An intelligent water and fertilizer integrated control method and system for vegetable production

Abstract

This application provides an intelligent water and fertilizer integrated control method and system for vegetable production, belonging to the field of water and fertilizer control technology. The method includes: acquiring temporal multispectral images of each preset grid unit within the vegetable planting area, and constructing a stress map characterizing the crop nutrient stress type within each preset grid unit, as well as a soil condition map characterizing soil texture and spatial distribution of moisture; performing spatial overlay analysis of the stress map and soil condition map to identify heterogeneous stress patches; determining customized nutrient element ratios, irrigation baseline amounts, and fertilization durations for the current growth stage based on the heterogeneous stress patches, the nutrient stress types in the stress map, and the soil texture and moisture distribution information represented by the soil condition map; and providing differentiated water and fertilizer supply to the heterogeneous stress patches based on the customized nutrient element ratios, irrigation baseline amounts, and fertilization durations. The technical solution provided by this application achieves accurate decision-making regarding water and fertilizer supply under structural soil spatial variations.

Description

Technical Field

[0001] This application relates to the field of water and fertilizer control technology, and more specifically, to an intelligent integrated water and fertilizer control method and system for vegetable production. Background Technology

[0002] In greenhouse vegetable production, water and fertilizer management is a key factor determining yield and quality. Traditional irrigation and fertilization methods often rely on experience, resulting in low water and fertilizer utilization, resource waste, and environmental pollution. With the development of precision agriculture, intelligent water and fertilizer integration technology has emerged. Intelligent water and fertilizer integration uses sensors to monitor soil moisture, nutrient content, and environmental parameters in real time. Combined with crop growth models, it automatically controls irrigation and fertilization to achieve precise supply of water and nutrients on demand. It can also dynamically adjust according to demand at different growth stages, improve water and fertilizer utilization efficiency, reduce labor input, and ensure high-quality and stable vegetable yields.

[0003] Existing integrated water and fertilizer systems generally employ a control principle combining closed-loop feedback and timed / quantitative control. These systems use soil moisture, conductivity, and pH sensors embedded in the root zone to collect water and fertilizer status parameters in real time, uploading this data to the controller for analysis and decision-making to achieve a balanced water and fertilizer supply. However, in vegetable production scenarios, there are often variations in soil texture and micro-topography at the microscale within the field. This structural spatial variation in soil leads to two intertwined processes: first, the spatial differentiation of soil water-holding capacity and water infiltration / evaporation rates directly causes local differences in water stress; second, these differences in water stress affect… Nutrient flow and diffusion indirectly induce spatially heterogeneous nutrient stress phenotypes, meaning that crops passively exhibit nutrient deficiency symptoms due to insufficient water absorption. This makes it easy for traditional diagnostic methods based on single-period multispectral images or simple vegetation indices to misjudge "physiological nutrient deficiency induced by drought" as "absolute soil nutrient deficiency." Consequently, it is impossible to make correct decisions at the implementation level of fertigation, leading to further aggravation of root salt accumulation in water-stressed areas due to limited root water absorption, creating a vicious cycle of stress. Therefore, how to achieve accurate decision-making on water and fertilizer supply under structural soil spatial variations has become a challenge for the industry. Summary of the Invention

[0004] This application provides an intelligent water and fertilizer integrated control method and system for vegetable production, which can achieve accurate decision-making on water and fertilizer supply under structural soil spatial variations.

[0005] In a first aspect, this application provides an intelligent water and fertilizer integrated control method for vegetable production, comprising the following steps:

[0006] Before irrigation and fertilization operations, time-series multispectral images of the crop canopy on each preset grid unit in the vegetable planting area are acquired;

[0007] Based on the coefficient of variation of reflectance of each spectral band in the time series of the time-series multispectral image, a stress map characterizing the crop nutrient stress type in each preset grid unit is constructed, and a soil condition map characterizing soil texture and spatial distribution of moisture is generated by the soil conductivity data of each preset grid unit in the planting area.

[0008] By performing spatial overlay analysis of the stress map and the soil condition map, heterogeneous stress patches composed of adjacent grid cells with different nutrient stress types but consistent soil conditions are identified.

[0009] Based on the heterogeneous stress patches and the nutrient stress types in the stress map, the customized nutrient element ratio for the current growth stage is determined. The customized irrigation baseline and fertilization duration for the current growth stage are calculated by combining the heterogeneous stress patches with the soil texture and moisture distribution information represented by the soil baseline map.

[0010] The integrated water and fertilizer control system delivers differentiated water and fertilizer to the heterogeneous stress patches according to the customized nutrient element ratio, the customized irrigation baseline amount, and the fertilization duration.

[0011] In some embodiments, acquiring temporal multispectral images of the crop canopy on each preset grid unit within the vegetable planting area before irrigation and fertilization operations specifically includes:

[0012] Multispectral images of the crop canopy within each preset grid unit were collected at fixed time intervals over the vegetable planting area.

[0013] Radiometric and geometric corrections are performed sequentially on the multispectral images at each acquisition time to obtain the corrected multispectral images at each acquisition time.

[0014] The calibrated multispectral images at each acquisition time are registered to align all images in space.

[0015] Using the vector boundaries of the preset grid units, spatial cropping is performed on the corrected multispectral images acquired at all times after registration to obtain temporal multispectral images of the crop canopy on each preset grid unit.

[0016] In some embodiments, the preset grid units within the vegetable planting area are divided in the following manner:

[0017] Obtain boundary vector data of the vegetable planting area, and generate an initial grid covering the entire vegetable planting area based on the boundary vector data. The initial grid is composed of multiple square cells with the same side length.

[0018] The initial mesh is spatially intersected with the boundary vector data, and square cells that are completely or partially located within the boundary vector data are retained to obtain the initial mesh.

[0019] The area of ​​the square cells that intersect with the boundary vector data in the initial grid is determined. If the proportion of the area of ​​the square cell located in the boundary vector data to its total area is lower than a preset area threshold, it is merged with the adjacent square cells whose area proportion is not lower than the preset area threshold to obtain the boundary optimized grid.

[0020] Each square cell in the boundary-optimized mesh is used as a preset mesh cell.

[0021] In some embodiments, constructing a stress map characterizing the crop nutrient stress type within each preset grid cell based on the coefficient of variation of reflectance of each spectral band in the time-series multispectral image over time specifically includes:

[0022] The reflectance of each spectral band is extracted pixel by pixel from the temporal multispectral image of the crop canopy on each preset grid unit. The coefficient of variation of the reflectance of each spectral band is calculated pixel by pixel on the time series to obtain the multi-band coefficient of variation image of each preset grid unit.

[0023] The coefficient of variation of each pixel in the multi-band coefficient of variation image in each spectral band is matched with the established nutrient stress spectral coefficient of variation feature library, which stores the feature band coefficient of variation threshold range corresponding to different nutrient stress types.

[0024] The nutrient stress type of each pixel is assigned to the pixel whose coefficient of variation meets the matching condition with the threshold range of the coefficient of variation of the characteristic band, and a nutrient stress classification map of each preset grid unit is generated.

[0025] The nutrient stress classification maps of each preset grid unit are spatially stitched together to obtain a stress map characterizing the crop nutrient stress type within each preset grid unit.

[0026] In some embodiments, generating a soil baseline map characterizing soil texture and spatial distribution of moisture using soil electrical conductivity data from each preset grid cell within the planting area specifically includes:

[0027] Conductivity measuring points are set up in each preset grid cell, and soil conductivity data of each measuring point are collected. The soil conductivity data of all measuring points in each preset grid cell are spatially interpolated to obtain the soil conductivity distribution grid of each preset grid cell.

[0028] Unsupervised classification of soil electrical conductivity distribution grids in each preset grid unit is performed, and the electrical conductivity values ​​are divided into multiple base condition categories that characterize different soil textures and moisture combinations.

[0029] The classification result of the base condition category is assigned to each raster cell in the corresponding preset grid unit to generate a soil base condition classification raster for each preset grid unit;

[0030] The soil condition classification grids of each preset grid unit are spatially stitched together to obtain a soil condition map that characterizes the spatial distribution of soil texture and moisture.

[0031] In some embodiments, determining a customized nutrient element ratio for the current reproductive stage based on the heterogeneous stress patches and the nutrient stress type in the stress map specifically includes:

[0032] Extract all nutrient stress types contained in the stress map for each heterogeneous stress patch, and use the area ratio of each nutrient stress type in the heterogeneous stress patch as the stress weight;

[0033] The nutrient deficiency correction amounts corresponding to each type of nutritional stress are weighted and fused according to the stress weights to obtain the multi-element composite correction amount of the heterogeneous stress plaques in the basic nutrient solution formula at the current reproductive stage.

[0034] Obtain the standard nutrient solution formula for crops at the current growth stage, and superimpose the multi-element composite correction amount with the baseline content of each nutrient element in the standard nutrient solution formula for crops to generate a customized nutrient element ratio for the current growth stage.

[0035] In some embodiments, a drone equipped with a multispectral sensor is used to collect multispectral images of the crop canopy within each preset grid cell at fixed time intervals over a vegetable growing area.

[0036] Secondly, this application provides an intelligent water and fertilizer integrated control system for vegetable production, used to execute an intelligent water and fertilizer integrated control method for vegetable production, the system comprising:

[0037] The acquisition module is used to acquire temporal multispectral images of the crop canopy on each preset grid unit within the vegetable planting area before irrigation and fertilization operations.

[0038] The processing module is used to construct a stress map characterizing the crop nutrient stress type in each preset grid unit based on the coefficient of variation of reflectance of each spectral band in the time series of the time-series multispectral image, and to generate a soil condition map characterizing soil texture and spatial distribution of moisture through soil conductivity data of each preset grid unit in the planting area.

[0039] The processing module is also used to perform spatial overlay analysis of the stress map and the soil condition map to identify heterogeneous stress patches composed of adjacent grid cells with different nutrient stress types but consistent soil conditions.

[0040] The processing module is also used to determine the customized nutrient element ratio for the current growth stage based on the heterogeneous stress patches and the nutrient stress type in the stress map, and to calculate the customized irrigation baseline and fertilization duration for the current growth stage by combining the heterogeneous stress patches with the soil texture and water distribution information represented by the soil baseline map.

[0041] The execution module is used to control the integrated water and fertilizer execution terminal to provide differentiated water and fertilizer supply to the heterogeneous stress patches according to the customized nutrient element ratio, the customized irrigation benchmark amount and fertilization duration.

[0042] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described intelligent water and fertilizer integrated control method for vegetable production.

[0043] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent water and fertilizer integrated control method for vegetable production.

[0044] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0045] The intelligent water and fertilizer integrated control method and system for vegetable production provided in this application firstly acquires temporal multispectral images of the crop canopy on each preset grid unit within the vegetable planting area before irrigation and fertilization operations; secondly, based on the coefficient of variation of reflectance of each spectral band in the temporal multispectral images over time, a stress map characterizing the type of crop nutrient stress within each preset grid unit is constructed, and a soil condition map characterizing soil texture and spatial distribution of moisture is generated using soil conductivity data of each preset grid unit within the planting area; furthermore, the stress map and the soil condition map are spatially overlaid for analysis. The process involves identifying heterogeneous stress patches composed of adjacent grid cells with different nutrient stress types but consistent soil conditions. Then, based on the heterogeneous stress patches and the nutrient stress types in the stress map, a customized nutrient element ratio for the current growth stage is determined. The customized irrigation baseline and fertilization duration for the current growth stage are calculated using the heterogeneous stress patches combined with soil texture and moisture distribution information represented by the soil condition map. Finally, the integrated water and fertilizer management system is controlled to provide differentiated water and fertilizer supply to the heterogeneous stress patches according to the customized nutrient element ratio, the customized irrigation baseline, and the fertilization duration.

[0046] Therefore, this application demonstrates that it can achieve accurate decision-making regarding water and fertilizer supply under structural soil spatial variations. In summary: First, by acquiring temporal multispectral images of the crop canopy on each preset grid unit before irrigation and fertilization operations, single spectral acquisition is expanded into time-series monitoring, laying a data foundation for subsequent identification of nutrient stress based on temporal fluctuation characteristics. Second, a stress map is constructed based on the coefficient of variation of reflectance in each band of the temporal multispectral image over time. Using the coefficient of variation instead of absolute reflectance as the discriminant feature effectively distinguishes between spectral changes during normal growth and those caused by nutrient deficiency. This study analyzes abnormal fluctuations to achieve precise identification and spatial localization of various nutrient stress types, including nitrogen, phosphorus, and potassium. Soil conductivity data from each grid cell is used to generate soil baseline maps via spatial interpolation and unsupervised classification. This allows for the simultaneous characterization of the spatial combination of soil texture and moisture using a single, easily measurable parameter, avoiding the complex process of separately measuring multiple soil physical indicators. Furthermore, spatial overlay analysis of the stress map and soil baseline map is performed to generate heterogeneous stress patches through heterogeneous stress adjacency pair identification and region growth algorithms. This accurately pinpoints the cause of stress while excluding differences in soil background conditions. This study investigates heterogeneous stress regions caused by nutrient deficiency, enabling spatial analysis of different nutrient stress zones within the same soil area for zoned treatment. Then, based on the heterogeneous stress patches and the nutrient stress types in the stress map, customized nutrient element ratios are determined, avoiding formulation deviations caused by a single stress assumption. The irrigation baseline and fertilization duration are calculated by combining the texture and moisture information from the soil baseline map for the heterogeneous stress patches. This ensures that the irrigation amount accurately matches the actual water-holding capacity and current moisture status of the soil within the patch, guaranteeing the timing of water and fertilizer supply. This effectively prevents further exacerbation of root stress in water-stressed areas due to limited root water absorption. Salt accumulation in the soil creates a vicious cycle of stress. Finally, the integrated water and fertilizer management system implements differentiated water and fertilizer supply based on a zoned supply prescription map and a grid-level supply control sequence. This transforms customized parameters at the patch level into fertilizer uptake ratio and injection timing instructions, which are then executed sequentially along the walking path. This avoids inaccurate water and fertilizer supply decisions caused by structural soil spatial variations, enabling automated and precise operation of multi-channel online mother liquor mixing and patch-level variable irrigation and fertilization. The technical solution provided in this application can achieve accurate water and fertilizer supply decisions under structural soil spatial variations. Attached Figure Description

[0047] Figure 1 This is an exemplary flowchart of an intelligent water and fertilizer integrated control method for vegetable production, according to some embodiments of this application;

[0048] Figure 2 This is a flowchart illustrating the division of grid cells within a vegetable planting area according to some embodiments of this application;

[0049] Figure 3 This is an exemplary flowchart illustrating the determination of a soil condition map according to some embodiments of this application;

[0050] Figure 4 This is a schematic diagram of the structure of an intelligent water and fertilizer integrated control system for vegetable production, as shown in some embodiments of this application;

[0051] Figure 5 This is a schematic diagram of the structure of a computer device for implementing an intelligent water and fertilizer integration control method for vegetable production, according to some embodiments of this application. Detailed Implementation

[0052] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] refer to Figure 1 The figure is an exemplary flowchart of an intelligent water and fertilizer integrated control method for vegetable production according to some embodiments of this application. The figure mainly includes the following steps:

[0054] In step S101, before irrigation and fertilization operations, time-series multispectral images of the crop canopy on each preset grid unit within the vegetable planting area are acquired.

[0055] In some embodiments, acquiring temporal multispectral images of the crop canopy on each preset grid unit within the vegetable planting area before irrigation and fertilization operations is achieved through the following steps:

[0056] Multispectral images of the crop canopy within each preset grid unit were collected at fixed time intervals over the vegetable planting area.

[0057] Radiometric and geometric corrections are performed sequentially on the multispectral images at each acquisition time to obtain the corrected multispectral images at each acquisition time.

[0058] The calibrated multispectral images at each acquisition time are registered to align all images in space.

[0059] Using the vector boundaries of the preset grid units, spatial cropping is performed on the corrected multispectral images acquired at all times after registration to obtain temporal multispectral images of the crop canopy on each preset grid unit.

[0060] In practice, firstly, a drone equipped with a multispectral sensor is used to collect multispectral images of the crop canopy within each preset grid unit at fixed time intervals over the vegetable planting area. Each spectral band corresponds to the reflection intensity of the crop canopy at different center wavelengths. The fixed time interval refers to the equal duration between two consecutive acquisitions, for example, once every two days. Secondly, the original multispectral images at each acquisition time are sequentially subjected to radiometric and geometric corrections. Radiometric correction uses a diffuse reflection standard reference board on the drone, which collects reference images before takeoff, to convert the digital count values ​​of each pixel in the original image into surface reflectance, eliminating radiance distortion introduced by solar altitude angle, atmospheric scattering, and sensor response differences. Geometric correction combines the positioning and orientation parameters recorded in the drone's flight log and the coordinates of control points deployed on the ground, using a collinearity equation model to solve for the geographic coordinates pixel by pixel, resampling the image to a unified map projection coordinate system to obtain a corrected multispectral image with accurate spatial location and true reflectance values. Then, the images at each acquisition time are... The corrected multispectral images are registered using a reference image acquired at any given time. An image registration algorithm based on maximizing mutual information is employed to search for corresponding image points with the highest similarity to the reference image feature points in the corrected multispectral images acquired at other times. A spatial mapping relationship between the remaining images and the reference image is established through an affine transformation model, and the images are resampled to achieve pixel-level alignment of the corrected multispectral images acquired at all times. Finally, the vector boundaries of a preset grid unit are used to spatially crop the registered corrected multispectral images acquired at all times. The vector polygons corresponding to each preset grid unit are used as masks, and a subset of pixels within the polygon range is extracted from the aligned corrected multispectral images acquired at each time. These subsets are then stacked in chronological order of acquisition to form a temporal multispectral image of the crop canopy on the preset grid unit. Here, the vector boundary refers to the spatial range planar vector data of each square unit determined in the early stage by the boundary of the planting area, after grid division and boundary optimization.

[0061] It should be noted that, in this application, the temporal multispectral image refers to a raster time series dataset composed of surface reflectance images of each spectral band at multiple consecutive acquisition times within the same preset grid cell. This temporal multispectral image can provide a basic data source for subsequent pixel-by-pixel calculation of the coefficient of variation of spectral band reflectance in the time series and construction of stress maps.

[0062] In some embodiments, the preset grid units within the vegetable planting area are divided in the following manner:

[0063] Obtain boundary vector data of the vegetable planting area, and generate an initial grid covering the entire vegetable planting area based on the boundary vector data. The initial grid is composed of multiple square cells with the same side length.

[0064] The initial mesh is spatially intersected with the boundary vector data, and square cells that are completely or partially located within the boundary vector data are retained to obtain the initial mesh.

[0065] The area of ​​the square cells that intersect with the boundary vector data in the initial grid is determined. If the proportion of the area of ​​the square cell located in the boundary vector data to its total area is lower than a preset area threshold, it is merged with the adjacent square cells whose area proportion is not lower than the preset area threshold to obtain the boundary optimized grid.

[0066] Each square cell in the boundary-optimized mesh is used as a preset mesh cell.

[0067] In practice, firstly, a positioning device is used to measure points along the actual boundary of the vegetable planting area to obtain the geodetic coordinate sequence of the boundary points. These boundary point coordinates are then imported into a geographic information system (GIS) software. A point set to surface generation tool is used to construct boundary vector data for the vegetable planting area. This boundary vector data refers to closed polygonal surface features enclosed by ordered boundary points, used to describe the spatial extent of the planting area. Subsequently, based on the circumscribed rectangle of this boundary vector data, the grid side length is determined according to the effective coverage diameter of the crop canopy and the ground sampling distance of the UAV multispectral imagery. For example, the side length is taken as an integer multiple of the crop row spacing or a fraction of the coverage width of a single frame image. A net creation tool is used to generate an initial grid within the circumscribed rectangle, consisting of multiple square units with identical side lengths. This initial grid refers to a regular grid covering the minimum bounding rectangle of the planting area, and a square unit refers to a surface division unit with equal sides and area. Secondly, a spatial intersection analysis tool is used to overlay the initial grid layer with the boundary vector data layer to determine the spatial relationship between each square unit and the boundary vector data. Square units completely located within the boundary vector data and those related to the boundary vector data are retained. Intersecting square cells are discarded, and those completely outside the boundary vector data are removed to obtain the initial grid. The spatial intersection analysis refers to the topological operation that determines whether the features of two layers share a spatial region. The initial grid refers to the set of valid square cells retained after boundary constraint filtering. Then, the area of ​​each square cell intersecting with the boundary vector data in the initial grid is judged one by one. The ratio of the area of ​​each intersecting cell within the boundary vector data to the total area of ​​the cell is calculated. Cells with a ratio lower than a preset area threshold are marked as fragment cells. The preset area threshold is set according to the ratio of the minimum effective working width of the water and fertilizer integration execution terminal to the area of ​​the square cell. For example, it can be 0.3 or 0.5. Each fragment cell is merged with the adjacent cells whose area ratio is not lower than the preset area threshold. The merging operation is performed by a fusion tool according to the shared boundary elimination rule of adjacent surface features, so that the fragment cell and its adjacent valid cells are merged into the same surface feature, resulting in the boundary optimized grid. The boundary optimized grid refers to the set of regular square cells after eliminating the narrow fragment cells at the edge of the planting area. Finally, each square cell in the boundary optimized grid is used as a preset grid cell.

[0068] refer to Figure 2 The figure is a flowchart of the grid cell division in a vegetable planting area according to some embodiments of this application. The figure first delineates the boundary of the irregular vegetable planting area, and then generates a regular initial square grid in the area. Subsequently, fragmented grids with insufficient area at the boundary are marked, and boundary optimization is completed by merging these fragmented units to form a regular boundary optimization grid without scattered units as the final grid unit, which provides a unified minimum spatial unit for subsequent image acquisition, stress analysis and differentiated water and fertilizer operations.

[0069] It should be noted that the preset grid unit in this application refers to the smallest spatial unit after the vegetable planting area is spatially divided. It consists of square units with the same side length in the boundary optimization grid. Since the crop growth and soil properties in the planting area are not uniformly distributed in space, if the entire planting area is treated as a single operation unit for uniform water and fertilizer supply, the differentiated needs caused by differences in soil texture or nutrient deficiency in local areas will be ignored, resulting in low efficiency of water and fertilizer resource utilization. After dividing the planting area into multiple preset grid units, the temporal multispectral image is cropped with each grid unit as the basic unit. This can accurately extract the spectral information of the crop canopy within the spatial range, so that the coefficient of variation of reflectance of each band and the constructed stress map have clear spatial assignments.

[0070] In step S102, based on the coefficient of variation of reflectance of each spectral band in the time series of the time-series multispectral image, a stress map characterizing the crop nutrient stress type in each preset grid unit is constructed, and a soil condition map characterizing soil texture and spatial distribution of moisture is generated by the soil electrical conductivity data of each preset grid unit in the planting area.

[0071] In some embodiments, constructing a stress map characterizing the crop nutrient stress type within each preset grid cell based on the coefficient of variation of reflectance of each spectral band in the time-series multispectral image over time is achieved through the following steps:

[0072] The reflectance of each spectral band is extracted pixel by pixel from the temporal multispectral image of the crop canopy on each preset grid unit. The coefficient of variation of the reflectance of each spectral band is calculated pixel by pixel on the time series to obtain the multi-band coefficient of variation image of each preset grid unit.

[0073] The coefficient of variation of each pixel in the multi-band coefficient of variation image in each spectral band is matched with the established nutrient stress spectral coefficient of variation feature library, which stores the feature band coefficient of variation threshold range corresponding to different nutrient stress types.

[0074] The nutrient stress type of each pixel is assigned to the pixel whose coefficient of variation meets the matching condition with the threshold range of the coefficient of variation of the characteristic band, and a nutrient stress classification map of each preset grid unit is generated.

[0075] The nutrient stress classification maps of each preset grid unit are spatially stitched together to obtain a stress map characterizing the crop nutrient stress type within each preset grid unit.

[0076] In practice, firstly, a time-series multispectral image of the crop canopy is read from each preset grid unit. This time-series multispectral image contains surface reflectance raster data of multiple spectral bands, including blue, green, red, red-edge, and near-infrared bands, acquired at consecutive acquisition times. A raster calculation tool is used to extract reflectance layer by layer according to band, and these are stacked along the time dimension to form a reflectance time-series vector at each pixel location. For each pixel, the ratio of the standard deviation to the mean of the time-series vector is calculated band by band to obtain the coefficient of variation. The coefficient of variation measures the relative dispersion of the time-series data, reflecting the strength of the fluctuation of the crop canopy spectral reflectance at that pixel over time. The coefficients of variation of all pixels are combined into a multi-band raster file with the same spatial resolution as the original image, resulting in a multi-band coefficient of variation image for each preset grid unit. The multi-band coefficient of variation image refers to a multi-channel image composed of raster layers of coefficients of variation corresponding to each spectral band. Next, the multi-band coefficient of variation image is matched pixel-by-pixel with an established nutrient stress spectral coefficient of variation feature library. This feature library is stored in the form of a relational data table, where each record corresponds to a nutrient stress type and includes the lower and upper threshold values ​​of the coefficient of variation for the characteristic bands of that stress type. For example, the characteristic bands corresponding to nitrogen stress are the red-edge band and the near-infrared band, with their coefficient of variation threshold ranges set to 0.15 to 0.30 and 0.12 to 0.25, respectively; the characteristic bands corresponding to phosphorus stress are the green band and the red band, with threshold ranges set to 0.10 to 0.20 and 0.08 to 0.18, respectively; and the characteristic bands corresponding to potassium stress are the near-infrared band and the short-wave infrared band, with threshold ranges set to 0.18 to 0.35 and 0.14 to 0.28. During matching, a multi-condition discrimination rule is adopted. This involves extracting the measured coefficient of variation (COP) values ​​of the current pixel in the specified feature bands of the feature library and comparing them one by one with the COP threshold range for each nutrient stress type. If the COP of the pixel falls within the threshold range corresponding to a certain nutrient stress type in all feature bands, then the matching condition is satisfied. Then, pixels that meet the matching condition are assigned a category code corresponding to the nutrient stress type. For example, no stress is assigned a value of 0, nitrogen stress is assigned a value of 1, phosphorus stress is assigned a value of 2, and potassium stress is assigned a value of 3. For pixels that simultaneously meet the matching conditions for multiple stress types, they are sorted according to the degree of deviation of the COP from the median threshold value, and the stress type with the largest deviation is taken as the classification of the pixel. The results are then analyzed. All pixels that do not meet any stress type matching condition are assigned a normal type category code, generating a nutrient stress classification map for each preset grid unit. This nutrient stress classification map refers to a single-band raster image where the category code is used as the pixel value, and each pixel value represents the category of nutrient stress experienced by the crop at that location. Finally, the nutrient stress classification maps of each preset grid unit are stitched together according to their spatial position within the planting area. Since the grid units are adjacent to each other and have unified spatial coordinates, the pixel row and column numbers at the junction of adjacent units are continuous and non-overlapping. These are spatially arranged and merged into a complete raster dataset covering the entire planting area, resulting in a stress map characterizing the crop nutrient stress type within each preset grid unit.

[0077] It should be noted that the stress map in this application refers to a single-band raster image of the spatial distribution of crop nutrient stress types within a vegetable growing area. Existing technologies mostly rely on the spectral characteristics of single or a small number of temporal images, and use vegetation index thresholds or spectral matching for stress diagnosis. Such methods are prone to misjudging normal spectral changes caused by the progression of crop growth as nutrient stress, and are difficult to distinguish subtle differences in spectral response caused by different nutrient deficiencies. This solution calculates the ratio of the standard deviation to the mean of the reflectance of each band in the temporal multispectral image pixel by pixel, and extracts the coefficient of variation image reflecting the intensity of temporal fluctuations in the spectrum. The dynamic change information in the time dimension is explicitly introduced into the stress discrimination process, rather than simply relying on the level of the absolute value of the spectrum. On this basis, the multi-band coefficient of variation image is matched pixel by pixel with the nutrient stress spectral coefficient of variation feature library, so that the discrimination rules directly point to the temporal fluctuation characteristics caused by different nutrient deficiencies, thereby achieving a fine distinction between various nutrient stress types.

[0078] In some embodiments, reference Figure 3 As shown, this figure is an exemplary flowchart of determining a soil baseline map according to some embodiments of this application. In this embodiment, the generation of a soil baseline map characterizing soil texture and spatial distribution of moisture by using soil electrical conductivity data of each preset grid cell within the planting area is achieved through the following steps:

[0079] In step S1021, conductivity measuring points are set up in each preset grid cell, soil conductivity data of each conductivity measuring point are collected, and the soil conductivity data of all conductivity measuring points in each preset grid cell are spatially interpolated to obtain the soil conductivity distribution grid of each preset grid cell.

[0080] In step S1022, the soil electrical conductivity distribution grid of each preset grid unit is unsupervised and classified into multiple base condition categories that characterize different soil textures and moisture combinations.

[0081] In step S1023, the classification result of the basis condition category is assigned to each grid cell in the corresponding preset grid unit to generate a soil basis condition classification grid for each preset grid unit.

[0082] In step S1024, the soil condition classification grids of each preset grid unit are spatially spliced ​​to obtain a soil condition map that characterizes the spatial distribution of soil texture and moisture.

[0083] In practice, firstly, conductivity measurement points are set up and data is collected within each preset grid cell. The measurement points are arranged in a regular, equidistant pattern within the grid. The spacing between adjacent measurement points is set according to the side length of the preset grid cell and the spatial variation of soil properties. For example, when the side length of the preset grid cell is 5 meters, the measurement point spacing is 1 meter. A 5x5 matrix of measurement points is formed within each grid cell. At each measurement point, a portable soil conductivity meter is used to collect soil conductivity data. Then, the soil conductivity data of all measurement points within each preset grid cell are spatially analyzed. Interpolation generates a continuous surface. Spatial interpolation employs inverse distance weighted interpolation. This method uses the coordinates of the center point of the raster cell to be interpolated as a reference, searches for all known conductivity measurement points within a specified search radius around the cell, and weights the conductivity values ​​of the measurement points using the reciprocal of the distance from each measurement point to the center point of the cell to be interpolated as a weight. The closer the measurement point, the greater its weight. The search radius is set to twice the distance between the measurement points to ensure the continuity of the interpolated surface and the preservation of local details. The resolution of the raster cell output is set to half the distance between the measurement points. For example, when the distance between the measurement points is 1 meter, the output raster cell size is 0.For each preset grid cell, the above interpolation operation is performed sequentially to obtain the soil conductivity distribution raster of each preset grid cell. The soil conductivity distribution raster refers to a single-band raster image with continuously changing conductivity values ​​as pixel values, reflecting the spatial continuous distribution pattern of soil conductivity within the grid cell. Next, unsupervised classification is performed on the soil conductivity distribution raster of each preset grid cell. The unsupervised classification adopts an iterative self-organizing data analysis algorithm. Before execution, the maximum number of classification categories is set to a preset value, for example, 5 categories, corresponding to the various soil texture and moisture combinations to be distinguished. The algorithm implementation process is as follows: randomly initialize within the conductivity value range. The cluster center values ​​for each category are determined by assigning the conductivity value of each raster cell to the category of the nearest cluster center according to the minimum distance principle. The mean conductivity value of all cells within each category is recalculated as the new cluster center. This assignment and update process is repeated until the displacement of the cluster center between two adjacent iterations is less than a preset convergence threshold. At output, each category is assigned a value from 1 to 5 based on the mean conductivity value, from smallest to largest. Category 1 corresponds to low conductivity combinations, such as sandy soil with low water content, while Category 5 corresponds to high conductivity combinations, such as clay soil with high water content or salt accumulation. This method divides continuous conductivity values ​​into categories representing different soil textures and water conditions. Multiple baseline condition categories are defined in a combined state. Each baseline condition category refers to a discrete category obtained by segmenting and classifying soil electrical conductivity values. Each category represents a type of basic soil condition with similar soil texture and moisture characteristics. Then, the classification results of the baseline condition categories are assigned to each raster cell within a corresponding preset grid unit. The assignment method involves directly writing the category label value output from the unsupervised classification into the raster cell at the corresponding row and column position, replacing the continuous values ​​in the original soil electrical conductivity distribution raster with discrete baseline condition category codes. This generates a soil baseline condition classification raster for each preset grid unit. The soil baseline condition classification raster refers to a single-band raster image with baseline condition category codes as cell values. Pixel values ​​1 to 5 represent different soil textures and moisture combinations. Finally, the soil condition classification grids of each preset grid unit are spatially stitched together. The stitching is based on the spatial arrangement of the preset grid units in each planting area. The soil condition classification grids of each preset grid unit are read sequentially, and the pixel values ​​are written to the corresponding spatial positions according to the row and column offsets and spatial reference coordinates of each grid unit within the planting area. Since the boundaries of each grid unit are adjacent and the classification system is consistent, the condition categories at the edges of adjacent grid units maintain spatial continuity. After stitching, a soil condition map covering the entire vegetable planting area and representing the spatial distribution of soil texture and moisture is obtained.

[0084] It should be noted that the soil baseline map in this application refers to the distribution of basic soil conditions in the planting area caused by differences in soil texture and uneven spatial distribution of moisture. This soil baseline map uses the soil apparent electrical conductivity, a comprehensive parameter that can be collected quickly and at high density on-site, as the data source. Through spatial interpolation, the electrical conductivity values ​​of discrete measuring points are transformed into a grid surface that reflects the continuous distribution of electrical conductivity within the grid cell. Then, an iterative self-organizing data analysis algorithm is used to perform unsupervised classification of the electrical conductivity values, automatically dividing the continuous changes in electrical conductivity into a finite number of baseline categories. Each baseline category corresponds to a coupling state of soil texture and moisture content. Thus, the spatial differences in soil texture and moisture content are characterized by a single data source and an automated classification process, avoiding the complex process and error accumulation problem of obtaining the texture map and moisture content map separately and then overlaying them in traditional methods.

[0085] In step S103, the stress map and the soil condition map are spatially overlaid to identify heterogeneous stress patches composed of adjacent grid cells with different nutrient stress types but consistent soil conditions.

[0086] In some embodiments, spatial overlay analysis of the stress map and the soil condition map is performed to identify heterogeneous stress patches composed of adjacent grid cells with different nutrient stress types but consistent soil conditions. This is achieved through the following steps:

[0087] The stress map and the soil condition map are spatially overlaid. Using a preset grid cell as the analysis unit, the nutrient stress type with the largest proportion in each preset grid cell is extracted as the dominant stress type of that grid cell. At the same time, the soil condition category with the largest proportion in each preset grid cell is extracted as the dominant condition category of that grid cell.

[0088] Traverse all adjacent grid cell pairs and select adjacent grid cell pairs with the same dominant base case category but different dominant stress types as heterogeneous stress adjacent pairs;

[0089] Using the grid cells in the heterogeneous stress adjacent pair as seed cells, perform region growth on all adjacent grid cells with the same dominant stress type as the seed cell, and include grid cells that are directly adjacent to the seed cell or connected through grid cells of the same stress type, belong to the same heterogeneous stress adjacent pair and have the same dominant base case category into the same connected domain.

[0090] The spatial extent enclosed by the boundaries of each connected domain is identified as a heterogeneous stress patch.

[0091] In practice, firstly, the raster layer of the stress map and the raster layer of the soil baseline map are spatially overlaid under a unified spatial reference system. Spatial overlay refers to establishing attribute correspondence between the two raster layers pixel by pixel according to the same pixel row and column positions, so that each pixel simultaneously has two attribute values: nutrient stress type code and soil baseline category code. Using the vector boundary of the preset grid cell as the spatial statistical range, the frequency of each nutrient stress type and each soil baseline category appearing in all pixels covered by each grid cell is counted. The nutrient stress type with the highest frequency is extracted as the dominant stress type of the grid cell, and the soil baseline category with the highest frequency is extracted as the dominant baseline category of the grid cell. The dominant stress type refers to the number of pixels occupied in a grid cell. The most numerous nutrient stress type is the dominant soil condition category, which refers to the soil condition category that occupies the most pixels within a grid cell. If multiple types have the same frequency, the type with the largest spatial connectivity area is taken as the dominant type. Secondly, a spatial adjacency matrix is ​​constructed based on the adjacency relationship of grid cells. The spatial adjacency matrix is ​​an undirected graph data structure constructed with grid cells as nodes and the length of the shared boundary between cells as the edge weight. The adjacency relationship is determined using the four-neighbor rule, that is, only two grid cells sharing a complete boundary are defined as adjacent grid cell pairs. All adjacent grid cell pairs are traversed, and the dominant soil condition category code and dominant stress type code of the two grid cells in each pair are compared. If the dominant soil condition category code is the same and the dominant stress type is... If the encoding is different, the adjacent grid cell pair is marked as a heterogeneous stress adjacent pair. A heterogeneous stress adjacent pair refers to a pairing of adjacent cells with the same soil basic conditions but different nutrient stress types. Then, using each grid cell in the heterogeneous stress adjacent pair as a seed cell, a seed-filling algorithm is used to perform region growth on grid cells with the same dominant base condition category and dominant stress type. The region growth process is as follows: initialize an empty set of connected components; take an unvisited seed cell from the seed cell set; create a new connected component and add the seed cell; use the dominant stress type of the seed cell as the target stress type; use the dominant base condition category of the seed cell as the target base condition category; and check all four neighboring regions of the seed cell. If a neighboring cell is not visited and its dominance stress type is equal to the target stress type and its dominance base case category is equal to the target base case category, then the neighboring cell is added to the current connected component and pushed onto the growth stack. Cells are popped from the growth stack and the four-neighbor pair is checked and added again until the growth stack is empty, thus completing the generation of a connected component. All grid cells that have been pushed onto the growth stack during the connected component generation process are marked as visited. If the connected component contains at least one pair of grid cells from the set of heterogeneous stress neighbor pairs, then the connected component is retained. A connected component refers to a set of grid cells that have the same dominance stress type and the same dominance base case category and are spatially connected by the four-neighbor adjacency relationship. The above growth process is repeated until all seed cells have been visited.Finally, the spatial extent enclosed by the boundaries of each retained connected domain is identified as a heterogeneous stress patch. This identification method involves extracting the outer boundary contours of all mesh cells within the connected domain, eliminating these contours at the inner boundaries to obtain the outer envelope polygon of the connected domain. This polygon represents the spatial extent of the heterogeneous stress patch.

[0092] It should be noted that, in this application, heterogeneous stress patches refer to independent operational zones within a vegetable planting area that require differentiated water and fertilizer supply due to differences in crop nutrient deficiency types. The spatial range of the heterogeneous stress patches is stored in the form of a vector polygon layer. By spatially overlaying stress maps with soil condition maps, heterogeneous stress patches can identify areas where nutrient stress differences are indeed caused by insufficient nutrient supply rather than uneven soil conditions, while excluding the impact of differences in soil texture and moisture background on crop growth. This effectively avoids the waste of resources or improper supply caused by confusing soil factors with nutrient factors and applying the same water and fertilizer treatment to areas with different soil conditions but similar stress performance in traditional methods.

[0093] In step S104, the customized nutrient element ratio for the current growth stage is determined based on the heterogeneous stress patches and the nutrient stress type in the stress map. The customized irrigation baseline and fertilization duration for the current growth stage are calculated by combining the heterogeneous stress patches with the soil texture and moisture distribution information represented by the soil baseline map.

[0094] In some embodiments, determining the customized nutrient element ratio for the current reproductive stage based on the heterogeneous stress patches and the nutrient stress type in the stress map is achieved through the following steps:

[0095] Extract all nutrient stress types contained in the stress map for each heterogeneous stress patch, and use the area ratio of each nutrient stress type in the heterogeneous stress patch as the stress weight;

[0096] The nutrient deficiency correction amounts corresponding to each type of nutritional stress are weighted and fused according to the stress weights to obtain the multi-element composite correction amount of the heterogeneous stress plaques in the basic nutrient solution formula at the current reproductive stage.

[0097] Obtain the standard nutrient solution formula for crops at the current growth stage, and superimpose the multi-element composite correction amount with the baseline content of each nutrient element in the standard nutrient solution formula for crops to generate a customized nutrient element ratio for the current growth stage.

[0098] In specific implementation, firstly, the vector polygon layer of the heterogeneous stress patch is spatially overlaid with the stress map raster layer. Using the spatial extent of each heterogeneous stress patch as the clipping boundary, all pixel values ​​of the stress map within that patch's area are extracted. These pixel values ​​contain the category code for the nutrient stress type. All pixels within the heterogeneous stress patch are traversed, and the number of pixels appearing for each nutrient stress type is counted. The number of pixels for each nutrient stress type is multiplied by the area of ​​the stress map raster pixels to obtain the coverage area of ​​that nutrient stress type within the heterogeneous stress patch. This coverage area is then divided by the total area of ​​the heterogeneous stress patch to obtain the area proportion of that nutrient stress type. This area proportion is used as the stress weight for that nutrient stress type. The stress weight refers to... The contribution coefficient of nutrient stress type to the overall nutrient deficiency state of the heterogeneous stress patch is calculated, with the sum of stress weights for all nutrient stress types being 1. Then, the stress weight of each nutrient stress type is weighted and fused with a pre-defined nutrient deficiency correction table. The nutrient deficiency correction table is a two-dimensional relational table stored in the system database, with the horizontal axis representing nutrient stress type and the vertical axis representing each nutrient element. Each cell in the table records the nutrient deficiency correction per unit area for each nutrient element under the current growth stage, in kilograms per hectare. For example, during the fruit-setting period, the nitrogen nutrient deficiency correction corresponding to the nitrogen stress type is 15 kilograms per hectare, the phosphorus nutrient deficiency correction is 5 kilograms per hectare, the potassium nutrient deficiency correction is 8 kilograms per hectare, and the phosphorus nutrient deficiency correction is... The nitrogen deficiency correction amount corresponding to the stress type is 3 kg / ha, the phosphorus deficiency correction amount is 18 kg / ha, and the potassium deficiency correction amount is 6 kg / ha. The nitrogen deficiency correction amount corresponding to the potassium stress type is 4 kg / ha, the phosphorus deficiency correction amount is 4 kg / ha, and the potassium deficiency correction amount is 20 kg / ha. Weighted fusion is performed using each nutrient element as the calculation object. The deficiency correction amount of each nutrient element under each nutrient stress type is multiplied by the corresponding stress weight and summed to obtain the multi-element composite correction amount of the heterogeneous stress patch in the current growth stage of the basic nutrient solution formula. The multi-element composite correction amount refers to the additional application amount of each nutrient element per unit area for multiple nutrient stress types within the heterogeneous stress patch. The process involves: First, obtaining the standard nutrient solution formula for the crop at its current growth stage. This formula refers to the combination of baseline contents of various nutrients that meet the normal growth requirements of the crop at that stage, obtained from agricultural technology extension departments or crop cultivation manuals, based on the crop type and current growth stage. The baseline contents of nitrogen, phosphorus, and potassium are expressed in kilograms per hectare. For example, the standard nutrient solution formula for tomatoes during the fruit-setting stage is 120 kg / ha of nitrogen, 60 kg / ha of phosphorus, and 150 kg / ha of potassium. The correction values ​​of each nutrient element in the multi-element composite correction are added element-by-element to the baseline contents of the corresponding nutrient elements in the standard nutrient solution formula to obtain the customized nutrient element ratio for the heterogeneous stress patch at its current growth stage.

[0099] It should be noted that the customized nutrient element ratio in this application refers to the final combination of the application amount of each nutrient element after adjusting for the actual nutrient deficiency in a specific heterogeneous stress patch, output in the form of mass concentration or application amount per unit area, providing nutrient solution preparation parameters for the differentiated fertilization at the subsequent fertigation execution end.

[0100] In some embodiments, the following steps are used to calculate the customized irrigation baseline and fertilization duration for the current growth stage by combining the soil texture and moisture distribution information represented by the soil baseline map with the heterogeneous stress patches:

[0101] The spatial extent of each heterogeneous stress patch is overlaid with the soil condition map to extract the soil texture type and moisture state level corresponding to each soil condition category within the coverage area of ​​the heterogeneous stress patch.

[0102] The soil texture water holding coefficient and initial water content of the heterogeneous stress patches are calculated by weighting the area proportion of each soil condition category within the heterogeneous stress patches.

[0103] Obtain the standard value of daily evapotranspiration of crops at the current growth stage, correct the standard value of daily evapotranspiration with the soil texture water holding coefficient, and calculate the irrigation baseline amount required to raise the soil moisture content to the target moisture content at the current growth stage by combining the area of ​​the heterogeneous stress patch and the initial moisture content.

[0104] Based on the total mass of each nutrient element added in the customized nutrient element ratio of the heterogeneous stress patch and the irrigation baseline, combined with the rated fertilization flow rate and rated irrigation flow rate of the integrated water and fertilizer application terminal, the fertilization duration of the heterogeneous stress patch is calculated.

[0105] Specifically, the vector polygon layer of each heterogeneous stress patch is spatially overlaid with the soil baseline map raster layer under a unified spatial reference system. Spatial overlay refers to using the spatial extent of the heterogeneous stress patch as a clipping mask to extract all raster pixels of the soil baseline map within the patch's coverage area. Each pixel carries a baseline category code. Based on the mapping relationship table between the baseline category code and soil texture type and moisture status level established in the previous unsupervised classification, the baseline category code is converted into the corresponding soil texture type and moisture status level for each pixel. Soil texture type refers to the category divided according to the international soil texture classification standard, such as sandy soil, loam, and clay. Moisture status level refers to the soil moisture content grade expressed as field capacity percentage, such as low (below 40% field capacity), medium (40% to 70% field capacity), and high (above 70% field capacity). All combinations of soil texture type and moisture status level appearing within the coverage area of ​​the heterogeneous stress patch are extracted, and the pixel area ratio of each combination category within the patch is calculated.

[0106] Specifically, the soil texture water-holding coefficient and initial moisture content of the heterogeneous stress patch are calculated by weighting the area proportion of each soil condition category within the patch. The soil texture water-holding coefficient is a dimensionless parameter characterizing the difference in water retention capacity among different soil texture types, with a value ranging from 0 to 1, corresponding to 0.4 for sandy soil, 0.7 for loam, and 1.0 for clay. The initial moisture content refers to the weighted average of the soil volumetric moisture content before irrigation, expressed as a percentage. It is calculated by multiplying the median moisture content of each soil condition category by its area proportion and then summing the results. For example, the median moisture content for low moisture state is 20%, for medium moisture state it is 55%, and for high moisture state it is 85%. At the same time, the soil texture water-holding coefficient of each soil condition category is multiplied by its area proportion and then summed to obtain the soil texture water-holding coefficient of the patch.

[0107] Specifically, the standard value of daily evapotranspiration for the crop at its current growth stage is obtained. This standard value refers to the daily water consumption per unit area at the current growth stage, obtained by multiplying the reference crop evapotranspiration by the crop coefficient under standard conditions. The actual daily evapotranspiration is then obtained by correcting the standard value using the soil texture water-holding coefficient. This correction method reflects the impact of different soil textures on the actual water consumption rate of the crop due to differences in water-holding capacity. The difference in water content required to replenish the crop is obtained by subtracting the initial water content from the target water content at the current growth stage. The target water content refers to the water content required to replenish the crop at the current growth stage. The upper limit of soil volumetric water content required for normal growth of crops during the growing stage is determined. For example, the target water content for tomatoes during the fruit-setting period is set at 80% of field capacity. The difference in water content is multiplied by the area of ​​the heterogeneous stress patch and the depth of the crop root layer to obtain the total volume of water that needs to be replenished in the patch. The total volume of water that needs to be replenished is then divided by the irrigation water utilization coefficient to obtain the irrigation baseline. The irrigation baseline refers to the total amount of irrigation water required to raise the soil water content in the heterogeneous stress patch from the current state to the target water content. The irrigation water utilization coefficient is taken as 0.9 to account for evaporation and deep seepage losses during the irrigation process.

[0108] Specifically, the total volume of the required mother liquor is calculated based on the total mass of each nutrient element in the customized nutrient element ratio for heterogeneous stress patches. The total mass of each nutrient element is obtained by multiplying the application rate per unit area of ​​each element in the customized nutrient element ratio by the area of ​​the heterogeneous stress patch. The required volume of the mother liquor for each nutrient element is obtained by dividing the total mass of each nutrient element by the mass concentration of its corresponding mother liquor. The total volume of the water-fertilizer mixture is obtained by summing the required volumes of all nutrient element mother liquors and adding the volume of dilution water. The irrigation reference flow rate is used as the control total volume of the water-fertilizer mixture. The total irrigation duration is obtained by dividing the irrigation reference flow rate by the rated irrigation flow rate of the water-fertilizer integration actuator. The rated irrigation flow rate refers to the time when the irrigation pipeline at the actuator reaches the specified value. The water output per unit time under standard working pressure is defined as the fertilization time, which is the time required to inject the portion of the nutrient element mother liquor from the total volume of the water-fertilizer mixture into the main irrigation pipeline at the rated fertilization flow rate. The rated fertilization flow rate refers to the volume of mother liquor drawn by the Venturi injector or metering pump at the actuator per unit time. The fertilization time is calculated by dividing the sum of the required volumes of each nutrient element mother liquor by the rated fertilization flow rate. The start time of the fertilization time is set at the beginning of the total irrigation time to ensure sufficient flushing time after the mother liquor is injected, so as to ensure that there is no fertilizer residue in the pipeline. The calculated irrigation baseline volume and fertilization time are output in units of this heterogeneous stress patch, providing operational parameters for differentiated water and fertilizer supply at the water-fertilizer integration actuator.

[0109] In step S105, the integrated water and fertilizer control unit provides differentiated water and fertilizer supply to the heterogeneous stress patches according to the customized nutrient element ratio, the customized irrigation benchmark amount, and the fertilization duration.

[0110] In some embodiments, controlling the integrated water and fertilizer application terminal to provide differentiated water and fertilizer supply to the heterogeneous stress patches according to the customized nutrient element ratio, the customized irrigation baseline amount, and the fertilization duration is achieved through the following steps:

[0111] Based on the customized nutrient element ratio, a fertilizer absorption ratio instruction corresponding to each nutrient element mother liquor channel is generated. A fertilizer injection timing instruction is generated based on the irrigation baseline amount and fertilizer injection duration. The fertilizer absorption ratio instruction and the fertilizer injection timing instruction are jointly encapsulated as differentiated supply control parameters.

[0112] The differentiated supply control parameters are grouped into a partitioned supply prescription map according to the spatial location of the patches, and the partitioned supply prescription map is spatially associated with the boundary vector of the preset grid unit to generate a supply control sequence for the walking path of the water and fertilizer integration execution terminal.

[0113] The integrated water and fertilizer control unit moves sequentially to each preset grid cell according to the supply control sequence. When it enters the grid cell covered by the heterogeneous stress patch, it calls the corresponding differentiated supply control parameters to drive the fertilizer absorption channel and irrigation valve to perform water and fertilizer ratio and quantitative supply according to the fertilizer absorption ratio command and the fertilizer injection timing command.

[0114] In practice, firstly, the customized nutrient element ratio for each heterogeneous stress patch is read. This ratio is expressed as the application rate of each nutrient element per unit area. The application rate of each nutrient element is then converted into the nutrient uptake ratio of the corresponding mother liquor channel. The conversion method is to divide the application rate of each nutrient element by its mother liquor mass concentration to obtain the required mother liquor volume per unit area. Then, the required mother liquor volume of each nutrient element is normalized to a ratio value relative to the maximum mother liquor volume, generating a set of nutrient uptake ratio instructions with values ​​ranging from 0 to 1. The nutrient uptake ratio instructions refer to analog signals that control the opening of the flow regulating valves of each nutrient uptake channel, such as nitrogen channel opening 0.8, phosphorus channel opening 0.5, and potassium channel opening 1.0. At the same time, the irrigation reference amount and fertilization duration of the heterogeneous stress patch are read. The irrigation reference amount is divided by the rated irrigation flow rate of the water and fertilizer integration execution terminal to obtain the total Irrigation duration and fertilization duration are the time required for the entire mother liquor to be injected. This generates fertilization timing instructions, which define the timing control parameters for the duration of irrigation valve opening and the opening time windows of each fertilization channel. These include the total irrigation start time, fertilization start time, fertilization end time, and total irrigation stop time. The fertilization start time is set to the 5th second after the total irrigation start time to allow the pipeline to fill with water and establish a stable flow rate. The fertilization end time is set to the fertilization start time plus the fertilization duration. The total irrigation stop time is set to the total irrigation start time plus the total irrigation duration. The fertilization ratio instruction and fertilization timing instructions are combined and encapsulated into a differentiated supply control parameter data package according to a custom data frame format. The differentiated supply control parameters refer to a complete set of water and fertilizer supply execution parameters corresponding to a single heterogeneous stress patch.Secondly, the differentiated supply control parameters of all heterogeneous stress patches are sorted and grouped according to the spatial coordinates of the patch center point to generate a zonal supply prescription map. This zonal supply prescription map refers to the mapping between the spatial range of heterogeneous stress patches and their corresponding differentiated supply control parameters, stored as a vector polygon layer. Each polygon feature's attribute table contains a patch number and a binary encoding field for the differentiated supply control parameters. The zonal supply prescription map is spatially correlated with the boundary vectors of preset grid units. The spatial correlation uses a polygon overlay analysis tool to perform intersection operations. For each preset grid unit, it is determined whether its spatial range intersects with any heterogeneous stress patch. If they intersect, the differentiated supply control parameters of that heterogeneous stress patch are assigned to that preset grid unit; otherwise, the default water and fertilizer supply parameters are assigned. The default water and fertilizer supply parameters refer to the preset control parameters for uniform supply using the current growth stage's standard nutrient solution formula and standard irrigation volume. Finally, based on the predetermined walking path of the integrated water and fertilizer application terminal within the planting area, the walking path refers to the path along the preset grid units. The ordered spatial trajectory formed by the serpentine or reciprocating movement of the grid cells arranges each preset grid cell into a linear sequence according to the order in which they are traversed along the path. The differentiated supply control parameters or default water and fertilizer supply parameters of each grid cell in the sequence are packaged with the coordinates of the grid cell's center point and the trigger geofence radius for entering that grid cell, generating a supply control sequence for the fertigation execution terminal's path. This supply control sequence refers to a queue of fertilizer and irrigation control instructions arranged according to the execution terminal's movement order. Each record in the queue contains trigger position coordinates, trigger distance thresholds, and execution parameter data packets. Finally, the fertigation execution terminal is controlled to move sequentially to each preset grid cell according to the supply control sequence. When entering a grid cell covered by a heterogeneous stress patch, the corresponding differentiated supply control parameters are invoked, driving the fertilizer absorption channel and irrigation valve to execute water and fertilizer ratio and quantitative supply according to the fertilizer absorption ratio command and the fertilizer injection timing command. After the current grid cell is completed, the controller retrieves the next record from the supply control sequence and repeats the above process until all records are completed.

[0115] In another aspect, in some embodiments, this application provides an intelligent integrated water and fertilizer control system for vegetable production, see reference. Figure 4 The figure is a schematic diagram of the structure of an intelligent water and fertilizer integrated control system for vegetable production according to some embodiments of this application. The intelligent water and fertilizer integrated control system for vegetable production includes: an acquisition module 201, a processing module 202, and an execution module 203, which are described below:

[0116] The acquisition module 201 in this application is mainly used to acquire temporal multispectral images of the crop canopy on each preset grid unit in the vegetable planting area before irrigation and fertilization operations.

[0117] Processing module 202, in this application, is mainly used to construct a stress map characterizing the crop nutrient stress type in each preset grid unit based on the coefficient of variation of reflectance of each spectral band in the time series of the time-series multispectral image, and to generate a soil condition map characterizing soil texture and spatial distribution of water through the soil conductivity data of each preset grid unit in the planting area.

[0118] The processing module 202 is also used to perform spatial overlay analysis of the stress map and the soil condition map to identify heterogeneous stress patches composed of adjacent grid cells with different nutrient stress types but consistent soil conditions.

[0119] In addition, the processing module 202 is also used to determine the customized nutrient element ratio for the current growth stage based on the heterogeneous stress patches and the nutrient stress type in the stress map, and to calculate the customized irrigation baseline and fertilization duration for the current growth stage by combining the soil texture and water distribution information represented by the soil baseline map with the heterogeneous stress patches.

[0120] The execution module 203 in this application is mainly used to control the water and fertilizer integration execution terminal to provide differentiated water and fertilizer supply to the heterogeneous stress patches according to the customized nutrient element ratio, the customized irrigation benchmark amount and the fertilization duration.

[0121] In addition, this application also provides a computer device, which includes a memory and a processor. The memory stores code, and the processor is configured to acquire the code and execute the above-described intelligent water and fertilizer integrated control method for vegetable production.

[0122] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device for implementing an intelligent water and fertilizer integrated control method for vegetable production, according to some embodiments of this application. The intelligent water and fertilizer integrated control method for vegetable production in the above embodiments can be implemented through... Figure 5 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0123] The processor 301 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the intelligent water and fertilizer integration control method for vegetable production in this application.

[0124] The communication bus 302 can be used to transmit information between the aforementioned components.

[0125] The memory 303 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 303 may exist independently and be connected to the processor 301 via the communication bus 302. The memory 303 may also be integrated with the processor 301.

[0126] The memory 303 stores program code for executing the scheme of this application, and its execution is controlled by the processor 301. The processor 301 executes the program code stored in the memory 303. The program code may include one or more software modules. The determination of the intelligent water and fertilizer integration control method for vegetable production in the above embodiments can be achieved by the processor 301 and one or more software modules in the program code in the memory 303.

[0127] Communication interface 304 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0128] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0129] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0130] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent water and fertilizer integrated control method for vegetable production.

[0131] Although preferred embodiments of this application 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.

[0132] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application.

Claims

1. A smart water and fertilizer integrated control method for vegetable production, characterized in that, Includes the following steps: Before irrigation and fertilization operations, time-series multispectral images of the crop canopy on each preset grid unit in the vegetable planting area are acquired; Based on the coefficient of variation of reflectance of each spectral band in the time series of the time-series multispectral image, a stress map characterizing the crop nutrient stress type in each preset grid unit is constructed, and a soil condition map characterizing soil texture and spatial distribution of moisture is generated by the soil conductivity data of each preset grid unit in the planting area. By performing spatial overlay analysis of the stress map and the soil condition map, heterogeneous stress patches composed of adjacent grid cells with different nutrient stress types but consistent soil conditions are identified. Based on the heterogeneous stress patches and the nutrient stress types in the stress map, the customized nutrient element ratio for the current growth stage is determined. The customized irrigation baseline and fertilization duration for the current growth stage are calculated by combining the heterogeneous stress patches with the soil texture and moisture distribution information represented by the soil baseline map. The integrated water and fertilizer control system delivers differentiated water and fertilizer to the heterogeneous stress patches according to the customized nutrient element ratio, the customized irrigation baseline amount, and the fertilization duration.

2. The method as described in claim 1, characterized in that, Before irrigation and fertilization operations, acquiring temporal multispectral images of the crop canopy on each preset grid unit within the vegetable planting area specifically includes: Multispectral images of the crop canopy within each preset grid unit were collected at fixed time intervals over the vegetable planting area. Radiometric and geometric corrections are performed sequentially on the multispectral images at each acquisition time to obtain the corrected multispectral images at each acquisition time. The calibrated multispectral images at each acquisition time are registered to align all images in space. Using the vector boundaries of the preset grid units, spatial cropping is performed on the corrected multispectral images acquired at all times after registration to obtain temporal multispectral images of the crop canopy on each preset grid unit.

3. The method as described in claim 1, characterized in that, The pre-defined grid units within the vegetable planting area are divided in the following manner: Obtain boundary vector data of the vegetable planting area, and generate an initial grid covering the entire vegetable planting area based on the boundary vector data. The initial grid is composed of multiple square cells with the same side length. The initial mesh is spatially intersected with the boundary vector data, and square cells that are completely or partially located within the boundary vector data are retained to obtain the initial mesh. The area of ​​the square cells that intersect with the boundary vector data in the initial grid is determined. If the proportion of the area of ​​the square cell located in the boundary vector data to its total area is lower than a preset area threshold, it is merged with the adjacent square cells whose area proportion is not lower than the preset area threshold to obtain the boundary optimized grid. Each square cell in the boundary-optimized mesh is used as a preset mesh cell.

4. The method as described in claim 1, characterized in that, Based on the coefficient of variation of reflectance of each spectral band in the time-series multispectral image, a stress map characterizing the crop nutrient stress type within each preset grid cell is constructed, specifically including: The reflectance of each spectral band is extracted pixel by pixel from the temporal multispectral image of the crop canopy on each preset grid unit. The coefficient of variation of the reflectance of each spectral band is calculated pixel by pixel on the time series to obtain the multi-band coefficient of variation image of each preset grid unit. The coefficient of variation of each pixel in the multi-band coefficient of variation image in each spectral band is matched with the established nutrient stress spectral coefficient of variation feature library, which stores the feature band coefficient of variation threshold range corresponding to different nutrient stress types. The nutrient stress type of each pixel is assigned to the pixel whose coefficient of variation meets the matching condition with the threshold range of the coefficient of variation of the characteristic band, and a nutrient stress classification map of each preset grid unit is generated. The nutrient stress classification maps of each preset grid unit are spatially stitched together to obtain a stress map characterizing the crop nutrient stress type within each preset grid unit.

5. The method as described in claim 1, characterized in that, Generating a soil condition map characterizing soil texture and spatial distribution of moisture using soil electrical conductivity data from each preset grid cell within the planting area specifically includes: Conductivity measuring points are set up in each preset grid cell, and soil conductivity data of each measuring point are collected. The soil conductivity data of all measuring points in each preset grid cell are spatially interpolated to obtain the soil conductivity distribution grid of each preset grid cell. Unsupervised classification of soil electrical conductivity distribution grids in each preset grid unit is performed, and the electrical conductivity values ​​are divided into multiple base condition categories that characterize different soil textures and moisture combinations. The classification result of the base condition category is assigned to each raster cell in the corresponding preset grid unit to generate a soil base condition classification raster for each preset grid unit; The soil condition classification grids of each preset grid unit are spatially stitched together to obtain a soil condition map that characterizes the spatial distribution of soil texture and moisture.

6. The method as described in claim 1, characterized in that, Determining the customized nutrient element ratio for the current reproductive stage based on the heterogeneous stress patches and the nutrient stress types in the stress map specifically includes: Extract all nutrient stress types contained in the stress map for each heterogeneous stress patch, and use the area ratio of each nutrient stress type in the heterogeneous stress patch as the stress weight; The nutrient deficiency correction amounts corresponding to each type of nutritional stress are weighted and fused according to the stress weights to obtain the multi-element composite correction amount of the heterogeneous stress plaques in the basic nutrient solution formula at the current reproductive stage. Obtain the standard nutrient solution formula for crops at the current growth stage, and superimpose the multi-element composite correction amount with the baseline content of each nutrient element in the standard nutrient solution formula for crops to generate a customized nutrient element ratio for the current growth stage.

7. The method as described in claim 1, characterized in that, Using drones equipped with multispectral sensors, multispectral images of the crop canopy within each preset grid unit are collected at fixed time intervals over the vegetable growing area.

8. An intelligent water and fertilizer integrated control system for vegetable production, used to execute the intelligent water and fertilizer integrated control method for vegetable production as described in any one of claims 1 to 7, characterized in that, The system includes: The acquisition module is used to acquire temporal multispectral images of the crop canopy on each preset grid unit within the vegetable planting area before irrigation and fertilization operations. The processing module is used to construct a stress map characterizing the crop nutrient stress type in each preset grid unit based on the coefficient of variation of reflectance of each spectral band in the time series of the time-series multispectral image, and to generate a soil condition map characterizing soil texture and spatial distribution of moisture through soil conductivity data of each preset grid unit in the planting area. The processing module is also used to perform spatial overlay analysis of the stress map and the soil condition map to identify heterogeneous stress patches composed of adjacent grid cells with different nutrient stress types but consistent soil conditions. The processing module is also used to determine the customized nutrient element ratio for the current growth stage based on the heterogeneous stress patches and the nutrient stress type in the stress map, and to calculate the customized irrigation baseline and fertilization duration for the current growth stage by combining the heterogeneous stress patches with the soil texture and water distribution information represented by the soil baseline map. The execution module is used to control the integrated water and fertilizer execution terminal to provide differentiated water and fertilizer supply to the heterogeneous stress patches according to the customized nutrient element ratio, the customized irrigation benchmark amount and fertilization duration.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the intelligent water and fertilizer integrated control method for vegetable production as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent water and fertilizer integrated control method for vegetable production as described in any one of claims 1 to 7.

Images