Crop planting method and system based on big data

Abstract

The invention discloses a crop planting method and system based on big data, and relates to the technical field of agricultural planting. According to the method, soil and historical crop data of a to-be-planted area are collected, and a stress-resistant disturbance characteristic field is constructed by clustering and adding Laplacian noise, so that environmental stress is simulated and privacy is protected; calculating a reverse neighbor count and an environment weight based on a local neighborhood, further calculating a weighted normalized centroid offset value, and identifying an ecologically unstable grid by using a dual statistical threshold; constructing a comprehensive evaluation function to optimize a dominant planting kernel, expanding a stable clustering region by adopting a sequential region growing algorithm limited by ecological niche, and forcibly merging broken plaques to a dominant cluster; finally, a differentiated planting method is applied to the core planting area and the ecological buffer zone. According to the method, the micro-ecological heterogeneity of the farmland is effectively identified, the crushing of an operation plot is eliminated, and the crop yield and the ecological stability are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural planting technology, and in particular to crop planting methods and systems based on big data. Background Technology

[0002] Agriculture is the foundation of the national economy. With the deep penetration of new-generation information technologies such as the Internet of Things, cloud computing, and artificial intelligence, traditional agriculture is accelerating its transformation towards digitalization and intelligence. In modern crop planting and management, it has become an industry consensus to use big data technology to monitor and analyze the entire agricultural production process. By extensively collecting multi-source heterogeneous data such as meteorological and hydrological data, soil physicochemical properties, crop growth, and historical yields in the planting area, and using data mining algorithms to reveal the coupling relationship between crop growth and environmental factors, it is possible to assist agricultural practitioners in formulating scientific strategies for sowing, fertilization, irrigation, and pest and disease control. This is of great practical significance for improving the efficiency of agricultural resource utilization, ensuring food security, and enhancing the market competitiveness of agricultural products.

[0003] To achieve the above goals, existing agricultural data analysis technologies have begun to utilize historical data models to guide current planting decisions. Chinese invention patent CN117033810B discloses an agricultural data analysis management system and method based on big data. This technology primarily acquires historical agricultural data, analyzes the average yield differences of the same crop in different regions, and combines this with the specific impact of temperature changes, pest and disease occurrence, and planting methods on yield to match the optimal planting conditions for the current location. However, existing technologies mostly focus on horizontal comparisons of yield and environmental averages across different locations, lacking a refined identification of the spatial heterogeneity of the soil ecological environment within a plot. Farmland soil properties are often not uniformly distributed, but rather... Complex spatial variations exist, such as significant differences in ecological stability between soil abrupt change zones and adjacent ditch areas. Existing technologies, when utilizing environmental data, mostly rely directly on correlation analysis based on raw observations, lacking robustness verification of data under simulated environmental stresses. They also fail to fully consider privacy protection and feature resilience construction during multi-source data fusion. Furthermore, in delineating planting zones, existing technologies mostly rely on fixed physical boundaries or simple threshold cutting, lacking a mechanism that can automatically optimize regions and dynamically grow based on ecological feature similarity and spatial compactness. This makes it difficult to form contiguous, regular, and uniform standardized operating units, limiting the implementation effect of large-scale automated agricultural machinery operations and the depth of refined management. Summary of the Invention

[0004] The technical problem solved by this invention is that existing technologies lack refined identification of the spatial heterogeneity of soil ecological environment within a plot, lack robustness verification of data under simulated environmental stress, and do not fully consider privacy protection and feature resilience construction in the process of multi-source data fusion. In addition, existing technologies also lack a mechanism that can automatically perform regional optimization and dynamic growth based on ecological feature similarity and spatial compactness, making it difficult to form contiguous, regular and uniform standardized operation units, which limits the implementation effect of large-scale agricultural machinery automation operations and the depth of refined management.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a crop planting method based on big data, comprising the following steps: Step S1: Collect soil data and historical crop data for the area to be planted; Step S2: Perform cluster analysis on the soil data and historical crop data to generate an anti-disturbance feature field, and calculate the reverse nearest neighbor count, environmental weight and local geometric centroid of the operation grid in the planting area based on the local spatial neighborhood. Step S3: Combine local geometric centroid, environmental weight, and reverse nearest neighbor count to calculate the weighted normalized centroid offset value of each work grid. Based on the statistical distribution of the weighted normalized centroid offset value and the reverse nearest neighbor count, set a judgment threshold to identify ecologically unstable work grids. Step S4: Construct a comprehensive evaluation function and select the dominant planting kernels. Use the region growing algorithm to expand the dominant planting kernels into stable clustering regions, and configure differentiated planting methods for the stable clustering regions and the ecologically unstable operation grids respectively.

[0006] In step S1, the area to be planted is divided into multiple work grids, and soil data and historical crop data are collected in each work grid. The soil data includes soil organic matter, pH, and electrical conductivity. The historical crop data includes the plant height and leaf area index of historical crops; Step S1 also involves collecting the spatial coordinates of the work grid; Soil data, historical crop data, and spatial coordinates were normalized.

[0007] In step S2, the process of generating the anti-reverse disturbance characteristic field specifically includes: Soil data and historical crop data in each work grid within the area to be planted are concatenated into a feature vector, and a clustering algorithm is used to generate several soil feature center vectors from the feature vector; Laplace noise is added to the soil feature center vector to generate an anti-disturbance feature field.

[0008] The process of calculating the reverse nearest neighbor count of the job grid specifically includes: Select any work grid within the planting area as the target grid, denote other work grids within the local spatial neighborhood of the target grid as the first neighborhood grid set, and denote other work grids within the local spatial neighborhood of the first neighborhood grid set as the second candidate grid set; Calculate the Euclidean distance between the feature vectors of the first neighboring grid set and the feature vectors of the second candidate grid set. Sort the second candidate grid set in ascending order based on the Euclidean distance, and select several second candidate grid sets with the smallest Euclidean distance to form a neighbor set. The total frequency of the target grid appearing in all neighbor sets is counted, and the total frequency is the reverse nearest neighbor count of the target grid; The environmental weight of the target grid is inversely proportional to the reverse nearest neighbor count of the target grid; The method for calculating the local geometric centroid of the target mesh is as follows: calculate the mean vector of the eigenvectors of the target mesh and the eigenvectors of the first neighboring mesh set, and use it as the local geometric centroid of the target mesh.

[0009] The process of calculating the weighted normalized centroid offset value specifically includes: Calculate the Euclidean distance between the eigenvectors of the target mesh and the local geometric centroid to obtain the original centroid offset of the target mesh; The mean of the Euclidean distance between the feature vector of the target grid and the feature vector of the first neighboring grid set is used as a normalization factor. The original centroid offset is normalized using the normalization factor to obtain the normalized centroid offset value. The normalized centroid offset value is corrected by using reverse nearest neighbor counting to obtain the weighted normalized centroid offset value of the target mesh.

[0010] The process of identifying the ecologically unstable work grid specifically includes: Calculate the average and standard deviation of the weighted normalized centroid offset values ​​of all work grids within the planting area, and use the average and standard deviation to construct an offset determination threshold; Based on the statistical distribution of reverse nearest neighbor counting, an isolation degree threshold is set. Traverse each work grid within the planting area. If the weighted normalized centroid offset value of the work grid is greater than the offset determination threshold or the reverse nearest neighbor count is less than the isolation determination threshold, then the work grid is determined to be an ecologically unstable work grid.

[0011] The process of constructing the comprehensive evaluation function and selecting the dominant planting kernels specifically includes: From the work grids in the area to be planted, remove the work grids marked as ecologically unstable, and form a candidate set of the remaining work grids; Calculate the Euclidean distance between the feature vector of each job grid in the candidate set and the soil feature center vector in the anti-disturbance feature field, and take the minimum value of the Euclidean distance as the feature distance component; A comprehensive evaluation function is constructed based on the aforementioned feature distance components. The mathematical expression of the comprehensive evaluation function is as follows: ; in, For the job grid Overall score For the job grid eigenvectors, The soil feature center vector in the anti-disturbance feature field. The index of the soil feature center vector. and These are the spatial coordinates of the grid. For the job grid A set of job grids within a local spatial neighborhood. and The weighting coefficients for balancing features and space; The job grids in the candidate set are sorted according to the comprehensive score, and the job grids with the smallest comprehensive score are selected as the dominant planting cores.

[0012] The region growing algorithm specifically includes: The dominant planting core is used as the seed region for regional growth and iteratively grows to the adjacent operation grid; Calculate the difference in growth resistance between adjacent work grids and the current seed region. If the difference in growth resistance is less than a preset growth tolerance threshold and the adjacent work grid is not marked as an ecologically unstable work grid, then merge the adjacent work grid with the current seed region and continue regional growth using the adjacent work grid as the seed region until growth converges. The area of ​​each connected region formed after growth convergence is counted. If the area of ​​the connected region is less than the preset minimum operation granularity threshold, the connected region is determined to be a broken patch. If the area of ​​the connected region is greater than or equal to the preset minimum task granularity threshold, the connected region is determined to be the dominant cluster. Calculate the Euclidean distance between the feature vector of the dominant planting kernel corresponding to the fragmented patch and the feature vector of the dominant planting kernel corresponding to each dominant cluster. Merge the fragmented patches into the dominant cluster with the smallest Euclidean distance, update the area covered by the dominant cluster, and the area covered by all dominant clusters constitutes a stable clustering region. The growth resistance difference is the absolute value of the difference between the weighted normalized centroid offset of the adjacent working grid and the weighted normalized centroid offset of the seed region.

[0013] The specific configuration of the differentiated planting method includes: The stable clustering areas were identified as core planting areas, and high-density planting and high-fertility input methods were adopted. Ecologically unstable work grids are identified as ecological buffer zones, and low-density planting and conservation tillage methods are implemented accordingly.

[0014] A big data-based crop cultivation system includes a data collection module, a classification module, an identification module, and a planting module. The data acquisition module is used to collect soil data and historical crop data of the area to be planted. The classification module is used to perform cluster analysis on the soil data and historical crop data, generate a feature field resistant to adverse disturbances, and calculate the reverse nearest neighbor count, environmental weight, and local geometric centroid of the operation grid in the planting area based on the local spatial neighborhood. The identification module is used to calculate the weighted normalized centroid offset value of each work grid by combining the local geometric centroid, environmental weight, and reverse nearest neighbor count, and to set a judgment threshold based on the statistical distribution of the weighted normalized centroid offset value and the reverse nearest neighbor count to identify ecologically unstable work grids. The planting module is used to construct a comprehensive evaluation function and select advantageous planting kernels. It uses a region growth algorithm to expand the advantageous planting kernels into stable clustering regions and configures differentiated planting methods for stable clustering regions and ecologically unstable operation grids respectively.

[0015] The beneficial effects of this invention are as follows: Existing technologies often struggle to accurately distinguish between gradual and abrupt transition zones in soil properties. However, this invention utilizes reverse nearest neighbor counting to characterize the centrality of the work grid, enabling it to keenly capture, from a topological perspective, those ecologically adjacent but different characteristic zones. It quantifies the degree to which each work grid deviates from the local ecological core, thereby accurately identifying and isolating ecologically unstable growth barrier zones. The stress-resistant perturbation feature field constructed in this invention, by superimposing Laplace noise on the soil feature center vector, not only protects the privacy of farmers' sensitive data but also innovatively introduces a simulation mechanism for environmental stress. By constructing a comprehensive evaluation function that includes feature distance and spatial compactness, it selects advantageous planting cores that are both stress-resistant and contiguous as seed areas and allows them to grow competitively according to priority. For the small, fragmented patches formed after growth, this invention uses Euclidean distance to forcibly merge them into the most similar dominant cluster, eliminating small, fragmented plots that are difficult for agricultural machinery to operate on. Attached Figure Description

[0016] Figure 1This is a flowchart illustrating the steps of a big data-based crop cultivation method, as provided in one embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example 1, referring to Figure 1 This paper provides a big data-based crop cultivation method, which includes the following steps: Step S1: Collect soil data and historical crop data for the area to be planted; Step S2: Perform cluster analysis on soil data and historical crop data to generate a resilient perturbation feature field, and calculate the reverse nearest neighbor count, environmental weight, and local geometric centroid of the operation grid in the planting area based on the local spatial neighborhood. Step S3: Combine local geometric centroid, environmental weight, and reverse nearest neighbor count to calculate the weighted normalized centroid offset value of each work grid. Based on the statistical distribution of the weighted normalized centroid offset value and the reverse nearest neighbor count, set a judgment threshold to identify ecologically unstable work grids. Step S4: Construct a comprehensive evaluation function and select the dominant planting kernels. Use the region growing algorithm to expand the dominant planting kernels into stable clustering regions, and configure differentiated planting methods for the stable clustering regions and the ecologically unstable operation grids respectively.

[0019] This invention introduces a stability determination mechanism based on topology, which can accurately identify ecological ecotones where soil properties undergo abrupt changes and core stable advantageous regions from massive operational grids. By first locking unstable growth barrier zones as logical isolation boundaries and then using a region growth algorithm to expand the core area, this invention achieves a refined deconstruction of the spatial heterogeneity within farmland, providing a solid decision-making basis for the subsequent implementation of differentiated planting methods and effectively improving the agricultural input-output ratio and ecological sustainability.

[0020] In step S1, the area to be planted is divided into multiple work grids, and soil data and historical crop data are collected in each work grid. Soil data includes soil organic matter, pH, and electrical conductivity; Historical crop data includes plant height and leaf area index of historical crops; Step S1 also involves collecting the spatial coordinates of the work grid; Soil data, historical crop data, and spatial coordinates were normalized.

[0021] In a specific embodiment of the present invention, the area to be planted is divided into several rows and several columns of grid units. Based on geographic information GIS software, the plots of land to be planted are automatically divided into grids, and the latitude and longitude spatial coordinates of the center point of each grid are automatically calculated. In traditional agricultural management, the entire plot of land is often treated as a whole for uniform fertilization and irrigation. However, the soil properties within the land are spatially heterogeneous. By using gridding to discretize the continuous land into independent operational units, the algorithm can capture subtle changes in the soil within a small area.

[0022] In step S2, the process of generating the anti-inverse perturbation characteristic field specifically includes: Soil data and historical crop data in each work grid within the area to be planted are concatenated into a soil feature vector. Then, a clustering algorithm (K-means++ algorithm) is used to generate several soil feature center vectors from the soil feature vector. Laplace noise is added to the soil feature center vector to generate an anti-perturbation feature field.

[0023] Each soil feature center vector represents the average characteristic of a typical plot, and the number of soil feature center vectors is determined by the elbow rule to determine the optimal number of categories.

[0024] The generation of a robust perturbation characteristic field specifically includes: Calculate the fluctuation range of soil data and historical crop data for each dimension, including organic matter, pH and electrical conductivity. The fluctuation range is the difference between the maximum and minimum values ​​of soil data and historical crop data for each dimension. The scale parameter of the Laplace noise is determined. The scale parameter is derived by multiplying the preset sensitivity coefficient and the fluctuation range. Each dimension is configured with an independently calculated scale parameter. The sensitivity coefficient is an adjustable dimensionless parameter. Its value is preset according to the intensity requirement of the environmental stress simulation of the planting area and the privacy protection level. It is used to achieve a balance between maintaining the data distribution characteristics and introducing random disturbances. It is set to be large when strong noise is needed and small when weak noise is needed. In this embodiment, the optimal balance of the sensitivity coefficient is 0.05. By using the probability density function of the Laplace distribution, a noise distribution for each dimension is constructed based on the scale parameter of each dimension. Independent Laplace random noise values ​​generated by the features of each dimension are obtained. The location parameter is configured to have zero mean to ensure that the introduced perturbation does not cause a shift in the overall distribution of the original data. Noise injection is performed by element-wise superposition. The generated Laplace random noise values ​​are accumulated onto the corresponding dimensional components of the soil feature center vector to obtain the constructed anti-disturbance feature field. The anti-disturbance feature field, as the final generated set of noisy feature center vectors, is presented in the data structure as a superposition of the original agronomic attributes and the random disturbances of simulated environmental stress. The anti-disturbance feature field retains the core agronomic attributes of the area to be planted and contains the random disturbances of simulated environmental stress. It will then be sent to the central server to participate in global calculation.

[0025] To verify the stress resistance and stability of the planting area in subsequent steps and to protect farmers' privacy, Laplace noise was added to the soil feature center vector.

[0026] Traditional cluster centers only represent the average level, while this invention artificially simulates environmental stresses in nature, such as sudden climate change, by calculating the fluctuation range of each dimension and superimposing Laplace noise that conforms to the probability distribution. This makes the final generated feature center vector more robust. Using this as a standard for calculation means that the selected core area not only performs well in normal years, but also maintains feature stability under simulated harsh environments. At the same time, under the federated computing architecture, the introduction of noise effectively masks the precise values ​​of the original sampling points, which can protect farmers' production privacy while realizing joint modeling of multi-plot data.

[0027] The process of calculating the reverse nearest neighbor count of the job grid specifically includes: Select any work grid within the planting area as the target grid. Based on the calculated range of the local spatial neighborhood, select the remaining work grids (excluding itself) within the local spatial neighborhood of the target grid and denote them as the first neighborhood grid set. The remaining job grids within the local spatial neighborhood of each of the first neighboring grids included in the first neighboring grid set are selected and denoted as the second candidate grid set (including the target grid) corresponding to each first neighboring grid. Calculate the Euclidean distance between each soil feature vector corresponding to the first neighborhood grid set and each soil feature vector corresponding to the second candidate grid set. Sort the second candidate grid sets in ascending order based on the Euclidean distance. Select the second candidate grid sets whose calculated Euclidean distances are less than a preset Euclidean distance threshold (the value range in this embodiment is 0.5~0.8) corresponding to each first neighborhood grid set to form the neighbor set corresponding to each first neighborhood grid set. The total frequency of the target grid appearing in all neighbor sets is counted, and the total frequency is the reverse nearest neighbor count of the target grid. The environmental weight of the target grid is calculated based on the reverse nearest neighbor count. The environmental weight is obtained by calculating the reverse nearest neighbor count using an inverse proportional function. The mathematical expression of the inverse proportional function is: ; in, To determine the weight, For reverse nearest neighbor counting; The method for calculating the local geometric centroid of the target grid is as follows: calculate the mean vector of the soil feature vector and the feature vector of the first neighboring grid set in the target grid, and use it as the local geometric centroid of the target grid.

[0028] Calculating the range of a local spatial neighborhood specifically includes: The grid whose physical spatial coordinates of the target task grid are within a preset search radius (50m in this embodiment) is defined as a local spatial neighborhood; In this embodiment, when the reverse nearest neighbor count is low, the calculated judgment weight is high, which indicates that the job grid is isolated or marginal in the feature space and is identified as a potential growth barrier zone or ecologically unstable area, implying that there is significant spatial heterogeneity in soil properties. When the reverse nearest neighbor count is high, the judgment weight is low, indicating that the operation grid is the core cluster point in the feature space, and the soil properties are uniform and stable.

[0029] This invention utilizes reverse nearest neighbor counting to calculate environmental weights, which can accurately identify outliers and core points in the data space. By converting reverse nearest neighbor counting into environmental weights that are inversely proportional, it can automatically assign higher risk weights to grids located at the edge or in transitional zones, i.e., grids with low reverse nearest neighbor counts. This amplifies the characteristic signals of the ecotone at the mathematical level and identifies hidden barrier zones that are physically adjacent but have different ecological properties.

[0030] The process of calculating the weighted normalized centroid offset value specifically includes: Calculate the Euclidean distance between the eigenvectors of the target mesh and the local geometric centroid to obtain the original centroid offset of the target mesh; The mean of the Euclidean distance between the feature vector of the target grid and the feature vector of the first neighboring grid set is used as a normalization factor. The original centroid offset is normalized by the normalization factor (the original centroid offset is divided by the normalization factor) to obtain the normalized centroid offset value. The normalized centroid offset value is nonlinearly corrected by reverse nearest neighbor counting to obtain the weighted normalized centroid offset value of the target mesh.

[0031] In a specific embodiment of the present invention, the mathematical expression for calculating the weighted normalized centroid offset value is as follows: ; in, For the job grid The weighted normalized centroid offset of the target mesh. For the job grid The normalized centroid offset value, For the job grid Reverse nearest neighbor counting, It is the maximum value among the reverse nearest neighbor counts of all job grids within the area to be planted.

[0032] This invention eliminates calculation biases caused by different sparsity of data in different regions by introducing the mean of local feature distance as a normalization factor, and realizes a unified standard for measuring offset under different geological conditions. It uses reverse nearest neighbor counting to perform secondary correction on the normalized offset value, mathematically suppressing the small fluctuations in the core area while retaining the significant bias in the edge area. This dual weighting mechanism greatly reduces the misjudgment rate of the core planting area.

[0033] The process of identifying ecologically unstable work grids specifically includes: The weighted normalized centroid offset values ​​of all work grids in the planting area are statistically analyzed, and their arithmetic mean and standard deviation are calculated. The offset judgment threshold is constructed using the mean and standard deviation. In this embodiment, the offset judgment threshold is set as the sum of the mean and twice the standard deviation. Based on the statistical distribution of reverse nearest neighbor counts, an isolation degree threshold is set. In this embodiment, the reverse nearest neighbor counts are sorted in ascending order, and the reverse nearest neighbor counts corresponding to the top 5% of positions are selected as the isolation degree threshold. Traverse each work grid within the planting area. If the weighted normalized centroid offset value of the work grid is greater than the offset judgment threshold or the reverse nearest neighbor count is less than the isolation judgment threshold, then the work grid is determined to be an ecologically unstable work grid.

[0034] Unlike traditional methods that use fixed constants as thresholds, this invention utilizes the weighted sum of the mean and standard deviation, as well as quantile statistics, to dynamically adjust the judgment criteria based on data fluctuations in different plots. This results in greater universality. By using OR logic judgment based on excessive centroid shift or reverse nearest neighbor counting, it can screen out mutation points whose characteristics significantly deviate from the population, as well as isolated points whose characteristics are not extreme but lack similar support. This ensures that the final determined growth barrier zone covers all potential ecological risk areas.

[0035] The process of constructing a comprehensive evaluation function and selecting dominant planting kernels specifically includes: From the work grids in the area to be planted, remove the work grids marked as ecologically unstable, and form a candidate set of the remaining work grids; Calculate the Euclidean distance between the feature vector of each job grid in the candidate set and the soil feature center vector in the anti-disturbance feature field, and take the minimum value of the Euclidean distance as the feature distance component; A comprehensive evaluation function is constructed based on the feature distance components. The mathematical expression of the comprehensive evaluation function is as follows: ; in, For the job grid Overall score For the job grid eigenvectors, The soil feature center vector in the anti-disturbance feature field. The index of the soil feature center vector. and These are the spatial coordinates of the grid. For the job grid A set of job grids within a local spatial neighborhood. and The weighting coefficients for balancing features and space (in this embodiment) Take 0.7, Take 0.3); Based on the comprehensive score, the job grids in the candidate set are sorted in ascending order, and the top 15% of job grids in the comprehensive score are selected as the dominant planting cores; After removing obstacle zones, this invention not only requires candidate grids to be highly matched with the federal stress resistance standard in terms of characteristics, that is, to have stress resistance and high yield potential, but also requires them to be located in the center of a contiguous area through spatial density components. This screening mechanism avoids selecting those plots that are scattered but have good soil quality as dominant planting cores, laying the best foundation for the subsequent generation of regular and contiguous suitable operation areas.

[0036] The specific algorithms for region growing include: The dominant planting core is used as the seed region for regional growth and iteratively grows to the adjacent operation grid; Calculate the difference in growth resistance between adjacent work grids and the current seed region. If the difference in growth resistance is less than the preset growth tolerance threshold and the adjacent work grid is not marked as an ecologically unstable work grid, then merge the adjacent work grid with the current seed region and continue the regional growth with the adjacent work grid as the seed region until the growth converges. The area of ​​each connected region formed after growth convergence is counted. If the area of ​​the connected region is less than the preset minimum operation particle size threshold (in this embodiment, the value is 80m, which corresponds to the minimum operation unit of agricultural machinery), then the connected region is determined to be a broken patch. If the area of ​​the connected region is greater than or equal to the preset minimum task granularity threshold, the connected region is determined to be the dominant cluster. Calculate the Euclidean distance between the feature vector of the dominant planting kernel corresponding to the fragmented patch and the feature vector of the dominant planting kernel corresponding to each dominant cluster. Merge the fragmented patches into the dominant cluster with the smallest Euclidean distance, update the area covered by the dominant cluster, and the area covered by all dominant clusters constitutes a stable clustering region. The growth resistance difference is the absolute value of the difference between the weighted normalized centroid offset of adjacent work grids and the weighted normalized centroid offset of the seed region. When initial growth begins with a dominant planting core, this growth resistance difference is the absolute value of the difference between the weighted normalized centroid offset of adjacent work grids and the weighted normalized centroid offset of the dominant planting core. During iterative growth, the growth resistance difference is the absolute value of the difference between the currently growing work grid and the weighted normalized centroid offset of its adjacent work grids. In a specific embodiment of the present invention, the advantageous planting kernels are sorted by priority according to the comprehensive evaluation function value, and the job grid with the smallest comprehensive score has the highest priority. The job grid with the highest priority is selected as the initial seed region. Examine the eight adjacent job grids surrounding the seed region and calculate the difference in growth resistance between the adjacent job grids and the current seed region; The growth resistance difference is the absolute value of the difference between the weighted normalized centroid offset value of the adjacent working grid and the weighted normalized centroid offset value of the seed region; When the initial growth begins with a dominant planting core, the growth resistance difference is the absolute value of the difference between the adjacent working grid and the dominant planting core. During the iterative growth process, it is the absolute value of the difference between the current growth front grid and its adjacent grids. If the growth resistance difference is less than the preset growth tolerance threshold (0.2 in this embodiment), and the adjacent work grid is not marked as an ecologically unstable work grid, then the adjacent work grid is merged with the current seed region, and the adjacent work grid is marked as occupied, and iterative growth continues as a new growth front. If the growth resistance difference is greater than or equal to the preset growth tolerance threshold, or the adjacent work grid is an ecologically unstable work grid, then growth in that direction is stopped. The area of ​​each connected region formed after growth convergence is counted. If the area of ​​the connected region is less than the preset minimum operation particle size threshold (80m in this embodiment, corresponding to the minimum operation unit of agricultural machinery), then the connected region is determined to be a broken patch. If the area of ​​a connected region is greater than or equal to the preset minimum task granularity threshold, then the connected region is determined to be the dominant cluster. If the entire plot is very fragmented, resulting in no connected area being larger than the minimum operation granularity threshold, then the fragmented patch with the largest area will be upgraded to the dominant cluster or an error message will be issued. Calculate the Euclidean distance between the feature vectors of the dominant planting kernels corresponding to the fragmented patches and the feature vectors of the dominant planting kernels corresponding to each dominant cluster. Merge the fragmented patches into the dominant cluster with the smallest Euclidean distance, update the area covered by the dominant cluster, and the area covered by all dominant clusters constitutes a stable clustering region.

[0037] This invention prioritizes the expansion of the highest-scoring dominant cores through priority ranking, ensuring that the best soil resources dominate the land parcel division pattern. During the growth process, it strictly avoids ecologically unstable areas, preventing the core area from blindly encroaching on inferior land. In addition, for fragmented patches formed after growth with an area smaller than the minimum operational particle size, they are forcibly merged into the most similar dominant clusters based on characteristic Euclidean distance, eliminating fragmented plots that agricultural machinery cannot operate on. Under the premise of ensuring ecological rationality, it maximizes land utilization and mechanized operation efficiency.

[0038] The specific methods for configuring differentiated planting include: The stable clustering areas were identified as core planting areas, and high-density planting and high-fertility input methods were adopted. Ecologically unstable work grids are identified as ecological buffer zones, and low-density planting and conservation tillage methods are implemented accordingly.

[0039] This invention defines stable clustering areas as core planting areas and implements high-density, high-fertility strategies, which can fully tap the yield potential of high-quality soil and maximize resource utilization efficiency. Ecologically unstable areas are defined as buffer zones and protected tillage is implemented, which not only avoids the loss and waste of chemical fertilizers and pesticides in soils with low retention capacity, but also plays a role in conserving water sources and preventing soil erosion. This site-specific and precise management strategy transforms digital analysis results into direct productivity, reflecting the core value of precision agriculture.

[0040] A big data-based crop cultivation system includes a data collection module, a classification module, an identification module, and a planting module. The data acquisition module is used to collect soil data and historical crop data for the area to be planted. The classification module is used to perform cluster analysis on soil data and historical crop data, generate a feature field that is resistant to adverse disturbances, and calculate the reverse nearest neighbor count, environmental weight and local geometric centroid of the operation grid in the area to be planted based on the local spatial neighborhood. The identification module is used to calculate the weighted normalized centroid offset value of each work grid by combining the local geometric centroid, environmental weight, and reverse nearest neighbor count. Based on the statistical distribution of the weighted normalized centroid offset value and the reverse nearest neighbor count, a judgment threshold is set to identify ecologically unstable work grids. The planting module is used to construct a comprehensive evaluation function and select the dominant planting kernels. It uses a region growing algorithm to expand the dominant planting kernels into stable clustering regions and configures differentiated planting methods for stable clustering regions and ecologically unstable operation grids respectively.

[0041] Existing technologies often struggle to accurately distinguish between gradual and abrupt transition zones in soil properties. However, this invention utilizes reverse nearest neighbor counting to characterize the centrality of the work grid. This allows for the keen capture of ecologically adjacent but distinct characteristic zones from a topological perspective, such as the boundary between sandy and clay soils and the edges of ditches. By combining the centroid offset value corrected by environmental weights, the invention quantifies the degree to which each work grid deviates from the local ecological core. This precisely identifies and isolates ecologically unstable growth barrier zones, preventing the mixed management of good and bad lands due to blind zoning. This provides solid micro-data support for subsequent differentiated planting. The stress-resistant feature field constructed in this invention, by superimposing Laplace noise on the soil feature center vector, not only protects the privacy of farmers' sensitive data, but also innovatively introduces a simulation mechanism of environmental stress. This invention simulates random environmental disturbances in nature through Laplace noise, making the generated feature center vector more robust. The superior planting core selected based on this standard not only has excellent soil fertility and crop growth characteristics, but also has stress resistance stability under simulated harsh environments, ensuring that the selected planting core area has the potential for long-term high yield. This invention constructs a comprehensive evaluation function that includes characteristic distance and spatial compactness. It selects advantageous planting cores that are both resilient and contiguous as seed areas and promotes competitive growth based on priority. It strictly prohibits the growth process from crossing ecologically unstable areas, thereby ensuring a high degree of uniformity of properties within the delineated core planting areas. At the same time, for the small and fragmented patches formed after growth, this invention forces them to be merged into the most similar dominant clusters based on Euclidean distance, eliminating small and fragmented plots that are difficult for agricultural machinery to operate on. Under the premise of ensuring ecological rationality, it maximizes land utilization and mechanized operation efficiency.

[0042] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0043] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A crop cultivation method based on big data, characterized in that, Includes the following steps: Step S1: Collect soil data and historical crop data for the area to be planted; Step S2: Perform cluster analysis on the soil data and historical crop data to generate an anti-disturbance feature field, and calculate the reverse nearest neighbor count, environmental weight and local geometric centroid of the operation grid in the planting area based on the local spatial neighborhood. Step S3: Combine local geometric centroid, environmental weight, and reverse nearest neighbor count to calculate the weighted normalized centroid offset value of each work grid. Based on the statistical distribution of the weighted normalized centroid offset value and the reverse nearest neighbor count, set a judgment threshold to identify ecologically unstable work grids. Step S4: Construct a comprehensive evaluation function and select the dominant planting kernels. Use the region growing algorithm to expand the dominant planting kernels into stable clustering regions, and configure differentiated planting methods for the stable clustering regions and the ecologically unstable operation grids respectively.

2. The crop cultivation method based on big data as described in claim 1, characterized in that, In step S1, the area to be planted is divided into multiple work grids, and soil data and historical crop data are collected in each work grid. The soil data includes soil organic matter, pH, and electrical conductivity. The historical crop data includes the plant height and leaf area index of historical crops; Step S1 also involves collecting the spatial coordinates of the work grid; Soil data, historical crop data, and spatial coordinates were normalized.

3. The crop cultivation method based on big data as described in claim 2, characterized in that, In step S2, the process of generating the anti-reverse disturbance characteristic field specifically includes: Soil data and historical crop data in each work grid within the area to be planted are concatenated into a feature vector, and a clustering algorithm is used to generate several soil feature center vectors from the feature vector; Laplace noise is added to the soil feature center vector to generate an anti-disturbance feature field.

4. The crop cultivation method based on big data as described in claim 3, characterized in that, The process of calculating the reverse nearest neighbor count of the job grid specifically includes: Select any work grid within the planting area as the target grid, denote other work grids within the local spatial neighborhood of the target grid as the first neighborhood grid set, and denote other work grids within the local spatial neighborhood of the first neighborhood grid set as the second candidate grid set; Calculate the Euclidean distance between the feature vectors of the first neighboring grid set and the feature vectors of the second candidate grid set. Sort the second candidate grid sets in ascending order based on the Euclidean distance. Select the k second candidate grid sets with the smallest calculated Euclidean distances corresponding to each first neighboring grid set to form the neighbor set corresponding to each first neighboring grid set. The total frequency of the target grid appearing in all neighbor sets is counted, and the total frequency is the reverse nearest neighbor count of the target grid; The environmental weight of the target grid is inversely proportional to the reverse nearest neighbor count of the target grid; The method for calculating the local geometric centroid of the target mesh is as follows: calculate the mean vector of the eigenvectors of the target mesh and the eigenvectors of the first neighboring mesh set, and use it as the local geometric centroid of the target mesh.

5. The crop cultivation method based on big data as described in claim 4, characterized in that, The process of calculating the weighted normalized centroid offset value specifically includes: Calculate the Euclidean distance between the eigenvectors of the target mesh and the local geometric centroid to obtain the original centroid offset of the target mesh; The mean of the Euclidean distance between the feature vector of the target grid and the feature vector of the first neighboring grid set is used as a normalization factor. The original centroid offset is normalized using the normalization factor to obtain the normalized centroid offset value. The normalized centroid offset value is corrected by using reverse nearest neighbor counting to obtain the weighted normalized centroid offset value of the target mesh.

6. The crop cultivation method based on big data as described in claim 5, characterized in that, The process of identifying the ecologically unstable work grid specifically includes: Calculate the average and standard deviation of the weighted normalized centroid offset values ​​of all work grids within the planting area, and use the average and standard deviation to construct an offset determination threshold; Based on the statistical distribution of reverse nearest neighbor counting, an isolation degree threshold is set. Traverse each work grid within the planting area. If the weighted normalized centroid offset value of the work grid is greater than the offset determination threshold or the reverse nearest neighbor count is less than the isolation determination threshold, then the work grid is determined to be an ecologically unstable work grid.

7. The crop cultivation method based on big data as described in claim 6, characterized in that, The process of constructing the comprehensive evaluation function and selecting the dominant planting kernels specifically includes: From the work grids in the area to be planted, remove the work grids marked as ecologically unstable, and form a candidate set of the remaining work grids; Calculate the Euclidean distance between the feature vector of each job grid in the candidate set and the soil feature center vector in the anti-disturbance feature field, and take the minimum value of the Euclidean distance as the feature distance component; A comprehensive evaluation function is constructed based on the aforementioned feature distance components. The mathematical expression of the comprehensive evaluation function is as follows: ； in, For the job grid Overall score For the job grid eigenvectors, The soil feature center vector in the anti-disturbance feature field. The index of the soil feature center vector. and These are the spatial coordinates of the grid. For the job grid A set of job grids within a local spatial neighborhood. and The weighting coefficients for balancing features and space; The job grids in the candidate set are sorted according to the comprehensive score, and the job grids with the smallest comprehensive score are selected as the dominant planting cores.

8. The crop cultivation method based on big data as described in claim 7, characterized in that, The region growing algorithm specifically includes: The dominant planting core is used as the seed region for regional growth and iteratively grows to the adjacent operation grid; Calculate the difference in growth resistance between adjacent work grids and the current seed region. If the difference in growth resistance is less than a preset growth tolerance threshold and the adjacent work grid is not marked as an ecologically unstable work grid, then merge the adjacent work grid with the current seed region and continue regional growth using the adjacent work grid as the seed region until growth converges. The area of ​​each connected region formed after growth convergence is counted. If the area of ​​the connected region is less than the preset minimum operation granularity threshold, the connected region is determined to be a broken patch. If the area of ​​the connected region is greater than or equal to the preset minimum task granularity threshold, the connected region is determined to be the dominant cluster. Calculate the Euclidean distance between the feature vector of the dominant planting kernel corresponding to the fragmented patch and the feature vector of the dominant planting kernel corresponding to each dominant cluster. Merge the fragmented patches into the dominant cluster with the smallest Euclidean distance, update the area covered by the dominant cluster, and the area covered by all dominant clusters constitutes a stable clustering region. The growth resistance difference is the absolute value of the difference between the weighted normalized centroid offset of the adjacent working grid and the weighted normalized centroid offset of the seed region.

9. The crop cultivation method based on big data as described in claim 8, characterized in that, The specific configuration of the differentiated planting method includes: The stable clustering areas were identified as core planting areas, and high-density planting and high-fertility input methods were adopted. Ecologically unstable work grids are identified as ecological buffer zones, and low-density planting and conservation tillage methods are implemented accordingly.

10. A big data-based crop cultivation system, applied in the big data-based crop cultivation method as described in any one of claims 1-9, characterized in that, It includes a data collection module, a classification module, an identification module, and a planting module; The data acquisition module is used to collect soil data and historical crop data of the area to be planted. The classification module is used to perform cluster analysis on the soil data and historical crop data, generate a feature field resistant to adverse disturbances, and calculate the reverse nearest neighbor count, environmental weight, and local geometric centroid of the operation grid in the planting area based on the local spatial neighborhood. The identification module is used to calculate the weighted normalized centroid offset value of each work grid by combining the local geometric centroid, environmental weight, and reverse nearest neighbor count, and to set a judgment threshold based on the statistical distribution of the weighted normalized centroid offset value and the reverse nearest neighbor count to identify ecologically unstable work grids. The planting module is used to construct a comprehensive evaluation function and select advantageous planting kernels. It uses a region growth algorithm to expand the advantageous planting kernels into stable clustering regions and configures differentiated planting methods for stable clustering regions and ecologically unstable operation grids respectively.

Images