Intelligent management method and system for water and fertilizer of planting land

Abstract

The invention relates to the technical field of agricultural irrigation, in particular to a planting land water and fertilizer intelligent management method and system. The method comprises the steps of constructing a planting field three-dimensional model of divided regions, designing a sensor sensing scheme based on the planting field three-dimensional model of the divided regions, establishing a prediction model to predict a future change trend of a decision variable in combination with sensor sensing data, evaluating a crop growth risk based on a prediction result, and formulating an advanced intervention strategy. Intelligent water and fertilizer management of the planting land is realized. Precise planting and scientific management are achieved, and the overall production benefit is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural irrigation, and particularly to a method and system for intelligent management of water and fertilizer in planting areas. Background Art

[0002] Traditional agricultural production models rely heavily on manual labor. Not only does it require a large amount of time and energy, but the production efficiency is extremely limited. Traditional planting models often rely on farmers' own planting experience, lacking quantitative and scientific technical support. There is a lack of real-time data for planted crops, making it difficult to adjust nutrient supply according to actual needs. In the field of crop planting, supplying water and fertilizer as needed plays a decisive role in plant growth. Once crops lack water and fertilizer supply for a long time, their metabolism will be disordered, and their growth trend will gradually decline.

[0003] In recent years, with the rapid development of information technology, agricultural digital technologies have gradually emerged. In terms of irrigation, intelligent irrigation systems use sensor networks, Internet of Things technology, and cloud computing platforms to monitor soil moisture, meteorological conditions, and crop growth information in real time, accurately determine the water requirements of crops, and achieve intelligent irrigation management. Such systems can automatically adjust the irrigation volume and irrigation time according to real-time monitoring data, improve water resource utilization efficiency, and reduce waste. The current development of intelligent irrigation systems is not yet mature, and there is room for optimization in terms of sensor perception accuracy, energy consumption of sensor perception, and credibility of irrigation management decision-making suggestions.

[0004] How to scientifically deploy sensors in different planting areas and save costs to a certain extent, and how to combine sensor data to propose reasonable and scientific decision-making suggestions based on accurate goals are the precise bases for the normal and accurate operation of intelligent irrigation systems and also key factors in agricultural production and operation. Summary of the Invention

[0005] (I) Technical Problems to be Solved

[0006] The purpose of the present invention is to provide a method and system for intelligent management of water and fertilizer in planting areas to solve the above technical problems.

[0007] (II) Technical Solutions

[0008] To achieve the above purpose, the present invention provides a method for intelligent management of water and fertilizer in planting areas, including the steps of:

[0009] S1, constructing a three-dimensional model of the planting area with divided regions; referring to the land area size and the crop density relationship, dividing the initial planting regions of the three-dimensional model and numbering them to obtain the three-dimensional model of the planting area with divided regions;

[0010] S2, designing a sensor perception scheme based on the three-dimensional model of the planting area with divided regions, including:

[0011] S21, determine the sensor category; use the analytic hierarchy process to quantify the influence weights of key environmental factors in different growth stages of crops, so as to select sensors that sense the corresponding environmental factors.

[0012] S22, determine the sensor sensing positions, including:

[0013] S221, evenly set a preset number of fixed points in each planting area divided in the step S1, and measure the comprehensive value of the key environmental factors of the fixed points.

[0014] S222, compare whether the dispersion degree of the comprehensive values of all fixed points within each planting area exceeds a preset dispersion degree threshold to judge whether to perform regional subdivision.

[0015] S223, if so, perform regional subdivision, subdivide the planting area into multiple planting areas, evenly set a preset number of fixed points in the multiple planting areas, and return to step S222; if not, do not perform regional subdivision; stop until the dispersion degree of the comprehensive values of the fixed points in all the planting areas is within the preset dispersion degree threshold.

[0016] S224, judge whether the dispersion degree of the comprehensive values of all fixed points in two adjacent planting areas is less than the preset dispersion degree threshold.

[0017] S225, if so, merge the two adjacent planting areas, evenly set a preset number of fixed points in the merged planting area, and return to step S224; if not, return to step S224; stop until adjacent areas cannot be merged or there are no adjacent areas.

[0018] S226, set the sensor sensing positions according to all the planting areas and the positions of the fixed points obtained after processing.

[0019] S23, determine the sensor frequency; according to the change characteristics of the key environmental factors, the sensor dynamically adjusts the data acquisition frequency according to the data fluctuation.

[0020] S3, combine the sensor sensing data, establish a prediction model to predict the future change trend of the decision variable, evaluate the crop growth risk based on the prediction result, formulate an early intervention strategy, and realize the intelligent management of water and fertilizer in the planting area.

[0021] Preferably, the step S1 further includes:

[0022] S11. Analyze the difference in the degree of change in the height of crops in the planting area; construct an analysis system for the degree of change in height data, input the model coordinates, analyze the change in height data corresponding to the model, and set the range of the degree of change in height data. The expression of the analysis system for the degree of change in height data is as follows:

[0023] H = f(Z)

[0024] Z = z * d 模 / d 实

[0025] where Z represents the difference in the height of a certain point coordinate compared to the lowest height in the model, z represents the value of the coordinate in the z-axis direction, H represents the actual degree of change in height, and d 实 represents the actual length data between any two points on the same height plane measured on the ground, and d 模 represents the coordinate difference between any two points on the same height plane in the reconstructed model;

[0026] S12. Identify the data of the boundary range of the planting area to obtain the size of the land area of the planting area; based on the three-dimensional model of the planting area, obtain the height value of the model, analyze the degree of change in the height value according to the analysis system for the degree of change in height data. If the degree of change in the height value exceeds the set range of the degree of change in height data, it is initially determined as the planting area range. For the area initially determined as the planting area range, analyze whether there are no points with a degree of change in height value exceeding the set range of the degree of change in height data within a certain range in the horizontal plane in at least one direction, and if so, it is identified as the boundary crops. According to the boundary crop information, determine the boundary value in the three-dimensional model to obtain the size of the planting area;

[0027] S13. Based on the image data, construct a three-dimensional model of the planting area, obtain the crop coordinates, and obtain the crop density relationship;

[0028] S14. Using the size of the land area and the crop density relationship as a reference, divide the initial planting area and number it to obtain the three-dimensional model of the planting area of the divided area.

[0029] Preferably, step S1 further includes: constructing a three-dimensional model of the planting area based on the image data; specifically including the steps:

[0030] S101. Obtain the image data; the drone takes 360° circumferential photos of the planting area to collect the video data of the planting area;

[0031] S102. Extract frames from the video data to obtain an image data set; perform frame extraction on the video data of the planting area based on a time interval to obtain an image data set with an overlap rate of more than 40% between adjacent frames;

[0032] S103, Pose analysis and model construction; use the Sfm algorithm for pose analysis, reference the hash encoding in Instant-NGP and perform model construction based on the NeRf model.

[0033] Preferably, the step S21 includes:

[0034] S211, Construct a judgment matrix; determine the key environmental factors E of the main decision variables of water and fertilizer i and the importance of each factor, use the 1-9 scale method to quantify this relative importance, make pairwise comparisons, and construct an n×x judgment matrix A=(a ij ), where a ij represents the importance degree of the i-th environmental factor relative to the j-th environmental factor, and satisfies

[0035] S212, Calculate the eigenvector; calculate the maximum eigenvalue λ max of the judgment matrix A and its corresponding eigenvector W; the eigenvector W is normalized to obtain the weight vector of the key environmental factors;

[0036] S213, Based on the obtained weight vector of the key environmental factors, comprehensively consider the sensors that can better sense the corresponding environmental factors;

[0037] S214, Based on the conditions of the actual planting area, determine the measurement method of the sensor, and the measurement methods of the sensor include mobile measurement, fixed measurement, or a combined measurement scheme combining mobile and fixed measurements.

[0038] Preferably, the step S22 includes: At the fixed-point measurement of the comprehensive value X of the key environmental factors i ;

[0039] Calculate the dispersion degree of the comprehensive values of different fixed points through the formula , where μ represents the average value of different fixed-point arrays.

[0040] Preferably, the step S23 includes:

[0041] S231, Set the initial acquisition frequency and the acquisition frequency limit range; estimate the change range of relevant environmental factors, combine the sensor characteristics, and set the initial acquisition frequency as the lower limit acquisition frequency; determine the limit frequency of the sensor through the sensor specification sheet, and set the upper limit acquisition frequency f max ;

[0042] S232, Analyze the data fluctuation degree and predict the future change trend; use the collected data to fill in the missing points to form a smooth curve; select continuous points near the current data point, calculate the slope k by taking the derivative, and observe the slope change to predict the future data trend;

[0043] S233. Adjust the data acquisition frequency in real time according to the actual degree of fluctuation;

[0044] Design an adjustment formula for the acquisition frequency f:

[0045] f = f min *(1 + m*|k|)

[0046] where m is an adjustment coefficient representing the adjustment amplitude for controlling the acquisition frequency; k is the current slope representing the degree of data fluctuation, f represents the subsequent data acquisition frequency, and f min represents the minimum acquisition frequency. When the calculated f is greater than f max , adjust f to the upper limit value f max ; when the calculated f is less than f min , adjust f to the lower limit value f min to ensure that the acquisition frequency is within the limited range.

[0047] Preferably, the step S3 includes:

[0048] S31. Establish a prediction model between the target variable and relevant environmental factors to predict the future change trend of the decision variable; determine the relevant environmental factors of the target variable from the key environmental factors in step S21, monitor the data in real time through corresponding sensors, collect the historical data of the target variable and relevant environmental factors, and record the corresponding time data; determine the mathematical relationship between the target variable and relevant environmental factors by the least squares method and predict future data points in time series analysis;

[0049] S32. Compare the result predicted by the prediction model in step S31 with the subsequent actual situation to improve the prediction model; collect the actual data of the target variable and relevant environmental factors during the prediction period, compare it with the prediction result, analyze the magnitude of the prediction error, and take corresponding measures for model adjustment according to the analysis result;

[0050] S33. Evaluate the risk degree of the future growth of the crop according to the prediction model; the risk coefficient is a comprehensive index considering the degree of deviation of the decision variable from the threshold, the volatility of environmental factors, and the urgency of the crop's nutrient demand; it is divided into three levels: low risk coefficient, medium risk coefficient, and high risk coefficient according to the magnitude of the risk coefficient;

[0051] S34. Define the intervention target based on the prediction model and the risk coefficient;

[0052] S35, based on the appropriate growth condition threshold of the target variable, different operation levels G(x) are divided, where x represents the specific level number; a relationship model L(x) between intervention measures and variables is established, where x represents the specific level number; the impact of intervention measures on the target variable is clarified, and the operation plan is determined according to the operation level G(x) combined with the intervention model L(x) to assist in water and fertilizer management.

[0053] Preferably, the step S32 further includes:

[0054] By calculating the arithmetic mean of the absolute error, the change of the prediction error over time is plotted into a curve graph. According to the analysis results, corresponding measures are taken to adjust the model. If the prediction error is large, a new data set is used to retrain the model to calibrate the parameters. If the impact of a specific environmental factor on the target variable is not fully reflected in the prediction, the specific environmental factor is added to the model. On the contrary, if the impact of a specific environmental factor on the prediction result is not significant, it is deleted from the model.

[0055] Preferably, step S33 includes: the degree to which the determining variable of the risk coefficient deviates from the threshold, the volatility of environmental factors, and the urgency of the crop's nutrient demand account for a%, b%, and c%, respectively; the evaluation scores of each part are set to x, y, and z, respectively, and the value range is 110 points; the risk coefficient β is obtained according to the formula β=a%x+b%y+c%z; when 1≤β≤3, it is a low risk; when 3<β≤7, it is a medium risk; when 7<β≤10, it is a high risk; according to the risk level, the basic goal of early intervention is determined.

[0056] The present invention also provides an intelligent water and fertilizer management system for planting areas, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the steps of the intelligent water and fertilizer management method for planting areas as described in any one of the above are implemented.

[0057] (III) Beneficial effects

[0058] The present invention constructs a three-dimensional model of the planting site with divided areas, rigorously designs a sensor perception scheme, quickly and efficiently divides and numberes the planting fields, and achieves efficient and energy-saving use while ensuring accurate and stable sensor measurement perception, laying a solid foundation for subsequent planting management work. At the same time, it effectively avoids excessive consumption of equipment resources and manpower due to complex measurement data, accurately and quickly obtains land area and crop density, realizes precise planting, scientific management and improves overall production efficiency.

[0059] The present invention divides the sensing area according to the key environmental factors for the growth of crops, determines the sensor positions by subdividing and merging areas, and realizes the efficient and accurate sensing of the growth of the planting area, laying an objective foundation for subsequent intelligent water and fertilizer management.

[0060] The present invention combines the sensor sensing data, establishes a prediction model to predict the future change trend of the decision variables, evaluates the crop growth risk based on the prediction results, formulates an early intervention strategy, realizes the intelligent management of water and fertilizer in the planting area, and achieves the precise water and fertilizer management of the planting area through digital means. It integrates functions such as data collection, model analysis, visual display, and intelligent decision-making to form a comprehensive agricultural management system, improving agricultural production efficiency and resource utilization rate. Brief Description of the Drawings

[0061] Figure 1 It is a schematic flowchart of a method for intelligent management of water and fertilizer in a planting area provided in this embodiment;

[0062] Figure 2 It is a schematic flowchart of automatically obtaining the area of the planting area and the density of crops provided in this embodiment;

[0063] Figure 3 It is a schematic diagram of the fixed-point distribution in a single area provided in this embodiment;

[0064] Figure 4 It is a schematic diagram of area subdivision and fixed-point distribution provided in this embodiment;

[0065] Figure 5 It is a schematic diagram of area merging and fixed-point distribution provided in this embodiment;

[0066] Figure 6 It is a schematic diagram of the light intensity change curve provided in this embodiment;

[0067] Figure 7 It is a schematic flowchart of determining the risk level by the prediction model provided in this embodiment;

[0068] Figure 8 It is a schematic diagram of the structure of the corn water and fertilizer management system provided in this embodiment. Detailed Description of the Embodiment

[0069] In order to better explain the present invention for easy understanding, the present invention will be described in detail below with reference to the accompanying drawings through specific embodiments.

[0070] It should be noted that all the directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship and movement conditions between components in a specific posture (as shown in the drawings). If the specific posture changes, the directional indications will also change accordingly.

[0071] In addition, in the present invention, descriptions such as "first" and "second" are for descriptive purposes only, and should not be construed as indicating or implying their relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0072] In the present invention, unless otherwise clearly specified and defined, terms such as "connection" and "fixation" shall be understood in a broad sense. For example, "fixation" may be a fixed connection, a detachable connection, or integrated; "connection" may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and may be the internal communication of two components or the interaction relationship between two components, unless otherwise clearly defined. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0073] As Figure 1 shown, in this embodiment, an intelligent management method for water and fertilizer in a planting area is provided, and the method includes the steps of:

[0074] S1. Construct a three-dimensional model of the planting area; with reference to the land area size and the crop density relationship, divide the initial planting area of the three-dimensional model and number it to obtain the three-dimensional model of the planting area with divided regions;

[0075] Construct a three-dimensional model of the planting area based on image data, analyze the difference in the degree of change in the height of crops in the planting area, identify the data of the boundary range of the planting area, and obtain the land area size of the planting area; obtain the crop coordinates based on the image data to construct a three-dimensional model of the planting area to obtain the crop density relationship; divide the planting area of the three-dimensional model and number it with reference to the land area size and the crop density relationship; upload the three-dimensional model to the intelligent water and fertilizer management platform and display it to the user in a visual form, and the user can intuitively view the crop growth status, environmental factor changes, and equipment operation conditions on the platform interface to grasp the farmland dynamics in real time.

[0076] S2. Based on the three-dimensional model of the planting area with divided regions, design a sensor perception scheme, including:

[0077] S21. Determine the sensor category; use the analytic hierarchy process to quantify the influence weights of key environmental factors in different growth stages of crops, so as to select sensors that sense the corresponding environmental factors. Generally, a monitoring system may only rely on a single type of sensor to obtain data on a certain environmental factor, but single data can only reflect the current crop growth environment and cannot predict the crop's demand situation. In this embodiment, multiple different types of sensors are integrated to synchronously monitor multiple environmental factors, more comprehensively and accurately reflecting the true situation of the crop growth environment, predicting the water and fertilizer requirements of the crops, and providing a scientific basis for precise irrigation and fertilization.

[0078] S22. Determine the sensor sensing positions, including:

[0079] S221. Uniformly set a preset number of fixed points within each planting area divided in the step S1, and measure the comprehensive value of the key environmental factors at the fixed points.

[0080] S222. Compare whether the dispersion degree of the comprehensive values of all fixed points within each planting area exceeds a preset dispersion degree threshold to determine whether to perform regional subdivision.

[0081] S223. If so, perform regional subdivision, subdivide the planting area into multiple planting areas, uniformly set a preset number of fixed points within the multiple planting areas, and return to step S222; if not, do not perform regional subdivision; stop until the dispersion degree of the comprehensive values of the fixed points within all the planting areas is within the preset dispersion degree threshold. In other embodiments, when the number of subdivided areas is too large to facilitate operation and observation, a certain accuracy can be sacrificed, and the dispersion degree threshold can be appropriately adjusted and increased according to the actual situation, and the judgment of the dispersion degree and the subdivision of the area are performed again.

[0082] S224. Judge whether the dispersion degree of the comprehensive values of all fixed points within two adjacent planting areas is less than the preset dispersion degree threshold.

[0083] S225. If so, merge the two adjacent planting areas, uniformly set a preset number of fixed points within the merged planting area, and return to step S224; if not, return to step S224; stop until adjacent areas cannot be merged or there are no adjacent areas.

[0084] S226. Set the sensor sensing position according to all the obtained planting areas and the positions of the fixed points. In other embodiments, when setting the sensor sensing position, the position of the fixed point should also avoid positions with abnormal data (for example, when measuring the soil moisture content of a certain area, pay attention to avoiding low-lying areas and near irrigation water sources), and select a position that can more accurately sense. After the sensor measures the corresponding environmental factors at the fixed point and processes the data, the data is input into the corresponding area of the three-dimensional model of the water and fertilizer intelligent management platform for the planting area to complete data visualization. When the divided area has not been merged or subdivided, display each piece of data on the original divided area of the model on the platform; if the divided area has been subdivided, specifically display the corresponding data of each small area on the originally divided local area of the model; if the divided area has been merged, display the jointly measured data on the original divided areas it contains respectively.

[0085] S23. Determine the sensor frequency. According to the change characteristics of the key environmental factors, the sensor dynamically adjusts the data acquisition frequency with the data fluctuation; so as to realize that when the data fluctuates greatly, collect data more frequently to capture these data changes to avoid losing important information; when the data is relatively stable, the acquisition frequency can be reduced to save resources and improve efficiency.

[0086] S3. Combine the sensor sensing data, establish a prediction model to predict the future change trend of the decision variable, evaluate the crop growth risk based on the prediction result, and formulate an early intervention strategy to realize the intelligent management of water and fertilizer in the planting area.

[0087] Preferably, in other embodiments, the planting area is a corn planting area and the crop is corn.

[0088] As Figure 2 shown, specifically, in this embodiment, the step S1 further includes:

[0089] S11. Analyze the difference in the change degree of the crop height in the planting area; construct an analysis system for the change degree of the height data, input the model coordinates, and analyze the change situation of the height data corresponding to the model, and set the range of the change degree of the height data. In other embodiments, by locating the lowest point of the area and using this as the origin, set a horizontal plane with the x-axis and y-axis directions to construct a three-dimensional space rectangular coordinate system. Among them, the expression of the analysis system for the change degree of the height data is:

[0090] H = f(Z)

[0091] Z = z * d 模 / d 实

[0092] Among them, Z represents the difference in the coordinate height of a certain point compared to the lowest height in the model, which is used to quantify the degree of change in the coordinate height; z represents the value of this coordinate in the z-axis direction; H represents the actual degree of height change, which directly reflects the undulation of the terrain or crop height; d 实 represents the actual length data between any two points on the same height plane measured in the field, d 模 represents the coordinate difference between any two points on the same height plane in the reconstructed model; in other embodiments, in order to effectively distinguish between crop and non-crop areas, a critical height change value Hz is also set. When the H value calculated by the expression of the height data change degree analysis system is greater than Hz, it can be determined that the height change at this point exceeds the normal range, and then it can be inferred that the coordinate point corresponding to this value is very likely to be the location of the crop, and the coordinate data of this crop is immediately collected and recorded. After determining the crop coordinate point, it is necessary to further analyze this point in detail. Specifically, it is to judge whether there is a sector area with an angle not less than γ and a radius of α on the x and y planes centered on this coordinate point, and there are no other points with too large a height change degree in this sector area. If there is no such sector area, then it can be determined that this coordinate point belongs to the coordinate point of the crop on the boundary of the planting area; conversely, if there is such a sector area, it can be determined that this coordinate point is the coordinate point of the non-boundary crop inside the planting area. After completing the judgment and collection of all crop coordinate points, all the coordinate points determined to be the crops on the boundary of the planting area are extracted. Then, the maximum value Xmax and the minimum value Xmin in the X-axis direction, and the maximum value Ymax and the minimum value Ymin in the Y-axis direction are selected from these boundary crop coordinate points. Finally, through these extreme points, using the corresponding geometric calculation method, the boundary length of the planting area can be accurately calculated. In particular, considering the differences in growth forms, height distributions, etc. of different crop types, a specific parameter w needs to be set according to the specific crop type. When multiple points with too large a height change are detected within a range with a radius of w centered on a certain coordinate point, in order to avoid misjudging different parts of the same crop as multiple crops, special treatment methods need to be taken: regarding the coordinates of these multiple points as different sampling points on the same crop, discarding the remaining points, and finally uniformly selecting the point with the largest height change degree within the range as the representative point to represent the actual coordinates of this crop.

[0093] S12. Identify the data of the boundary range of the planting area to obtain the size of the land area of the planting area; based on the three-dimensional model of the planting area, obtain the model height value, and analyze the change degree of the height value according to the height data change degree analysis system. If the change degree of the height value exceeds the set range of the height data change degree, it is initially determined as the planting area range; for the area initially determined as the planting area range, analyze that if there are no other points with a change degree of the height value exceeding the set range of the height data change degree within a certain range of the horizontal plane in at least one direction, it is determined as the boundary crop of the planting; according to the boundary crop information, determine the boundary value in the three-dimensional model to obtain the size of the planting area; in other embodiments, if the change degree of the height value does not exceed the set range of the height data change degree, it is determined as the external range of the unplanted area or the position of the crop gap. If the change degree of the height value exceeds the set range of the height data change degree, it is determined as the planting area range. After it is determined as the planting area range, if there are no other points with a too large change degree of the height value within a certain range of the horizontal plane in at least one direction, then the data model is determined as the boundary crop of the planting, otherwise it is determined as a non-boundary crop. According to the information of all boundary crops obtained, the boundary information is obtained, the boundary value of the model is obtained, and then based on the proportional relationship between the actual area and the modeling, the actual boundary value of the planting area and its related area are obtained.

[0094] S13. Based on the image data, construct a three-dimensional model of the planting area, obtain the crop coordinates, and get the crop density relationship; and upload the data to the intelligent water and fertilizer management platform, and use red dots to visually represent the crop coordinates in the planting area.

[0095] S14. Taking the size of the land area and the crop density relationship as a reference, divide the initial planting area and number it to obtain the three-dimensional model of the planting area of the divided area. The three-dimensional model is uploaded to the intelligent water and fertilizer management platform to achieve an intuitive visual display.

[0096] Furthermore, in this embodiment, the step S1 further includes: constructing a three-dimensional model of the planting area based on the image data; specifically including the steps:

[0097] S101. Obtain image data; the drone takes a 360° circular photograph of the planting area to collect the video data of the planting area;

[0098] S102. Obtain an image data set by frame extraction of the video data; perform frame extraction processing on the video data of the planting area based on the time interval to obtain an image data set with an overlap rate of more than 40% between adjacent frames;

[0099] S103, pose analysis and model construction: Use the SFM algorithm to perform pose analysis, feature point detection, and extraction. First, find overlapping scenes in the input image. Find overlapping images in the input scene, mark the projections of common points, and output a set of geometrically verified image pairs and the image projection of each point.

[0100] The input scene is defined as I = {Ii|i=1...NI}, and a feature set fi is established, Fi = {(xj,fj)|j=1...NFi}, where xj represents the position, fj is the specific feature associated with each specific position xj, and for each I, fi represents the feature set that detects I at position xj.

[0101] Perform feature matching and geometric verification, using the feature set fi as the appearance description of the image to find images of the same scene part seen, and overlap the test scene for each image.

[0102] Then, a similarity metric is established based on fi, and for each feature in image Ib, the most similar feature in image Ia is found to search for the corresponding feature points, thereby obtaining the feature correspondence of this image pair.

[0103] Output a set of possibly overlapping image pairs C = {{Ia, Ib}|Ia, Ib∈I, a<b} and their associated feature correspondence matrix Mab∈Fa×Fb, and verify the feature correspondence associated with the possibly overlapping image pairs C.

[0104] Output the camera pose. Initialize the camera pose by selecting two viewpoints with the largest visible area between the cameras in the planting area. Leveraging symmetric geometric relationships, calculate the fundamental matrix or essential matrix, which is further decomposed to obtain the camera pose, and finally triangulate to obtain 3D points.

[0105] Select a new image, and then use the triangulated 3D points and the 2D points of the new image to get the camera pose of the new image.

[0106] The model is constructed based on the NeRf model and hash coding in Instant-NGP. For a sampling point x in the planting area image, nearby voxels are found at different resolution levels and hash indices are assigned to the vertices of these voxels through hash mapping.

[0107] A hash table T of dimension f is maintained for each mesh of different resolutions. For all generated mesh vertex indices, the corresponding f-dimensional feature vectors are searched from the corresponding hash table.

[0108] The obtained eigenvectors are linearly interpolated according to the relative position of x in grids of different resolutions.

[0109] The linearly interpolated feature vectors at each resolution are combined with the camera pose information ζ∈R encoded using spherical harmonics E Connected together as a small neural network The input y∈R LF+E .

[0110] The gradient of the neural network loss value is back-propagated, and the network weights of the small MLP are iteratively updated. Finally, numerical accumulation is achieved during the feature vector search process, completing encoding training. The reconstructed scene gradually approaches the actual shape of the cornfield, completing the training of the cornfield scene, outputting the color value c = (R, G, B) and volume density σ of the sampling point. The coordinates and color information of each sampling point are then mapped to the world coordinate system to generate a dense 3D point cloud, thereby realizing 3D reconstruction of the cornfield.

[0111] Optionally, step S21 includes:

[0112] S211, construct a judgment matrix; determine the key environmental factors E that determine the main variables of water and fertilizer i The importance of each factor is determined by relevant data, and the 1-9 scale is used to quantify the relative importance. The 1-9 scale is used to quantify the relative importance. 1 means that the two factors are equally important, 3 means that the former is slightly more important than the latter, 5 means that the former is obviously more important than the latter, 7 means that the former is strongly more important than the latter, 9 means that the former is extremely more important than the latter, and 2, 4, 6, and 8 are the intermediate values of adjacent judgments. Perform pairwise comparisons and construct an n×x judgment matrix A=(a ij ), where a ij Indicates the importance of the i-th environmental factor relative to the j-th environmental factor, and satisfies

[0113] S212, calculate the eigenvector; calculate the maximum eigenvalue λ of the judgment matrix A max and its corresponding eigenvector W; the eigenvector W is normalized to obtain the weight vector of the key environmental factors; in other embodiments, each column element of the judgment matrix A is normalized to obtain a matrix B = (b), where Then add the normalized matrix B row by row to get the vector M=m i ,in Normalize the vector m to get the weight vector W=w i ,in Each element w in the weight vector W i It represents the weight of the impact of the i-th environmental factor on crop growth, yield and quality at this growth stage.

[0114] S213, based on the obtained weight vectors of the key environmental factors, comprehensively considering sensors that better perceive the corresponding environmental factors;

[0115] S214, based on the conditions of the actual planting site, determine the measurement method of the sensor, which includes mobile measurement, fixed measurement, or a combination measurement scheme combining mobile and fixed methods. By deploying sensors on movable devices such as drones and unmanned vehicles, the equipment can patrol and measure environmental factors at multiple fixed points in the planting site, which belongs to mobile measurement. By installing fixed measurement stations in the planting site, installing distributed fixed sensors in local areas of the planting site, etc., it is possible to measure environmental factors in a fixed range of the orchard, which belongs to fixed measurement. This embodiment preferably adopts a combination measurement scheme with mobile measurement as the main method and fixed measurement as the auxiliary method.

[0116] Preferably, if Figure 3 , Figure 4 as well as Figure 5 As shown, the step S22 includes: when the discrete degree of the comprehensive values of different fixed points in the area is greater than the set discrete degree threshold, performing area subdivision processing;

[0117] In the area divided in step S1, evenly distributed fixed points are set and the comprehensive value of key environmental factors X is measured at the fixed points. i ;

[0118] By formula To calculate the discreteness of the integrated value of different fixed points, where μ represents the average value of different fixed-point arrays,

[0119] It is determined whether to perform regional subdivision by comparing the degree of dispersion of the comprehensive values of different fixed points within the region. When the degree of dispersion of the comprehensive values of different fixed points within the region is greater than the dispersion degree threshold, it is necessary to subdivide the region, and the fixed points also increase correspondingly during subdivision; the dispersion degree threshold can be adjusted according to the normal change range of relevant environmental factors, the actual environmental situation, and the accuracy requirements; specifically, in this embodiment, the dispersion degree threshold is 0.1; and the operation of determining whether to perform regional subdivision by comparing the degree of dispersion of the comprehensive values of different fixed points within the region is repeated until the degree of dispersion of the comprehensive values of the fixed points in all regions is within the set dispersion degree threshold; then, it is determined whether to perform regional merging by comparing the degree of dispersion of the comprehensive values of fixed points in adjacent different regions. When the degree of dispersion of the comprehensive values of different fixed points within the region is less than the dispersion degree threshold, it is necessary to merge the adjacent different regions, and the fixed points also decrease correspondingly during merging, and the operation of determining whether to perform regional merging by comparing the degree of dispersion of the comprehensive values of fixed points in adjacent different regions is repeated until the adjacent regions cannot be merged or there are no adjacent regions; after processing, the fixed point positions of the region are set as the sensor sensing positions to guide the setting of the sensors. In other embodiments, in order to avoid data contingency, the regional subdivision or merging is also inspected. A new set of point positions is randomly selected, the environmental factors at the corresponding fixed point positions are measured, and through the formula Calculate the degree of dispersion between the calculated data and the average value calculated previously, and determine whether it meets the dispersion degree threshold.

[0120] Specifically, the step S23 includes:

[0121] S231, set the initial acquisition frequency and the acquisition frequency limit range; estimate the change range of relevant environmental factors, and combine the sensor characteristics to set an initial frequency, and the initial slope is defaulted to 0, that is, the data does not change, and set the initial acquisition frequency as the lower limit acquisition frequency f min ; Determine the limit frequency of the sensor through the sensor specification and actual test, and set the upper limit acquisition frequency f max ;

[0122] S232, analyze the data fluctuation degree and predict the future change trend; use the collected data to fill in the missing points to form a smooth curve; select continuous points near the current data point and calculate the slope k by taking the derivative; when the slope is positive, it means the data is rising, and when the slope is negative, it means the data is falling; the greater the absolute value of the slope, the greater the data fluctuation; observe the change of the slope to predict the future data trend; observe the change of the slope, and predict the possible future change trend of the data. If the slope remains unchanged or slowly increases or decreases in one change direction, the future data may continue to change along the current trend; if the slope changes significantly, it means that the data change trend has changed.

[0123] S233. Adjust the data acquisition frequency in real time according to the actual degree of fluctuation;

[0124] Design an adjustment formula for the acquisition frequency f. Let the slope of the current data point be k, and the initial acquisition frequency be f min :

[0125] f = f min *(1 + m*k)

[0126] Where m is an adjustment coefficient, representing the adjustment amplitude for controlling the acquisition frequency. When the current slope k is greater than the initial slope 0, increase the acquisition frequency to capture the trend of data change. The greater the slope, that is, the greater the degree of fluctuation, the greater the acquisition frequency; conversely, the smaller the acquisition frequency. The magnitude of the adjustment coefficient m can be adjusted according to actual application requirements. k is the current slope, representing the degree of data fluctuation. To avoid missing data mutation points, once a significant change in the slope is detected, the acquisition frequency is set to the upper limit value until the slope basically remains unchanged or slowly increases or decreases in one direction of change, and then the frequency is adjusted to the calculated acquisition frequency f. f represents the next data acquisition frequency, and f min represents the minimum acquisition frequency. When the calculated f is greater than f max , adjust f to the upper limit value f max ; when the calculated f is less than f min , adjust f to the lower limit value f min , ensuring that the acquisition frequency is within the limited range. Specifically, in other embodiments, such as Figure 6 shown, taking the monitoring of the light intensity on a corn planting ground as an example, dynamically adjust the data acquisition frequency of the sensor according to the data fluctuation of the light intensity. Before point A, the sunlight intensity is stable at 0, and according to the adjustment formula of the acquisition frequency f, the acquisition frequency is always the lower limit value; near point A, it is detected that the degree of illumination fluctuation changes significantly, and the acquisition frequency is adjusted to the upper limit value; from point A to point B, it is detected that the slope slowly increases, and the acquisition frequency is adjusted to the value calculated by the adjustment formula of the acquisition frequency f; from B to D, the slope remains basically unchanged, that is, the fluctuation situation remains the same, and the acquisition frequency should also be relatively stable, with only minor adjustments; near point D, it is detected that the degree of illumination fluctuation changes significantly, and the same adjustment as in the above situation is taken.

[0127] Preferably, the planting ground in this embodiment is a corn field, and the crop is corn.

[0128] As a preferred implementation manner of the present invention, the step S3 includes:

[0129] Such as Figure 7As shown in the figure, in step S31, a prediction model is established between the target variable and relevant environmental factors to predict the future change trend of the decision variable. The relevant environmental factors of the target variable can be determined from the key environmental factors in step S21. The historical data of the target variable and relevant environmental factors are collected by real-time monitoring data of corresponding sensors, and the corresponding time data is recorded. The mathematical relationship between the target variable and relevant environmental factors is determined by the least squares method, and future data points are predicted in time series analysis.

[0130] The following illustrates step S31 in this embodiment through a specific operation process. A prediction model is established between the target variable of fertilizer in the corn planting area and relevant environmental conditions. The fertilizer content is mainly related to environmental factors such as temperature, precipitation, soil clay content, and soil pH value. The target variable is the fertilizer content, and the environmental factors are temperature, precipitation, soil clay content, and soil pH value. The data of the environmental factors are measured regularly, and it is ensured that the data cover different seasons, climate conditions, and soil types. To predict the future fertilizer content in the planting area, first calculate the current fertilizer content and the soil fertilizer reduction rate. Let the soil fertilizer reduction rate be R, which can be represented by the reduction amount of a certain fertilizer (such as nitrogen, phosphorus, potassium) content in the soil per unit time, with the unit of mg / kg·d, indicating how many milligrams of nutrient (fertilizer) content are reduced per kilogram of soil per day. mg represents milligram, kg represents kilogram, and d represents day. Assume a linear regression model, where the fertilizer reduction rate R is a linear combination of the independent variables (i.e., the above-mentioned relevant factors).

[0131] Assume that there is the following linear relationship between the soil fertilizer reduction rate and each environment:

[0132] R = β0 + β1T + β2P + β3C + β4pH + ε

[0133] where T, P, C, and pH represent the data values of each relevant environmental factor respectively, β0 is the intercept term, β1, β2, β3, and β4 are regression coefficients, and ε is the error term, which follows a normal distribution with a mean of 0 and a variance of σ 2 of the normal distribution.

[0134] Using the collected historical data, the parameters of the linear regression model are solved by the least squares method to complete parameter estimation. Then, a goodness-of-fit test is performed, and the coefficient of determination R 2 is calculated to evaluate the fitting degree of the model to the data. R 2The closer it is to 1, the better the fitting effect of the model. Then, a significance test is conducted: a t-test is performed on the regression coefficients to determine whether the influence of each independent variable on the dependent variable is significant. If the regression coefficient of an independent variable is not statistically significant, it is considered to be removed from the model. At the same time, residual analysis is carried out to check whether the residuals follow a normal distribution, whether there is heteroscedasticity or autocorrelation. If the residuals do not meet the assumed conditions, the model is corrected. For example, weighted least squares can be used to handle heteroscedasticity or autoregressive models can be used to handle autocorrelation.

[0135] Based on the monitoring data of current environmental factors and the prediction of future environmental changes, relevant environmental factors within a future period are estimated, and the data is input into the fitted model to predict the reduction rate of soil fertilizer and obtain the values of fertilizer content at future time points.

[0136] As Figure 7 shown, in S32, the result predicted by the prediction model in step S31 is compared with the subsequent actual situation to improve the prediction model; the actual data of the target variable and relevant environmental factors within the prediction period are collected, compared with the prediction result, the magnitude of the prediction error is analyzed, and corresponding measures are taken to adjust the model according to the analysis result; in other embodiments, time series analysis or regression analysis methods are also used to predict the error trend within a future period to adjust and optimize the model.

[0137] As Figure 7 shown, in S33, the risk degree of the future growth of the crop is evaluated according to the prediction model; the risk coefficient is a comprehensive index considering the degree of deviation of the decision variable from the threshold, the volatility of environmental factors, and the urgency of the crop's nutrient requirements; it is divided into three levels: low risk coefficient, medium risk coefficient, and high risk coefficient according to the magnitude of the risk coefficient.

[0138] As Figure 7 shown, in S34, based on the prediction model and the risk coefficient, the goal of intervention is clarified, and a scientific and reasonable advance intervention operation plan is formulated, which can enable the crop to receive timely water and nutrient supplementation when facing potential growth risks, thus providing an additional guarantee for its healthy and stable growth.

[0139] S35. In this embodiment, different operation levels G(x) are divided based on the appropriate growth condition thresholds of the set target variables, where x represents the specific level number; the growth condition thresholds are based on the different water and fertilizer requirements of different crops at different times. By referring to relevant materials to obtain the appropriate growth conditions of crops at different times, the appropriate growth conditions are used as thresholds. A relationship model L(x) between the intervention measures and the variables is established, where x represents the specific level number; the impact of the intervention measures on the target variables is clarified. Different intervention measures will have different impacts on the variables. Therefore, a relationship model between the intervention measures and the variables can be established. After dividing the operation levels, corresponding operation plans can be provided for each level according to the intervention model, that is, each G(x) will correspond to a corresponding measure L(x). When there are too many relevant factors affecting the variables, the relevant factors are selected to establish the intervention model according to the environmental variable impact weights determined by S21. The corresponding environmental factors are collected for data, and then a prediction model for the change in variable content is trained according to the machine learning method. The model training method can be set according to the actual situation. For example, the random forest method can be used for model training, and the coefficient of determination index is used to evaluate the model performance.

[0140] Determine the operation plan according to the operation level G(x) in combination with the intervention model L(x); specifically, after dividing the operation level G(x), the intervention model L(x) can provide a corresponding operation plan for each level. That is to say, each operation level G(x) will correspond to a corresponding intervention measure L(x). From the hierarchical division logic of the previous steps, it can be seen that when dividing the levels, the higher the degree of deviation of the conditions from the threshold, the higher the corresponding level number. This means that in the case of a higher level number, the environment lacks the corresponding variables (such as water) more. Based on the results calculated by the intervention model, the required intervention intensity is also higher at this time. However, since excessive intervention at one time will cause waste of resources, the total amount of intervention variables calculated by the intervention model needs to be split into multiple executions. The number of executions is set to the number of levels to be operated, so as to reasonably arrange the intervention operations.

[0141] The following illustrates step S35 in this embodiment through a specific operation process. Maize is a crop that requires a large amount of fertilizer and water, and its demand characteristics for water and fertilizer are different at different growth stages. By referring to relevant materials, it is obtained that:

[0142] Seedling stage: The growth is slow, the amount of water and fertilizer required is small, but the sensitivity to nutrient supply is high, and the growth of the root system requires a certain amount of phosphorus to promote.

[0143] Jointing stage to heading stage: The growth is rapid, which is a period of concurrent vegetative growth and reproductive growth. The demand for water and fertilizer increases sharply, especially the need for sufficient nitrogen supply. At the same time, it is also more sensitive to water, and water shortage will affect ear differentiation.

[0144] Filling stage: It is a crucial period for yield formation, requiring sufficient water and nutrients to ensure the plumpness of grains, and the demand for potassium increases.

[0145] Then set the appropriate soil water content: Assume the normal value of field capacity is x, then the threshold for the seedling stage of maize is approximately 0.6x to 0.65x, the threshold for the jointing-heading stage is 0.7x to 0.8x, and the threshold for the filling stage is approximately 0.75x - 0.85x.

[0146] Taking soil water content as an example, first divide the soil water content into different operation levels according to a certain method: According to the threshold T(th), the water content range above the threshold is divided into the non-operation first level G(0), and the system does not need to perform the water replenishment task at this level; the water content range below the threshold T(th) is divided into the to-be-operated G(1) - G(n), and the system needs to perform the corresponding water replenishment task at this level. Different levels have corresponding water replenishment operation plans.

[0147] Specifically, in the embodiment of the maize seedling stage, perform level division:

[0148] (1) G(1) suitable water content level (60% - 80% field capacity)

[0149] (2) G(2) slightly water-deficient level (40% - 60% field capacity)

[0150] (3) G(3) severely water-deficient level (<40% field capacity)

[0151] Build a model for irrigation in the maize field. The soil water content is related to the intensities of irrigation, precipitation, evaporation, transpiration, deep percolation, etc. Then, according to the water balance, the change in soil water content = irrigation water volume + precipitation amount - evaporation amount - leakage amount - crop transpiration amount. Each factor can be calculated according to the corresponding environmental factors. For example, the soil evaporation intensity is obtained according to E(t) = k E ×E0(t), where k E is the soil evaporation coefficient, which is related to factors such as soil texture and vegetation cover; E0(t) is the reference crop evapotranspiration, which can be calculated through meteorological data (such as air temperature, humidity, wind speed, sunshine hours, etc.).

[0152] The irrigation intensity can be determined according to the type of irrigation system and the irrigation plan. For example, the drip irrigation system can calculate the irrigation intensity through the flow rate of the drippers and the irrigation time. Assume the water volume for each irrigation is V, the irrigation area is A, and the irrigation time is t, then the irrigation water volume can be calculated according to the formula I(t) = W min (t) - W(t), where W min(t) is the lower limit of the soil water content threshold corresponding to the growth stage, and the irrigation area can be obtained according to S2-4. Then the irrigation intensity I(t) can be expressed as:

[0153]

[0154] After calculating the irrigation intensity, divide the levels according to the seedling stage and formulate a sprinkler irrigation plan. For example: it is divided into the current G(x) layer. After calculating I(t), each time an irrigation task is executed to supplement moisture.

[0155] (1) Irrigation operation plan for the level G(0) to be operated: no task to execute;

[0156] (2) Irrigation operation plan for the level G(1) to be operated: the total irrigation volume is V1, the number of irrigation times is 1 time, and the single irrigation volume is V1;

[0157] (3) Irrigation operation plan for the level G(2) to be operated: the total irrigation volume is V2, the number of irrigation times is 2 times, and the single irrigation volume is V2 / 2;

[0158] (4) Irrigation operation plan for the level G(3) to be operated: the total irrigation volume is V3, the number of irrigation times is 3 times, and the single irrigation volume is V3 / 3.

[0159] Optionally, the step S32 further includes:

[0160] By calculating the arithmetic mean of the absolute errors, plot the change of the prediction error over time as a curve graph. For the analysis results, take corresponding measures to adjust the model; if the prediction error is large, retrain the model with a new dataset to calibrate the parameters; if the influence of specific environmental factors on the target variable is not fully reflected in the prediction, add the specific environmental factors to the model; on the contrary, if the influence of specific environmental factors on the prediction result is not significant, delete it from the model.

[0161] As a preferred embodiment of the present invention, step S33 includes: the degree of deviation of the decision variable of the risk coefficient from the threshold, the volatility of environmental factors, and the proportion of the importance of the urgency of crop nutrient requirements are a%, b%, and c% respectively. The evaluation scores for each part are set as x, y, and z, and the value ranges are all 1-10 points. According to the formula β = a%x + b%y + c%z, the risk coefficient β is obtained. When 1 ≤ β ≤ 3, it belongs to low risk; when 3 < β ≤ 7, it belongs to medium risk; when 7 < β ≤ 10, it belongs to high risk. According to the risk level, the basic goals of early intervention are determined. The basic goals can be selected according to actual situations. For example, in the low-risk level, the goal is to maintain the stability of the crop growth environment; in the medium-risk level, the goal is to adjust the nutrient supply to meet the crop needs; in the high-risk level, the goal is to quickly improve the crop growth conditions to reduce the risk.

[0162] The following illustrates step S33 in this embodiment through a specific operation process. During the maize seedling stage, since the root system has not yet fully developed and the ability to absorb nutrients and water is limited, even a small environmental fluctuation may lead to greater growth risks.

[0163] By collecting real-time data on relevant environmental factors such as soil humidity, nutrient content (especially nitrogen), temperature, and light, and obtaining predictions of future changes. Comparing the future prediction data with the growth threshold of the maize seedling stage, identifying the critical time points and deviation degrees at which the nutrient content may be lower than the growth threshold of the maize seedling stage, and judging the urgency of nutrient requirements to obtain the risk coefficient.

[0164] The risk coefficient mainly consists of three parts, namely, the distance of the critical time point when the decision variable exceeds the threshold from the current time, the degree of deviation of the decision variable from the growth threshold, and the urgency of the early intervention measures. The evaluation scores for each part are set as x, y, and z, and the value ranges are all 1-10 points.

[0165] When the critical time point when the decision variable exceeds the threshold is relatively far from the current time, the score x is smaller; otherwise, the score x is larger. When the degree of deviation of the decision variable from the growth threshold is smaller, the score y is smaller; otherwise, the score y is larger. When it is predicted that the urgency of nutrient requirements is low and the top-dressing plan can be implemented after a period of time in the future, the evaluation score z for this item is smaller. If it is predicted that emergency fertilization is required and only by starting the fertilization plan as soon as possible can the decision variable be ensured to maintain the normal value, the evaluation score z for this item is larger.

[0166] According to the actual situation, the proportion of the importance of the three parts is set as a%, b%, and c% respectively. According to the formula:

[0167] β = a%x + b%y + c%z

[0168] The risk coefficient β can be obtained. If other risk coefficient components actually need to be added, they can be added to the calculation formula of β according to the above method.

[0169] According to the risk coefficient, it can be roughly divided into three risk levels: when 1 ≤ β ≤ 3, it belongs to low risk; when 3 < β ≤ 7, it belongs to medium risk; when 7 < β ≤ 10, it belongs to high risk.

[0170] Specifically, the risk assessment form is shown in Table 1 (the importance degrees of the 3 components are 30%, 30%, and 40% respectively, which can be adjusted according to the actual situation).

[0171] Table 1 Risk Assessment Form

[0172]

[0173] In the case of low risk, the growth environment of maize seedlings is relatively stable, and the nutrient content in the soil remains within an appropriate range. At this time, the existing management strategy should be temporarily maintained unchanged, and the changes in soil nutrient content and related environmental factors should be continuously monitored closely to ensure that maize can continuously obtain suitable growth conditions; in the case of medium risk, the growth environment of maize seedlings may show certain fluctuations, and appropriate intervention measures need to be taken to supplement the nutrients required for maize growth. At the same time, the changes in related environmental factors should be closely monitored, and the fertilization plan should be adjusted in a timely manner to avoid adverse effects caused by overuse; in the case of high risk, the growth environment of maize seedlings may face serious challenges, such as severe nutrient deficiency, etc. Immediate emergency measures need to be taken. At the same time, the growth status of maize should be closely monitored, problems should be discovered in a timely manner, and the intervention plan should be adjusted to minimize the risk of maize growth to the greatest extent.

[0174] Furthermore, in this embodiment, step S3 further includes:

[0175] S36. Upload the operation plan of step S35 to the water and fertilizer intelligent management platform, and the water and fertilizer intelligent management platform executes operation tasks according to user terminal instructions or system intelligent instructions; after the operation is completed, at a preset specified time point, the value of the decision variable is fed back, whether it returns to the normal range is judged, and further adjustment instructions are given according to the actual influence situation of the intervention measures; when the sensor sensed data is seriously lower than the threshold value, the water and fertilizer irrigation equipment in the planting area is automatically started;

[0176] The following illustrates step S36 in this embodiment through a specific operation process:

[0177] When the user views the warning issued by the platform and the provided solution and issues an instruction to the platform to directly execute the operation plan obtained by the model; if the user discovers a problem by observing the real-time monitoring data, the user can also send a custom instruction through the user terminal and feedback the problem to the platform. After receiving the instruction, the system will immediately start the operating equipment and complete the operation according to the received instruction.

[0178] During the monitoring process, if an intervention instruction fails to be issued in time, resulting in an emergency where the decision variable continuously deviates severely from the normal value, the system will automatically start the field equipment and intervene according to the operation plan. Specifically, if the soil nitrogen content is severely insufficient and the user fails to issue a fertilization instruction in time, the system will automatically start the fertilization equipment and apply emergency fertilization to the corn field according to the preset fertilization amount and fertilization method to avoid growth retardation caused by nitrogen deficiency.

[0179] After the operation is completed, if it is found through data comparison and analysis after a certain period of time that the decision variable does not reach the appropriate range, the platform will issue an adjustment instruction again, suggesting that the user increase the amount of water and fertilizer application or adjust the time of water and fertilizer application; if the decision variable has returned to the normal range and the growth condition of the corn is stable, the platform will automatically cancel the warning, indicating that the intervention task has been completed.

[0180] The present invention also provides a smart water and fertilizer management system for planting land, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it realizes the steps of the smart water and fertilizer management method for planting land as described in any one of the above.

[0181] As Figure 8 shown, the following illustrates a smart water and fertilizer management system for planting land in this embodiment through a specific system structure.

[0182] A corn water and fertilizer management system includes a data collection and management center, a data analysis and decision center, and an operation feedback center.

[0183] The data collection and management center. This center is composed of a planting land visualization module and a regional data monitoring module.

[0184] The planting area visualization module is divided into an image acquisition sub-module, a map reconstruction sub-module, and an image information processing sub-module. The image acquisition sub-module mainly uses a drone to take 360° panoramic photos of the cornfield, collecting complete and high-definition videos of the cornfield, providing a clear and accurate information source for subsequent image processing; the planting area reconstruction sub-module mainly uses image processing technology to perform pose analysis on the collected image data, generating a corresponding three-dimensional model, enabling users to more intuitively understand the corn planting area; the image information processing sub-module mainly automatically obtains the planting area and crop density of the planting area based on the reconstructed model of the corn planting area, and conducts scientific regional division and numbering of the planting area, providing a solid foundation for subsequent data monitoring and operations.

[0185] The regional data monitoring module is divided into a measurement point distribution sub-module and a sensor measurement sub-module. The measurement point distribution sub-module mainly determines the local measurement points of the sensor based on the regions divided in the reconstructed model, combined with the characteristics of environmental factors and the actual terrain of the corn planting area, ensuring the comprehensiveness and accuracy of the data. The sensor measurement sub-module is mainly used to determine the sensor measurement scheme, including the selection of the sensor and its measurement method, the deployment of the sensor, and the adjustment of the data acquisition frequency, ensuring the real-time and reliability of the data. The sensor detects the regional data in real time, feeds the data back to the data acquisition and management center, and visually displays it in the planting area model. On the premise of ensuring the measurement accuracy of the data, the monitoring efficiency is greatly improved.

[0186] The data analysis and decision-making center. This center consists of a data analysis module, a mathematical model establishment module, and an operation suggestion generation module.

[0187] The data analysis module is mainly responsible for analyzing and visualizing the real-time data of the collected decision variables and related environmental factors, facilitating subsequent data processing, and enabling users to clearly understand the ecological environment and crop growth status of the current corn planting area.

[0188] The mathematical model establishment module is divided into an intervention model establishment sub-module and a prediction model establishment sub-module. The intervention model establishment sub-module is mainly used to establish a simulation mathematical model between the intervention measures and the decision variables, clarifying the different impacts of different intervention measures on the decision variables. The data of the prediction model establishment sub-module is mainly used to establish a simulation model between the target variables and related environmental factors to predict the future change trend of the decision variables, providing a forward-looking decision-making basis for users. In addition, the mathematical model can be visualized using charts, making the prediction results of the model more intuitive and easy to understand, facilitating managers to make decisions quickly.

[0189] The operation suggestion generation module mainly generates operation plans by dividing the current operation level according to the intervention model and prediction model established in the mathematical model establishment module, or determines the future crop growth risk level and generates early intervention operation plans. These suggestions not only help users take targeted agricultural operation measures in a timely manner to improve the yield and quality of crops, but also effectively prevent potential risks during crop growth and reduce losses in agricultural production.

[0190] Operation feedback center. This center consists of an early warning module, a comprehensive operation module, and a feedback module.

[0191] The early warning module mainly relies on the accurate analysis results provided by the data analysis module. Once the conclusion drawn by the data analysis module shows that the measured value is lower than the ideal value, the early warning module will be quickly activated. Through a sound alarm, a prominent prompt window will pop up on the user's device screen to attract the user's attention.

[0192] The comprehensive operation module mainly includes a user-defined sub-module and an application self-start sub-module. In the user-defined sub-module, the user can formulate an operation plan based on the model shown in the planting area visualization and their own experience, or directly execute the plan provided by the operation plan generation module with one key. To avoid losses caused by the user's neglect of operations, the system has a special self-start module. Once the analysis result of the data analysis module shows that the measured value is seriously lower than the ideal value and has damaged the plant growth, the self-start module will automatically execute relevant operations according to the established operation plan to ensure that the plant can be properly maintained and managed in a timely manner.

[0193] The feedback module mainly includes a model improvement sub-module and an operation adjustment sub-module. When initially constructing the model in the mathematical model establishment module, due to heavy reliance on ideal estimates and current limited data values, there are inevitably certain limitations. The model improvement sub-module conducts in-depth analysis of these initially established models by leveraging a large amount of data real-time feedback from sensors. Through precise interpretation and in-depth mining of the feedback data, it can accurately identify the deficiencies in the model, and then targetedly adjust and improve the model, enabling the model to continuously optimize with the changes in the actual situation, thus better fitting the actual scenario and improving the accuracy and practicality of the model. The operation adjustment sub-module closely cooperates with and complements the model improvement sub-module. As the model improvement sub-module continuously optimizes the model, the operation adjustment sub-module comprehensively evaluates and analyzes the existing operation plan based on the improved model. It will identify possible unreasonable points in the original operation plan according to the updated results of the model and promptly make corresponding changes and optimizations to the operation plan. In this way, the operation adjustment sub-module can ensure that the operation plan always matches the actual situation, provide more scientific and reasonable operation guidance for users, effectively improve the efficiency and quality of operations, and contribute to the healthy growth and management of plants.

[0194] The above are only specific application examples of the present invention and do not constitute any limitation to the protection scope of the present invention. In addition to the above embodiments, the present invention may also have other implementation manners. All technical solutions formed by equivalent replacement or equivalent transformation fall within the scope of protection required by the present invention.

Claims

1. An intelligent management method for water and fertilizer in a planting area, characterized in that, Including the steps: S1. Construct a three-dimensional model of the planting land with divided areas; with reference to the land area size and the crop density relationship, divide the initial planting areas of the three-dimensional model and number them to obtain the three-dimensional model of the planting land with divided areas; S2. Based on the three-dimensional model of the planting land with divided areas, design a sensor sensing scheme, including: S21. Determine the sensor category; use the analytic hierarchy process to quantify the influence weights of key environmental factors in different growth stages of crops to select sensors for sensing corresponding environmental factors; S22. Determine the sensor sensing positions, including: S221. Uniformly set a preset number of fixed points in each planting area divided in step S1, and measure the comprehensive value of the key environmental factors of the fixed points; S222. Compare whether the dispersion degree of the comprehensive values of all fixed points within each planting area exceeds a preset dispersion degree threshold to judge whether to perform regional subdivision; S223. If so, perform regional subdivision, subdivide the planting area into multiple planting areas, and uniformly set a preset number of fixed points in the multiple planting areas, and return to step S222; if not, do not perform regional subdivision; until the dispersion degree of the comprehensive values of the fixed points in all the planting areas is within the preset dispersion degree threshold and then stop; S224. Judge whether the dispersion degree of the comprehensive values of all fixed points in two adjacent planting areas is less than the preset dispersion degree threshold; S225. If so, merge the two adjacent planting areas, and uniformly set a preset number of fixed points in the merged planting area, and return to step S224; if not, return to step S224; until adjacent areas cannot be merged or there are no adjacent areas; S226. Set the sensor sensing positions according to all the planting areas obtained after processing and the positions of the fixed points; S23. Determine the sensor frequency; according to the change characteristics of the key environmental factors, the sensor dynamically adjusts the data acquisition frequency according to the data fluctuation; S3. Combine the sensor sensing data, establish a prediction model to predict the future change trend of the decision variables, evaluate the crop growth risk based on the prediction results, formulate an early intervention strategy, and realize the intelligent management of water and fertilizer in the planting land.

2. The intelligent management method for water and fertilizer in a planting area according to claim 1, wherein, The step S1 further includes: S11. Analyze the difference in the change degree of the crop height in the planting land; construct an analysis system for the change degree of height data, input the model coordinates, analyze the change of the corresponding height data of the model, and set the range of the change degree of height data; among them, the expression of the analysis system for the change degree of height data is: H = f(Z) Z = z * d 模 / d 实 Among them, Z represents the difference in the coordinate height of a certain point compared to the lowest height in the model, z represents the value of the coordinate in the z-axis direction, H represents the actual height change degree, d 实 represents the actual length data between any two points on the same height plane measured on the ground, d 模 represents the coordinate difference between any two points on the same height plane in the reconstructed model; S12, identifying the boundary range data of the planting site to obtain the land area size of the planting site; based on the three-dimensional model of the planting site, obtaining the model height value, analyzing the degree of change of the height value according to the height data change degree analysis system, if the degree of change of the height value exceeds the set height data change degree range, it is preliminarily determined to be the planting site range; for the area preliminarily determined to be the planting site range, if there is no other point whose degree of change of the height value exceeds the set height data change degree range within a certain range of the horizontal plane in at least one direction, it is determined to be a planting boundary crop; based on the boundary crop information, the boundary value in the three-dimensional model is determined to obtain the area size of the planting site; S13, constructing a three-dimensional model of the planting site based on the image data, obtaining crop coordinates, and determining the density relationship of the crops; S14, dividing the initial planting area into regions and numbering them based on the land area size and the crop density relationship, so as to obtain a three-dimensional model of the planting area in the divided regions.

3. The intelligent management method for water and fertilizer in a planting area according to claim 2, characterized in that, The step S1 further includes: constructing a three-dimensional model of the planting site based on the image data; specifically including the steps of: S101, acquiring image data; the drone performs 360° surround photography of the planting site to collect video data of the planting site; S102, extracting frames from the video data to obtain an image dataset; extracting frames from the video data of the planting site based on a time interval to obtain an image dataset with an overlap rate of more than 40% between adjacent frames; S103, posture analysis and model construction; use the Sfm algorithm for posture analysis, quote the hash code in Instant-NGP and build the model based on the NeRf model.

4. The intelligent management method for water and fertilizer in a planting area according to claim 1, wherein The step S21 includes: S211. Construct a judgment matrix; determine the key environmental factors E of the main decision variables of water and fertilizer, and the importance of each factor. Use the 1-9 scale method to quantify this relative importance, make pairwise comparisons, and construct an n×x judgment matrix A=(a i )), where a ij represents the importance degree of the i-th environmental factor relative to the j-th environmental factor, and satisfies ij ​ S212, calculate the eigenvector; calculate the maximum eigenvalue λ of the judgment matrix A max and its corresponding eigenvector W; the eigenvector W is normalized to obtain the weight vector of the key environmental factors; S213, based on the obtained weight vectors of the key environmental factors, comprehensively considering sensors that better perceive the corresponding environmental factors; S214, determining a measurement method of the sensor based on the conditions of the actual planting site, wherein the measurement method of the sensor includes a mobile measurement, a fixed measurement, or a combined measurement solution of the mobile and fixed measurement methods.

5. The intelligent management method for water and fertilizer in a planting area according to claim 1, wherein The step S22 includes: measuring the comprehensive value X of key environmental factors at the fixed point i ; Calculate the degree of dispersion of the comprehensive values of different fixed points through the formula where μ represents the average value of different fixed point arrays.

6. The intelligent management method for water and fertilizer in a planting area according to claim 1, characterized in that, The step S23 includes: S231, Set the initial acquisition frequency and the acquisition frequency limit range; Estimate the change range of relevant environmental factors, and combine with the sensor characteristics to set the initial acquisition frequency as the lower limit acquisition frequency; Determine the limit frequency of the sensor through the sensor specification sheet, and set the upper limit acquisition frequency f max ; S232, analyzing the degree of data fluctuation and predicting future trends; using the collected data to fill in gaps and form a smooth curve; selecting consecutive points near the current data point, calculating the slope k by differentiation, and observing the slope change to predict future data trends; S233, adjusting the data collection frequency in real time according to the actual fluctuation level; Design the adjustment formula for acquisition frequency f: f = f min *(1 + m * |k|) Among them, m is the adjustment coefficient, representing the adjustment amplitude for controlling the acquisition frequency; k is the current slope, representing the degree of data fluctuation, f represents the next data acquisition frequency, f min represents the minimum acquisition frequency. When the calculated f is greater than f max , f is adjusted to the upper limit value f max ; when the calculated f is less than f min , f is adjusted to the lower limit value f min , ensuring that the acquisition frequency is within the limited range.

7. The intelligent management method for water and fertilizer in a planting area according to claim 1, wherein The step S3 comprises: S31, establishing a prediction model between the target variable and related environmental factors to predict the future change trend of the variable; determining the related environmental factors of the target variable based on the key environmental factors in step S21, monitoring data in real time through corresponding sensors, collecting historical data of the target variable and related environmental factors, and recording the corresponding time data; determining the mathematical relationship between the target variable and related environmental factors through the least squares method, and predicting future data points in time series analysis; S32. Compare the result predicted by the prediction model in step S31 with the subsequent actual situation to improve the prediction model; collect the actual data of the target variable and relevant environmental factors during the prediction period, compare it with the prediction result, analyze the magnitude of the prediction error, and take corresponding measures for model adjustment according to the analysis result. S33. Evaluate the risk degree of the future growth of the crop according to the prediction model; the risk coefficient considers a comprehensive index that determines the degree of deviation of the variable from the threshold, the volatility of environmental factors, and the urgency of the crop's nutrient requirements; it is divided into three levels: low risk coefficient, medium risk coefficient, and high risk coefficient according to the magnitude of the risk coefficient. S34. Based on the prediction model and the risk coefficient, clarify the intervention target. S35. Divide different operation levels G(x) based on the appropriate growth condition threshold of the set target variable, where x represents the specific level number; establish a relationship model L(x) between the intervention measure and the variable, where x represents the specific level number; clarify the impact of the intervention measure on the target variable, and determine the operation plan according to the operation level G(x) combined with the intervention model L(x) to assist in water and fertilizer management.

8. The intelligent management method for water and fertilizer in a planting area according to claim 7, wherein The step S32 further includes: By calculating the arithmetic mean of the absolute error, plot the change of the prediction error over time as a curve graph, and take corresponding measures for model adjustment according to the analysis result; if the prediction error is large, retrain the model with a new data set to calibrate the parameters; if the impact of a specific environmental factor on the target variable is not fully reflected in the prediction, add the specific environmental factor to the model; on the contrary, if the impact of a specific environmental factor on the prediction result is not significant, delete it from the model.

9. The intelligent management method for water and fertilizer in a planting area according to claim 7, characterized in that, The step S33 includes: the proportion of the degree of deviation of the decision variable of the risk coefficient from the threshold, the volatility of environmental factors, and the importance of the urgency of the crop's nutrient requirements are a%, b%, and c% respectively, and the evaluation scores of each part are set as x, y, and z respectively, and the value range is all 1 - 10 points. The risk coefficient β is obtained according to the formula β = a%x + b%y + c%z. When 1 ≤ β ≤ 3, it belongs to low risk; when 3 < β ≤ 7, it belongs to medium risk; when 7 < β ≤ 10, it belongs to high risk; according to the risk level, determine the basic target of early intervention.

10. An intelligent water and fertilizer management system for planting land, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent water and fertilizer management method for the planting land as described in any one of claims 1 to 9.

Images