Route collaborative scheduling optimization method for corn straw returning and fertilization operation

Abstract

The invention relates to the field of agricultural fertilization, in particular to a corn straw returning-to-field and fertilization operation path collaborative scheduling optimization method, which comprises the following steps: performing standardization processing on farmland multi-source original data to obtain a field piece unit standardization feature vector; performing imbalance degree analysis on the feature dimension distance component to obtain a feature contribution balance factor; performing activeness evaluation on neighborhood membership distribution of the field units to obtain a spatial membership smoothing factor; obtaining adaptive distance measurement by combining a feature contribution balance factor and a space membership smoothing factor; and performing hardening processing on the final membership matrix to obtain an agricultural machinery path collaborative scheduling instruction set, thereby solving the problems of misalignment of a zoning result, fragmentation of a management boundary and unreasonable agricultural machinery path scheduling in an abnormal straw accumulation region based on a standard FCM clustering algorithm in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural fertilization, in particular to a path coordination scheduling optimization method for corn straw returning to field and fertilization operation. BACKGROUND

[0002] In modern precision agriculture practice, in order to realize the differentiation and refinement management of large area farmland, it is usually necessary to divide the farmland into several management partitions, so that the land units in the same area have similar characteristics in soil properties, crop growth potential and the like, so that uniform field operation measures can be applied. In the prior art, the fuzzy C-means (FCM) clustering algorithm is a commonly used method for realizing farmland management partitioning. This method collects multi-source agronomic data of multiple sampling points in the farmland area, such as historical yield data, soil apparent conductivity (ECa) data and digital elevation model (DEM) data, abstracts each sampling point as a multi-dimensional feature vector, and calculates the similarity between different land units by Euclidean distance. On this basis, the FCM algorithm divides similar land units into corresponding clustering categories through iterative optimization, and then forms a management partition map that can reflect the spatial heterogeneity of the farmland. This partitioning method provides important data support and decision basis for variable fertilization, variable seeding and other precision agricultural activities. With the popularization of straw returning to field technology in agricultural production, the farmland partitioning method gradually needs to introduce new feature data such as straw coverage or straw density into clustering analysis, so as to reflect the influence of straw distribution on soil nutrient cycling and crop growth in the partitioning results. However, after the fusion of straw density features and traditional agronomic feature data, the existing partitioning method based on the standard FCM algorithm has certain defects. Since the calculation process of Euclidean distance adopts equal weight accumulation for each feature dimension, when the straw density value in the abnormal straw accumulation area is much higher than that in the surrounding area due to operation overlap, land turning and other reasons, the numerical difference of this single dimension will dominate in the distance calculation, destroying the balanced contribution of the original multi-dimensional features. As a result, these abnormal areas are often incorrectly separated, forming broken boundaries or enclaves that do not match the actual agronomic needs, thereby distorting the management partitioning results. More importantly, this distorted partitioning result will cause low efficiency and resource waste problems in subsequent agricultural machinery path planning and operation scheduling. Therefore, how to effectively suppress the distance pollution effect caused by a single operation feature in the farmland partitioning process, and obtain a geometrically regular and executable partitioning result while ensuring the authenticity of the agronomic features, has become a technical problem that needs to be solved in the field. SUMMARY

[0003] Therefore, the present application aims to provide a corn straw returning to field and fertilization operation path coordination scheduling optimization method to solve the problems of inaccurate partitioning results, broken management boundaries and unreasonable agricultural machinery path scheduling in the existing standard FCM clustering algorithm in the abnormal straw accumulation area.

[0004] To achieve the above purpose, the technical scheme of the present application is as follows:

[0005] A corn straw returning to field and fertilization operation path coordination scheduling optimization method, comprising:

[0006] Step S1: obtaining a field unit standardized feature vector by standardizing the multi-source original data of the farmland;

[0007] Step S2: obtaining a feature contribution balance factor by analyzing the unbalance degree of the feature dimension distance component;

[0008] Step S3: obtaining a spatial membership degree smoothing factor by evaluating the activity degree of the field unit neighborhood membership degree distribution;

[0009] Step S4: obtaining an adaptive distance measure by combining the feature contribution balance factor and the spatial membership degree smoothing factor;

[0010] Step S5: obtaining an agricultural machinery path coordination scheduling instruction set by hardening the final membership degree matrix.

[0011] Further, the method of obtaining a field unit standardized feature vector by standardizing the multi-source original data of the farmland comprises:

[0012] Grid sampling is performed on the target farmland area to obtain all field units of the target farmland area; historical yield data of the field units are obtained by a combine harvester equipped with a yield monitor; soil apparent conductivity data of the field units are obtained by a soil conductivity scanning device; digital elevation model data of the field units are obtained by an airborne or spaceborne remote sensing device; straw density data of the field units are obtained by a multi-spectral or high-resolution visible light camera carried by a UAV combined with image analysis technology; the historical yield data, the soil apparent conductivity data, the digital elevation model data and the straw density data are combined to form a multi-dimensional original data matrix, and each feature dimension of the matrix is subjected to Z-score standardization processing, thereby obtaining a standardized feature vector corresponding to each field unit.

[0013] Further, the method of obtaining a feature contribution balance factor by analyzing the unbalance degree of the feature dimension distance component comprises:

[0014] Obtaining a cluster center in the clustering process; obtaining a distance contribution imbalance index by performing energy distribution analysis on feature dimension difference data between the field block unit and the cluster center; and obtaining a feature contribution balance factor by performing nonlinear response processing on the distance contribution imbalance index.

[0015] Further, the distance contribution imbalance index is obtained by performing energy distribution analysis on feature dimension difference data between the field block unit and the cluster center, comprising:

[0016] For any target field block unit in all field block units and any target cluster center in all cluster centers, the square of the difference between the target field block unit and the target cluster center in any target feature dimension is taken as the distance energy component of the target feature dimension; the distance energy components of all feature dimensions are accumulated to obtain a total energy distribution evaluation; and the fourth power of the difference between the target field block unit and the target cluster center in all target feature dimensions is accumulated to obtain an energy concentration evaluation;

[0017] The calculation result of multiplying the number of feature dimensions and the energy concentration evaluation is taken as the numerator; the square of the total energy distribution evaluation is taken as the denominator, and the calculation result of subtracting the corresponding fraction from the constant 1 is taken as the distance contribution imbalance index.

[0018] Further, the feature contribution balance factor is obtained by performing nonlinear response processing on the distance contribution imbalance index, comprising:

[0019] A gain coefficient of the imbalance response is set; an amplified imbalance response value is obtained by performing gain coefficient amplification processing on the distance contribution imbalance index; a smooth correction increment between zero and a constant one is obtained by performing an inverse tangent nonlinear mapping processing on the amplified imbalance response value and normalizing by the ratio of a constant 2 to a constant pi; and the calculation result of adding the smooth correction increment to the constant 1 is taken as the feature contribution balance factor.

[0020] Further, the spatial membership degree smoothing factor is obtained by performing activity evaluation on the neighborhood membership degree distribution of the field block unit, comprising:

[0021] The neighborhood membership degree mean value data is obtained by performing statistical processing on the neighborhood membership degree data of the field block unit;

[0022] The neighborhood membership degree activity index is obtained by performing variance analysis processing on the neighborhood membership degree data of the field block unit;

[0023] The spatial membership degree smoothing factor is obtained by performing adaptive ratio processing on the membership degree and neighborhood mean deviation data of the field block unit.

[0024] Further, the obtaining of the neighborhood membership mean value data by statistically processing the neighborhood membership data of the field block units comprises:

[0025] The neighborhood range of the field block units is set, and for any target field block unit in all the field block units, the neighborhood membership set of the target field block unit is obtained through the neighborhood range of the field block units.

[0026] In the clustering process, the membership values of all the field block units in the neighborhood membership set to the same target cluster center are accumulated, and the obtained accumulated value is divided by the number of the field block units in the neighborhood to obtain the neighborhood membership mean value data of the target field block unit to the target cluster center.

[0027] Further, the obtaining of the neighborhood membership activity index by performing variance analysis on the neighborhood membership data of the field block units comprises:

[0028] In the neighborhood range centered on the target field block unit, the membership values of all the field block units to all the cluster centers are obtained, and the membership values of all the field block units to all the cluster centers in the neighborhood are taken as the overall sample to calculate the variance value of the sample, and the variance value is taken as the neighborhood membership activity index.

[0029] Further, the obtaining of the spatial membership smoothing factor by performing adaptive ratio processing on the membership deviation data of the field block unit and the neighborhood mean value comprises:

[0030] The membership value of the target field block unit to the target cluster center is obtained, and the difference between the membership value and the neighborhood membership mean value data is calculated. The difference is squared to obtain the deviation energy of the membership of the target field block unit. The calculation result of adding the deviation energy and the neighborhood membership activity index is taken as the neighborhood corrected membership fluctuation value. The spatial membership smoothing factor is obtained by performing ratio calculation on the neighborhood corrected membership fluctuation value and the neighborhood membership activity index.

[0031] Further, the obtaining of the adaptive distance measure by combining the feature contribution balancing factor and the spatial membership smoothing factor comprises:

[0032] The original Euclidean distance between the target field block unit and the target cluster center is obtained. The distance value corrected based on the feature dimension distance component is obtained by multiplying the original Euclidean distance by the feature contribution balancing factor. The adaptive distance measure between the target field block unit and the target cluster center is obtained by multiplying the distance value corrected based on the feature dimension distance component by the spatial membership smoothing factor again.

[0033] Compared with the prior art, the present application has the following advantages:

[0034] The corn straw returning to field and fertilization operation path collaborative scheduling optimization method can effectively inhibit the excessive dominant effect of single operation characteristics such as straw density on the partition result, while maintaining the reasonable weight distribution of the agronomic background characteristics in the partition process. In this way, not only can the high-straw load area with special management value be accurately identified, but also the partition result deviation caused by single-dimensional feature abnormalities can be avoided, so that the generated management partition is more in line with the actual heterogeneity distribution of the farmland, and the scientificity and pertinence of the variable fertilization decision are improved. At the same time, by introducing the neighborhood space smoothing mechanism in the membership calculation, the management partition boundary is more continuous and regular in geometric form, avoiding the unreasonable enclaves and jagged boundaries on the partition map. Combined with the agricultural operation scene, the optimized partition result can be directly converted into path scheduling instructions, so that the agricultural machinery can adaptively adjust the operation speed and fertilization mode when entering the high-straw load area, realizing the synchronous optimization of fertilization operation and straw returning operation, and significantly improving the precision and efficiency of field operation. BRIEF DESCRIPTION OF DRAWINGS

[0035] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application, illustrate the preferred embodiments of the application, and assist in the explanation of the application. In the drawings, the same reference numbers represent the same elements throughout the several views of the drawings:

[0036] Figure 1 A method flowchart of a corn straw returning to field and fertilization operation path collaborative scheduling optimization method according to an embodiment of the present application. DETAILED DESCRIPTION

[0037] The present application will be described in detail below with reference to the accompanying drawings and embodiments.

[0038] Reference Figure 1 A method flowchart of a corn straw returning to field and fertilization operation path collaborative scheduling optimization method according to an embodiment of the present application is shown in FIG. 1. The corn straw returning to field and fertilization operation path collaborative scheduling optimization method can include: Figure 1

[0039] Step S1: obtaining a field unit standardized feature vector by standardizing the multi-source original data of the farmland.

[0040] This step aims to establish a unified and comparable data basis for subsequent management partition analysis, so as to ensure that farmland feature data of different sources can be measured and calculated in the same scale space.

[0041] ​Firstly, the target farmland area is grid-divided, and the farmland is divided into a plurality of field block units, the number of field block units is determined according to the size of the farmland in the actual scene, and a corresponding spatial position is set for each field block unit. For each field block unit, the following multi-source original data is obtained: historical yield data collected by a combine harvester equipped with a yield monitor, used to reflect the yield level of crops in the historical planting process; apparent soil conductivity data collected by a soil conductivity scanning device, used to reflect the differences in soil texture and fertility of the farmland; digital elevation model data obtained by airborne or spaceborne remote sensing equipment, used to reflect the topographic relief and hydrological environment characteristics of the farmland area; straw density data obtained by a multi-spectral or high-resolution visible light camera carried by a UAV combined with image processing technology, used to reflect the spatial distribution of straw coverage and accumulation after farm work.

[0042] After the above data collection is completed, the multi-source data corresponding to each field block unit forms a multi-dimensional feature vector, and a multi-dimensional original data matrix is formed with all field block units as samples. Since different types of data come from different sources, their physical dimensions and numerical ranges differ significantly, and if used directly for clustering calculation, it will lead to uneven feature contribution, so standardization processing is needed. The present application adopts Z-score standardization method, that is, the mean and standard deviation of each feature dimension of the original data matrix are normalized respectively: each value is subtracted from the arithmetic mean of the feature dimension, and divided by the standard deviation of the feature dimension. After this processing, each field block unit corresponds to a standardized feature vector, and the values of each feature dimension are converted into dimensionless standardized results with a mean of zero and a variance of one.

[0043] At this point, the field block unit standardized feature vector is obtained by standardizing the multi-source original data of the farmland.

[0044] Step S2, obtain a feature contribution balance factor by analyzing the imbalance degree of the feature dimension distance component.

[0045] The step aims to solve the single feature distance pollution problem existing in the prior art. The root cause of the problem is that when quantifying the multi-dimensional feature similarity between a field unit and a cluster center, the standard Euclidean distance equally weights the square sum of the differences in all feature dimensions, which fails to identify and adjust the balance of the contribution of each dimension to the total distance. In the actual operation scene of straw returning to the field, the data feature vector of the field unit is usually composed of two types of information: one is composed of multiple dimensions such as yield and soil conductivity, which can reflect the inherent agronomic properties of the field; the other is introduced by factors such as the overlapping of the harvester operation path, which presents a narrow-band high-amplitude form in the single dimension of straw density. When a field unit located in the straw abnormal accumulation zone and a cluster center located in the normal area perform distance calculation, only the huge difference in the straw density dimension will occupy a dominant proportion in the cumulative sum of the Euclidean distance. At this time, the result of distance calculation is no longer a true reflection of the overall similarity, but is polluted by a single operation feature. Therefore, the logic of the step is that a truly reliable similarity measure must be balanced in the composition of the final value. To this end, the application constructs a feature contribution balance factor. The factor analyzes the internal quantitative relationship between the single-dimensional distance components that constitute a certain specific distance calculation. It can quantify the degree to which the result of this distance calculation is dominated by a single or a few feature dimensions. When a serious imbalance in contribution is identified, i.e. distance pollution, the factor will generate a correction value significantly greater than 1 and act on the original distance calculation result. It is equivalent to adding a feature dominance analysis to each distance calculation. By punishing and enlarging the distance value dominated by a single operation feature, the difference between these special regions (such as high straw belt) and all other conventional agronomic regions in the feature space is significantly increased. This enhanced difference enables the FCM clustering algorithm to identify and separate these regions with significant features in operation load, forming clusters with specific management significance, rather than mistakenly mixing them with regions similar in agronomic background.

[0046] In summary, first, the cluster center is obtained in the clustering process, the distance contribution imbalance index is obtained by energy distribution analysis on the feature dimension difference data between the field unit and the cluster center. Specifically, for any target field unit in all field units and any target cluster center in all cluster centers, the square of the difference value of the target field unit and the target cluster center in any target feature dimension is taken as the distance energy component of the target feature dimension; the distance energy components of all feature dimensions are accumulated to obtain the total energy distribution evaluation; the fourth power of the difference value of the target field unit and the target cluster center in all target feature dimensions is accumulated to obtain the energy concentration evaluation; the calculation result of multiplying the number of feature dimensions and the energy concentration evaluation is taken as the numerator; the square of the total energy distribution evaluation is taken as the denominator, and the calculation result of subtracting the corresponding fraction from the constant 1 is taken as the distance contribution imbalance index.

[0047] After obtaining the distance contribution imbalance index, the feature contribution balance factor is obtained by further processing the distance contribution imbalance index through nonlinear response. Specifically, the gain coefficient of the imbalance response is set, and in the embodiment of the present application, the gain coefficient of the imbalance response is set to 1.5. The coefficient is used to adjust the sensitivity of the inverse tangent function to the imbalance degree. The greater the value of the gain coefficient, the more sensitive the feature contribution balance factor to the imbalance of the distance contribution. The gain coefficient can be adjusted according to the actual scene, and is not required; the gain coefficient amplification processing is performed on the distance contribution imbalance index to obtain the amplified imbalance response value; the inverse tangent nonlinear mapping processing is performed on the amplified imbalance response value, and the ratio of constant 2 to the constant π is normalized to obtain the smooth correction increment between zero and constant one; the calculation result of adding the smooth correction increment and the constant 1 is taken as the feature contribution balance factor.

[0048] In an embodiment, it is assumed that the feature value of the first field unit in the first feature dimension is ; the feature value of the first cluster center in the first dimension is ; the number of feature dimensions is ; the gain coefficient of the imbalance response is , then the calculation expression of the feature contribution balance factor between the first field unit and the first cluster center is:

[0049]

[0050] Wherein, represents the first field unit and the first The feature contribution balance factor between cluster centers; represents pi; represents the inverse tangent function; The gain factor representing the imbalance response; Indicates the number of feature dimensions; Indicates the The field unit is in The eigenvalues ​​on the feature dimensions; Indicates the The cluster center is in the The eigenvalues ​​in the dimensions.

[0051] It should be noted that the present invention has constructed a method to solve the problem of single feature distance pollution. This distance contribution imbalance index is constructed. This construction aims to establish a quantitative index that can reflect the distance component contributed by each feature dimension and its distribution in the overall distance structure. In the calculation of multidimensional distance, the square of each single-dimensional distance can be used to calculate the distance component contributed by each feature dimension. is considered as the energy contributed by this dimension, and is the total energy. This construction determines whether the energy distribution is over-concentrated by comparing the relationship between the sum of the fourth power of the energy in each dimension and the square of the total energy. Its working principle is that the numerical value of the core construction directly reflects the concentration of the energy distribution of the single-dimensional distance. In a conventional distance calculation, for example, when calculating the distance between two field units with similar agronomic properties, both located in the center of the field. Although each single-dimensional distance There are differences in the values ​​of , but their magnitudes are comparable after data normalization, and the energy distribution is relatively dispersed. In this case, the calculation result of the core structure will be a small positive value close to zero. This represents a distance composition state with a relatively balanced contribution. At this time, after the distance contribution imbalance index is mapped by the inverse tangent function, the value of the feature contribution balance factor will be close to 1, and only a negligible correction will be applied to the original distance, thereby ensuring the stability of the similarity measure in most conventional cases. However, when a single feature distance contamination phenomenon occurs, for example, when calculating the distance between a field unit located in a straw accumulation belt formed by overlapping harvester operations and a cluster center located in a normal area, the energy of only the straw density dimension is used. will exceed all other agronomic characteristic dimensions in magnitude, which leads to an extremely concentrated peak in the energy distribution. Since the amplification effect of the fourth power operation on large values ​​is much stronger than that of the second power, this will lead to The value of is dominated by the peak, making the calculation result of the core structure a significant positive number greater than zero. The size of this positive value accurately quantifies the degree of imbalance in the energy distribution in this distance calculation. Finally, by using the inverse tangent function to perform nonlinear mapping on the distance contribution imbalance index, for distance calculations with relatively balanced distance contribution imbalance index, the distance contribution imbalance index of the core structure is close to zero. After the inverse tangent function mapping, its correction effect is weak, ensuring that the feature contribution balance factor is close to , thus protecting the stability of conventional distance calculations. When the energy distribution shows a significant concentration, indicating that the distance calculation is contaminated, the value of the distance contribution imbalance index will increase, and the characteristic contribution balance factor will increase smoothly, applying a correction corresponding to the degree of imbalance to the distance. When the distance contribution imbalance index reaches an extremely high level, the saturation characteristic of the inverse tangent function will cause the growth of the characteristic contribution balance factor to slow down and saturate at a preset upper limit (in this embodiment, ). The bounded penalty mechanism ensures that contaminated distances are adequately but not excessively corrected, fundamentally avoiding algorithm instability issues that could result from excessive penalties. This allows the feature contribution balancing factor to adaptively amplify and correct contaminated distances, effectively suppressing distortion introduced by single operation traces during the FCM clustering iteration process and guiding the clustering results back to the correct partitioning, which is driven by multi-dimensional features and better reflects the inherent differences in the fields.

[0052] At this point, the feature contribution balance factor is obtained by performing imbalance analysis on the feature dimension distance component.

[0053] Step S3, obtaining a spatial membership smoothing factor by performing an activity evaluation on the neighborhood membership distribution of the field unit.

[0054] After completing step S2, by introducing the feature contribution balancing factor, the present application has been able to identify and separate the areas with significant work features (high straw load) into independent management partitions. However, the inherent defects of the FCM algorithm itself still exist. The defect is reflected in that when the algorithm calculates the membership of each field unit to each cluster center, its decision-making process occurs completely in the multi-dimensional feature space without analyzing the location of the field unit in the real geographical space and its neighborhood relationship. This spatial blind characteristic can lead to an isolated field unit (for example, a small saline-alkali spot surrounded by a large area of ordinary soil) that is similar in agronomic features to a distant cluster center being assigned a membership that is completely different from its geographical neighborhood. When such isolated points or elongated strips composed of them appear on the partition map, they form enclaves or jagged boundaries that are not continuous and irregular in geometric shape. These broken boundaries not only do not meet the actual needs of large-scale farmland management, but also bring unnecessary complex turns and inefficient movement for subsequent automatic path planning of agricultural machinery based on the partition map, reducing the overall work efficiency. Therefore, the final attribution of a field unit should not only be determined by its own features, but also be coordinated with its geographical spatial neighborhood environment. A reasonable management partition should present a smooth and continuous transition in space in terms of the membership of its internal members. To this end, the present application further constructs a spatial membership smoothing factor, which quantifies the harmony between the membership state of a field unit and the macro membership trend of its neighborhood, to make a secondary correction to the distance calculation. When the membership state of a field unit is inconsistent with its neighborhood environment, indicating that an irregular boundary shape may be formed, the factor will generate a correction value to increase the cost of such isolated allocation, thereby guiding the membership distribution to evolve towards a more spatially smooth and geometrically regular shape in the clustering iteration.

[0055] In summary, first, the neighborhood membership mean value data is obtained by statistical processing of the neighborhood membership data of the field unit. Specifically, the neighborhood range of the field unit is set, and for any target field unit in all field units, the neighborhood membership set of the target field unit is obtained through the neighborhood range of the field unit. In the clustering process, the membership values of all field units in the neighborhood membership set to the same target cluster center are accumulated, and the obtained accumulated value is divided by the number of field units in the neighborhood to obtain the neighborhood membership mean value data of the target field unit on the target cluster center.

[0056] After obtaining the mean neighborhood membership data of the target plot unit on the target cluster center, continue to perform variance analysis on the neighborhood membership data of the plot unit to obtain the neighborhood membership activity index. Specifically, within the neighborhood range centered on the target plot unit, obtain the membership values ​​of each plot unit on all cluster centers; take the membership values ​​of each plot unit in the neighborhood on all cluster centers as the overall sample, calculate the variance value of the sample, and use the variance value as the neighborhood membership activity index.

[0057] Finally, the spatial membership smoothing factor is obtained by adaptively ratio processing the deviation data of the field unit membership and the neighborhood mean. Specifically, the membership value of the target field unit on the target cluster center is obtained, and the difference between it and the neighborhood membership mean data is calculated; the difference is squared to obtain the deviation energy of the target field unit membership; the calculation result of adding the deviation energy to the neighborhood membership activity index is used as the neighborhood corrected membership fluctuation value; the spatial membership smoothing factor is obtained by performing ratio calculation on the neighborhood corrected membership fluctuation value and the neighborhood membership activity index.

[0058] In one embodiment, assuming that The neighborhood membership activity index of a field unit is ;No. The field unit is The membership degree of a cluster center is ;No. The field unit is in The mean data of neighborhood membership on the cluster center is ;No. The neighborhood set of a field unit is , then The field unit is The calculation expression of the spatial membership smoothing factor of a cluster center is:

[0059]

[0060] in, Indicates the The field unit is The spatial membership smoothing factor of each cluster center; Indicates the Neighborhood membership activity index of each field unit; Indicates the The field unit is The membership degree of each cluster center; Indicates the The field unit is in The mean neighborhood membership data on the cluster center.

[0061] It should be noted that the above formula does not penalize all points whose membership is different from the mean of their neighborhood, but introduces the membership activity of the neighborhood. As an adaptive adjustment benchmark, it can intelligently distinguish the fluctuation of membership. Its working mechanism is that the denominator in the formula Represents a point The basic membership of the neighborhood is chaotic level, and the molecule This means that the field unit is taken into consideration. After the membership state of the neighborhood is determined, the total membership chaos level of the neighborhood is calculated. Therefore, the membership smoothing factor of the entire space quantifies the field unit. The relative increment of the membership disorder of its neighbors. In an internal area of ​​a cluster with highly consistent membership, the membership activity of its neighbors is will be very small. In this stable environment, if a and its neighborhood mean The deviation term of a plot unit (i.e., an enclave) with significant differences The numerator will be much larger than the denominator. At this time, the spatial membership smoothing factor will obtain a value significantly greater than 1, which will impose a strong correction on the allocation behavior that destroys regional stability, thereby suppressing the formation of isolated points. On the contrary, in a fuzzy boundary area that is at the intersection of multiple clusters, the membership of each point in its neighborhood to different clusters is quite different, resulting in membership activity. In this environment, it is normal for the membership of a field unit to deviate from the mean of its neighborhood. The deviation term The impact on the final ratio is weakened accordingly, and the value of the spatial membership smoothing factor is maintained at a level close to 1. This makes the algorithm tolerant to this normal boundary transition and avoids the rigidity of the management zone boundary caused by excessive smoothing.

[0062] In summary, the spatial membership smoothing factor achieves intelligent smoothing of the spatial membership distribution through this adaptive adjustment mechanism. It accurately identifies and corrects membership assignments that may lead to irregular partitioning patterns, while also protecting the fuzzy boundaries of normal gradients. Ultimately, it guides the algorithm to generate optimized partitioning results with more regular and continuous geometric forms that better meet the needs of actual farmland management and agricultural machinery operations.

[0063] At this point, the spatial membership smoothing factor is obtained by evaluating the activity of the neighborhood membership distribution of the field unit.

[0064] Step S4, obtaining the adaptive distance metric by combining the feature contribution balancing factor and the spatial membership degree smoothing factor.

[0065] After obtaining the feature contribution balancing factor and the spatial membership degree smoothing factor, the adaptive distance metric is obtained by combining the feature contribution balancing factor and the spatial membership degree smoothing factor. Specifically, the original Euclidean distance between the target field unit and the target cluster center is obtained. The distance value based on the feature dimension distance component correction is obtained by multiplying the original Euclidean distance by the feature contribution balancing factor. The adaptive distance metric between the target field unit and the target cluster center is obtained by multiplying the distance value based on the feature dimension distance component correction by the spatial membership degree smoothing factor again.

[0066] After obtaining the adaptive distance metric between the target field unit and the target cluster center, the target cluster number of the FCM clustering process is set. In the embodiment of the present application, the target cluster number is set to 8. The target cluster number can be adjusted according to the fuzzy division coefficient and the division entropy in the actual scene, and is not required. The adaptive distance metric is used for the FCM clustering process and the clustering iteration is completed to obtain the corresponding FCM clustering result.

[0067] Step S5, obtaining the agricultural machinery path coordination scheduling instruction set by hardening the final membership matrix.

[0068] This step aims to convert the FCM clustering result obtained in step S4 into path coordination scheduling instructions that can guide agricultural operations. First, based on the membership matrix output after the FCM clustering iteration converges, the clustering membership of each field unit is hardened, i.e. the field unit is divided into the target management partition with the highest membership. After hardening, a geometrically continuous and boundary regular farmland management partition map is obtained. The partition map not only reflects the differences in yield, soil conductivity and terrain and other agronomic properties of farmland, but also identifies high-load areas formed by straw accumulation, providing spatial guidance for subsequent differentiated operations.

[0069] Subsequently, differentiated operation strategies are formulated according to the characteristic attributes of each management partition. For the identified high-straw-load area, a complex operation prescription containing additional fertilization measures and operation speed constraints is generated, for example, appropriate additional application of quick-acting nitrogen fertilizer in the fertilization operation, and the recommended low-speed running value of the agricultural machine in the area is set to ensure smooth completion of the operation in the field with thick straw coverage. For the conventional management area, the corresponding conventional fertilization prescription is generated according to its agronomic characteristics, maintaining normal operation speed and fertilization intensity. It should be noted that in the embodiment of the present application in the high-straw-load area, the running speed of the agricultural machine is set to 3-4 km / h, which is significantly lower than the operation speed of 6-7 km / h in the conventional management area. The setting of this speed can maintain the continuity of the operation while avoiding the shaking or blocking of the machine due to straw accumulation, thereby ensuring the smooth execution of the fertilization operation and straw returning operation.

[0070] Finally, the above-mentioned partitioned map and differentiated operation prescription are compiled into a path coordination scheduling instruction set that can be recognized by the automatic control system of the agricultural machine. During the travel of the agricultural machine along the preset operation path, its positioning system and control unit can determine the management partition to which the current operation position belongs in real time, and call the corresponding operation prescription according to the instruction set. When the agricultural machine enters the high-straw-load area, the automatic triggering of the fertilization amount adjustment and speed control is realized, achieving the linkage optimization of straw returning operation and fertilization operation, thereby improving the precision and efficiency of field operation and reducing the waste of resources caused by unreasonable path or mismatched operation strategy.

[0071] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for coordinating and optimizing the paths of corn straw return and fertilization operations, characterized in that: The method comprises: Step S1: Obtaining the standardized feature vector of the field unit by standardizing the multi-source original data of the farmland; Step S2: Obtaining a feature contribution balance factor by performing an imbalance analysis on the feature dimension distance component; Step S3: Obtaining the spatial membership smoothing factor by performing activity evaluation on the neighborhood membership distribution of the field unit; Step S4: Obtaining an adaptive distance metric by combining the feature contribution equalization factor and the spatial membership smoothing factor; Step S5: Obtain the agricultural machinery path collaborative scheduling instruction set by hardening the final membership matrix.

2. The method for optimizing the coordinated scheduling of paths for corn straw return to field and fertilization operations according to claim 1, characterized in that: The method of obtaining the standardized feature vector of the field unit by performing standardization processing on the original data of multiple sources of farmland includes: Grid sampling is performed on the target farmland area to obtain all field units in the target farmland area; historical yield data of the field units are obtained by a combine harvester equipped with a yield monitor; soil apparent conductivity data of the field units are obtained by a soil conductivity scanning device; digital elevation model data of the field units are obtained by airborne or satellite-borne remote sensing equipment; straw density data of the field units are obtained by using a multispectral or high-resolution visible light camera equipped with an unmanned aerial vehicle in combination with image analysis technology; the historical yield data, soil apparent conductivity data, digital elevation model data and straw density data are constructed into a multidimensional raw data matrix, and each characteristic dimension of the matrix is ​​subjected to Z-score normalization processing to obtain a standardized characteristic vector corresponding to each field unit.

3. The method for coordinated scheduling and optimization of paths for corn straw return to field and fertilization operations according to claim 1, characterized in that: The obtaining of the feature contribution balance factor by performing imbalance analysis on the feature dimension distance component includes: The cluster center is obtained in the clustering process; the distance contribution imbalance index is obtained by performing energy distribution analysis on the characteristic dimension difference data between the field unit and the cluster center; the characteristic contribution balance factor is obtained by performing nonlinear response processing on the distance contribution imbalance index.

4. The method for coordinated scheduling and optimization of paths for corn straw return to field and fertilization operations according to claim 3, characterized in that: The energy distribution analysis of the characteristic dimension difference data between the field unit and the cluster center is performed to obtain the distance contribution imbalance index, including: For any target field unit in all field units and any target cluster center in all cluster centers, the square of the difference between the target field unit and the target cluster center in any target feature dimension is used as the distance energy component of the target feature dimension; the distance energy components of all feature dimensions are accumulated to obtain the total energy distribution evaluation; the fourth power of the difference between the target field unit and the target cluster center in all target feature dimensions is accumulated to obtain the energy concentration evaluation; The result of multiplying the number of feature dimensions by the energy concentration evaluation is used as the numerator; the square of the total energy distribution evaluation is used as the denominator, and the result of subtracting the corresponding fraction from the constant 1 is used as the distance contribution imbalance indicator.

5. The method for coordinated scheduling and optimization of paths for corn straw return to field and fertilization operations according to claim 3, characterized in that: The method of performing nonlinear response processing on the distance contribution imbalance index to obtain the feature contribution balance factor includes: The gain coefficient of the imbalance response is set; the distance contribution imbalance index is amplified by the gain coefficient to obtain the amplified imbalance response value; the amplified imbalance response value is processed by inverse tangent nonlinear mapping and normalized by the ratio of constant 2 and pi to obtain a smooth correction increment between zero and constant 1; the result of adding the smooth correction increment and constant 1 is used as the characteristic contribution equalization factor.

6. The method for coordinated scheduling and optimization of paths for corn straw return to field and fertilization operations according to claim 1, characterized in that: The spatial membership smoothing factor is obtained by performing activity evaluation on the neighborhood membership distribution of the field unit, including: By statistically processing the neighborhood membership data of the field units, the neighborhood membership mean data is obtained; By performing variance analysis on the neighborhood membership data of the field units, the neighborhood membership activity index is obtained; The spatial membership smoothing factor is obtained by performing adaptive ratio processing on the field unit membership and neighborhood mean deviation data.

7. The method for coordinated scheduling and optimization of paths for corn straw return to field and fertilization operations according to claim 6, characterized in that: The method of performing statistical processing on the neighborhood membership data of the field units to obtain the neighborhood membership mean data includes: Set the neighborhood range of the field unit, and for any target field unit among all the field units, obtain the neighborhood membership set of the target field unit through the neighborhood range of the field unit; During the clustering process, the membership values ​​of all field units in the neighborhood membership set to the same target cluster center are accumulated, and the corresponding accumulated values ​​are divided by the number of field units in the neighborhood to obtain the neighborhood membership mean data of the target field unit on the target cluster center.

8. The method for coordinated scheduling and optimization of paths for corn straw return to field and fertilization operations according to claim 6, characterized in that: The neighborhood membership activity index is obtained by performing variance analysis on the neighborhood membership data of the field unit, including: In the neighborhood centered on the target field unit, the membership values ​​of each field unit on all cluster centers are obtained; the membership values ​​of each field unit in the neighborhood on all cluster centers are used as the overall sample, the variance value of the sample is calculated, and the variance value is used as the neighborhood membership activity index.

9. The method for coordinated scheduling and optimization of paths for corn straw return to field and fertilization operations according to claim 6, characterized in that: The method of obtaining a spatial membership smoothing factor by performing adaptive ratio processing on the field unit membership and the neighborhood mean deviation data includes: The membership value of the target plot unit on the target cluster center is obtained, and the difference between it and the neighborhood membership mean data is calculated; the difference is squared to obtain the deviation energy of the target plot unit membership; the calculation result of adding the deviation energy to the neighborhood membership activity index is used as the neighborhood corrected membership fluctuation value; the spatial membership smoothing factor is obtained by calculating the ratio of the neighborhood corrected membership fluctuation value to the neighborhood membership activity index.

10. The method for coordinated scheduling and optimization of paths for corn straw return to field and fertilization operations according to claim 1, characterized in that: The method of obtaining the adaptive distance metric by combining the feature contribution equalization factor and the spatial membership smoothing factor includes: The original Euclidean distance between the target field unit and the target cluster center is obtained; a distance value corrected based on the feature dimension distance component is obtained by multiplying the original Euclidean distance by the feature contribution equalization factor; and an adaptive distance metric between the target field unit and the target cluster center is obtained by multiplying the distance value corrected based on the feature dimension distance component by the spatial membership smoothing factor.

Images