Corn adaptive variable seeding method based on soil spatial difference analysis

Abstract

The present application relates to the technical field of data processing, and especially relates to a corn adaptive variable seeding method based on soil spatial difference analysis, which comprises the following steps: obtaining the first soil attribute correlation degree between all sampling points in a target planting area, and obtaining the correlation degree sequence of each sampling point; obtaining the first difference degree of the correlation degree sequence between the sampling points as the clustering space distance to cluster the sampling points, and obtaining the correlation degree fitting curve of each category; obtaining the interpolation weight of each sampling point to the to-be-estimated point according to the spatial distance between the to-be-estimated point and the sampling point and the correlation degree fitting curve of the category to which the sampling point belongs, and performing weighted summation on the first soil attribute time series data of each sampling point to obtain the first soil attribute time series data of the to-be-estimated point, which is used to construct a soil attribute spatial distribution model of the target planting area and realize adaptive variable seeding of corn. The method can effectively avoid the serious waste of seeds and agricultural materials.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an adaptive variable sowing method for maize based on soil spatial variability analysis. Background Technology

[0002] As modern agriculture develops towards intelligence and precision, the planting density and sowing method of maize, a major food crop in my country, have a crucial impact on crop growth, nutrient absorption, and final yield. The development of precision agriculture technology has made it possible to conduct refined zoning management and intelligent decision-making based on soil properties (such as soil nutrient content, moisture, pH, and temperature). When implementing adaptive variable sowing for maize, the primary prerequisite is obtaining accurate spatial distribution information of soil properties within the farmland.

[0003] Since the number and density of sensors or sampling points in actual farmland are limited, existing technologies generally use spatial interpolation methods (such as inverse distance weighted interpolation (IDW) and Kriging interpolation) to estimate soil properties in unknown areas. This transforms discrete point data into a continuous spatial distribution model of soil properties, and generates a sowing prescription map based on this model to guide variable seeders to adaptively adjust sowing parameters at different locations.

[0004] However, existing methods for maize variable sowing based on spatial interpolation to construct soil models have inherent limitations. Existing interpolation algorithms typically rely on a unified "global distance assumption," which assumes that things closer together are more similar, and apply a fixed mathematical model to the entire plot. However, in actual agricultural production, soil properties within farmland often exhibit strong "spatial heterogeneity" due to the combined effects of micro-topographical undulations, water flow convergence, and localized historical fertilization or cultivation differences. This means that the decay and change patterns of soil properties with increasing spatial distance vary significantly across different local areas. Current techniques, when calculating interpolation weights for unknown points, lack in-depth analysis of the temporal correlation data of sampling points, failing to identify and extract these differences in spatial variation patterns.

[0005] In summary, existing adaptive variable sowing methods for maize mainly suffer from the following technical problems: When constructing the core spatial model of soil properties, existing methods cannot accurately identify and classify soil heterogeneity categories with similar spatial variation patterns, nor can they adaptively adjust the interpolation weights based on the spatial attenuation patterns specific to different geological categories. This "one-size-fits-all" approach to interpolation weight allocation leads to large estimation errors in the constructed spatial distribution model of soil properties, failing to accurately depict the complex microscopic heterogeneity of the plot. Variable sowing strategies based on this spatially biased model deviate significantly from the actual physical conditions of the plot, resulting in insufficient adaptation of sowing parameters to the soil environment at different locations. This not only leads to serious waste of seeds and agricultural inputs but also results in uneven emergence and growth of maize seedlings, ultimately making it difficult to achieve stable and high-yield planting goals.

[0006] Therefore, how to construct a more accurate spatial distribution model of soil properties and avoid serious waste of seeds and agricultural inputs has become an urgent problem to be solved. Summary of the Invention

[0007] In view of this, embodiments of the present invention provide a maize adaptive variable sowing method based on soil spatial variability analysis to solve the problems of large errors in the constructed soil property spatial distribution model and serious waste of seeds and agricultural materials when sowing maize using adaptive variable methods.

[0008] This invention provides an adaptive variable sowing method for maize based on soil spatial variability analysis, which includes the following steps: Obtain the first soil attribute time series data of each sampling point in the target planting area, and obtain the first soil attribute correlation degree between every two sampling points based on the first soil attribute time series data; Based on the correlation degree of the first soil attribute between every two sampling points and the corresponding spatial distance, the correlation degree sequence of each sampling point is obtained. The correlation degree sequence of any sampling point is composed of the correlation degree between any sampling point and any sampling point at a location with a different spatial distance from any sampling point. The first degree of difference in the correlation sequence between sampling points is used as the clustering distance to cluster the sampling points and obtain at least one category; the relationship between the correlation and spatial distance of sampling points in the same category is fitted to obtain the correlation fitting curve for each category; Based on the spatial distance between the point to be estimated and the sampling point, and the correlation fitting curve of the category to which the sampling point belongs, the interpolation weight of each sampling point to be estimated is obtained, and the sum of the interpolation weights of each sampling point to be estimated is 1; the first soil attribute time series data of each sampling point is weighted and summed according to the interpolation weights to obtain the first soil attribute time series data of the point to be estimated. Based on the time series data of the first soil properties of all sampling points and all points to be estimated, a spatial distribution model of soil properties in the target planting area is constructed, and adaptive variable sowing of maize is implemented based on the spatial distribution model of soil properties.

[0009] Preferably, the step of obtaining the correlation degree of the first soil attribute between every two sampling points based on the time series data of the first soil attribute includes: Obtain the grey relational coefficient sequence between the first soil attribute time series data of any two sampling points. For the i-th grey relational coefficient in the grey relational coefficient sequence, calculate the absolute value of the difference between the i-th grey relational coefficient and other grey relational coefficients. Sort all the absolute values ​​of difference in ascending order. The first n absolute values ​​of difference in the sorted absolute values ​​of difference are denoted as the target absolute value of difference. The other grey relational coefficients corresponding to the target absolute value of difference are used as the comparison objects of the i-th grey relational coefficient. Other grey relational coefficients refer to the grey relational coefficients in the grey relational coefficient sequence other than the i-th grey relational coefficient. The overall correlation difference value of the i-th grey correlation coefficient is obtained based on the absolute value of the difference between the i-th grey correlation coefficient and each of the comparison objects. Based on the overall correlation difference value of each grey correlation coefficient in the grey correlation coefficient sequence, the weight of each grey correlation coefficient is obtained. Based on the weight, the grey correlation coefficients in the grey correlation coefficient sequence are weighted and summed to obtain the first soil attribute correlation degree between any two sampling points.

[0010] Preferably, the step of obtaining the correlation sequence of each sampling point based on the first soil attribute correlation degree and the corresponding spatial distance between every two sampling points includes: Obtain the spatial distance set corresponding to each sampling point. The spatial distance set corresponding to the a-th sampling point is the set composed of the spatial distances between the a-th sampling point and all other sampling points. Take the union of the spatial distance sets corresponding to all sampling points to obtain the overall distance set. The method involves obtaining the first soil attribute correlation degree corresponding to each sampling point at each distance in the overall distance set, and sorting the first soil attribute correlation degrees in ascending order of distance to form a correlation degree sequence for each sampling point; wherein, the method for obtaining the first soil attribute correlation degree corresponding to the a-th sampling point at each distance in the overall distance set includes: The difference set is obtained by subtracting the overall distance set from the spatial distance set corresponding to the a-th sampling point; a two-dimensional plane coordinate system is constructed, where the vertical axis represents the correlation degree of the first soil attribute between sampling points and the horizontal axis represents the spatial distance between sampling points; the spatial distance between the a-th sampling point and other sampling points and the corresponding correlation degree of the first soil attribute are mapped to the two-dimensional plane coordinate system in the form of scattered points, and the scattered points are interpolated using a cubic spline interpolation algorithm to obtain the correlation degree of the first soil attribute at each spatial distance in the difference set, thereby obtaining the correlation degree corresponding to the a-th sampling point at each distance in the overall distance set.

[0011] Preferably, the method for obtaining the first degree of difference in the correlation sequence between sampling points includes: For any two sampling points, the stability index of the correlation sequence of the two sampling points is obtained respectively, and the absolute value of the difference between the two stability indices is obtained. The absolute value of the difference is normalized to obtain the deviation. The correlation between the correlation sequences of the two sampling points is calculated, and the correlation is normalized to obtain the comprehensive correlation. Based on the deviation and the comprehensive correlation, the first degree of difference of the correlation sequence between the two sampling points is obtained.

[0012] Preferably, obtaining the stability index of the correlation sequence between any two sampling points includes: For any one of the two sampling points, calculate the curvature of each data point in the correlation sequence of the sampling point to obtain a curvature sequence; perform first-order difference on the curvature sequence to obtain a difference sequence; and calculate the mean of the absolute values ​​of each difference value in the difference sequence as a stability index of the correlation sequence of the sampling point.

[0013] Preferably, obtaining the first degree of difference in the correlation sequence between any two sampling points based on the deviation and the overall correlation includes: Obtain the difference between the constant 1 and the comprehensive correlation degree, calculate the mean between the difference and the deviation, and obtain the first degree of difference of the correlation degree sequence between any two sampling points.

[0014] Preferably, fitting the relationship between the correlation degree and spatial distance of sampling points within the same category to obtain a correlation degree fitting curve for each category includes: For the j-th category, based on the correlation sequence of each sampling point, the relationship between the correlation and spatial distance of the sampling points in the j-th category is fitted multiple times to obtain the fitting curve after each fitting. After each fitting, the corresponding fitting loss is calculated based on the fitting curve. The fitting loss is positively correlated with the deviation between the fitted value and the true value corresponding to each correlation in the correlation sequence. The corresponding fitting effect evaluation index is calculated based on the fitting loss. The fitting effect evaluation index is negatively correlated with the fitting loss. The fitting curve with the largest fitting effect evaluation index is selected as the correlation fitting curve for the j-th category.

[0015] Preferably, the formula for calculating the fitting loss is: ; In the formula, L represents the fitting loss, and z represents the number of sampling points in the j-th category. This represents the length of the correlation sequence at the a-th sampling point. This represents the correlation degree of the h-th first soil attribute in the correlation degree sequence of the a-th sampling point in the j-th category. This represents the correlation fit value (calculated from the fitted curve) corresponding to the h-th first soil attribute correlation in the correlation sequence of the a-th sampling point in the j-th category. This represents the mean of the first degree of difference between the association sequence of the a-th sampling point in the j-th category and the association sequences of other sampling points in the j-th category. Represents the absolute value symbol.

[0016] Preferably, the fitting performance evaluation index calculated based on the fitting loss includes: After the k-th fitting, the correlation fitting value corresponding to each spatial distance in the overall distance set is obtained based on the fitting curve, and the correlation fitting values ​​are sorted in order of spatial distance from smallest to largest to obtain the correlation fitting value sequence. Obtain the first-order difference sequence of the correlation fitting value sequence. If any difference value in the first-order difference sequence is less than or equal to 0, the monotonic discriminant value of the difference value is 1. If any difference value in the first-order difference sequence is greater than 0, the monotonic discriminant value of the difference value is 0. Take the mean of the monotonic discriminant values ​​of each difference value in the first-order difference sequence to obtain the monotonicity evaluation index of the k-th fitting.

[0017] Preferably, the step of obtaining the interpolation weight of each sampling point for the point to be estimated based on the spatial distance between the point to be estimated and the sampling point, and the correlation fitting curve of the category to which the sampling point belongs, includes: The formula for calculating the interpolation weight of the a-th sampling point for the point to be estimated x is: ; in, This represents the interpolation weight of the a-th sampling point for the point x to be estimated. This represents the spatial distance between the a-th sampling point and the point x to be estimated. This represents the correlation curve fitted based on the category to which the a-th sampling point belongs, at a spatial distance of... The corresponding correlation fitting value at time, denoted by , represents the mean of the sum of the absolute values ​​of the differences between the correlation sequence of the a-th sampling point and its fitted correlation sequence, and v represents the number of sampling points.

[0018] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: This invention breaks through the limitations of traditional spatial interpolation algorithms that rely on a single absolute physical distance for correlation assumptions. By clustering sampling points with similar spatial heterogeneity patterns (the first degree of difference in the correlation sequence) and fitting a dedicated "distance-correlation" decay curve for each category of sampling points, the unique geological spatial heterogeneity evolution patterns of the region can be fully considered when calculating the interpolation weights for unknown points. This weighted interpolation method based on dynamic heterogeneity patterns can extremely accurately characterize local soil property mutations caused by micro-topography, historical cultivation, etc., in large-scale farmland, significantly reducing the estimation error of continuous soil spatial models. Therefore, using the method of this invention for adaptive variable sowing of maize can effectively avoid serious waste of seeds and agricultural materials. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a maize adaptive variable sowing method based on soil spatial difference analysis provided in Embodiment 1 of the present invention. Detailed Implementation

[0021] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0022] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0023] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0024] See Figure 1 This is a flowchart of a maize adaptive variable sowing method based on soil spatial variability analysis provided in Embodiment 1 of the present invention, as follows: Figure 1 As shown, this maize adaptive variable sowing method based on soil spatial variability analysis may include: Step S101: Obtain the first soil attribute time series data of each sampling point in the target planting area, and obtain the first soil attribute correlation degree between each pair of sampling points based on the first soil attribute time series data.

[0025] In maize cultivation, soil properties significantly influence planting density and seed quantity. Differences in nutrient and water requirements under different soil conditions necessitate targeted and adaptable planting strategies. For instance, in areas with high soil fertility and good water retention, planting density can be appropriately increased to enhance yield per unit area; conversely, in areas with poor soil or limited water resources, planting density should be reduced to avoid intensified competition among crops and negatively impacting growth. Therefore, developing reasonable planting strategies based on the specific soil conditions within a plot is crucial for achieving stable and high maize yields.

[0026] Therefore, existing technologies typically obtain plot attribute distribution information through soil sampling and spatial modeling to serve as the basis for planting strategies. However, in actual farmland environments, soil attributes not only exhibit continuous spatial variations but are also subject to local fluctuations due to factors such as irrigation and tillage activities. This makes it difficult for a single spatial distribution result based on discrete sampling to fully reflect the true state of the plot. For example, within the same management zone, some areas may experience water accumulation due to low-lying terrain, resulting in significant differences from surrounding areas; or uneven historical fertilization may cause fine-grained variations in nutrient distribution. These situations directly impact planting decisions.

[0027] Therefore, in this embodiment of the invention, spatial modeling of soil properties during the corn planting process is used to improve the accuracy of subsequent adaptive variable sowing decisions. To this end, multiple types of sensors are deployed within the target corn planting area using a combination of sparse and locally dense deployment based on plot size and soil differences, achieving effective coverage of the area. Specifically, sampling points are determined within the target planting area based on plot size and soil differences using a combination of sparse and locally dense deployment, and sensors are deployed at each sampling point to collect time-series data of the first soil property, thereby achieving effective coverage of the target planting area. For example, the deployment density is appropriately increased in areas with significant differences, while sparse deployment is used in relatively uniform areas to balance accuracy and cost. The sensors continuously collect the first soil property data at each sampling point to obtain time-series data of the first soil property, and historical monitoring data from one consecutive month is selected as the analysis sample.

[0028] It should be noted that in this embodiment, the first soil attribute can be soil moisture. In other embodiments, the first soil attribute can also be soil attributes such as pH value, soil temperature, and light intensity. To address issues such as inconsistent sampling frequencies among multi-source data, time-series data of the collected first soil attribute can be aligned using time series processing, including timestamp unification, resampling, and missing value completion, to ensure the consistency and usability of different types of first soil attribute data. This provides a reliable data foundation for subsequent soil spatial model construction and variable-based sowing decisions.

[0029] After obtaining the time series data of the first soil attribute of each sampling point, the time series data of the first soil attribute of different sampling points are analyzed to obtain the correlation degree of the first soil attribute between different sampling points. By calculating the correlation degree of the first soil attribute between each pair of sampling points, the relationship between the correlation degree of the first soil attribute of the sampling point and other sites and the corresponding spatial distance can be analyzed in the future. This helps to build a distribution characteristic model that can more accurately represent the actual situation of soil space, thereby providing more reliable data support and decision-making basis for maize variable planting.

[0030] In this embodiment, the correlation degree between the first soil attributes of two sampling points can be calculated using the grey relational analysis method based on the time series data of the first soil attributes of the two sampling points. The time series data of the first soil attributes of one sampling point can be used as a reference sequence, and the time series data of the first soil attributes of the other sampling point can be used as a comparison sequence. The reference sequence and the comparison sequence are then dimensionless, and the grey relational coefficient and the correlation degree of the first soil attribute are calculated.

[0031] In addition, the correlation between the first soil attributes of two sampling points can also be measured by calculating the Euclidean distance or cosine similarity between the time series data of the first soil attributes of the two sampling points. The larger the Euclidean distance, the smaller the correlation between the first soil attributes; the larger the cosine similarity, the greater the correlation between the first soil attributes.

[0032] Preferably, in this embodiment, taking sampling point A and sampling point B as examples, the time series data of the first soil attribute of sampling point A and the time series data of the first soil attribute of sampling point B are obtained respectively. Then, based on the time series data of the first soil attribute of sampling point A and the time series data of the first soil attribute of sampling point B, the correlation degree of the first soil attribute between sampling point A and sampling point B is obtained. The specific method for obtaining this correlation degree is as follows: Based on the time-series data of the first soil attribute of sampling point A and sampling point B, the gray-level correlation coefficients at each sampling time are obtained, thus obtaining the gray-level correlation coefficient sequence between sampling point A and sampling point B, denoted as . ,in, This represents the sequence of grey relational coefficients between sampling point A and sampling point B. This represents the m-th grey relational coefficient in the grey relational coefficient sequence.

[0033] To obtain a stable and reliable correlation measure of the first soil attribute between sampling point A and sampling point B, and to avoid abnormal disturbances in the local correlation during the calculation of the correlation degree, which could affect the stability of the comprehensive correlation analysis of the first soil attribute corresponding to sampling point A and sampling point B over the entire time series, this embodiment of the invention, for the i-th grey correlation coefficient in the grey correlation coefficient sequence, utilizes the KNN idea to select the n grey correlation coefficients with the smallest absolute difference from the i-th grey correlation coefficient as the comparison objects of the i-th grey correlation coefficient. Here, n is the value of 0.05 times the length of the grey correlation coefficient sequence rounded up. Then, the absolute difference between the i-th grey correlation coefficient and other grey correlation coefficients is calculated. All absolute difference values ​​are sorted in ascending order. The top n absolute difference values ​​are denoted as the target absolute difference values. The other grey correlation coefficients corresponding to the target absolute difference values ​​are used as the comparison objects of the i-th grey correlation coefficient. Here, other grey correlation coefficients refer to the grey correlation coefficients in the grey correlation coefficient sequence other than the i-th grey correlation coefficient.

[0034] After obtaining n comparison objects for the i-th grey relational coefficient, the overall relational difference value of the i-th grey relational coefficient is obtained based on the absolute value of the difference between the i-th grey relational coefficient and each of the comparison objects. The formula for calculating the overall relational difference value of the i-th grey relational coefficient is as follows: In the formula, represents the overall correlation difference value of the i-th grey relational coefficient, and n represents the number of comparison objects for the i-th correlation coefficient. This represents the value of the i-th grey relational coefficient. Let | represent the value of the r-th comparison object of the i-th grey relation coefficient, where | represents the absolute value sign. This represents the mean of the sum of the absolute values ​​of the differences between the i-th grey relational coefficient and its comparison object. The larger this value is, the greater the difference between the grey relational coefficient and its comparison object, and the weaker the ability of the grey relational coefficient to characterize the association between the first soil attribute of sampling point A and sampling point B may be.

[0035] For the i-th grey relational coefficient, by using the KNN idea, several grey relational coefficients with the smallest absolute difference from the i-th grey relational coefficient are selected as the comparison objects of the i-th grey relational coefficient. This can avoid the abnormal disturbance of local grey relational coefficients affecting the stability of the correlation calculation of the first soil attribute between sampling point A and sampling point B over the entire time series when calculating the correlation degree between the first soil attribute between sampling point A and sampling point B.

[0036] Similarly, the overall correlation difference value of each gray correlation coefficient in the gray correlation coefficient sequence is obtained. Then, based on the overall correlation difference value of each gray correlation coefficient in the gray correlation coefficient sequence, the weight of each gray correlation coefficient is obtained. Based on the weight, the gray correlation coefficients in the gray correlation coefficient sequence are weighted and summed to obtain the first soil attribute correlation degree between sampling point A and sampling point B.

[0037] In one embodiment, the expression for calculating the correlation degree of the first soil attribute between sampling point A and sampling point B is: In the formula, This represents the correlation degree of the first soil attribute between sampling point A and sampling point B, where m represents the length of the grey relational coefficient sequence, and 1 represents a constant. This represents the weight of the i-th grey relational coefficient. This represents the weighted average of all grey relational coefficients between sampling point A and sampling point B. This value characterizes the degree of correlation between the first soil attribute and sampling point A and sampling point B. This represents the overall correlation difference value of the i-th grey relational coefficient. This represents the value of the i-th grey relational coefficient.

[0038] The method for calculating the correlation degree of the first soil attribute between two sampling points in this embodiment effectively suppresses data jump interference caused by local accidental environmental disturbances or instantaneous sensor anomalies in time-series data. By calculating the degree of difference between each grey relational coefficient and its comparison object, and assigning a negative correlation weight to the grey relational coefficient according to the degree of difference, the weights of those abnormal instantaneous grey relational coefficients that deviate from the mainstream trend are automatically weakened when calculating the comprehensive correlation degree in the final weighted summation. This ensures that the final output correlation degree can truly and stably reflect the essential physical connection between the two sampling points in the long-term time dimension, improves the robustness of basic data analysis, provides refined data support for subsequent spatial modeling and maize variable planting strategy formulation, and improves the pertinence and scientific nature of planting decisions.

[0039] Step S102: Based on the first soil attribute correlation degree between every two sampling points and the corresponding spatial distance, obtain the correlation degree sequence of each sampling point.

[0040] The above steps yield the first soil property correlation degree between every two sampling points. Traditional methods such as inverse distance weighting and Kriging interpolation typically build models based on the assumption that "closer distances mean stronger correlations" when interpolating spatial data. However, in actual planting environments, factors such as soil structure, water content, and local microenvironment often lead to significant differences in the relationship between soil properties and spatial distance at different locations. Using a uniform model for interpolation can easily overlook the heterogeneity of soil properties in local areas, thus reducing the accuracy and adaptability of soil spatial modeling. Therefore, in this embodiment of the invention, based on the first soil property correlation degree between every two sampling points and the corresponding spatial distance, the correlation degree sequence for each sampling point is obtained. This allows for the capture and quantification of the patterns of soil property changes with spatial distance, providing a basis for subsequently developing customized spatial interpolation strategies.

[0041] In this embodiment, taking the a-th sampling point as an example, the method for obtaining the correlation sequence of the a-th sampling point is as follows: First, obtain the spatial distance between the a-th sampling point and each other sampling point to form a spatial distance set. Similarly, obtain the spatial distance set of each sampling point. Then, take the union of the spatial distance sets corresponding to all sampling points to obtain the overall distance set.

[0042] Then, the correlation degree of the first soil attribute corresponding to each sampling point at each distance in the overall distance set is obtained, and the correlation degrees of the first soil attribute are sorted in ascending order of distance to form a correlation degree sequence of each sampling point. The method for obtaining the correlation degree of the a-th sampling point at each distance in the overall distance set includes: subtracting the overall distance set from the spatial distance set corresponding to the a-th sampling point to obtain the difference set; constructing a two-dimensional plane coordinate system, where the vertical axis is the correlation degree of the first soil attribute between sampling points and the horizontal axis is the spatial distance between sampling points; mapping the spatial distance between the a-th sampling point and other sampling points and the corresponding first soil attribute correlation degree to the two-dimensional plane coordinate system in the form of scattered points, and using a cubic spline interpolation algorithm to interpolate the scattered points to obtain the correlation degree of the first soil attribute at each spatial distance in the difference set, thereby obtaining the correlation degree of the first soil attribute corresponding to each distance in the overall distance set of the a-th sampling point.

[0043] It should be noted that when mapping the spatial distance and the corresponding first soil attribute correlation degree to a two-dimensional plane coordinate system in the form of scatter points, since the value range of the first soil attribute correlation degree is 0 to 1, there is no need to normalize it. However, the horizontal axis is the spatial distance, which has a large value and needs to be normalized. The normalization method can be maximum and minimum value normalization or other normalization methods. That is, the spatial distances that appear later are all normalized spatial distances.

[0044] Step S103: Obtain the first degree of difference in the correlation sequence between sampling points as the clustering distance to cluster the sampling points and obtain at least one category; fit the relationship between the correlation of sampling points in the same category and the spatial distance to obtain the correlation fitting curve for each category.

[0045] In order to obtain the spatial heterogeneity characteristics of the soil in the target planting area and provide a basis for subsequent classification models based on spatial relationships, in this embodiment of the invention, firstly, the sampling points are clustered by obtaining the first degree of difference of the correlation sequence between sampling points as the clustering distance to obtain at least one category. Then, the relationship between the correlation degree and spatial distance of sampling points in the same category is fitted to obtain the correlation degree fitting curve for each category.

[0046] Specifically, the sampling points are clustered by using the first difference in the correlation sequence between sampling points as the clustering distance to obtain at least one category. The specific process is as follows: Typically, the correlation between other sites and the sampling point in terms of the first soil property usually changes with the spatial distance between the other sites and the sampling point. The smaller the first difference between the correlation sequences of two sampling points, the higher the similarity of the spatial heterogeneity of the two sampling points. The more similar the correlation between the first soil property and the spatial distance changes, the more similar the process is. Therefore, sampling points with similar correlation sequences can be classified into one category.

[0047] For any two sampling points, the curvature of the correlation degree of each first soil attribute in the correlation degree sequence of that sampling point is calculated to obtain a curvature sequence. The curvature sequence is then subjected to first-order differencing, and the absolute value of the differencing results is taken to form a difference absolute value sequence. The mean of each difference absolute value in the difference absolute value sequence is calculated to characterize the stability of the correlation degree sequence of that sampling point as a function of spatial distance. This mean is then used as a stability index of the correlation degree sequence of that sampling point, effectively reflecting the degree of fluctuation and consistency in the spatial correlation change patterns of different sampling points. Similarly, the stability index of the correlation degree sequence of each of the two sampling points is obtained, and the absolute value of the difference between these two stability indices is calculated. This absolute value of the difference is then normalized to obtain the deviation.

[0048] It should be noted that the stability index can extremely sensitively extract the nonlinear, minute fluctuations in soil properties during spatial evolution. By calculating the curvature of the correlation sequence to characterize the bending shape of the spatial decay curve, and using the absolute value of the first-order difference to capture the abrupt change rate of this shape, a mean smoothing process is finally applied, effectively filtering out random jumps at individual spatial nodes. This series of data processing steps enables the final output stability index to macroscopically and accurately characterize the global stability of soil properties in the region as spatial distance changes.

[0049] Furthermore, based on this, following the calculation method of the first soil attribute correlation degree between the two sampling points, the correlation degree between the correlation degree sequences of any two sampling points is calculated. The correlation degree is then normalized to obtain the comprehensive correlation degree. That is, firstly, the grey correlation coefficient sequence between the correlation degree sequences of any two sampling points is calculated, then the overall correlation difference value of each grey correlation coefficient in the grey correlation coefficient sequence is calculated, and finally, the weight of the grey correlation coefficient is calculated based on the overall correlation difference value of the grey correlation coefficient, and the grey correlation coefficient is weighted and summed to obtain the correlation degree between the correlation degree sequences of any two sampling points.

[0050] Finally, based on the deviation and the comprehensive correlation degree, a first degree of difference in the correlation degree sequence between any two sampling points is obtained. This first degree of difference is positively correlated with the deviation and negatively correlated with the comprehensive correlation degree. Specifically, the calculation method for obtaining the first degree of difference in the correlation degree sequence between any two sampling points based on the deviation and the comprehensive correlation degree is as follows: The difference between a constant 1 and the comprehensive correlation degree is obtained; the mean of this difference and the deviation is calculated to obtain the first degree of difference in the correlation degree sequence between the two sampling points.

[0051] In one embodiment, for the a-th sampling point and the b-th sampling point, the calculation expression corresponding to the first degree of difference in the correlation sequence between these two sampling points is: in, This represents the first degree of difference in the correlation sequence between the a-th sampling point and the b-th sampling point. This represents the correlation between the correlation sequence of the a-th sampling point and the correlation sequence of the b-th sampling point. The stability index of the correlation sequence at the a-th sampling point is represented. The stability index representing the correlation sequence of the b-th sampling point. Represents a linear normalization function. 1 represents the absolute value symbol, and 1 represents a constant.

[0052] It should be noted that, The smaller, that is The larger the value, the weaker the correlation and the more obvious the difference between the two sampling points; This represents the absolute value of the difference between the stability index of sampling point a and sampling point b. The larger this value is, the more significant the difference in the spatial correlation change pattern between the two sampling points.

[0053] It is worth noting that the first degree of difference between the correlation sequences of two sampling points can also be obtained by other methods, such as calculating the Euclidean distance, Manhattan distance or Chebyshev distance between the correlation sequences of two sampling points, and using the corresponding calculation results as the first degree of difference between the correlation sequences of two sampling points.

[0054] Thus, by obtaining the first degree of difference, the first degree of difference in the correlation sequence between every two sampling points can be obtained. Then, the first degree of difference is used as the clustering distance to cluster all sampling points and obtain at least one category. It is worth noting that the clustering method can be K-means clustering algorithm, DBSCAN clustering algorithm, HDBSCAN clustering algorithm, fuzzy C-means clustering algorithm or other suitable clustering algorithm. Preferably, the clustering method in this embodiment adopts HDBSCAN clustering algorithm.

[0055] When measuring the degree of difference between sampling points, not only is the similarity of the time series data of the two sampling points examined (the correlation degree of the correlation sequence), but the consistency of their spatial decay patterns radiating outwards is also deeply integrated (the deviation of the stability index). This dual-dimensional distance measurement mechanism, which takes into account both "temporal fluctuation similarity" and "spatial divergence consistency," enables subsequent clustering operations to go beyond surface distance and instead perform precise regional division based on deep geological spatial evolution mechanisms, greatly improving the scientific rigor and physical accuracy of soil heterogeneity regional division (clustering).

[0056] Furthermore, there are multiple sampling points in the same category, and each sampling point has a corresponding correlation degree sequence. The multiple elements in the correlation degree sequence are the first soil attribute correlation degree corresponding to different spatial distances. Since the functional relationship between the first soil attribute correlation degree and spatial distance of sampling points in the same category is relatively close (the first difference between the correlation degree sequences of sampling points is small), the functional relationship between the first soil attribute correlation degree and spatial distance of all sampling points in a category can be uniformly fitted to obtain the correlation degree fitting curve for each category. Taking the j-th category as an example, a two-dimensional plane coordinate system is established based on the correlation sequence of each sampling point in the j-th category. The vertical axis represents the correlation of the first soil attribute, and the horizontal axis represents the spatial distance. The correlation of the first soil attribute and the corresponding spatial distance in the correlation sequence of each sampling point are mapped to the two-dimensional plane coordinate system in scatter plot form. The vertical axis of the corresponding scatter plot is the value of the correlation of the first soil attribute, and the horizontal axis is the spatial distance corresponding to the correlation of the first soil attribute. Multiple fitting operations are performed on the scatter plots in the two-dimensional plane coordinate system to obtain the fitting curve after each fitting. When performing multiple fitting operations on the scatter plots in the two-dimensional plane coordinate system, the fitting method can be polynomial fitting, least squares fitting, linear regression, or sequence model fitting. Preferably, in this embodiment, a GRU model is used to fit the functional relationship between the correlation of the first soil attribute and the spatial distance of all sampling points in the j-th category.

[0057] During the fitting process, considering the different representativeness of different sampling points in the category and the different degrees of difference between their association sequence and other sampling points, a weighted loss function is further introduced to improve the fitting accuracy. Therefore, after each fitting, the corresponding fitting loss is calculated based on the fitting curve. The fitting loss is positively correlated with the deviation between the fitted value and the true value corresponding to each association degree in the association degree sequence. Then, the corresponding fitting effect evaluation index is calculated based on the fitting loss. The fitting effect evaluation index is negatively correlated with the fitting loss. The fitting curve with the largest fitting effect evaluation index is selected as the association degree fitting curve for the j-th category.

[0058] The weighted loss function is calculated as follows: In the formula, L represents the fitting loss, and z represents the number of sampling points in the j-th category. This represents the length of the correlation sequence at the a-th sampling point. This represents the correlation degree of the h-th first soil attribute in the correlation degree sequence of the a-th sampling point in the j-th category. This represents the correlation fit value (calculated from the fitted curve) corresponding to the h-th first soil attribute correlation in the correlation sequence of the a-th sampling point in the j-th category. This represents the mean of the first degree of difference between the association sequence of the a-th sampling point in the j-th category and the association sequences of other sampling points in the j-th category. Represents the absolute value symbol.

[0059] It should be noted that, The larger the value, the greater the difference in the correlation sequence between the a-th sampling point and other sampling points, and the smaller the weight of the correlation sequence of the a-th sampling point in the fitting evaluation. This represents the weighted difference between the correlation sequence of each sampling point in the j-th category and the fitted correlation sequence. The smaller the weighted difference, the smaller the fitting loss.

[0060] The method for calculating the fitting loss endows the curve fitting process with strong resistance to noise and heterogeneity interference. When calculating the single-step fitting loss, the mean of the first degree of difference between the sampled point and other sampled points of the same category is introduced as a penalty weight. This automatically reduces the weight of "marginal" sample points that are barely classified into the category and have high heterogeneity during model training, effectively preventing individual free samples from skewing the overall fitting curve of the category, and ensuring that the fitting result highly represents the mainstream geological evolution law of the category.

[0061] The method for obtaining the fitting effect evaluation index based on the fitting loss calculation is as follows: Taking the k-th fitting as an example, after the k-th fitting, the correlation fitting value corresponding to each spatial distance in the overall distance set is obtained according to the fitting curve, and the correlation fitting values ​​are sorted in order of spatial distance from smallest to largest to obtain the correlation fitting value sequence; considering that the change of soil spatial attributes with spatial distance usually satisfies the monotonicity law of "the correlation weakens as the distance increases", a monotonicity evaluation index is introduced to constrain and discriminate the result of the k-th fitting, that is, to obtain the first-order difference sequence of the correlation fitting value sequence. If any difference value in the first-order difference sequence is less than or equal to 0, the monotonicity discriminant value of the difference value is 1; if any difference value in the first-order difference sequence is greater than 0, the monotonicity discriminant value of the difference value is 0; the average of the monotonicity discriminant values ​​of each difference value in the first-order difference sequence is taken to obtain the monotonicity evaluation index of the k-th fitting.

[0062] In one embodiment, the expression for calculating the monotonicity evaluation index of the k-th fit is: in, The monotonicity evaluation index represents the fitted result. This represents the (h-1)th correlation fit value in the correlation fit value sequence. This represents the h-th correlation fit value in the correlation fit value sequence. Let M represent the monotonic discriminant value of the h-th difference value in the first-order difference sequence, and M represent the length of the first-order difference sequence.

[0063] Furthermore, based on the monotonicity evaluation index and the fitting loss of the k-th fitting, the fitting effect evaluation index of the k-th fitting is calculated. The fitting effect evaluation index is negatively correlated with the fitting loss and positively correlated with the monotonicity evaluation index. The expression for calculating the fitting effect evaluation index is as follows: In the formula, This represents the evaluation index for the fit effect, where L represents the fit loss. The smaller the value, the better the fit. The larger the value, the higher the fitting accuracy; H represents the monotonicity evaluation index after fitting, and the larger the value, the more the fitting result conforms to the laws of space physics. Therefore, the product of the two is used to represent the final fitting effect evaluation value.

[0064] It should be noted that the calculation method of the fitting effect evaluation index not only improves the fitting accuracy and stability, but also introduces monotonicity constraints to maintain consistency with soil physical laws, making spatial interpolation and soil property modeling more reliable, and providing a solid spatial information foundation for maize planting strategies.

[0065] By performing curvature analysis, cluster distance calculation, and HDBSCAN clustering on the correlation sequences of sampling points in different categories, and then combining this with GRU model fitting, a spatial distance-correlation fitting curve is constructed. This overcomes the technical bottlenecks of traditional low-order polynomials being unable to fit complex curves and high-order polynomials being prone to overfitting oscillations. By employing the GRU model, which excels at handling sequence dependencies, the nonlinear spatial decay law within each category is deeply explored. Multiple iterations are then performed based on the fitting loss, which is positively correlated with the deviation from the true value, to ensure that the selected optimal fitting result (the fitting curve corresponding to the maximum fitting effect evaluation index, i.e., the correlation fitting curve) can accurately reproduce the highly complex real spatial distance decay model within different heterogeneous soil regions.

[0066] Step S104: Based on the spatial distance between the point to be estimated and the sampling point, and the correlation fitting curve of the category to which the sampling point belongs, obtain the interpolation weight of each sampling point for the point to be estimated, and the sum of the interpolation weights of each sampling point for the point to be estimated is 1; based on the interpolation weights, perform weighted summation on the time series data of the first soil attribute of each sampling point to obtain the time series data of the first soil attribute of the point to be estimated.

[0067] Traditional methods such as inverse distance weighting and Kriging interpolation typically build models based on the assumption that "closer distances mean stronger correlations" when interpolating spatial data. However, in actual planting environments, factors such as soil structure, water content, and local microenvironments often lead to significant differences in the relationship between soil properties and spatial distance at different locations. Using a uniform model for interpolation can easily overlook the heterogeneity of soil properties in local areas, thus reducing the accuracy and adaptability of soil spatial modeling. Therefore, in this embodiment of the invention, the model focuses on the points to be estimated within the target planting area. Points to be estimated This refers to the location points within the target planting area where no time-series data for the first soil attribute has been collected and where it is necessary to estimate the time-series data for the first soil attribute. Based on the spatial distance between the point to be estimated and the sampling points, and the correlation fitting curve of the category to which the sampling points belong, the interpolation weights of each sampling point for the point to be estimated are obtained, as follows: For the a-th sampling point, obtain the correlation curve of the category to which the a-th sampling point belongs, and use this correlation curve to obtain the relationship between the a-th sampling point and the point to be predicted. The correlation fit value corresponding to the spatial distance between them is denoted as To obtain the reliability of the fitting result for the a-th sampling point, the mean of the sum of the absolute values ​​of the differences between the actual correlation degree sequence and the fitted correlation degree sequence of the a-th sampling point is calculated to characterize the fitting error level. This fitting error level can reflect the degree of fit of the a-th sampling point under the current fitting model. The larger the fitting error level value, the more obvious the fitting deviation, and the lower its reference reliability in the subsequent interpolation process should be.

[0068] Taking into account both spatial correlation strength and fitting error, the interpolation weight of the a-th sampling point can be obtained. The larger the correlation fitting value between the point to be estimated and the a-th sampling point, the greater the similarity of the time series data of the first soil attribute between the a-th sampling point and the point to be estimated. Therefore, when using spatial interpolation to calculate the time series data of the first soil attribute of the point to be estimated, the weight assigned to the a-th sampling point should be larger. That is, when calculating the interpolation weight of the a-th sampling point to the point to be estimated, the interpolation weight is positively correlated with the correlation fitting value between the a-th sampling point and the point to be estimated.

[0069] In one embodiment, the formula for calculating the interpolation weight of the a-th sampling point for the point to be estimated x is: in, This represents the interpolation weight of the a-th sampling point for the point x to be estimated. This represents the spatial distance between the a-th sampling point and the point x to be estimated. This indicates that the correlation curve fitted based on the category to which the a-th sampling point belongs is at a spatial distance of The corresponding correlation fitting value at time, denoted by , represents the mean of the sum of the absolute values ​​of the differences between the correlation sequence of the a-th sampling point and its fitted correlation sequence, and v represents the number of sampling points.

[0070] Since the larger the deviation between the actual correlation sequence and the fitted correlation sequence of the a-th sampling point, the lower the accuracy of the corresponding correlation fitting curve, in this embodiment, when calculating the interpolation weight of the a-th sampling point for the point to be estimated, not only is the magnitude of the correlation fitting value between the a-th sampling point and the point to be estimated taken into account, but also the deviation between the actual correlation sequence of the a-th sampling point and its fitted correlation sequence is further taken into account, and the interpolation weight is negatively correlated with the deviation between the actual correlation sequence and its fitted correlation sequence, thereby further improving the accuracy of the calculation result of the interpolation weight of the sampling point for the point to be estimated.

[0071] Similarly, the interpolation weight of each sampling point to the estimated point x is obtained. Based on this, the time series data of the first soil attribute of all sampling points are weighted and fused according to the corresponding weights to obtain the time series data of the first soil attribute at the estimated point x, thereby realizing a refined estimate of the soil state at the estimated point x.

[0072] Step S105: Construct a spatial distribution model of soil attributes for the target planting area based on the first soil attribute time series data of all sampling points and all points to be estimated, and implement adaptive variable sowing for maize based on the spatial distribution model of soil attributes.

[0073] Following the method described above for obtaining the time series data of the first soil attribute of the point to be estimated x, the time series data of the first soil attribute of each point to be estimated in the target planting area are obtained. Based on the time series data of the first soil attribute of all sampling points and all points to be estimated, a continuous spatial distribution model of soil attributes is constructed. When the first soil attribute is soil moisture, the spatial distribution model of soil attributes is a spatial distribution model of soil moisture. When the first soil attribute includes nutrient content, temperature, organic matter content, etc., the nutrient content, temperature, and organic matter content in the target planting area can be consistently modeled according to the construction method of the spatial distribution model of soil moisture, thereby forming a multi-dimensional spatial distribution model of soil attributes. Through unified modeling and fusion analysis of multi-source data, the spatial heterogeneity characteristics of actual cultivated land can be reflected more comprehensively and accurately.

[0074] Furthermore, planting decision analysis can be conducted based on the multi-dimensional spatial distribution model of soil properties constructed above. For example, existing precision agriculture methods can be combined to construct crop suitability evaluation functions or threshold discrimination rules based on indicators such as soil moisture and nutrient content at different locations, enabling zoned management of plots. Alternatively, based on empirical models, machine learning models, and other methods, a mapping relationship between soil properties and sowing density and fertilizer application can be established, thereby obtaining differentiated planting parameter configuration schemes for different regions. In practical applications, corn sowing density, seed quantity, row spacing, plant spacing, and corresponding water and fertilizer management strategies can be adaptively adjusted according to the soil property characteristics at each location, achieving variable sowing and precise input control, thereby improving resource utilization efficiency and crop growth consistency. It should be noted that the above planting decision-making process can be implemented using existing mature methods or custom optimization strategies, and the specific implementation method is not limited.

[0075] Ultimately, by combining the soil spatial model with the planting decision-making mechanism, a closed-loop control process of "data perception - spatial modeling - decision optimization" can be achieved, thereby effectively improving the yield stability and overall economic benefits of maize planting.

[0076] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A maize adaptive variable sowing method based on soil spatial variability analysis, characterized in that, include: Obtain the first soil attribute time series data of each sampling point in the target planting area, and obtain the first soil attribute correlation degree between every two sampling points based on the first soil attribute time series data; Based on the correlation degree of the first soil attribute between every two sampling points and the corresponding spatial distance, the correlation degree sequence of each sampling point is obtained. The correlation degree sequence of any sampling point is composed of the correlation degree between any sampling point and any sampling point at a location with a different spatial distance from any sampling point. The first degree of difference in the correlation sequence between sampling points is used as the clustering distance to cluster the sampling points and obtain at least one category. The relationship between the correlation degree and spatial distance of sampling points in the same category is fitted to obtain the correlation degree fitting curve for each category; Based on the spatial distance between the point to be estimated and the sampling point, and the correlation fitting curve of the category to which the sampling point belongs, the interpolation weight of each sampling point to be estimated is obtained, and the sum of the interpolation weights of each sampling point to be estimated is 1; the first soil attribute time series data of each sampling point is weighted and summed according to the interpolation weights to obtain the first soil attribute time series data of the point to be estimated. Based on the time series data of the first soil properties of all sampling points and all points to be estimated, a spatial distribution model of soil properties in the target planting area is constructed, and adaptive variable sowing of maize is implemented based on the spatial distribution model of soil properties.

2. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 1, characterized in that, The step of obtaining the correlation degree of the first soil attribute between every two sampling points based on the time series data of the first soil attribute includes: Obtain the grey relational coefficient sequence between the first soil attribute time series data of any two sampling points. For the i-th grey relational coefficient in the grey relational coefficient sequence, calculate the absolute value of the difference between the i-th grey relational coefficient and other grey relational coefficients. Sort all the absolute values ​​of difference in ascending order. The first n absolute values ​​of difference in the sorted absolute values ​​of difference are denoted as the target absolute value of difference. The other grey relational coefficients corresponding to the target absolute value of difference are used as the comparison objects of the i-th grey relational coefficient. Other grey relational coefficients refer to the grey relational coefficients in the grey relational coefficient sequence other than the i-th grey relational coefficient. The overall correlation difference value of the i-th grey correlation coefficient is obtained based on the absolute value of the difference between the i-th grey correlation coefficient and each of the comparison objects. Based on the overall correlation difference value of each grey correlation coefficient in the grey correlation coefficient sequence, the weight of each grey correlation coefficient is obtained. Based on the weight, the grey correlation coefficients in the grey correlation coefficient sequence are weighted and summed to obtain the first soil attribute correlation degree between any two sampling points.

3. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 1, characterized in that, The correlation sequence of each sampling point is obtained based on the first soil attribute correlation degree and the corresponding spatial distance between every two sampling points, including: Obtain the spatial distance set corresponding to each sampling point. The spatial distance set corresponding to the a-th sampling point is the set composed of the spatial distances between the a-th sampling point and all other sampling points. Take the union of the spatial distance sets corresponding to all sampling points to obtain the overall distance set. The method involves obtaining the first soil attribute correlation degree corresponding to each sampling point at each distance in the overall distance set, and sorting the first soil attribute correlation degrees in ascending order of distance to form a correlation degree sequence for each sampling point; wherein, the method for obtaining the first soil attribute correlation degree corresponding to the a-th sampling point at each distance in the overall distance set includes: The difference set is obtained by subtracting the overall distance set from the spatial distance set corresponding to the a-th sampling point; a two-dimensional plane coordinate system is constructed, where the vertical axis represents the correlation degree of the first soil attribute between sampling points and the horizontal axis represents the spatial distance between sampling points; the spatial distance between the a-th sampling point and other sampling points and the corresponding correlation degree of the first soil attribute are mapped to the two-dimensional plane coordinate system in the form of scattered points, and the scattered points are interpolated using a cubic spline interpolation algorithm to obtain the correlation degree of the first soil attribute at each spatial distance in the difference set, thereby obtaining the correlation degree corresponding to the a-th sampling point at each distance in the overall distance set.

4. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 1, characterized in that, The method for obtaining the first degree of difference in the correlation sequence between sampling points includes: For any two sampling points, the stability index of the correlation sequence of the two sampling points is obtained respectively, and the absolute value of the difference between the two stability indices is obtained. The absolute value of the difference is normalized to obtain the deviation. The correlation between the correlation sequences of the two sampling points is calculated, and the correlation is normalized to obtain the comprehensive correlation. Based on the deviation and the comprehensive correlation, the first degree of difference of the correlation sequence between the two sampling points is obtained.

5. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 4, characterized in that, Obtain the stability index of the correlation sequence between any two sampling points, including: For any one of the two sampling points, calculate the curvature of each data point in the correlation sequence of the sampling point to obtain a curvature sequence; perform first-order difference on the curvature sequence to obtain a difference sequence; and calculate the mean of the absolute values ​​of each difference value in the difference sequence as a stability index of the correlation sequence of the sampling point.

6. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 4, characterized in that, Based on the deviation and the overall correlation, the first degree of difference in the correlation sequence between any two sampling points is obtained, including: Obtain the difference between the constant 1 and the comprehensive correlation degree, calculate the mean between the difference and the deviation, and obtain the first degree of difference of the correlation degree sequence between any two sampling points.

7. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 3, characterized in that, The process of fitting the relationship between the correlation degree and spatial distance of sampling points within the same category to obtain a correlation degree fitting curve for each category includes: For the j-th category, based on the correlation sequence of each sampling point, the relationship between the correlation and spatial distance of the sampling points in the j-th category is fitted multiple times to obtain the fitting curve after each fitting. After each fitting, the corresponding fitting loss is calculated based on the fitting curve. The fitting loss is positively correlated with the deviation between the fitted value and the true value corresponding to each correlation in the correlation sequence. The corresponding fitting effect evaluation index is calculated based on the fitting loss. The fitting effect evaluation index is negatively correlated with the fitting loss. The fitting curve with the largest fitting effect evaluation index is selected as the correlation fitting curve for the j-th category.

8. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 7, characterized in that, The formula for calculating the fitting loss is: ； In the formula, L represents the fitting loss, and z represents the number of sampling points in the j-th category. This represents the length of the correlation sequence at the a-th sampling point. This represents the correlation degree of the h-th first soil attribute in the correlation degree sequence of the a-th sampling point in the j-th category. This represents the correlation fit value (calculated from the fitted curve) corresponding to the h-th first soil attribute correlation in the correlation sequence of the a-th sampling point in the j-th category. This represents the mean of the first degree of difference between the association sequence of the a-th sampling point in the j-th category and the association sequences of other sampling points in the j-th category. Represents the absolute value symbol.

9. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 7, characterized in that, The fitting performance evaluation index calculated based on the fitting loss includes: After the k-th fitting, the correlation fitting value corresponding to each spatial distance in the overall distance set is obtained based on the fitting curve, and the correlation fitting values ​​are sorted in order of spatial distance from smallest to largest to obtain the correlation fitting value sequence. Obtain the first-order difference sequence of the correlation fitting value sequence. If any difference value in the first-order difference sequence is less than or equal to 0, the monotonic discriminant value of the difference value is 1. If any difference value in the first-order difference sequence is greater than 0, the monotonic discriminant value of the difference value is 0. Take the mean of the monotonic discriminant values ​​of each difference value in the first-order difference sequence to obtain the monotonicity evaluation index of the k-th fitting.

10. The maize adaptive variable sowing method based on soil spatial variability analysis according to claim 1, characterized in that, The step of obtaining the interpolation weights for each sampling point based on the spatial distance between the point to be estimated and the sampling points, and the correlation fitting curve of the category to which the sampling points belong, includes: The formula for calculating the interpolation weight of the a-th sampling point for the point to be estimated x is: ； in, This represents the interpolation weight of the a-th sampling point for the point x to be estimated. This represents the spatial distance between the a-th sampling point and the point x to be estimated. This indicates that the correlation curve fitted based on the category to which the a-th sampling point belongs is at a spatial distance of The corresponding correlation fitting value at time, denoted by , represents the mean of the sum of the absolute values ​​of the differences between the correlation sequence of the a-th sampling point and its fitted correlation sequence, and v represents the number of sampling points.

Images