Irrigation water and fertilizer fine adjustment system and adjustment method

Abstract

The invention discloses an irrigation water and fertilizer fine adjustment system and method, and relates to the technical field of agricultural irrigation adjustment, and the method comprises the steps of information collection and storage, visual feature recognition and growth stage matching, irrigation area division, water application amount and fertilization amount calculation, irrigation instruction generation and irrigation operation execution. According to the invention, crop information is collected and visual feature identification is carried out to determine a growth stage, a crop growth stage spatial distribution map is generated, and irrigation area division is carried out according to the map; water and fertility calculation is carried out according to the soil moisture content and the soil conductivity of the soil in the irrigation area to obtain the water application amount and the fertilizer application amount, finally, an irrigation instruction is generated according to the water application amount and the fertilizer application amount, irrigation operation is executed, accurate water and fertilizer supply for crops in different growth stages is achieved, and the irrigation water and fertilizer adjustment fineness is improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural irrigation regulation technology, and in particular to a precision irrigation water and fertilizer regulation system and method. Background Technology

[0002] Irrigation fertigation refers to dissolving soluble fertilizers in irrigation water and delivering the nutrient solution directly to the crop root zone through an irrigation system. It combines fertilization and irrigation, using water to deliver fertilizer to where plants need it. Currently, the world faces widespread water scarcity, and agriculture is a major water user. Traditional large-scale, slow irrigation methods result in severe water waste, with significant losses due to evaporation and seepage. Furthermore, traditional fertilizer application methods have low utilization rates, with most fertilizer being lost or volatilized with the water, leading not only to high costs but also environmental problems such as soil compaction and eutrophication. Modern fertigation delivers water and fertilizer directly and accurately to the roots, minimizing deep seepage and surface evaporation. The fertilizer acts directly on the root system, significantly improving fertilizer utilization, directly reducing production costs, and mitigating agricultural non-point source pollution. Therefore, fertigation is an essential path for the upgrading of modern agriculture. Current agricultural practices treat large fields as a homogeneous whole for irrigation and fertilization, which can easily lead to localized deficiencies or excesses of fertilizer and water. Localized deficiencies can result in reduced yields, root damage, and decreased resistance to adverse conditions, while localized excesses can lead to excessive vegetative growth, decreased quality, increased pests and diseases, and nutrient loss. Both of these practices reduce resource utilization efficiency and bring environmental, economic, and regulatory risks. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a system for fine regulation of irrigation water and fertilizer, and also provides a method for fine regulation of irrigation water and fertilizer, in order to solve the aforementioned technical problems existing in the prior art.

[0004] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows: On the one hand, an irrigation water and fertilizer fine regulation system is provided, including: The information sensing module is used to collect and store crop information at fixed collection points. The crop information includes side and top views of the crops, soil electrical conductivity, and soil moisture content. At the same time, it stores a database of growth stages for various crops. The growth stage database includes standard feature vectors, target fertilizer requirements, target water holding capacity, and root depth for each type of crop at different growth stages. The stage recognition module is used to perform visual feature recognition and matching on the side and top views of crops to output the growth stage at each fixed collection point in the field, and generate a spatial distribution map of the crop growth stages in the field. The water and fertilizer calculation module is used to divide the field into several irrigation areas based on the spatial distribution map of crop growth stages, and to determine the amount of fertilizer and water to be applied based on the divided irrigation areas. The irrigation execution module is used to generate corresponding irrigation instructions and execute irrigation operations based on the amount of fertilizer and water applied to the irrigation area.

[0005] Preferably, the process of visual feature recognition and matching is as follows: S201: Adjust the side view and top view to a uniform resolution. Use a standard color chart or algorithm to eliminate color deviations caused by different lighting conditions. Then, apply a deep learning semantic segmentation model to separate the main crop from the side view and top view of the same location in the field, generating a clean image containing only the target crop. The clean image includes a clean side view and a clean top view. S202, identify and extract features from the clean side image to obtain plant features, including plant height, number of visible leaves and leaf tilt angle; S203 will identify and extract features from the clean aerial view to obtain vegetation features, including canopy coverage, mean vegetation group, and canopy uniformity. S204, based on parameters included in plant and vegetation features, constructs plant feature vectors and vegetation feature vectors, and performs similarity matching with standard feature vectors of crops at different growth stages to determine the growth stage; specifically: Plant feature vector A (x1, x2, x3) is constructed from plant height, number of visible leaves, and leaf tilt angle, corresponding to the plant characteristics. Here, x1, x2, and x3 are plant height, number of visible leaves, and leaf tilt angle, respectively. At the same time, vegetation feature vector B (y1, y2, y3) is constructed from canopy coverage, mean vegetation area, and canopy uniformity, corresponding to the vegetation characteristics. Here, y1, y2, and y3 are canopy coverage, mean vegetation area, and canopy uniformity, respectively. A pre-stored database of growth stages for various crops is constructed based on agronomic knowledge and historical data. This database defines standard feature vectors, target nutrient requirements, target water holding capacity, and root depth for each type of crop at different growth stages. Standard features include standard plant feature vectors and standard vegetation feature vectors for different growth stages. The standard plant feature vector HA (Hx1, Hx2, Hx3) includes standard plant height, standard number of visible leaves, and standard leaf angle. The standard vegetation feature vector HB (Hy1, Hy2, Hy3) includes standard canopy coverage, standard group vegetation mean, and standard canopy uniformity. The specific growth stages include: seedling stage, vegetative stage, reproductive stage, and maturity stage. Obtain the crop type and extract the standard feature vectors of different growth stages of the corresponding crop type from the growth stage database. Calculate the similarity Sim between the crop plant and vegetation status at the current location and each growth stage using cosine similarity and a linear weighted fusion formula. The calculation formula is as follows: (1) (2) (3) in Represents the dot product. The product represents the modular product; all parameters in vectors A, HA, B, and HB are non-negative, so the values ​​of S1 and S2 are in the range of [0,1]. Finally, S1 and S2 are linearly weighted and fused to calculate the similarity Sim between the crop plant and vegetation status at the current location and each growth stage. The closer the S value is to 1, the more consistent the direction of the multidimensional vector formed by the two vectors is, that is, the more similar the current crop status is to the typical feature pattern of the growth stage. The growth stage with the highest similarity is selected as the crop growth stage at the current location. Thus, the crop growth stages around each fixed collection point can be obtained.

[0006] Preferably, the plant feature extraction process is as follows: The two-dimensional image region of the plant in the clean side view is converted into a skeleton image with a width of only one pixel. In the skeleton image, the bottom endpoint is identified as the base point of the plant. All the pixels at the top of the skeleton image are drawn, and the point farthest from the base is determined as the vertex of the plant. The vertical pixel distance between the base point and the vertex is calculated to obtain the pixel height, which is then converted into the actual physical height to obtain the individual plant height. Thus, the individual plant heights of all plants in the clean side view can be obtained, and the average value is calculated to obtain the average plant height of this clean side view. In the skeleton image, the main stem of the plant is identified. Along the main stem skeleton, all points where branches occur are detected, which are the leaves. Starting from each leaf node, the branch skeleton extending outwards is traced to the endpoint. Each complete branch skeleton extending from the main stem to the endpoint represents a visible leaf. The number of visible leaves is obtained by counting all skeleton branches extending from the main stem and exceeding a certain threshold in length. Based on the leaf skeleton branches, the corresponding leaf regions are located in the lateral clean image. All pixel coordinates of the leaf regions are combined into a point set, and principal component analysis is performed on it to obtain the leaf inclination angle of a single plant. The leaf inclination angle of the lateral clean image is obtained by averaging the leaf inclination angles of all plants in the lateral clean image.

[0007] Preferably, the vegetation feature extraction process is as follows: After converting the overhead image to the HSV color space, it is converted into a binary image, namely the canopy mask, in which crop pixels are white and soil pixels are black. The total number of white pixels in the binary image and the total number of pixels in the binary image are counted, and then the total number of white pixels is divided by the total number of pixels to obtain the canopy coverage. The Hue channel values ​​are extracted from the white pixel locations in the canopy mask to obtain pixel values. Then, the average value of the group vegetation in the top-down clean image is calculated by averaging the pixel values ​​of all white pixel locations in the canopy mask. (1) After converting the clean overhead image into a grayscale image, divide it into several calculation windows. Within each calculation window, count the number of grayscale pairs (i,j) that appear in the grayscale image according to a predetermined spatial offset to obtain the grayscale co-occurrence counting matrix P(i,j), where i,j∈{0,…,L−1}, L is the length of the calculation window, and i and j represent the grayscale level indices within the calculation window. Divide P(i,j) by the total number of all counted pixel pairs in the window to obtain the normalized grayscale co-occurrence probability matrix p(i,j). Traverse each element P(i,j) in the grayscale co-occurrence matrix and first calculate the square of its corresponding grayscale level difference (ij), i.e. (ij). 2 Then multiply it by the probability p(i,j) of the element's appearance to obtain the contribution value of this element's P(i,j) to the contrast of the calculation window. Finally, sum the contribution values ​​of all elements in the gray-level co-occurrence count matrix of the calculation window to obtain the contrast. The specific calculation formula is as follows: Iterate through the contrast of all computed windows and calculate the standard deviation of the contrast of all computed windows, using it as the canopy uniformity of the overhead clean image.

[0008] Preferably, the process of generating a spatial distribution map of crop growth stages: A global map of the field is obtained, and the locations of each fixed collection point and its corresponding growth stage are marked on the global map. The global map of the entire field is then rasterized into several regular grids, where each grid is a single pixel. A search radius is set, and a circle is drawn around the grid of each unknown growth stage to obtain the search range. The growth stages of known grids within the search range are identified, and the probability of each growth stage within the search range is statistically calculated. The growth stage with the highest probability is taken as the growth stage of this unknown grid. If two or more growth stages have the highest probability simultaneously, the stage with the highest probability is taken as the candidate stage. Within the search range, the closest distance between each candidate stage and this unknown grid is selected, and the candidate stage corresponding to the smallest closest distance is taken as the growth stage of this unknown grid. In this way, all grids in the global map can be assigned growth stages to obtain a continuous spatial distribution map of crop growth stages covering the entire field.

[0009] Preferably, the irrigation zone division process is as follows: A spatial distribution map of crop growth stages is extracted. A connected component labeling algorithm from image processing is used to traverse all grids to group all spatially adjacent pixels with the same growth stage into the same region. This initially divides the crop growth stage spatial distribution map into several pixel regions, which are then converted into a vector polygon. A minimum irrigation area is set, and pixel regions with areas smaller than the minimum irrigation area are selected and marked as micro-regions. These micro-regions are merged into adjacent pixel regions sharing the longest boundary and with the same growth stage. If there are no pixel regions with the same growth stage adjacent to a micro-region, it is merged into the adjacent pixel region sharing the longest boundary and with the closest growth stage. The boundaries of the merged pixel regions are smoothed to obtain several irrigation regions. The area of ​​each irrigation region is calculated, and the growth stage with the highest proportion within the irrigation region is taken as the final growth stage of the irrigation region. Finally, the area and growth stage of the irrigation region are used as the attribute information of the irrigation region.

[0010] Preferably, the calculation process for fertilizer application in the irrigated area is as follows: The growth stages of the irrigated area are extracted and compared with the corresponding crops in the growth stage database to match the corresponding target nutrient requirements. The soil electrical conductivity at several fixed collection points in the irrigated area is extracted and averaged to obtain the average electrical conductivity of the irrigated area. Based on the average electrical conductivity of the irrigated area, the amount of effective nutrients that can be directly utilized by crops in the irrigated area is estimated. A fertilizer utilization rate is preset, and the fertilizer application amount for the irrigated area is obtained according to: Fertilizer application amount = (Target nutrient requirement - Effective nutrient storage) / Fertilizer utilization rate.

[0011] Preferably, the calculation process for the amount of water applied to the irrigation area is as follows: The growth stages of the irrigated area are extracted and compared with the corresponding crops in the growth stage database to match the target water holding capacity and root depth. Soil moisture content at several fixed sampling points within the irrigated area is extracted and averaged to obtain the mean moisture content of the irrigated area. A water use efficiency is preset. The target water holding capacity, root depth, mean moisture content, and water use efficiency are then used to calculate the irrigation amount for the irrigated area using a formula: .

[0012] Preferably, irrigation instructions and irrigation operations: The area of ​​each irrigation region is extracted and sorted in descending order to obtain an irrigation region list. The fertilizer and water application amounts for each irrigation region are then extracted sequentially from this list, and corresponding irrigation instructions are generated to complete the irrigation operation for that region. This process continues until all irrigation regions in the list have been processed. The irrigation instructions include fertilizer application instructions, water application instructions, and integrated water and fertilizer application instructions. Specifically: If both the amount of fertilizer applied and the amount of water applied are ≤0, then no irrigation equipment will be started, and the next irrigation area will continue to be treated. If the amount of fertilizer applied is ≤0 and the amount of water applied is >0, a water application instruction is generated. The specific execution process of the water application instruction is as follows: start the water pump and open the water supply valves of each irrigation point in the irrigation area, keep the fertilizer injection device closed, monitor the water flow rate and cumulative water volume, and when the cumulative water volume reaches the application amount, close the water pump and water supply valves, and continue to process the next irrigation area. If the amount of fertilizer applied is greater than 0 and the amount of water applied is less than or equal to 0, a fertilizer application instruction is generated. The fertilizer application instruction execution process is as follows: start the water pump, open the water supply valves of each irrigation point in the corresponding irrigation area, but only provide the minimum amount of water to the mixing tank. At the same time, start the fertilizer injection device to put the fertilizer corresponding to the amount of fertilizer applied into the mixing tank. The water and fertilizer are mixed evenly in the mixing tank and then transported to the buffer tank. From the buffer tank, the water pump is used to transport the fertilizer to each irrigation point in the irrigation area through the pipeline. After completion, the water pump and water supply valve are closed, and the next irrigation area is processed. If the amount of fertilizer applied is greater than 0 and the amount of water applied is greater than 0, a fertigation instruction is generated. The execution process of the fertigation instruction is as follows: start the water pump, open the water supply valves of each irrigation point in the corresponding irrigation area, start the fertilizer injection device to put the fertilizer corresponding to the amount of fertilizer applied into the mixing tank, and at the same time inject the same amount of water as the amount of water applied into the mixing tank to mix the water and fertilizer evenly in the mixing tank. Then, it is transported to the buffer tank, and from the buffer tank, it is transported through the water pump and pipeline to each irrigation point in the irrigation area. After completion, the water pump and water supply valve are closed, and the next irrigation area is processed.

[0013] On the other hand, the present invention provides a method for fine regulation of irrigation water and fertilizer, applied to the above-mentioned fine regulation system of irrigation water and fertilizer, the method specifically including the following steps: S100 collects and stores crop information at fixed collection points. The crop information includes side and top views of the crops, soil electrical conductivity, and soil moisture content. At the same time, it stores a database of growth stages for various crops. The growth stage database includes standard feature vectors, target fertilizer requirements, target water holding capacity, and root depth for each type of crop at different growth stages. S200 performs visual feature recognition and matching on the side and top views of crops to output the growth stage at each fixed collection point in the field, and generates a spatial distribution map of the crop growth stage in the field accordingly. S300 divides the field into several irrigation areas based on the spatial distribution map of crop growth stages, and determines the amount of fertilizer and water to be applied based on the divided irrigation areas. The S400 generates corresponding irrigation instructions and executes irrigation operations based on the amount of fertilizer and water applied to the irrigated area.

[0014] The beneficial effects of this invention are: 1. By performing deep analysis on the original side and top views to extract plant and vegetation features, and then using these features to identify and match growth stages to construct a spatial distribution map of crop growth stages in the field, the accurate identification results at discrete points can be reasonably extrapolated to the entire continuous space of the farmland, clearly showing the growth stages of crops at different locations in the field. 2. By identifying and dividing the spatial distribution of field growth stages, actual operable irrigation areas are generated. Unlike traditional zoning based on soil properties or topography, this method divides the fields based on growth stages that directly reflect the differences in crop needs, ensuring that the water and fertilizer requirements of crops in each irrigation area are highly consistent. Then, based on the soil conductivity and soil moisture content in the irrigation area, the water and fertilizer requirements are calculated to obtain the amount of water and fertilizer to be applied, ensuring the accuracy and scientific nature of water and fertilizer recommendations, and reducing resource waste and environmental pollution. 3. By converting the amount of fertilizer and water applied to each irrigation area into specific irrigation instructions, and executing corresponding irrigation operations on each irrigation area according to the irrigation instructions, precise water and fertilizer supply to irrigation areas at different growth stages can be achieved, improving the precision of irrigation water and fertilizer regulation. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the system module connections of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will now be clearly and completely described in conjunction with the accompanying drawings and specific embodiments.

[0017] Current agricultural practices typically treat large fields as a homogeneous whole for irrigation and fertilization, resulting in either insufficient or excessive amounts of fertilizer and water in certain areas. Insufficient fertilizer and water can lead to decreased yield, root damage, and reduced resistance to adverse conditions, while excessive fertilizer and water can cause excessive vegetative growth, reduced quality, increased pests and diseases, and nutrient loss. Both of these factors reduce resource utilization efficiency and bring environmental, economic, and regulatory risks.

[0018] To achieve this technical solution, irrigation points are densely distributed throughout the field, covering the entire area. The field is equipped with mixing tanks and buffer tanks. The mixing tanks thoroughly mix water and fertilizer according to a set ratio, while the buffer tanks stabilize flow and pressure, preventing stress to crops caused by fluctuations in water and fertilizer levels. A fertilizer injection device is located above the mixing tanks and can quantitatively release fertilizer according to control commands. After mixing with a specified amount of water (application volume) in the mixing tank, a water-fertilizer solution of a certain concentration is formed. This solution is pressurized by a water pump and transported through pipelines to the irrigation points within the irrigation area. Each irrigation point is equipped with a solenoid valve, which can be independently controlled to open or close, thereby achieving zoned fertilization and irrigation of each irrigation area, meeting the needs of precision agriculture's integrated water and fertilizer operation.

[0019] like Figure 1 As shown, an irrigation water and fertilizer fine regulation system of the present invention includes: The information sensing module is used to collect and store crop information at fixed collection points. The crop information includes side and top views of the crops, soil electrical conductivity, and soil moisture content. At the same time, it stores a database of growth stages for various crops. The growth stage database includes standard feature vectors, target fertilizer requirements, target water holding capacity, and root depth for each type of crop at different growth stages. The stage recognition module is used to perform visual feature recognition and matching on the side and top views of crops to output the growth stage at each fixed collection point in the field, and generate a spatial distribution map of the crop growth stages in the field. The water and fertilizer calculation module is used to divide the field into several irrigation areas based on the spatial distribution map of crop growth stages, and to determine the amount of fertilizer and water to be applied based on the divided irrigation areas. The irrigation execution module is used to generate corresponding irrigation instructions and execute irrigation operations based on the amount of fertilizer and water applied to the irrigation area; specifically: The area of ​​each irrigation region is extracted and sorted in descending order to obtain an irrigation region list. The fertilizer and water application amounts for each irrigation region are then extracted sequentially from this list, and corresponding irrigation instructions are generated to complete the irrigation operation for that region. This process continues until all irrigation regions in the list have been processed. The irrigation instructions include fertilizer application instructions, water application instructions, and integrated water and fertilizer application instructions. Specifically: If both the amount of fertilizer applied and the amount of water applied are ≤0, then no irrigation equipment will be started, and the next irrigation area will continue to be treated. If the amount of fertilizer applied is ≤0 and the amount of water applied is >0, a water application instruction is generated. The specific execution process of the water application instruction is as follows: start the water pump and open the water supply valves of each irrigation point in the irrigation area, keep the fertilizer injection device closed, monitor the water flow rate and cumulative water volume, and when the cumulative water volume reaches the application amount, close the water pump and water supply valves, and continue to process the next irrigation area. If the amount of fertilizer applied is greater than 0 and the amount of water applied is less than or equal to 0, a fertilizer application instruction is generated. The fertilizer application instruction execution process is as follows: start the water pump, open the water supply valves of each irrigation point in the corresponding irrigation area, but only provide the minimum amount of water to the mixing tank. At the same time, start the fertilizer injection device to put the fertilizer corresponding to the amount of fertilizer applied into the mixing tank. The water and fertilizer are mixed evenly in the mixing tank and then transported to the buffer tank. From the buffer tank, the water pump is used to transport the fertilizer to each irrigation point in the irrigation area through the pipeline. After completion, the water pump and water supply valve are closed, and the next irrigation area is processed. If the amount of fertilizer applied is greater than 0 and the amount of water applied is greater than 0, a fertigation instruction is generated. The execution process of the fertigation instruction is as follows: start the water pump, open the water supply valves of each irrigation point in the corresponding irrigation area, start the fertilizer injection device to put the fertilizer corresponding to the amount of fertilizer applied into the mixing tank, and at the same time inject the same amount of water as the amount of water applied into the mixing tank to mix the water and fertilizer evenly in the mixing tank. Then, it is transported to the buffer tank, and from the buffer tank, it is transported through the water pump and pipeline to each irrigation point in the irrigation area. After completion, the water pump and water supply valve are closed, and the next irrigation area is processed.

[0020] like Figure 2 As shown, an irrigation water and fertilizer fine regulation system of the present invention includes the following steps: S100, Crop Information Integration and Storage: Fixed collection points are set up in the fields, and various sensors are installed at these points, supplemented by drones, to regularly collect crop information, including crop images, soil volumetric water content, and soil electrical conductivity. The specific collection process is as follows: In the field, those skilled in the art set up fixed sampling points at representative locations. It should be noted that these fixed sampling points are not uniformly distributed, but rather selected by those skilled in the art from key locations that represent different fertility, moisture, and growth characteristics within the field, based on historical yield maps, soil conductivity scan maps, and topographic maps. An integrated sensor set is installed at each fixed sampling point, specifically including a soil moisture content sensor, a soil conductivity sensor, and a 360° rotating high-definition camera. The soil moisture content sensor and soil conductivity sensor are buried in the main active layer of the crop root system (e.g., at depths of 20 cm and 40 cm) to collect soil volumetric water content and soil conductivity. It should be noted that the sensor contact surface with the soil must be tightly backfilled and protected with the original soil sealant. Soil compaction is used to prevent gaps that could lead to lower or delayed readings. A 360° rotating high-definition camera captures side views of the crops, primarily to capture key aspects (such as leaf quantity, leaf area, stem thickness, and flower spike morphology). A drone equipped with a high-resolution camera is used for aerial photography to obtain overhead views of the field crops, reflecting the canopy structure, color uniformity, and coverage. It should be noted that to maintain consistent quality between the side and overhead views, images are captured during periods of stable lighting (e.g., 10 AM to 2 PM) each day to minimize the impact of shadows and light variations. The shooting angle and relative height for both side and overhead views at each location in the field are kept constant to ensure image comparability. During shooting, the sky or soil is kept as a background as possible to minimize interference from weeds and other obstructions. S200, image preprocessing, and analysis based on side and top views of the crop to identify its stage, specifically: S201: Adjust all images (side view and top view) to a uniform resolution, and use a standard color chart or algorithm to eliminate color deviations caused by different lighting conditions to ensure the authenticity and comparability of image colors; use a deep learning semantic segmentation model (such as U-Net) to accurately separate the main crop in the side view and top view of the same location in the field from the soil, sky background and weeds and other interference objects, and generate a clean image containing only the target crop. The clean image includes a clean side view and a clean top view. S202, identify and extract features from the clean side view image to obtain plant characteristics, including plant height, number of visible leaves, and leaf tilt angle; specifically: (1) Perform contrast stretching and other operations on the clean side image to make the plant outline and details clearer. Use morphological operations (such as thinning algorithms) to convert the two-dimensional image region of the plant in the clean side image into a skeleton image with a width of only one pixel. The skeleton image clearly shows the structural relationship between the main stem, petiole and midrib of the leaf. In the skeleton image, the bottommost endpoint is identified as the base of the plant. Usually, this point is located near the bottom boundary of the image and is the starting point of the main stem. Scan all the pixels at the top of the skeleton image, and the point farthest from the base is identified as the plant's... Vertex; for crops in the vegetative growth stage, the vertex is usually the tip of the highest leaf; for crops in the reproductive growth stage, the vertex may be the tip of the spike or the top of the highest fruit; calculate the vertical pixel distance between the base point and the vertex to obtain the pixel height. By placing a marker of known height next to the plant during shooting, establish the conversion relationship between pixels and actual length. Based on the conversion relationship, convert the pixel height to the real physical height to obtain the individual plant height. From this, the individual plant heights of all plants in the side clean image can be obtained, and the average value is calculated to obtain the average plant height of this side clean image. (2) In the skeleton image, the thickest and most coherent central skeleton is identified as the main stem. Along the main stem skeleton, all points where branches occur are detected. These points are leaf nodes (i.e., the connection points between the leaf and the main stem). Starting from each leaf node, trace the branch skeleton that extends outward to the endpoint. Each complete branch skeleton extending from the main stem to the endpoint represents a visible leaf. Count all skeleton branches that extend from the main stem and have a length exceeding a certain threshold (to avoid miscounting tiny protrusions) to obtain the number of visible leaves. (3) For each identified leaf, the corresponding leaf region is located in the original side pure image according to its skeleton branch. All pixel coordinates of the leaf region are combined into a point set and principal component analysis is performed on it: the direction of the first principal component is the direction of leaf elongation. The skeleton of the leaf is fitted with a straight line, and the angle of the fitted straight line is the orientation of the leaf. 0° is defined as vertically upward (parallel to the main stem) and 90° is horizontal. The angle between the leaf direction vector and the vertically upward direction is calculated to obtain the leaf tilt angle of a single plant. The leaf tilt angle of the side pure image is obtained by averaging the leaf tilt angles of all plants in the side pure image. S203 involves identifying and extracting features from a clean, overhead view of the vegetation map to obtain vegetation characteristics. Specific vegetation characteristics include canopy coverage, mean vegetation area, and canopy uniformity. Specifically: (2) Convert the image from the default RGB color space to the HSV color space. The HSV color space separates the value, saturation and hue of the color. It is more robust to changes in light than RGB and can more stably reflect the physiological characteristics of vegetation. Convert the overhead clean image into a binary image, i.e., the canopy mask, in which crop pixels are white and soil pixels are black. Count the total number of white pixels in the binary image and the total number of pixels in the binary image. Then divide the total number of white pixels by the total number of pixels to get the canopy coverage, which directly reflects the degree of shading of the crop canopy on the ground. (3) Using the Hue channel in the color-converted overhead pure image (HSV), healthy green vegetation will be concentrated in a specific value range in the Hue channel (usually around 60° to 120°, corresponding to a normalized value of 0.16-0.33). Extract the value of the Hue channel at the white pixel position (i.e., the crop position) in the canopy mask to obtain the pixel value; the higher this pixel value, the more yellow the green is (possibly due to aging or stress), and the lower the value, the purer and healthier the green is; calculate the mean value of the vegetation in the overhead pure image by averaging the pixel values ​​of all white pixel positions in the canopy mask. The mean value of the vegetation represents the degree of canopy greenness. (4) Convert the original top-down clean image (RGB) to a grayscale image, and divide the entire top-down clean image (RGB) into several small, potentially overlapping computation windows (using a 16x16 or 32x32 pixel sliding window). Within each computation window, count the number of grayscale pairs (i,j) that appear in the grayscale image according to a predetermined spatial offset, and obtain the grayscale co-occurrence counting matrix P(i,j), where i,j∈{0,…,L−1}, and L is the length of the computation window. For example, if a 16x16 pixel sliding window is used, then L =16, if a 32x32 pixel sliding window is used, then L=32; i and j represent the gray level indices within the calculation window. Divide P(i,j) by the total number of all counted pixel pairs in the window to obtain the normalized gray-level co-occurrence probability matrix p(i,j), where p(i,j) represents the probability that a randomly selected pixel pair is gray level i and gray level j under this calculation window and this offset relationship; traverse each element P(i,j) in the gray-level co-occurrence matrix, first calculate the square of its corresponding gray level difference (ij), i.e. (ij) 2 Then multiply it by the probability p(i,j) of the element's appearance to obtain the contribution value of this element's P(i,j) to the contrast of the calculation window. Finally, sum the contribution values ​​of all elements in the gray-level co-occurrence count matrix of the calculation window to obtain the contrast. The specific calculation formula is as follows: Contrast is a value that measures the degree of grayscale change in a local area. It should be noted that in agricultural scenes, the overhead image of a crop area with vigorous growth, large leaves, and a dense canopy should have smooth texture and low contrast. Conversely, the image of an area with uneven seedling emergence, diseased or pest-infested patches, or sparse leaves that allow soil to be seen will contain a large number of alternating bright and dark spots, resulting in high contrast. The contrast of all calculation windows is traversed, and the standard deviation of the contrast of all calculation windows is calculated. This standard deviation is then used as the canopy uniformity value of the overhead clean image (RGB). A larger canopy uniformity value indicates more obvious differences in canopy texture and poorer uniformity; a smaller canopy uniformity value indicates a more uniform canopy. S204, plant feature vector A (x1, x2, x3) is constructed from plant height, number of visible leaves, and leaf tilt angle corresponding to plant characteristics, where x1, x2, and x3 are plant height, number of visible leaves, and leaf tilt angle, respectively; at the same time, vegetation feature vector B (y1, y2, y3) is constructed from canopy coverage, mean of vegetation group, and canopy uniformity corresponding to vegetation characteristics, where y1, y2, and y3 are canopy coverage, mean of vegetation group, and canopy uniformity, respectively. A pre-stored database of growth stages for various crops is constructed based on agronomic knowledge and historical data. This database defines standard feature vectors, target nutrient requirements, target water holding capacity, and root depth for each type of crop at different growth stages. Standard features include standard plant feature vectors and standard vegetation feature vectors for different growth stages. The standard plant feature vector HA (Hx1, Hx2, Hx3) includes standard plant height, standard number of visible leaves, and standard leaf angle. The standard vegetation feature vector HB (Hy1, Hy2, Hy3) includes standard canopy coverage, standard group vegetation mean, and standard canopy uniformity. The specific growth stages include: seedling stage, vegetative stage, reproductive stage, and maturity stage. Obtain the crop type and extract the standard feature vectors of different growth stages of the corresponding crop type from the growth stage database. Calculate the similarity Sim between the crop plant and vegetation status at the current location and each growth stage using cosine similarity and a linear weighted fusion formula. The calculation formula is as follows: (4) (5) (6) in Represents the dot product. The product is represented by the modulus. All parameters in vectors A, HA, B, and HB are non-negative, so the values ​​of S1 and S2 are in the range of [0,1]. Finally, S1 and S2 are linearly weighted and fused to calculate the similarity Sim between the crop plant and vegetation status at the current location and each growth stage. The closer the S value is to 1, the more consistent the direction of the multidimensional vector formed by the two vectors is, that is, the more similar the current crop status is to the typical feature pattern of that growth stage. The growth stage with the highest similarity is selected as the crop growth stage at the current location. Thus, the crop growth stages around each fixed collection point can be obtained. S205: Obtain a global map of the field and mark the locations of each fixed sampling point and its corresponding growth stage on the global map. Then, divide the entire field's global map into a fine, regular grid, where each grid is a pixel. Set a search radius and draw a circle around each grid with an unknown growth stage (i.e., a place far from the fixed sampling point where no side view is captured) according to the search radius to obtain the search range. Identify the growth stages of known grids within the search range, statistically calculate the probability of each growth stage within the search range, and take the growth stage with the highest probability as the growth stage of this unknown grid. If two or more growth stages have the highest probability simultaneously, take the stage with the highest probability simultaneously as the candidate stage. Select the candidate stage with the smallest nearest distance from each candidate stage to this unknown grid within the search range as the growth stage of this unknown grid. In this way, all grids in the global map can be assigned growth stages to obtain a continuous spatial distribution map of crop growth stages covering the entire field, which can clearly show the growth stages of crops in different locations within the field. By performing deep analysis on the original side and top views to extract plant and vegetation features, and then using these features to identify and match growth stages, a spatial distribution map of crop growth stages in the field is constructed. The accurate identification results at discrete points are then reasonably extrapolated to the entire continuous space of the farmland, clearly showing the growth stages of crops at different locations within the field.

[0021] S300 involves dividing the farmland into several irrigation zones based on a spatial distribution map of crop growth stages, and determining the amount of fertilizer and water to be applied according to these zones; specifically: S301, extract the spatial distribution map of crop growth stages. Use the Connected Component Labeling (CCP) algorithm from image processing to traverse all grids of the entire crop growth stage spatial distribution map, merging all spatially adjacent (usually using 8-neighborhood connections) pixels with the same growth stage into the same region. This initially divides the crop growth stage spatial distribution map into several pixel regions. At this point, each pixel region represents an initial block with completely consistent growth stages. It should be noted that some of these pixel regions are too small to be considered independent irrigation areas, therefore further optimization and merging are needed. Set a minimum irrigation area (the minimum irrigation area is a minimum partition area threshold set by those skilled in the art based on farmland scale and management equipment operational capabilities), filter out pixel regions with areas smaller than the minimum irrigation area and record them as micro-patterns, merging them into the same region. If there are no adjacent pixel regions with the same growth stage that share the longest boundary, then the micro-patch is merged into the adjacent pixel region with the closest growth stage that shares the longest boundary. The concept of "closest growth stage" is defined by those skilled in the art based on crop phenology. The boundaries of the merged pixel regions are smoothed to obtain several irrigation areas, making them more in line with the outline of a natural field and avoiding jagged boundaries, which facilitates the planning of agricultural machinery operation paths. In this way, the entire field can be divided into several irrigation areas, the area of ​​each irrigation area is calculated, and the growth stage with the highest proportion in the irrigation area is taken as the final growth stage of the irrigation area. Finally, the area and growth stage of the irrigation area are used as the attribute information of the irrigation area. S302, extract the growth stage of the irrigated area and compare it with the corresponding crops in the growth stage database to match the corresponding target fertilizer requirement (the target fertilizer requirement is the total fertilizer requirement of crops in the irrigated area during this growth period, not the amount applied this time, in kilograms per acre); extract the soil electrical conductivity at several fixed sampling points in the irrigated area and calculate the average electrical conductivity of the irrigated area; based on the average electrical conductivity of the irrigated area, use a localized calibration model (machine learning model) obtained through laboratory soil sample calibration to estimate the effective nutrient stock that can be directly utilized by crops in the irrigated area (in kilograms per acre); since fertilizer does not... 100% absorbed by crops, but losses exist. A pre-set fertilizer utilization rate is used. The fertilizer application rate (kg / mu) is calculated as follows: Fertilizer application rate (kg / mu) = (Target fertilizer requirement - Available nutrient stock) / Fertilizer utilization rate. Specifically, the database shows that corn in the flowering stage requires 12 kg / mu of nitrogen. Sensor data and model estimation indicate that the current available nitrogen stock in the soil is 4 kg / mu. Based on past experience, the nitrogen fertilizer utilization rate in this irrigated area is approximately 50%. Therefore, the required nitrogen fertilizer is 12 kg / mu - 4 kg / mu = 8 kg / mu. Dividing the required nitrogen fertilizer (8 kg / mu) by the nitrogen fertilizer utilization rate of 50% yields a nitrogen fertilizer application rate of 16 kg / mu. S303: Extract the growth stage of the irrigation area and compare it with the corresponding crops in the growth stage database to match the corresponding target water holding capacity (target water holding capacity refers to the optimal soil moisture environment required by the crop at the current growth stage, that is, the soil moisture state that allows the crop to grow optimally and has the highest transpiration and absorption efficiency) and root layer depth (root layer depth refers to the depth that the main root layer of the crop can reach at this growth stage). Extract the soil moisture content at several fixed collection points in the irrigation area and calculate the average moisture content of the irrigation area. A water use efficiency is preset. Due to various losses (evaporation, runoff, deep seepage), water cannot be 100% fully utilized. Calculate the amount of water applied to the irrigation area according to the formula: ; By identifying and dividing the spatial distribution of crop growth stages in the field, practically operable irrigation areas are generated. Unlike traditional zoning based on soil properties or topography, this method uses growth stages that directly reflect the differences in crop needs to form zones, ensuring that the water and fertilizer requirements of crops in each irrigation area are highly consistent. Then, based on the soil conductivity and soil moisture content in the irrigation area, water and fertilizer requirements are calculated to obtain the amount of water and fertilizer to be applied, ensuring the accuracy and scientific nature of water and fertilizer recommendations, and reducing resource waste and environmental pollution.

[0022] S400 extracts the area (in mu) of the irrigated region, sorts the irrigated regions in descending order of area to obtain an irrigated region list, and extracts the fertilizer and water application amounts for each irrigated region according to the list. Based on this, it generates corresponding irrigation instructions to complete the irrigation operation for that region, until all irrigated regions in the list have been processed. The irrigation instructions include fertilizer application instructions, water application instructions, and integrated water and fertilizer application instructions; specifically: If the amount of fertilizer applied is ≤0 and the amount of water applied is ≤0, it means that no irrigation operation is required in this irrigation area. In this case, no irrigation equipment will be started, and the next irrigation area will be treated. If the amount of fertilizer applied is ≤0 and the amount of water applied is >0, it means that fertilizer needs to be applied to the irrigation area, and a fertilizer application instruction is generated. The specific execution process of the water application instruction is as follows: start the water pump and open the water supply valves of each irrigation point in the irrigation area, keep the fertilizer injection device closed, monitor the water flow rate and cumulative water volume, and when the cumulative water volume reaches the application amount, close the water pump and water supply valves, and continue to process the next irrigation area. If the amount of fertilizer applied is greater than 0 and the amount of water applied is less than or equal to 0, it indicates that water needs to be applied to the irrigation area. A fertilizer application instruction is then generated. The fertilizer application instruction execution process is as follows: Since the fertilizer needs to be dissolved in water and transported to the irrigation area, the water pump is started, and the water supply valves of each irrigation point in the corresponding irrigation area are opened. However, only the minimum amount of water (for fertilizer delivery) is provided to the mixing tank. At the same time, the fertilizer injection device is started to release the fertilizer corresponding to the amount of fertilizer applied into the mixing tank. The water and fertilizer are mixed evenly in the mixing tank and then transported to the buffer tank. From the buffer tank, the fertilizer is transported through the water pump and pipeline to each irrigation point in the irrigation area. After completion, the water pump and water supply valves are closed, and the process continues to process the next irrigation area. If the fertilizer application rate is greater than 0 and the water application rate is greater than 0, it indicates that integrated water and fertilizer irrigation is required for the irrigation area. A fertigation command is then generated. The execution process of the fertigation command is as follows: start the water pump, open the water supply valves at each irrigation point within the corresponding irrigation area, start the fertilizer injection device to release the fertilizer corresponding to the fertilizer application rate into the mixing tank, and simultaneously inject the same amount of water as the fertilizer application rate into the mixing tank. The water and fertilizer are then mixed evenly in the mixing tank and transported to the buffer tank. From the buffer tank, the fertilizer is pumped through pipelines to each irrigation point in the irrigation area. After completion, the water pump and water supply valves are closed. It should be noted that the constraint condition for the water application rate here is: water application rate ≥ minimum water application rate. If the water application rate is less than the minimum water application rate, the minimum water application rate is used. It should be noted that irrigation points are densely distributed throughout the field, covering the entire area. The field is equipped with mixing tanks and buffer tanks. The mixing tanks are used to thoroughly mix water and fertilizer according to a set ratio, while the buffer tanks stabilize flow and pressure, preventing stress to crops caused by fluctuations in water and fertilizer levels. The fertilizer injection device is located above the mixing tanks and can quantitatively release fertilizer according to control commands. After mixing with a specified amount of water (application volume) in the mixing tank, a water-fertilizer solution of a certain concentration is formed. This solution is pressurized by a water pump and transported through pipelines to the irrigation points within the irrigation area. Each irrigation point is equipped with a solenoid valve, which can be independently controlled to open or close, thus enabling zoned fertilization and irrigation of each irrigation area, meeting the needs of precision agriculture's integrated water and fertilizer operation.

[0023] By converting the amount of fertilizer and water applied to each irrigation area into specific irrigation instructions, and executing corresponding irrigation operations according to the irrigation instructions, precise water and fertilizer supply to irrigation areas at different growth stages can be achieved, improving the precision of irrigation water and fertilizer regulation.

[0024] The above are merely examples and descriptions of the present invention. Those skilled in the art can make various modifications, additions, or similar substitutions to the specific embodiments described above, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, they should all fall within the protection scope of the present invention.

Claims

1. A precision irrigation and fertilization system, characterized in that, include: The information sensing module is used to collect and store crop information at fixed collection points. The crop information includes side and top views of the crops, soil electrical conductivity, and soil moisture content. At the same time, it stores a database of growth stages for various crops. The growth stage database includes standard feature vectors, target fertilizer requirements, target water holding capacity, and root depth for each type of crop at different growth stages. The stage recognition module is used to perform visual feature recognition and matching on the side and top views of crops to output the growth stage at each fixed collection point in the field, and generate a spatial distribution map of the crop growth stages in the field. The water and fertilizer calculation module is used to divide the field into several irrigation areas based on the spatial distribution map of crop growth stages, and to determine the amount of fertilizer and water to be applied based on the divided irrigation areas. The irrigation execution module is used to generate corresponding irrigation instructions and execute irrigation operations based on the amount of fertilizer and water applied to the irrigation area.

2. The irrigation water and fertilizer fine regulation system according to claim 1, characterized in that, The process of visual feature recognition and matching is as follows: S201: Adjust the side view and top view to a uniform resolution. Use a standard color chart or algorithm to eliminate color deviations caused by different lighting conditions. Then, apply a deep learning semantic segmentation model to separate the main crop from the side view and top view of the same location in the field, generating a clean image containing only the target crop. The clean image includes a clean side view and a clean top view. S202, identify and extract features from the clean side image to obtain plant features, including plant height, number of visible leaves and leaf tilt angle; S203 will identify and extract features from the clean aerial view to obtain vegetation features, including canopy coverage, mean vegetation group, and canopy uniformity. S204. Based on the parameters included in the plant features and vegetation features, construct plant feature vectors and vegetation feature vectors, and perform similarity matching with the standard feature vectors of crops at different growth stages to determine the growth stage.

3. The irrigation water and fertilizer fine regulation system according to claim 2, characterized in that, Plant feature extraction process: The two-dimensional image region of the plant in the clean side view is converted into a skeleton image with a width of only one pixel. In the skeleton image, the bottom endpoint is identified as the base point of the plant. All pixels at the top of the skeleton image are scanned, and the point farthest from the base is identified as the vertex of the plant. The vertical pixel distance between the base point and the vertex is calculated to obtain the pixel height, which is then converted into the actual physical height to obtain the individual plant height. Thus, the individual plant heights of all plants in the clean side view can be obtained, and the average value is calculated to obtain the average plant height of this clean side view. In the skeleton image, the main stem of the plant is identified, and all points where branches occur along the main stem are detected, which are the leaves. Starting from each leaf node, the branch skeleton extending outwards is traced to the endpoint. Each complete branch skeleton extending from the main stem to the endpoint represents a visible leaf. The number of visible leaves is obtained by counting all skeleton branches extending from the main stem and exceeding a certain threshold in length. Based on the leaf skeleton branches, the corresponding leaf regions are located in the lateral clean image. All pixel coordinates of the leaf regions are combined into a point set, and principal component analysis is performed on it to obtain the leaf inclination angle of a single plant. The leaf inclination angle of the lateral clean image is obtained by averaging the leaf inclination angles of all plants in the lateral clean image.

4. The irrigation water and fertilizer fine regulation system according to claim 3, characterized in that, Vegetation feature extraction process: After converting the overhead image to the HSV color space, it is converted into a binary image, namely the canopy mask, in which crop pixels are white and soil pixels are black. The total number of white pixels in the binary image and the total number of pixels in the binary image are counted, and then the total number of white pixels is divided by the total number of pixels to obtain the canopy coverage. The Hue channel values ​​are extracted from the white pixel locations in the canopy mask to obtain pixel values. Then, the average value of the group vegetation in the top-down clean image is calculated by averaging the pixel values ​​of all white pixel locations in the canopy mask. After converting the clean overhead image to a grayscale image, it is divided into several calculation windows. Within each calculation window, the number of grayscale pairs (i,j) appearing in the grayscale image is counted according to a predetermined spatial offset, resulting in a grayscale co-occurrence count matrix P(i,j), where i,j∈{0,…,L−1}, L is the length of the calculation window, and i and j represent the grayscale level indices within the calculation window. P(i,j) is divided by the total number of all counted pixel pairs in the window to obtain a normalized grayscale co-occurrence probability matrix p(i,j). For each element P(i,j) in the grayscale co-occurrence matrix, the square of its corresponding grayscale difference value (ij) is calculated first, i.e., (ij). 2 Then multiply it by the probability p(i,j) of the element to obtain the contribution value of P(i,j) of this element to the contrast of the calculation window. Then sum the contribution values ​​of all elements in the gray-level co-occurrence count matrix of the calculation window to obtain the contrast. Iterate through the contrast of all calculation windows and calculate the standard deviation of the contrast of all calculation windows, and use it as the canopy uniformity of the overhead clean image.

5. The irrigation water and fertilizer fine regulation system according to claim 4, characterized in that, The process of generating a spatial distribution map of crop growth stages: A global map of the field is obtained, and the locations of each fixed collection point and its corresponding growth stage are marked on the global map. The global map of the entire field is then rasterized into several regular grids, where each grid is a single pixel. A search radius is set, and a circle is drawn around the grid of each unknown growth stage to obtain the search range. The growth stages of known grids within the search range are identified, and the probability of each growth stage within the search range is statistically calculated. The growth stage with the highest probability is taken as the growth stage of this unknown grid. If two or more growth stages have the highest probability simultaneously, the stage with the highest probability is taken as the candidate stage. Within the search range, the closest distance between each candidate stage and this unknown grid is selected, and the candidate stage corresponding to the smallest closest distance is taken as the growth stage of this unknown grid. In this way, all grids in the global map can be assigned growth stages to obtain a continuous spatial distribution map of crop growth stages covering the entire field.

6. The irrigation water and fertilizer fine regulation system according to claim 1, characterized in that, The process of dividing irrigation areas is as follows: A spatial distribution map of crop growth stages is extracted. A connected component labeling algorithm from image processing is used to traverse all grids to group all spatially adjacent pixels with the same growth stage into the same region. This initially divides the crop growth stage spatial distribution map into several pixel regions, which are then converted into a vector polygon. A minimum irrigation area is set, and pixel regions with areas smaller than the minimum irrigation area are selected and marked as micro-regions. These micro-regions are merged into adjacent pixel regions sharing the longest boundary and with the same growth stage. If there are no pixel regions with the same growth stage adjacent to a micro-region, it is merged into the adjacent pixel region sharing the longest boundary and with the closest growth stage. The boundaries of the merged pixel regions are smoothed to obtain several irrigation regions. The area of ​​each irrigation region is calculated, and the growth stage with the highest proportion within the irrigation region is taken as the final growth stage of the irrigation region. Finally, the area and growth stage of the irrigation region are used as the attribute information of the irrigation region.

7. The irrigation water and fertilizer fine regulation system according to claim 6, characterized in that, The calculation process for fertilizer application in irrigated areas: The growth stages of the irrigated area are extracted and compared with the corresponding crops in the growth stage database to match the corresponding target nutrient requirements. The soil electrical conductivity at several fixed collection points in the irrigated area is extracted and averaged to obtain the average electrical conductivity of the irrigated area. Based on the average electrical conductivity of the irrigated area, the amount of effective nutrients that can be directly utilized by crops in the irrigated area is estimated. A fertilizer utilization rate is preset, and the fertilizer application amount for the irrigated area is obtained according to: Fertilizer application amount = (Target nutrient requirement - Effective nutrient storage) / Fertilizer utilization rate.

8. The irrigation water and fertilizer fine regulation system according to claim 7, characterized in that, The calculation process for the amount of water applied in the irrigation area: The growth stages of the irrigated area are extracted and compared with the corresponding crops in the growth stage database to match the target water holding capacity and root depth. The soil moisture content at several fixed collection points in the irrigated area is extracted and averaged to obtain the average moisture content of the irrigated area. A water use efficiency is preset, and the target water holding capacity, root depth, average moisture content, and water use efficiency are calculated using a formula to obtain the amount of water applied to the irrigated area.

9. The irrigation water and fertilizer fine regulation system according to claim 8, characterized in that, Irrigation instructions and operations: The area of ​​each irrigation region is extracted and sorted in descending order to obtain an irrigation region list. The fertilizer and water application amounts for each irrigation region are then extracted sequentially from this list, and corresponding irrigation instructions are generated to complete the irrigation operation for that region. This process continues until all irrigation regions in the list have been processed. The irrigation instructions include fertilizer application instructions, water application instructions, and integrated water and fertilizer application instructions. Specifically: If both the amount of fertilizer applied and the amount of water applied are ≤0, then no irrigation equipment will be started, and the next irrigation area will continue to be treated. If the amount of fertilizer applied is ≤0 and the amount of water applied is >0, a fertilizer application instruction is generated. The specific execution process of the water application instruction is as follows: start the water pump and open the water supply valves of each irrigation point in the irrigation area, keep the fertilizer injection device closed, monitor the water flow rate and cumulative water volume, and when the cumulative water volume reaches the application amount, close the water pump and water supply valves, and continue to process the next irrigation area. If the amount of fertilizer applied is greater than 0 and the amount of water applied is less than or equal to 0, a fertilizer application instruction is generated. The fertilizer application instruction execution process is as follows: start the water pump, open the water supply valves of each irrigation point in the corresponding irrigation area, but only provide the minimum amount of water to the mixing tank. At the same time, start the fertilizer injection device to put the fertilizer corresponding to the amount of fertilizer applied into the mixing tank. The water and fertilizer are mixed evenly in the mixing tank and then transported to the buffer tank. From the buffer tank, the water pump is used to transport the fertilizer to each irrigation point in the irrigation area through the pipeline. After completion, the water pump and water supply valve are closed, and the next irrigation area is processed. If the amount of fertilizer applied is greater than 0 and the amount of water applied is greater than 0, a fertigation command is generated. The execution process of the fertigation command is as follows: start the water pump, open the water supply valves of each irrigation point in the corresponding irrigation area, start the fertilizer injection device to put the fertilizer corresponding to the amount of fertilizer applied into the mixing tank, and at the same time inject the same amount of water as the amount of water applied into the mixing tank to mix the water and fertilizer evenly in the mixing tank. Then, it is transported to the buffer tank, and from the buffer tank, it is transported through the water pump and pipeline to each irrigation point in the irrigation area. After completion, the water pump and water supply valve are turned off, and the next irrigation area is processed.

10. A method for fine regulation of irrigation water and fertilizer, using the fine regulation system for irrigation water and fertilizer as described in any one of claims 1-9, characterized in that, Includes the following steps: S100 collects and stores crop information at fixed collection points. The crop information includes side and top views of the crops, soil electrical conductivity, and soil moisture content. At the same time, it stores a database of growth stages for various crops. The growth stage database includes standard feature vectors, target fertilizer requirements, target water holding capacity, and root depth for each type of crop at different growth stages. S200 performs visual feature recognition and matching on the side and top views of crops to output the growth stage at each fixed collection point in the field, and generates a spatial distribution map of the crop growth stage in the field accordingly. S300 divides the field into several irrigation areas based on the spatial distribution map of crop growth stages, and determines the amount of fertilizer and water to be applied based on the divided irrigation areas. The S400 generates corresponding irrigation instructions and executes irrigation operations based on the amount of fertilizer and water applied to the irrigated area.

Images