A method for precise regulation of nitrogen in the corn root layer under drip irrigation conditions

Abstract

The present invention relates to the field of drip irrigation regulation, and in particular to a method for precise regulation of nitrogen in the root layer of corn under drip irrigation conditions. The method comprises: obtaining soil grid points by standardizing multi-source remote sensing soil data of a plot, and evaluating the adaptive relative grades of crop growth at the grid points; obtaining the soil potential grades of the grid points by factor modeling and adaptive grading of soil erodibility parameters; obtaining the adaptive spatial context weight factors of the grid points by fusing and evaluating global attribute fluctuations and spatial heterogeneity; obtaining the gradient consistency distance measurement between the grid points by fusing and evaluating the adaptive spatial context weight factors of the grid points with the soil potential grades of the grid points; and obtaining an optimized management zoning map and a differentiated drip irrigation strategy by performing fuzzy clustering analysis based on the gradient consistency distance measurement between the grid points, thereby reducing the control complexity of the drip irrigation network and improving the drip irrigation precision and fertilization efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of drip irrigation regulation, and in particular to a method for accurately regulating nitrogen in the root layer of corn under drip irrigation conditions. Background Art

[0002] In concentrated corn cultivation, drip irrigation technology is often used for integrated water and fertilizer management to improve water and fertilizer utilization efficiency. To achieve more refined management, existing technologies obtain a gridded soil type map and crop growth remote sensing map of the target plot. The soil type map contains soil particle composition information at each grid point, while the crop growth remote sensing map provides the Normalized Difference Vegetation Index (NDVI) value for each grid point. The fuzzy C-means clustering (FCM) algorithm then uses this soil and growth data as multidimensional feature inputs to perform cluster analysis on all grid points within the plot. Based on the clustering results, the plot is divided into several management zones. Finally, based on the average level of crop growth within each management zone, a unified drip irrigation and fertigation plan is developed and implemented for that zone. The FCM clustering algorithm uses iterative optimization to find a zone scheme that minimizes the sum of the squared Euclidean distances from all grid points to the center of their zone.

[0003] The core distance metric of the FCM clustering algorithm, when processing raw crop growth indices derived from remote sensing imagery, is affected by variations in the overall numerical range of crop growth data caused by macroeconomic factors such as seasonal light and temperature. Specifically, in years with generally good crop growth, numerical variations in the growth index are magnified, causing the algorithm to overemphasize growth characteristics when clustering, thereby reducing the role of other static characteristics, such as soil, in the clustering process. Conversely, in years with generally poor growth, numerical variations in the growth index are compressed, causing the algorithm to overemphasize soil characteristics when clustering, while ignoring potential key differences in growth. This imbalance in the influence of internal features within the algorithm, caused by changes in the absolute scale of the data, alters the classification basis of the final output. Furthermore, the algorithm's distance metric treats each grid point as an independent attribute vector when calculating the attribute, without considering the spatial relationship between that grid point and its geographically adjacent grid points. This approach contradicts the fundamental laws governing the distribution of field attributes. In real fields, the distribution of soil and crop properties is not isolated but generally follows the law of spatial autocorrelation, meaning that geographically adjacent areas tend to have similar properties. For example, the soil texture and water and fertilizer holding capacity of two adjacent grid points, and the resulting response patterns to drip irrigation and fertilizer application, should be highly similar or have a smooth transition.

[0004] The existing technology ignores the spatial relationship and classifies grid point A and distant grid point C into the same partition simply because their attribute values ​​are exactly the same; however, grid point B, which is closely adjacent to A and has only a small difference in attribute value, is divided into another completely different partition. This results in a result that is contrary to the actual physical process: two areas that are closely connected in space and should be managed similarly are forcibly divided. The partition map generated by this partitioning scheme does not conform to the natural gradual change law of field attributes. Therefore, providing a method for accurately regulating nitrogen in the root layer of corn under drip irrigation conditions that can overcome the above problems is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention aims to propose a method for precise regulation of nitrogen in the maize root layer under drip irrigation conditions to solve the problem that the zoning results are disturbed by growth scale fluctuations and lack of spatial continuity.

[0006] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0007] A method for accurately regulating nitrogen in the root zone of corn under drip irrigation conditions comprises the following steps:

[0008] Step S1: obtaining grid points for grid analysis by standardizing multi-source remote sensing soil data of the plot, and evaluating the relative level of crop growth adaptation at the grid points;

[0009] Step S2: Obtaining the soil potential level of the grid point by factor modeling and adaptive classification of soil erodibility parameters;

[0010] Step S3: Obtain the adaptive spatial context weight factor of the grid point by fusing and evaluating the global attribute fluctuation and spatial heterogeneity;

[0011] Step S4: Obtaining the gradient consistency distance metric between grid points by fusing and evaluating the adaptive spatial context weight factor of the grid points with the soil potential level of the grid points;

[0012] Step S5: Perform fuzzy cluster analysis using the gradient consistency distance metric between grid points to obtain the optimized management zoning map and differentiated drip irrigation strategy.

[0013] Furthermore, the method of obtaining grid points for grid analysis by standardizing multi-source remote sensing soil data of a plot and evaluating the relative level of crop growth adaptability of the grid points specifically includes:

[0014] Using UAV remote sensing technology to obtain multispectral remote sensing images covering the entire target plot, and extracting the normalized vegetation index as crop growth data;

[0015] Obtaining a gridded soil map of the target plot using a multispectral remote sensing image of the target plot, wherein the gridded soil map includes soil particle composition data and mass percentage data of soil organic carbon at each grid point in the target plot; the soil particle composition data at the grid point includes mass percentage data of sand, silt, and clay at the grid point; geo-registering the multispectral remote sensing image of the target plot with the gridded soil map, and unifying the spatial resolution of the data layers;

[0016] The number of growth grade divisions is set, and the percentile of the crop growth data is set according to the number of growth grade divisions; the crop growth data of the grid point is compared with the percentile of the crop growth data to obtain the adaptive relative grade of the crop growth of the grid point.

[0017] Furthermore, the soil potential level of the grid point is obtained by factor modeling and adaptive grading of soil erodibility parameters, including:

[0018] The soil erodibility factor is calculated using the soil particle composition data and the mass percentage data of soil organic carbon at each grid point in the target plot to obtain the soil erodibility K value of the grid point;

[0019] The soil potential index of the grid point is obtained by performing reciprocal conversion on the soil erodibility K value of the grid point;

[0020] The soil potential index of the grid points is graded using an adaptive grading method to obtain the soil potential level of the grid points.

[0021] Furthermore, the adaptive spatial context weight factor of the grid point is obtained by fusing and evaluating the global attribute fluctuation and spatial heterogeneity, including:

[0022] The two-dimensional attribute vector of the grid point is established through the soil potential level of the grid point and the adaptive relative level of the crop growth of the grid point;

[0023] By performing gradient analysis on the spatial neighboring grid points of the grid point, the local attribute gradient vector of the grid point is obtained;

[0024] The spatial heterogeneity intensity of the plot is analyzed by the local attribute gradient vector of the grid point to obtain the global gradient divergence;

[0025] The global heterogeneity of the plot where the grid point is located is analyzed through the global gradient divergence, the global soil potential level and the global crop growth adaptive relative level of the plot where the grid point is located, and the adaptive spatial context weight factor of the grid point is obtained.

[0026] Furthermore, the step of performing gradient analysis on the spatially neighboring grid points of the grid point to obtain the local attribute gradient vector of the grid point includes:

[0027] The neighborhood range of the grid point is set, and the result of subtracting the adaptive relative grade of crop growth of the grid point from the average adaptive relative grade of crop growth of the grid points in the neighborhood range of the grid point is used as the first gradient evaluation of the crop growth grade of the grid point; the result of subtracting the soil potential grade of the grid point from the average soil potential grade of the grid points in the neighborhood range of the grid point is used as the first gradient evaluation of the soil potential grade of the grid point; the two-dimensional vector formed by the first gradient evaluation of crop growth grade of the grid point and the first gradient evaluation of soil potential grade is used as the local attribute gradient vector of the grid point.

[0028] Furthermore, the spatial heterogeneity intensity analysis of the plot is performed through the local attribute gradient vector of the grid point to obtain the global gradient divergence, including:

[0029] The result of adding the square of the first gradient evaluation of the crop growth grade of the grid point and the square of the first gradient evaluation of the soil potential grade of the grid point is used as the modulus of the local attribute gradient vector of the grid point;

[0030] The square root of the mean of the modulus of the local attribute gradient vectors of all grid points in the target plot where the grid point is located is taken as the global gradient divergence.

[0031] Furthermore, the global heterogeneity analysis of the plot where the grid point is located is performed by using the global gradient divergence of the plot where the grid point is located, the global soil potential level of the plot where the grid point is located, and the global crop growth adaptive relative level of the plot where the grid point is located to obtain the adaptive spatial context weight factor of the grid point, including:

[0032] The variance of the soil potential grade of the target plot is obtained by the soil potential grade of all grid points in the target plot; the variance of the crop growth adaptive relative grade of the target plot is obtained by the crop growth adaptive relative grade of all grid points in the target plot;

[0033] The result of adding the variance of the soil potential grade of the target plot and the variance of the crop growth adaptive relative grade of the target plot is used as the attribute dispersion evaluation of the target plot;

[0034] The result of adding the constant 1 to the global gradient divergence is used as the denominator, the attribute discreteness evaluation of the target plot is used as the numerator, and the calculation result of the corresponding fraction is used as the adaptive spatial context weight factor of the grid point in the target plot.

[0035] Furthermore, the gradient consistency distance metric between grid points is obtained by fusing and evaluating the adaptive spatial context weight factor of the grid point with the soil potential level of the grid point, including:

[0036] The result of subtracting the adaptive relative levels of crop growth between grid points and performing square calculation is used as the first crop growth level difference between grid points; the result of subtracting the soil potential level between grid points and performing square calculation is used as the first soil potential level difference between grid points;

[0037] The result of multiplying the adaptive spatial context weight factor of the grid point in the target plot by the square of the modulus of the local attribute gradient vector difference between the grid points is used as a penalty evaluation of the spatial context difference between the grid points;

[0038] The calculation result of adding the first crop growth grade difference between the grid points, the first soil potential grade difference between the grid points and the spatial context difference penalty evaluation between the grid points is used as the gradient consistency distance measure between the grid points.

[0039] Furthermore, the fuzzy cluster analysis performed by using the gradient consistency distance measurement between grid points to obtain the optimized management partition map and differentiated drip irrigation strategy includes:

[0040] The Davis-Boulting index is used to optimize the number of management partitions and obtain the number of partitions for the clustering process;

[0041] The gradient consistency distance metric between the grid points is used as the distance metric between the grid points in the clustering process, and fuzzy C-means clustering is performed according to the number of partitions to obtain a management partition map;

[0042] Differentiated drip irrigation strategies are formulated by managing the average soil potential level and the average crop growth adaptive relative level of each zone in the zoning map.

[0043] Furthermore, the differentiated drip irrigation strategy is formulated by managing the average soil potential level and the average crop growth adaptive relative level of each partition in the partition map, specifically including:

[0044] For each zone in the management zone map, calculate the average soil fertility potential level and the average crop growth adaptability relative level of all grid points in the zone;

[0045] Set the level assessment threshold to classify the soil potential level and crop growth adaptability relative level into good and poor;

[0046] For sub-regions rated as good in both average soil potential and average crop growth adaptability, the drip irrigation fertigation strategy is based on maintenance fertilization, steadily supplying nitrogen according to the standard fertilizer requirement of the crop at its current growth stage.

[0047] For sub-areas where both the average soil fertility potential grade and the average crop growth adaptability relative grade were rated as poor, the drip irrigation fertigation strategy was based on economical fertilization, using a nitrogen supply level lower than the standard fertilizer requirement.

[0048] For sub-regions with an average soil potential grade of good and an average crop growth adaptability grade of poor, the drip irrigation fertigation strategy is mainly based on yield-increasing fertilization, using a nitrogen supply level higher than the standard fertilizer requirement.

[0049] For zones with a poor average soil potential grade and an excellent average crop growth adaptive relative grade, the drip fertigation strategy adopts a small amount and multiple times mode. On the basis of determining the total amount of fertilizer, the single fertilization task is broken down into multiple executions, significantly reducing the nitrogen concentration and irrigation time of a single drip fertigation, while increasing the frequency of drip irrigation.

[0050] Compared with the prior art, the present invention has the following advantages:

[0051] This method achieves in-depth optimization of drip irrigation fertigation management zoning by introducing dual-dimensional modeling of crop growth grade and soil fertility potential grade, combined with local attribute gradient and spatial heterogeneity modeling mechanisms. Compared with existing technologies, this method significantly enhances cluster analysis's dual perception of crop growth status and soil basic capacity, and effectively avoids the problem of clustering focus shift caused by data scale changes between different growing years in traditional clustering methods. In scenarios where environmental conditions such as light and moisture vary dramatically between years, this method can maintain stable feature expression and zoning logic, improving the continuity and repeatability of fertilization plans. Furthermore, the present invention introduces a gradient consistency distance metric method, incorporating spatial contextual features between grid points into the zoning decision process. This allows clustering to not only consider the attribute characteristics of each region itself but also integrate the divergent trends between it and the surrounding environment, thereby significantly reducing "island phenomena" and irregular boundaries in the zoning map. In actual field management, this optimization strategy can effectively match the physical continuity of the farming unit, reduce the control complexity of the drip irrigation network, and improve drip irrigation accuracy and fertilization efficiency. Especially when faced with complex field environments with large differences in soil properties and uneven crop growth distribution, this method can achieve differentiated and precise fertilization that is more in line with the actual needs of crops, improve resource utilization and enhance yield stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings, which constitute part of the present invention, are provided to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are provided to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0053] Figure 1 This is a flow chart of a method for precise regulation of nitrogen in the corn root layer under drip irrigation conditions according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0055] See also Figure 1 , is a method flow chart of a method for accurately controlling nitrogen in the root layer of corn under drip irrigation conditions provided by Example 1 of the present invention, such as Figure 1 As shown, a method for precise regulation of nitrogen in the corn root layer under drip irrigation conditions may include:

[0056] Step S1 : obtaining grid points for grid analysis by standardizing multi-source remote sensing soil data of a plot, and evaluating the relative levels of crop growth adaptation at the grid points.

[0057] In the corn root zone nitrogen drip irrigation control scenario, first, UAV remote sensing technology is used to obtain multispectral remote sensing images covering the entire target plot, and the Normalized Vegetation Index is extracted from it as crop growth data;

[0058] Then, a gridded soil map of the target plot is obtained through the multispectral remote sensing image of the target plot, wherein the gridded soil map includes soil particle composition data and mass percentage data of soil organic carbon at each grid point in the target plot; the soil particle composition data of the grid point includes mass percentage data of sand, silt and clay at the grid point; the multispectral remote sensing image of the target plot and the gridded soil map are geo-referenced, and the spatial resolution of the data layer is unified. In this embodiment, the spatial resolution is set to 10 meters by 10 meters, so that the target plot is divided into several grid points, and each grid point has corresponding original crop growth data, sand mass percentage data, silt mass percentage data, clay mass percentage data and organic carbon mass percentage data.

[0059] During drip irrigation and fertilization in the current season, the existing technology adopts the fuzzy C-means (FCM) algorithm, which directly uses the original and continuous crop growth index collected from each grid point in the plot. As core input features, this approach has limitations. This limitation lies in the fact that the distance measurement mechanism of the FCM algorithm is sensitive to the numerical scale and distribution variance of the input features, and the overall growth level of a plot is significantly affected by common macro factors (such as the light and temperature conditions of the year).

[0060] This sensitivity will directly lead to a technical problem in field practice: the influence of crop growth characteristics in cluster analysis will shift with changes in external conditions, thereby affecting the stability of the partitioning results.

[0061] Here we take the same target plot under two different growth conditions as an example:

[0062] When the growing conditions are favorable, the plot will grow well as a whole, and the growth data values ​​of each grid point will be distributed within a high value range (e.g. ).at this time, The data variance of this feature is large, and the difference in its value will be magnified in the distance calculation of FCM. As a result, the algorithm will become The clustering results will tend to be dominated by crop growth, while the role of other equally important static characteristics such as soil texture will be weakened accordingly.

[0063] Under poor growing conditions, the overall growth of the plot is inhibited, and the growth of each grid point is The values ​​are compressed into a low range (e.g. ).at this time, The data variance of this feature is small, and the difference in its value becomes too low to be considered in the distance calculation. As a result, The influence on the clustering results is greatly diluted, and the algorithm will now mainly partition based on other features with a larger numerical range, such as soil texture.

[0064] Existing techniques ignore the absolute numerical scale of input data, which directly interferes with the algorithm's internal feature weighting, resulting in a lack of consistency in the construction of partitioning schemes. The algorithm's attention fluctuates, making it impossible to consistently give comprehensive consideration to both soil and growth factors.

[0065] To solve this technical problem, the first step of optimization of the present invention is to process the original crop growth index before performing cluster analysis, and convert it from an absolute value with an uncertain numerical scale to a growth grade that only reflects its relative ranking position within the current plot. The core idea of ​​this step is to use the statistical distribution characteristics of the growth index set of all grid points in the current target plot to adaptively determine the grading standard. Through this step, regardless of whether the overall growth of the year is high or low, the transformed growth grade characteristics will have a constant numerical range and stable variance contribution, thereby completely eliminating the clustering offset problem caused by changes in the scale of the original data.

[0066] Set the number of growth grade divisions. In this embodiment, the total number of growth grade divisions is set to 5. Set the percentiles of crop growth data according to the number of growth grade divisions. Compare the crop growth data of the grid point with the percentiles of the crop growth data to obtain the adaptive relative grade of crop growth at the grid point. In this embodiment, grade 1 is set as the best and grade 5 is set as the worst. For the 5 growth grade divisions, set 4 percentiles respectively. It should be noted that the growth data at the percentile points are divided into smaller levels.

[0067] The present invention calculates The quantiles of the data set itself allow the grade boundaries to dynamically adapt to the overall growth level of the current field. For example, This boundary percentile, no matter how its absolute value changes, always objectively defines the relative position of the bottom 20% of the growth rate in the field.

[0068] The role of this transformation is that it performs a nonlinear normalization on the original data. Regardless of whether the overall growth of the year is high or low, the growth grade characteristics after the transformation will have a constant value range and relatively stable variance contribution. When the level is used as the input of the subsequent clustering algorithm, it completely eliminates the algorithm clustering deviation problem caused by the change of the original data scale. This means that the influence of the crop growth feature in the cluster analysis will become stable, and the algorithm can fairly evaluate difference.

[0069] At this point, the grid points for grid analysis are obtained by standardizing the multi-source remote sensing soil data of the plots, and the relative levels of crop growth adaptation at the grid points are evaluated.

[0070] Step S2: Obtain the soil potential level of the grid point by factor modeling and adaptively grading the soil erodibility parameters.

[0071] In step S1, the raw crop growth index at each grid point is converted into an adaptive relative grade of crop growth. However, directly inputting this single-dimensional grade and the raw soil data into the standard Fuzzy C-means (FCM) algorithm results in illogical zoning due to two flaws in the algorithm's distance measurement mechanism. First, this mechanism fails to connect and understand the deeper relationship between the inherent potential of the soil and the external performance of the crop. Second, it completely ignores the neighborhood relationships between data points in geographic space, resulting in fragmented zoning results and violating the spatial continuity of field attributes.

[0072] To systematically address these issues, this step introduces an objective evaluation of the soil's intrinsic physical properties. By using a recognized soil evaluation model, the soil potential index is calculated. This index is then classified using an adaptive grading method, converting it into a soil potential grade. This step elevates the decision-making basis from a one-dimensional observation of performance to a two-dimensional perspective of potential and performance.

[0073] Specifically, the soil erodibility factor is calculated by using the soil particle composition data and the mass percentage data of soil organic carbon at each grid point in the target plot to obtain the soil erodibility K value of the grid point; the soil erodibility K value of the grid point is converted inversely to obtain the soil potential index of the grid point; the soil potential index of the grid point is graded using an adaptive grading method to obtain the soil potential grade of the grid point.

[0074] First, to objectively evaluate the intrinsic production potential of each grid point that does not change with short-term factors, the present invention adopts a modified calculation formula for the soil erodibility factor based on Chinese soil characteristics, which is well known in the art. Based on the sand mass percentage data, silt mass percentage data, clay mass percentage data, and organic carbon mass percentage data extracted from the gridded soil map collected from the target plot, the soil erodibility K value of each grid point is calculated using the modified K factor calculation formula in the EPIC model. It should be noted that calculating the soil erodibility K value using the modified K factor calculation formula in the EPIC model is a well-known technique and will not be described in detail here.

[0075] After obtaining the soil erodibility K value of the grid point, the soil erodibility K value of the grid point is further converted into an inverse value to obtain the soil potential index of the grid point. Specifically, since the soil erodibility K value is negatively correlated with the soil's water and fertilizer holding potential, the inverse of the soil erodibility K value of the grid point is used as the soil potential index of the grid point;

[0076] Finally, the soil potential index of the grid point is graded by an adaptive grading method to obtain the soil potential level of the grid point. Specifically, the geographical potential index of the grid point is graded by the crop growth adaptive relative grading method of the grid point in step S1 to obtain the soil potential level of the grid point.

[0077] At this point, the soil potential level of the grid points is obtained by factor modeling and adaptive grading of soil erodibility parameters.

[0078] Step S3: Obtain the adaptive spatial context weight factor of the grid point by fusing and evaluating the global attribute fluctuation and spatial heterogeneity.

[0079] After obtaining the soil potential level of a grid point, a two-dimensional attribute vector can be constructed based on the adaptive relative level of crop growth and the soil potential level of the grid point. This vector can then be used to evaluate the distance between grid points in the subsequent clustering process. However, during the clustering process, the spatial distribution of plot attributes is not isolated but rather spatially correlated. Therefore, a spatial contextual difference penalty assessment must be incorporated into the distance measurement process to quantify the prominence or depression of a grid point relative to its local environment, transforming the invisible, unstructured spatial proximity relationship into a linear, structured feature.

[0080] When establishing a penalty assessment for spatial context discrepancy, the importance of spatial context information in the clustering process is not static but rather closely correlated with the spatial distribution of the target parcel's data. In a parcel with relatively homogeneous spatial structure and mostly smooth attribute transitions, spatial context information is reliable and should be given a higher weight to encourage more regular partitioning. Conversely, in a parcel with strong spatial heterogeneity and areas of attribute abrupt change, over-reliance on spatial context information for smoothing can mask true differences and should be reduced in this case.

[0081] To solve this problem, this step aims to construct an adaptive spatial context weight factor that can automatically calculate the grid point based on the data characteristics of the target plot itself. Based on this goal, it is necessary to construct a new indicator that can directly quantify the total intensity of the spatial heterogeneity of the entire plot, and then establish an evaluation method for the adaptive spatial context weight factor of the grid point based on this indicator. The present invention measures the overall unevenness of the spatial pattern of the entire plot by performing global statistics on the modulus of the gradient vectors of all grid points in the field, and obtains the adaptive spatial context weight factor of the grid point based on this. Specifically, a two-dimensional attribute vector of the grid point is established through the soil potential level of the grid point and the adaptive relative level of the crop growth of the grid point; the local attribute gradient vector of the grid point is obtained by performing gradient analysis on the spatial neighboring grid points of the grid point; the spatial heterogeneity intensity of the plot is analyzed by the local attribute gradient vector of the grid point to obtain the global gradient divergence; the global heterogeneity of the plot where the grid point is located is analyzed by the global gradient divergence of the plot where the grid point is located, the global soil potential level of the plot where the grid point is located, and the global crop growth adaptive relative level of the plot where the grid point is located, and the adaptive spatial context weight factor of the grid point is obtained.

[0082] Firstly, a two-dimensional attribute vector of the grid point is established through the soil potential level of the grid point and the adaptive relative level of crop growth at the grid point.

[0083] After obtaining the two-dimensional attribute vector of the grid point, continue to perform gradient analysis on the spatial neighboring grid points of the grid point to obtain the local attribute gradient vector of the grid point. Specifically, set the neighborhood range of the grid point. In the embodiment of the present invention, the neighborhood range of the grid point is selected as the 8 grid points directly adjacent to the target grid point. The calculation result of subtracting the crop growth adaptive relative level of the grid point from the average crop growth adaptive relative level of the grid points in the neighborhood range of the grid point is used as the first gradient evaluation of the crop growth level of the grid point; the calculation result of subtracting the soil potential level of the grid point from the average soil potential level of the grid points in the neighborhood range of the grid point is used as the first gradient evaluation of the soil potential level of the grid point; and the two-dimensional vector formed by the first gradient evaluation of the crop growth level of the grid point and the first gradient evaluation of the soil potential level is used as the local attribute gradient vector of the grid point.

[0084] In one embodiment, assuming that The soil potential level of each grid point is ;No. The relative level of crop growth adaptation at each grid point is , then The calculation expression of the local attribute gradient vector of a grid point is:

[0085]

[0086] in, Indicates the The local attribute gradient vector of each grid point; Indicates the The soil potential level of each grid point; Indicates the The relative level of crop growth adaptation at each grid point; The number of grid points representing the neighborhood of a grid point.

[0087] After obtaining the local attribute gradient vector of the grid point, the spatial heterogeneity intensity of the plot can be analyzed through the local attribute gradient vector of the grid point to obtain the global gradient divergence. Specifically, the calculation result of adding the square of the first gradient evaluation of the crop growth level of the grid point and the square of the first gradient evaluation of the soil potential level of the grid point is used as the modulus of the local attribute gradient vector of the grid point; the square root of the mean of the modulus of the local attribute gradient vectors of all grid points in the target plot where the grid point is located is used as the global gradient divergence.

[0088] In one embodiment, the global gradient divergence is calculated as:

[0089]

[0090] in, , represents the global gradient divergence; Indicates the The local attribute gradient vector of each grid point; Indicates the total number of grid points within the target plot; Indicates the The soil potential level of each grid point; Indicates the The relative level of crop growth adaptation at each grid point; Indicates the The average soil potential level of the grid points in the neighborhood of the grid point; Indicates the The average adaptive relative grade of crop growth at each grid point.

[0091] After obtaining the global gradient divergence, a global heterogeneity analysis of the plot where the grid point is located is performed through the global gradient divergence of the plot where the grid point is located, the global soil potential level of the plot where the grid point is located, and the global crop growth adaptive relative level of the plot where the grid point is located to obtain the adaptive spatial context weight factor of the grid point. Specifically, the soil potential level variance of the target plot is obtained through the soil potential level of all grid points in the target plot; the crop growth adaptive relative level variance of the target plot is obtained through the crop growth adaptive relative level of all grid points in the target plot; the calculation result of adding the soil potential level variance of the target plot and the crop growth adaptive relative level variance of the target plot is used as the attribute discreteness evaluation of the target plot; the calculation result of adding the constant 1 and the global gradient divergence is used as the denominator, the attribute discreteness evaluation of the target plot is used as the numerator, and the calculation result of the corresponding fraction is used as the adaptive spatial context weight factor of the grid point in the target plot.

[0092] In one embodiment, it is assumed that the adaptive relative level variance of crop growth in the target plot is ; The variance of the soil potential grade of the target plot is , then the calculation expression of the adaptive spatial context weight factor of the grid point in the target plot is:

[0093]

[0094] in, Represents the adaptive spatial context weight factor of the grid points in the target plot; Indicates the variance of soil potential grade of the target plot; Indicates the adaptive relative grade variance of crop growth in the target plot; represents the global gradient divergence.

[0095] It should be noted that the core of the above formula is to convert a hyperparameter used to adjust the behavior of the algorithm into a variable determined by the data characteristics of the plot itself. The global gradient divergence proposed in this invention is a measure of the spatial heterogeneity of the plot. This indicator is evaluated based on the local attribute gradient vector of the grid point. The modulus of the local attribute gradient vector has the physical meaning of the attribute difference intensity between the grid point and its surrounding neighborhood environment. By taking the square root of the difference intensity of all grid points in the plot, the global gradient divergence obtained can quantify the average degree of unevenness of the attribute distribution of the plot. The stronger the spatial heterogeneity of a plot and the more attribute mutation points it has, the larger the corresponding global gradient divergence value. In the calculation formula of the adaptive spatial context weight factor of the grid point, the result of adding the global gradient divergence and the constant 1 is used as the denominator, where the constant 1 prevents the denominator from being 0. The higher the value of the global gradient divergence, that is, when the overall spatial heterogeneity of the plot is strong, the adaptive spatial context weight factor of the grid point will decrease accordingly. In the subsequent clustering process, this will reduce the weight of the spatial context difference penalty evaluation, so that the partitioning results are more dominated by the basic attribute differences of each grid point. This is to avoid excessive spatial smoothing on the plots that are already relatively fragmented. On the contrary, the lower the value of the global gradient divergence, that is, when the spatial pattern of the plot is more homogeneous, the value of the adaptive spatial context weight factor of the grid point will increase accordingly. This will increase the weight of the spatial context difference penalty evaluation, so that the clustering process focuses more on maintaining the spatial continuity of the partition, thereby obtaining a partition boundary that is more in line with the natural gradient law in the area where the attributes are already smooth. At the same time, the sum of the variances of the soil potential level and the growth performance level Placed in the numerator, its function is to adjust the base scale of the adaptive spatial context weight factor of the grid point based on the degree of dispersion of the parcel attributes. In a parcel with very large attribute differences, a relatively large adaptive spatial context weight factor value is necessary to ensure that the spatial context term is not overwhelmed by the huge underlying attribute differences in the distance calculation.

[0096] At this point, the adaptive spatial context weight factors of grid points are obtained by fusing and evaluating the global attribute fluctuations and spatial heterogeneity.

[0097] Step S4, obtaining the gradient consistency distance metric between grid points by fusing and evaluating the adaptive spatial context weight factors of the grid points with the soil potential levels of the grid points.

[0098] Based on the two-dimensional attribute coordinates and local attribute gradient vectors of the grid points obtained in the above steps, the present invention performs distance measurement between the grid points through a gradient consistency adaptive distance metric, thereby calculating the true similarity between any two grid points. Specifically, the crop growth adaptive relative levels between the grid points are subtracted and the squared result is used as the first crop growth level difference between the grid points; the soil potential levels between the grid points are subtracted and the squared result is used as the first soil potential level difference between the grid points;

[0099] The result of multiplying the adaptive spatial context weight factor of the grid point in the target plot by the square of the modulus of the local attribute gradient vector difference between the grid points is used as a penalty evaluation of the spatial context difference between the grid points;

[0100] The calculation result of adding the first crop growth grade difference between the grid points, the first soil potential grade difference between the grid points and the spatial context difference penalty evaluation between the grid points is used as the gradient consistency distance measure between the grid points.

[0101] In one embodiment, the grid points and the The calculation expression of the gradient consistency distance measure between grid points is:

[0102]

[0103] in, Indicates the grid points and the Gradient consistency distance metric between grid points; Indicates the The soil potential level of each grid point; Indicates the The relative level of crop growth adaptation at each grid point; Indicates the The soil potential level of each grid point; Indicates the The relative level of crop growth adaptation at each grid point; Represents the adaptive spatial context weight factor of the grid points in the target plot; Indicates the The local attribute gradient vector of each grid point; Indicates the The local attribute gradient vector of each grid point; Indicates the The local attribute gradient vector of the grid point is The square of the magnitude of the vector obtained by subtracting the local attribute gradient vectors of the grid points.

[0104] It should be noted that, firstly, the present invention constructs the soil potential level and compare it with the growth performance level Combined, they form a two-dimensional attribute coordinate, expanding the analysis dimension from a single growth phenomenon to a two-dimensional level that can comprehensively reflect the inherent potential of the soil and current crop performance. Basic attribute distance term in the formula , is a direct measure of the difference in the coordinates of these two-dimensional attributes. This enables the subsequent clustering process to distinguish between areas with completely different meanings in agronomic management, such as high potential-low performance and low potential-low performance, providing the necessary information basis for scientific zoning. The core innovation of this invention is that the spatial distribution of field attributes is not isolated, but has spatial autocorrelation. To this end, in The spatial context difference penalty term is added to the formula This penalty term introduces the local attribute gradient vector This concept. The vector is calculated by each point The difference between the attribute level of a point and the average level of its neighborhood quantifies the prominence or depression of the point relative to its local environment, thereby transforming the implicit unstructured spatial neighborhood relationship into an explicit structured feature.

[0105] By calculating two points and The modulus of the difference between the gradient vectors of , quantifying the similarity of the local environmental change trends of the two points. The purpose of this design is to expand the comparison object of the clustering algorithm from isolated points to a comprehensive comparison of the points and their local environment. This makes the algorithm's similarity judgment criterion no longer a simple geometric distance, but upgraded to an understanding of the land environment. Only when two points are not only similar in their own attributes, but also in the similar trend of the local environment they are in, can they be considered to be truly similar. The distance will be calculated as the closest.

[0106] This design effectively addresses the island and hard boundary issues in existing partitioning. For an isolated point whose attributes differ significantly from those of its surroundings, the modulus of its gradient vector will be very large, resulting in a significant spatial contextual difference from any surrounding points, significantly pushing them apart. The algorithm tends to cluster all such points with similar island characteristics into one category, thus properly preserving these special areas of important management significance. Furthermore, for points located along a smooth gradient of attributes, their similar gradient vectors enable the algorithm to smoothly partition them, avoiding uneven boundaries.

[0107] At this point, the gradient consistency distance metric between grid points is obtained by fusion evaluation of the adaptive spatial context weight factor of the grid points and the soil potential level of the grid points.

[0108] Step S5: Perform fuzzy cluster analysis using the gradient consistency distance metric between grid points to obtain the optimized management zoning map and differentiated drip irrigation strategy.

[0109] After completing the calculations in the previous steps, the two-dimensional attribute coordinates of each grid point are obtained. , local attribute gradient vector , and an adaptive spatial context weighting factor for the entire plot After that, step S5 of the present invention performs the final optimized clustering operation to generate the management partition map. This step uses the Fuzzy C-means (FCM) algorithm as the basic framework, but replaces its core distance metric function with the gradient consistency distance metric between grid points in step S4 of the present invention. Through iterative optimization, the algorithm will find a distance that can make all grid points to the center of their own partition. Optimal partitioning scheme that minimizes the weighted sum of distances.

[0110] To determine the optimal number of management partitions The present invention uses the Davis-Boulding index, a clustering effectiveness index recognized in the field, to search for the optimal value. Within a preset partition number range (from 2 to 10), for each candidate partition number , perform a complete FCM clustering operation and calculate the DBI value corresponding to the partition result. Select the candidate partition number that makes the DBI value reach the minimum as the optimal partition number .

[0111] After determining the optimal number of partitions Finally, the present invention uses the Fuzzy C-Means (FCM) algorithm to perform the final clustering operation. This algorithm iteratively updates the membership matrix and cluster center matrix to minimize the objective function. The iterative solution process of this algorithm is well-known in the art, and the present invention applies it on this basis. After the algorithm converges, each grid point is assigned to the cluster with the highest membership, thus forming the final management partition map.

[0112] After obtaining the management zone map, the present invention enters the fertigation decision-making and execution phase. For each management zone, the average soil potential and average growth performance ratings of all grid points within it are calculated to determine the zone's core attributes. Based on these attributes, a differentiated fertigation prescription is formulated for each zone and executed through the intelligent drip irrigation control system.

[0113] The specific decision-making and implementation methods are as follows:

[0114] For subareas rated as excellent for both average soil potential and average growth performance, this indicates a good soil foundation and fully developed crop growth. At this time, drip irrigation fertigation prioritizes maintenance fertilization, providing a steady supply of nitrogen according to the crop's standard fertilizer requirements for its current growth stage to consolidate its growth advantage and ensure nutrient depletion in the later stages.

[0115] For subareas rated as poor in both average soil potential and average growth performance, soil physical properties are the primary limiting factor to yield. Directly increasing nitrogen fertilizer application is inefficient. Therefore, the drip irrigation strategy prioritizes economical fertilization, using a nitrogen supply level below the standard requirement to meet basic crop growth needs and avoid unnecessary fertilizer input.

[0116] Subareas with high soil potential but low growth performance are key management areas with the greatest potential for yield increases. This indicates that the soil foundation is excellent, but current growth is being inhibited by certain controllable factors. In this case, the drip irrigation strategy will prioritize yield-enhancing fertilization, using a nitrogen supply level above the standard fertilizer requirement. Simultaneously, this subarea will be marked by the system as requiring priority manual field inspections to identify any limiting factors besides nitrogen.

[0117] For all zones where the average soil potential grade is assessed as poor (indicating a high soil erodibility K value, mostly sandy soils), regardless of their growth status, their drip fertigation patterns will be forcibly adjusted. This is because these soils have poor water and fertilizer retention capacity, making nitrogen highly susceptible to leaching with the drip irrigation water flow. Therefore, their fertilization plan will adopt a small, multiple application model. That is, based on the total fertilizer application amount, a single fertilization task will be broken down into several executions, significantly reducing the nitrogen concentration and irrigation duration of a single drip fertigation, while increasing the frequency of drip irrigation. This model ensures that each applied nitrogen can be controlled within the shallow root zone of the corn, minimizing leaching losses caused by excessively deep infiltration, thereby greatly improving the efficiency of nitrogen fertilizer utilization in the current season.

[0118] Finally, these differentiated fertilization plans based on zoning characteristics are input into the intelligent drip irrigation control system in the field. By independently programming the pipe network valves and fertilizer injection pumps in different zones, precise regulation of the time, space and dosage of nitrogen supply to the corn root layer is achieved.

[0119] In summary, the embodiments of the present invention obtain soil grid points by standardizing multi-source remote sensing soil data of a plot, and evaluate the adaptive relative grade of crop growth at the grid points; obtain the soil potential grade of the grid points by factor modeling and adaptive grading of soil erodibility parameters; obtain the adaptive spatial context weight factor of the grid points by fusing and evaluating global attribute fluctuations and spatial heterogeneity; obtain the gradient consistency distance metric between grid points by fusing and evaluating the adaptive spatial context weight factor of the grid points with the soil potential grade of the grid points; and obtain the optimized management zoning map and differentiated drip irrigation strategy by performing fuzzy clustering analysis based on the gradient consistency distance metric between grid points. The present invention achieves in-depth optimization of drip irrigation and fertilization management zoning by introducing dual-dimensional modeling of crop growth grade and soil potential grade, combining local attribute gradient and spatial heterogeneity modeling mechanisms. Compared with the existing technology, this method significantly enhances the dual perception ability of cluster analysis on crop growth status and soil basic capacity, and effectively avoids the problem of clustering focus offset caused by data scale changes in different growing years in traditional clustering methods. In the face of dramatic changes in environmental conditions such as light and moisture between years, this method can maintain stable feature expression and partitioning logic, improving the continuity and repeatability of the fertilization plan. At the same time, the present invention introduces a gradient consistency distance measurement method to incorporate the spatial context features between grid points into the partitioning decision-making process, so that clustering not only considers the attribute characteristics of each region itself, but also integrates its difference trends with the surrounding environment, thereby greatly reducing the "island phenomenon" and irregular boundaries in the partition map. In actual field management, this optimization strategy can effectively match the physical continuity of the farming unit, reduce the complexity of drip irrigation network control, and improve drip irrigation accuracy and fertilization efficiency. Especially in the face of complex field environments with large differences in soil properties and uneven distribution of crop growth, this method can achieve differentiated precision fertilization that is more in line with the actual needs of crops, improve resource utilization and enhance yield stability.

[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for accurately regulating nitrogen in the root zone of corn under drip irrigation conditions, characterized in that: The method comprises the following steps: Step S1: obtaining grid points for grid analysis by standardizing multi-source remote sensing soil data of the plot, and evaluating the relative level of crop growth adaptation at the grid points; Step S2: Obtaining the soil potential level of the grid point by factor modeling and adaptive classification of soil erodibility parameters; Step S3: Obtain the adaptive spatial context weight factor of the grid point by fusing and evaluating the global attribute fluctuation and spatial heterogeneity; Step S4: Obtaining the gradient consistency distance metric between grid points by fusing and evaluating the adaptive spatial context weight factor of the grid points with the soil potential level of the grid points; Step S5: Perform fuzzy cluster analysis using the gradient consistency distance metric between grid points to obtain the optimized management zoning map and differentiated drip irrigation strategy; The method of obtaining the adaptive spatial context weight factor of the grid point by fusing and evaluating the global attribute fluctuation and spatial heterogeneity includes: establishing a two-dimensional attribute vector of the grid point through the soil potential level of the grid point and the adaptive relative level of crop growth of the grid point; obtaining the local attribute gradient vector of the grid point by performing gradient analysis on the spatial neighboring grid points of the grid point; performing spatial heterogeneity intensity analysis of the plot through the local attribute gradient vector of the grid point to obtain the global gradient divergence; and performing global heterogeneity analysis of the plot where the grid point is located through the global gradient divergence of the plot where the grid point is located, the global soil potential level of the plot where the grid point is located, and the global adaptive relative level of crop growth of the plot where the grid point is located to obtain the adaptive spatial context weight factor of the grid point.

2. The method for precise control of nitrogen in the corn root zone under drip irrigation conditions according to claim 1, characterized in that: The method of obtaining grid points for grid analysis by standardizing multi-source remote sensing soil data of a plot and evaluating the relative level of crop growth adaptability at the grid points specifically includes: Using UAV remote sensing technology to obtain multispectral remote sensing images covering the entire target plot, and extracting the normalized vegetation index as crop growth data; Obtaining a gridded soil map of the target plot using a multispectral remote sensing image of the target plot, wherein the gridded soil map includes soil particle composition data and mass percentage data of soil organic carbon at each grid point in the target plot; the soil particle composition data at the grid point includes mass percentage data of sand, silt, and clay at the grid point; geo-registering the multispectral remote sensing image of the target plot with the gridded soil map, and unifying the spatial resolution of the data layers; The number of growth grade divisions is set, and the percentile of the crop growth data is set according to the number of growth grade divisions; the crop growth data of the grid point is compared with the percentile of the crop growth data to obtain the adaptive relative grade of the crop growth of the grid point.

3. The method for precise control of nitrogen in the corn root zone under drip irrigation conditions according to claim 1, characterized in that: The soil potential level of the grid point is obtained by factor modeling and adaptive grading of soil erodibility parameters, including: The soil erodibility factor is calculated using the soil particle composition data and the mass percentage data of soil organic carbon at each grid point in the target plot to obtain the soil erodibility K value of the grid point; The soil potential index of the grid point is obtained by performing reciprocal conversion on the soil erodibility K value of the grid point; The soil potential index of the grid points is graded using an adaptive grading method to obtain the soil potential level of the grid points.

4. The method for precise control of nitrogen in the corn root zone under drip irrigation conditions according to claim 1, characterized in that: The method of performing gradient analysis on the spatially neighboring grid points of the grid point to obtain the local attribute gradient vector of the grid point includes: The neighborhood range of the grid point is set, and the result of subtracting the adaptive relative grade of crop growth of the grid point from the average adaptive relative grade of crop growth of the grid points in the neighborhood range of the grid point is used as the first gradient evaluation of the crop growth grade of the grid point; the result of subtracting the soil potential grade of the grid point from the average soil potential grade of the grid points in the neighborhood range of the grid point is used as the first gradient evaluation of the soil potential grade of the grid point; the two-dimensional vector formed by the first gradient evaluation of crop growth grade of the grid point and the first gradient evaluation of soil potential grade is used as the local attribute gradient vector of the grid point.

5. The method for precise control of nitrogen in the corn root zone under drip irrigation conditions according to claim 1, characterized in that: The spatial heterogeneity intensity analysis of the plot is performed through the local attribute gradient vector of the grid point to obtain the global gradient divergence, including: The result of adding the square of the first gradient evaluation of the crop growth grade of the grid point and the square of the first gradient evaluation of the soil potential grade of the grid point is used as the modulus of the local attribute gradient vector of the grid point; The square root of the mean of the modulus of the local attribute gradient vectors of all grid points in the target plot where the grid point is located is taken as the global gradient divergence.

6. The method for precise control of nitrogen in the corn root zone under drip irrigation conditions according to claim 1, characterized in that: The method of performing a global heterogeneity analysis on the plot where the grid point is located by using the global gradient divergence, the global soil potential level and the global crop growth adaptive relative level of the plot where the grid point is located to obtain the adaptive spatial context weight factor of the grid point includes: The variance of the soil potential grade of the target plot is obtained by the soil potential grade of all grid points in the target plot; the variance of the crop growth adaptive relative grade of the target plot is obtained by the crop growth adaptive relative grade of all grid points in the target plot; The result of adding the variance of the soil potential grade of the target plot and the variance of the crop growth adaptive relative grade of the target plot is used as the attribute dispersion evaluation of the target plot; The result of adding the constant 1 to the global gradient divergence is used as the denominator, the attribute discreteness evaluation of the target plot is used as the numerator, and the calculation result of the corresponding fraction is used as the adaptive spatial context weight factor of the grid point in the target plot.

7. The method for precise control of nitrogen in the corn root zone under drip irrigation conditions according to claim 1, characterized in that: The method of obtaining the gradient consistency distance metric between grid points by fusing and evaluating the adaptive spatial context weight factor of the grid points with the soil potential level of the grid points includes: The result of subtracting the adaptive relative levels of crop growth between grid points and performing square calculation is used as the first crop growth level difference between grid points; the result of subtracting the soil potential level between grid points and performing square calculation is used as the first soil potential level difference between grid points; The result of multiplying the adaptive spatial context weight factor of the grid point in the target plot by the square of the modulus of the local attribute gradient vector difference between the grid points is used as the penalty evaluation of the spatial context difference between the grid points; The calculation result of adding the first crop growth grade difference between the grid points, the first soil potential grade difference between the grid points and the spatial context difference penalty evaluation between the grid points is used as the gradient consistency distance measure between the grid points.

8. The method for precise control of nitrogen in the corn root zone under drip irrigation conditions according to claim 1, characterized in that: The fuzzy cluster analysis performed by measuring the gradient consistency distance between grid points to obtain the optimized management partition map and differentiated drip irrigation strategy includes: The Davis-Boulting index is used to optimize the number of management partitions and obtain the number of partitions for the clustering process; The gradient consistency distance metric between the grid points is used as the distance metric between the grid points in the clustering process, and fuzzy C-means clustering is performed according to the number of partitions to obtain a management partition map; Differentiated drip irrigation strategies are formulated by managing the average soil potential level and the average crop growth adaptive relative level of each zone in the zoning map.

9. The method for precise control of nitrogen in the corn root zone under drip irrigation conditions according to claim 8, characterized in that: The differentiated drip irrigation strategy is formulated by managing the average soil potential level and the average crop growth adaptive relative level of each zone in the zone map, specifically including: For each zone in the management zone map, calculate the average soil fertility potential level and the average crop growth adaptability relative level of all grid points in the zone; Set the level assessment threshold to classify the soil potential level and crop growth adaptability relative level into good and poor; For sub-regions rated as good in both average soil potential and average crop growth adaptability, the drip irrigation fertigation strategy is based on maintenance fertilization, steadily supplying nitrogen according to the standard fertilizer requirement of the crop at its current growth stage. For sub-areas where both the average soil fertility potential grade and the average crop growth adaptability relative grade were rated as poor, the drip irrigation fertigation strategy was based on economical fertilization, using a nitrogen supply level lower than the standard fertilizer requirement. For sub-regions with an average soil potential grade of good and an average crop growth adaptability grade of poor, the drip irrigation fertigation strategy is mainly based on yield-increasing fertilization, using a nitrogen supply level higher than the standard fertilizer requirement. For zones with a poor average soil potential grade and an excellent average crop growth adaptive relative grade, the drip fertigation strategy adopts a small amount and multiple times mode. On the basis of determining the total amount of fertilizer, the single fertilization task is broken down into multiple executions, significantly reducing the nitrogen concentration and irrigation time of a single drip fertigation, while increasing the frequency of drip irrigation.

Images