fertilizer applicator and the minimum compensation absolute accuracy of the passive flow regulation device. Obtain a preset baseline difference threshold and a target fertilizer application benchmark value corresponding to a preset target nutrient benchmark value, wherein the target fertilizer application benchmark value refers to the standard fertilizer application amount required to achieve the preset target nutrient benchmark value; Divide the minimum concentration adjustment absolute accuracy and the minimum compensation absolute accuracy by the target fertilization rate benchmark value to obtain the dimensionless concentration adjustment relative accuracy and the dimensionless compensation relative accuracy. The basic difference threshold, the global terrain variation coefficient, the concentration adjustment relative accuracy, and the compensation relative accuracy are input into the preset threshold calculation formula. After weighted summation, the irrigation management difference threshold used in this segmentation process is obtained.

7. The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients according to claim 1, characterized in that, Initial variable fertilization instructions are generated based on the nutrient deficit characteristics of each irrigation management area, and these initial variable fertilization instructions are dynamically optimized to generate optimized refined irrigation instructions, including: For each segmented irrigation management area, the nutrient deficit at all spatial locations within the irrigation management area is statistically analyzed, and the total cumulative nutrient deficit and area parameters of the irrigation management area are calculated. Based on the total cumulative nutrient deficit in the region, combined with the growth stage of the crops planted in the irrigation management area and the fertilizer requirement coefficient per unit area of ​​the growth stage, the total amount of topdressing required in the irrigation management area is calculated, and an initial variable fertilization instruction containing the target fertilization concentration and fertilization duration is generated. Acquire real-time monitoring data for each irrigation unit, wherein the real-time monitoring data includes at least the rate of change of soil moisture, soil solution conductivity, and plant nutrient absorption rate; The real-time monitoring data is input into a pre-trained nutrient leakage prediction model, and the nutrient leakage risk coefficient between adjacent irrigation units is output. When the nutrient leakage risk coefficient exceeds the preset risk threshold, with the goal of minimizing nutrient loss and maximizing fertilizer utilization, a multi-objective optimization algorithm is used to optimize and adjust the initial variable fertilization instruction, generating a refined irrigation instruction that includes the optimized start time, optimized start duration and optimized target fertilization concentration for each irrigation unit. The refined irrigation instructions for all irrigation management areas within the target field are integrated to form a refined irrigation operation instruction set covering the target field.

8. The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients according to claim 7, characterized in that, The process of constructing a nutrient leakage prediction model includes: Based on the field ditch network layout pre-constructed in the target field by the ridging and ditching seeder, as well as the soil texture parameters and groundwater level data of each irrigation unit, a digital twin space that completely corresponds to the geographic space of the target field is constructed. The digital twin space includes the spatial coordinates, adjacency relationships, hydraulic connectivity paths, and soil hydraulic parameters of each irrigation unit. In the digital twin space, a spatiotemporal correlation network describing the nutrient transport patterns between irrigation units is constructed, with each irrigation unit as a node and the water transport coefficient and nutrient diffusion coefficient between adjacent irrigation units as edge weights. Based on deep learning algorithms and the spatiotemporal correlation network, an initial nutrient leakage prediction model is constructed. Historical real-time monitoring data were collected from historical irrigation data to form a training sample set; Obtain the nutrient leakage event data actually observed within the time window corresponding to the training sample set, quantify and classify the nutrient leakage event data according to the degree of leakage, obtain discrete nutrient leakage risk level labels, and form a supervision label set corresponding to the training sample set; The training sample set and the supervision label set are input into the initial nutrient leakage prediction model for supervised training until the verification convergence, thus obtaining the pre-trained nutrient leakage prediction model.

9. The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients according to claim 1, characterized in that, The intelligent irrigation system, laid in the field ditch network, is controlled according to the refined irrigation instruction set to perform zoned differentiated irrigation for several irrigation management areas, including: According to the refined irrigation instructions, the fertilizer applicator is controlled to adjust the fertilizer concentration in the main delivery pipeline to the optimized target fertilizer concentration, and the adjusted fertilizer is continuously delivered to the main delivery pipeline. According to the optimized activation timing in the refined irrigation instruction, an activation signal is sent to the corresponding smart switch. The smart switch is installed at the opening of the flexible delivery pipe in the field ditch network pre-constructed during rice sowing. Each smart switch has a unique identification ID and corresponds to one irrigation unit. Fertilizer and water flow out from the corresponding openings of the flexible delivery pipes via the activated smart switch, naturally diffuse along the field ditch network, and infiltrate into the root zone of the corresponding irrigation unit. After the smart switch has been on for the optimized duration, a shutdown signal is sent to the smart switch to stop the supply of fertilizer and water to the irrigation unit until the differentiated irrigation of the target irrigation management area is completed.

10. A water and fertilizer integrated irrigation system based on dynamic tracking of plant nutrients, characterized in that, The water and fertilizer integrated irrigation method based on dynamic tracking of plant nutrients as described in any one of claims 1-9 includes: The nutrient spatial division module is used to obtain plant nutrient characterization parameters at multiple spatial locations within the target field, and to construct a nutrient spatial distribution model based on the plant nutrient characterization parameters. The nutrient deficit calculation module is used to calculate the nutrient deficit at each spatial location based on the nutrient spatial distribution model and the preset target nutrient benchmark value, and to generate a nutrient deficit distribution map based on the nutrient deficit. The irrigation management area division module is used to divide the target field into several irrigation management areas with different fertilizer requirements based on the nutrient deficit distribution map. The irrigation instruction generation module is used to generate initial variable fertilization instructions based on the nutrient deficit characteristics of each irrigation management area, and to dynamically optimize the initial variable fertilization instructions to generate an optimized refined irrigation instruction set. The differentiated irrigation module is used to control the intelligent irrigation system laid in the field ditch network according to the refined irrigation instruction set, and to perform zoned differentiated irrigation for several irrigation management areas. The field ditch network is pre-constructed by the ridging and furrowing seeder used during sowing. The intelligent irrigation system includes a fertilizer applicator, a main conveying pipeline, a flexible conveying pipeline laid in the field ditch network, and an intelligent switch with a unique identification ID installed at the opening of the flexible conveying pipeline.

Images