An agricultural planting soil parameter analysis system

Abstract

This invention discloses an agricultural planting soil parameter analysis system, belonging to the field of intelligent soil analysis technology. It includes a data acquisition module, a data measurement module, a quality evaluation module, and a dynamic optimization module. The data acquisition module is used to collect soil parameters and agricultural operation records from various planting areas. The data measurement module is used to analyze the differences in soil composition among planting areas and generate a first influence difference. The quality evaluation module is used to obtain a soil spatial distribution prediction map by using interactive correlation analysis of the spatial correlation of the first influence difference. The dynamic optimization module is used to dynamically adjust soil content based on the soil spatial distribution prediction map. This system protects the farmland ecological environment, promotes sustainable agricultural development, and improves the intelligent level of soil parameter regulation.

Description

Technical Field

[0001] This invention relates to the field of intelligent soil analysis technology, and in particular to an agricultural planting soil parameter analysis system. Background Technology

[0002] Soil, as the core carrier of terrestrial ecosystems, is the foundation of agricultural production and plays a vital role in maintaining biodiversity, purifying water quality, regulating climate, and protecting nature. Therefore, considering the health and ecological stability of farmland soil, it is urgent to construct a scientific soil parameter quality evaluation system to ultimately maintain land productivity and promote sustainable agricultural development.

[0003] A Chinese invention with application number 202510062875.2 discloses a soil fertility prediction method based on data mining. The method includes: introducing a partitioned and stratified sampling mechanism to obtain a soil sample set of the target soil, and extracting a first soil sample from the soil sample set; reading a predetermined fertility characteristic index, and performing fertility detection on the first soil sample based on the predetermined fertility characteristic index to obtain first detection data; retrieving a porosity pre-plan to perform porosity detection analysis on the first soil sample to obtain second detection data; using the first detection data and the second detection data as input information for an integrated fertility prediction model, and obtaining output information through the integrated fertility prediction model; taking the mean of a first predicted fertility index in the output information to obtain a target fertility index, wherein the target fertility index is used to characterize the fertility status of the target soil.

[0004] The aforementioned technologies do not quantitatively correlate soil chemical properties with spatial variability characteristics, fail to predict soil parameter block maps of spatial relationships, and lack spatial management of variability. Summary of the Invention

[0005] The technical problem solved by this invention is that: soil chemical properties are not quantitatively correlated with spatial variability characteristics, soil parameter block maps that fail to predict spatial relationships are not available, and spatial management of variability is lacking.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] An agricultural planting soil parameter analysis system includes a data acquisition module, a data measurement module, a quality evaluation module, and a dynamic optimization module;

[0008] The data acquisition module is used to collect soil parameters and agricultural operation records for each planting area;

[0009] The data measurement module is used to analyze the differences in soil composition in each planting area and generate the first influence difference.

[0010] The quality assessment module is used to extract the nugget value of the first impact difference and generate a spatial correlation coefficient by combining the coefficient of variation. The spatial correlation coefficient is used to filter the first impact difference to obtain the second impact difference. The non-stationary covariance function is used to integrate the structural equation model results, dynamically fit the nugget value of the second impact difference, and perform interpolation prediction through the improved kriging method to obtain a predicted map of soil spatial distribution.

[0011] The dynamic optimization module is used to dynamically adjust the soil content based on the soil spatial distribution prediction map.

[0012] Preferably, the data acquisition module includes:

[0013] Agricultural planting areas are divided into grids, and soil sensors are placed at the center of each grid. The soil sensors collect soil parameters for each planting area per unit time. The soil parameters include physical data, chemical data, and biological data. The physical data includes temperature, humidity, soil porosity, soil bulk density, soil particle size, topsoil thickness, and slope. The chemical data includes soil pH, electrical conductivity, nitrogen content, phosphorus content, potassium content, heavy metal content, and pesticide residues. The biological data includes microbial biomass carbon and nitrogen and urease activity. The soil sensors include temperature and humidity sensors, pH sensors, electrical conductivity sensors, organic matter sensors, and heavy metal sensors.

[0014] The historical planting records of agricultural planting areas are integrated with urban agricultural operation record data. The agricultural operation record data includes planting number, tillage data, fertilizer data, irrigation data, pesticide data, and harvest data. The tillage data includes tillage depth and tillage method. The fertilizer data includes fertilizer type, dosage, irrigation time, and fertilization method. The irrigation data includes irrigation method, irrigation time, and irrigation amount. The pesticide data includes pesticide type, concentration, and dosage. The harvest data includes harvest date and harvest quality.

[0015] The electrical conductivity is used to represent the soil salinity;

[0016] Soil porosity is used to indicate the degree of soil aeration.

[0017] Preferably, the data measurement module includes:

[0018] A three-dimensional soil grid space is constructed for the agricultural planting area. The soil grid space includes x-coordinates, y-coordinates, and z-coordinates. The x-coordinates and y-coordinates are used to define the extent of the agricultural planting area, and the z-coordinate represents the current soil state value. The soil spatial coordinates of each grid block are obtained, and the average value of the chemical data for each grid block is calculated. and standard deviation ;

[0019] The average phosphorus content in the chemical data is taken as the total phosphorus content of the soil. A Gaussian model is used to analyze the total phosphorus content of the soil and the soil spatial coordinates to obtain the difference in the influence of phosphorus content. The difference in the influence of phosphorus content is used to represent the linear difference correlation between the phosphorus content and the soil spatial coordinates.

[0020] Repeat the above operations to calculate the difference in influence of each component in the remaining chemical data, and save it as the first difference in influence along with the difference in influence of phosphorus content. The remaining chemical data includes soil pH, nitrogen content, potassium content, and heavy metal content.

[0021] Calculate the coefficient of variation for each sample in the soil parameters. The expression for calculating the coefficient of variation is:

[0022] ;

[0023] The first influence difference and coefficient of variation are rendered into the soil 3D mesh as z-coordinate values.

[0024] Preferably, the specific method for analyzing the total phosphorus content of the soil and the soil spatial coordinates using a Gaussian model to obtain the differences in phosphorus content is as follows:

[0025] The phosphorus content of different grid blocks was extracted, and each grid block was analyzed from... Number them, among which The phosphorus content of two grid blocks with a distance of h is randomly selected. The spatial variation coefficient of phosphorus content is quantified using a semi-variogram function, and a scatter plot is generated in the three-dimensional soil grid. The calculation expression of the semi-variogram function is as follows:

[0026] ;

[0027] in, For the semi-mutation function, Let h be the number of sample point pairs (i,j) between all two grid blocks with a distance of h. Let be the phosphorus content of the soil in the i-th block. Let be the phosphorus content of the soil in the j-th block;

[0028] A Gaussian model of soil phosphorus content is constructed, and the internal parameters are adjusted by the least squares method. The internal parameters are then fitted to the semivariogram function and matched to the scatter plot coordinates to generate the phosphorus content influence difference, which includes nugget value, sill value and range.

[0029] Preferably, the quality evaluation module includes a data fusion unit, an interactive analysis unit, and a spatial prediction unit;

[0030] The data fusion unit includes:

[0031] Extract the nugget value of the first impact difference, calculate the product of the nugget value of the first impact difference and the corresponding sample coefficient of variation, generate a spatial correlation coefficient, determine the spatial relationship between the chemical data corresponding to the nugget value and the soil, generate a second impact difference, the method for determining the spatial relationship includes retaining the chemical data corresponding to the nugget value when the spatial correlation coefficient is greater than a set threshold, deleting the chemical data corresponding to the nugget value when the spatial correlation coefficient is less than a set threshold, and generating state auxiliary interaction data by mapping the second impact difference to the soil parameters one by one.

[0032] Preferably, the interactive analysis unit includes:

[0033] Based on the state-assisted interaction data filtered and fused by the data fusion unit, a structural equation model is constructed. The second influence difference of the state-assisted interaction data is used as the dependent variable, and the chemical data corresponding to the nugget value of the second influence difference is used as the independent variable. A hypothetical causal relationship network is constructed. Path coefficients and model fit optimization are generated using maximum likelihood estimation. The path coefficients are used as edge weights, and the model fit optimization is used as a reference factor for drawing the hypothetical causal relationship network to generate a soil weight hypothetical path map.

[0034] Preferably, the spatial prediction unit includes:

[0035] A non-stationary covariance function is constructed to obtain the nugget value of the second influence difference. A structural equation model is fitted, and the corresponding soil parameters are substituted into the structural equation model to predict the nugget value of the soil parameters, obtaining a spatially varying nugget value distribution map. The non-stationary covariance function is redefined, which is used to multiply the global stationary covariance function by the local variance scalar and connect it to the structural equation model result. The local standard deviation is set to be proportional to the square root of the nugget value, and a scaling constant is set to bring the local standard deviation and the overall variance level of the nugget value to a balance. The Kriging system equation is modified, and the covariance obtained by the non-stationary covariance function is used as the covariance of the Kriging system equation, and spatial prediction is performed to generate a soil spatial distribution prediction map. The soil spatial distribution prediction map includes a pH value distribution map, an organic matter distribution map, an available nitrogen distribution map, an available phosphorus distribution map, a available potassium distribution map, and a heavy metal distribution map.

[0036] Preferably, the dynamic optimization module includes a decision-making unit, an execution unit, and an evaluation unit;

[0037] The decision-making unit includes:

[0038] Analyze the soil parameter values ​​of the pixels in the predicted soil spatial distribution map, and assign the corresponding operation values ​​to the corresponding pixels in the output image according to the rules to generate a variable fertilization prescription map. The rule judgment includes a first rule, a second rule, and a third rule. When the soil parameter value meets the first rule, a first fertilization schedule is triggered. When the soil parameter value meets the second rule, a second fertilization schedule is triggered. When the soil parameter value meets the third rule, a third fertilization schedule is triggered.

[0039] Preferably, the execution unit includes:

[0040] According to the variable fertilization prescription map, the fertilization path is transmitted to the variable control agricultural machinery, the work field is set, the automatic navigation system is turned on, and the variable fertilization prescription map is queried based on the GPS coordinates obtained per unit time to obtain the target fertilization amount corresponding to the current GPS coordinate point, so as to accurately control the amount of fertilizer dispensed.

[0041] Preferably, the evaluation unit includes:

[0042] After the execution unit completes its operation, it uses UAV multispectral remote sensing to monitor crop growth trends, check whether the crop response is uniform, and generate a yield distribution map. The yield map of this season is compared with the variable fertilization prescription map of the previous season. If the field has high fertilization but low yield, the current field area is marked as abnormal. If the field has normal fertilization but high yield, the execution unit adjusts the fertilization amount in the variable fertilization prescription map for the current field to reduce fertilizer input and generate the variable fertilization prescription map for the next season.

[0043] The beneficial effects of this invention are as follows: Based on the precision and digitalization of soil management, through multi-source data collection and geographic statistical analysis, the spatial variation in the field is accurately quantified, completely changing the traditional extensive and uniform management model and providing a scientific basis for variable operations. Secondly, the system demonstrates strong economic benefits. By generating variable prescription maps through the dynamic optimization module and guiding intelligent agricultural machinery to execute them, it can achieve precise application of inputs such as fertilizers and pesticides as needed, directly reducing production costs and increasing farmers' income. In addition, the system has outstanding environmental benefits, effectively reducing non-point source pollution and soil degradation caused by excessive fertilization and pesticide use, protecting the farmland ecological environment, and promoting sustainable agricultural development. Finally, its closed-loop optimization mechanism can continuously learn and improve the system's decision-making through continuous monitoring and effect feedback, thereby enhancing the intelligent level of soil parameter regulation. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the basic process of an agricultural planting soil parameter analysis system provided in one embodiment of the present invention. Detailed Implementation

[0045] 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.

[0046] Example, refer to Figure 1 This paper provides an agricultural planting soil parameter analysis system, including a data acquisition module, a data measurement module, a quality evaluation module, and a dynamic optimization module;

[0047] The data acquisition module is used to collect soil parameters and agricultural operation records for each planting area;

[0048] The data measurement module is used to analyze the differences in soil composition in each planting area and generate the first influence difference.

[0049] The quality assessment module is used to extract the nugget value of the first impact difference and generate a spatial correlation coefficient by combining the coefficient of variation. The spatial correlation coefficient is used to filter the first impact difference to obtain the second impact difference. The non-stationary covariance function is used to integrate the structural equation model results, dynamically fit the nugget value of the second impact difference, and perform interpolation prediction through the improved kriging method to obtain a predicted map of soil spatial distribution.

[0050] The dynamic optimization module is used to dynamically adjust the soil content based on the soil spatial distribution prediction map.

[0051] In this embodiment, based on the precision and digitalization of soil management, multi-source data collection and geostatistical analysis are used to accurately quantify the spatial variation in the field, completely changing the traditional extensive and uniform management model and providing a scientific basis for variable operations. Secondly, the system demonstrates strong economic benefits. By generating variable prescription maps through the dynamic optimization module and guiding intelligent agricultural machinery to execute them, it can achieve precise application of inputs such as fertilizers and pesticides as needed, directly reducing production costs and increasing farmers' income. In addition, the system has outstanding environmental benefits, effectively reducing non-point source pollution and soil degradation caused by excessive fertilization and pesticide use, protecting the farmland ecological environment, and promoting sustainable agricultural development. Finally, its closed-loop optimization mechanism can continuously learn and improve the system's decision-making through continuous monitoring and effect feedback, thereby enhancing the level of intelligence in soil parameter regulation.

[0052] The data acquisition module includes:

[0053] Agricultural planting areas are divided into grids, and soil sensors are placed at the center of each grid. The soil sensors collect soil parameters for each planting area per unit time. The soil parameters include physical data, chemical data, and biological data. The physical data includes temperature, humidity, soil porosity, soil bulk density, soil particle size, topsoil thickness, and slope. The chemical data includes soil pH, electrical conductivity, nitrogen content, phosphorus content, potassium content, heavy metal content, and pesticide residues. The biological data includes microbial biomass carbon and nitrogen and urease activity. The soil sensors include temperature and humidity sensors, pH sensors, electrical conductivity sensors, organic matter sensors, and heavy metal sensors.

[0054] The historical planting records of agricultural planting areas are integrated with urban agricultural operation record data. The agricultural operation record data includes planting number, tillage data, fertilizer data, irrigation data, pesticide data, and harvest data. The tillage data includes tillage depth and tillage method. The fertilizer data includes fertilizer type, dosage, irrigation time, and fertilization method. The irrigation data includes irrigation method, irrigation time, and irrigation amount. The pesticide data includes pesticide type, concentration, and dosage. The harvest data includes harvest date and harvest quality.

[0055] The electrical conductivity is used to represent the soil salinity;

[0056] Soil porosity is used to indicate the degree of soil aeration.

[0057] The planting area was divided into 10m×10m grids, and a multi-parameter soil sensor array was deployed at the center of each grid, including temperature and humidity sensors (accuracy ±0.5℃ / ±2%RH), pH sensors (measurement range 3-10), electrical conductivity sensors (0-20mS / cm), organic matter sensors (detection limit 0.1%), and heavy metal sensors (detection limit 0.01mg / kg). Soil physical data (temperature, humidity, porosity, bulk density, particle composition, topsoil thickness, slope), chemical data (pH value, electrical conductivity, nitrogen, phosphorus and potassium content, heavy metal content, pesticide residues), and biological data (microbial biomass carbon and nitrogen, urease activity) were collected in real time. The sampling frequency was set to once per hour. The data was uploaded to the edge gateway via the LoRaWAN protocol. Historical planting records were integrated through the IoT gateway to build an agricultural operation database. Data entries include planting number (unique identifier), tillage data (depth 20-30cm / rotary tillage / plowing), fertilizer data (organic fertilizer / compound fertilizer, application rate 50-200kg / mu, fertilization cycle 15-30 days), irrigation data (drip irrigation / sprinkler irrigation, time window 6:00-8:00 / 18:00-20:00, amount 30-50m³ / mu), pesticide data (insecticide / fungicide, concentration 0.1%-0.5%, application rate 100-300ml / mu), and harvest data (date accurate to the day, quality grade divided into three levels).

[0058] Employing a gridded sensor deployment strategy, farmland is divided into regular grids. Integrated sensor nodes are deployed at the center of each grid to continuously collect over ten soil parameters across three categories: physical (temperature, humidity, porosity, bulk density, etc.), chemical (pH, NPK, heavy metals, etc.), and biological (microbial biomass, enzyme activity). This achieves a comprehensive, real-time digital twin of the farmland's ecological environment. Simultaneously, the system integrates full-cycle agricultural operation records, structurally storing operations such as tillage, fertilization, irrigation, pesticide application, and harvesting, forming a complete data chain covering all soil parameters.

[0059] The data measurement module includes:

[0060] A three-dimensional soil grid space is constructed for the agricultural planting area. The soil grid space includes x-coordinates, y-coordinates, and z-coordinates. The x-coordinates and y-coordinates are used to define the extent of the agricultural planting area, and the z-coordinate represents the current soil state value. The soil spatial coordinates of each grid block are obtained, and the average value of the chemical data for each grid block is calculated. and standard deviation ;

[0061] The average phosphorus content in the chemical data is taken as the total phosphorus content of the soil. A Gaussian model is used to analyze the total phosphorus content of the soil and the soil spatial coordinates to obtain the difference in the influence of phosphorus content. The difference in the influence of phosphorus content is used to represent the linear difference correlation between the phosphorus content and the soil spatial coordinates.

[0062] Repeat the above operations to calculate the difference in influence of each component in the remaining chemical data, and save it as the first difference in influence along with the difference in influence of phosphorus content. The remaining chemical data includes soil pH, nitrogen content, potassium content, and heavy metal content.

[0063] Calculate the coefficient of variation for each sample in the soil parameters. The expression for calculating the coefficient of variation is:

[0064] ;

[0065] The first influence difference and coefficient of variation are rendered into the soil 3D mesh as z-coordinate values.

[0066] A three-dimensional grid space for the agricultural planting area was constructed. The three-dimensional cells were divided using a 10m × 10m planar grid (x, y coordinates) combined with a soil depth of 0-60cm (z coordinate, 10cm per layer). Sampling points were located using the GPS coordinates (error ≤ 0.5m) of the grid center point. The arithmetic mean of the chemical data (pH value, electrical conductivity, nitrogen, phosphorus, potassium content, and heavy metal content) collected for each grid block was calculated. The average phosphorus content was used as the total phosphorus content of the soil. This data was then input into a Gaussian model (RBF kernel function, bandwidth parameter σ = 5m) for spatial interpolation, generating a phosphorus content influence difference matrix. This matrix was quantified using the Pearson correlation coefficient (r) to determine the linear correlation between phosphorus content and spatial coordinates (|r| ≥ 0.6 indicates a strong correlation).

[0067] The above process is repeated to calculate the spatial differences in soil pH, nitrogen content, potassium content, and heavy metal content, and these are merged into a first impact difference dataset. Simultaneously, the coefficient of variation (CV = standard deviation / mean × 100%) for all soil parameters (physical, chemical, and biological) is calculated. Finally, the first impact difference (r value) and the coefficient of variation (CV value) are mapped onto a 3D mesh using voxel rendering technology, forming a visual model representing soil state with z-coordinate values. This embodiment constructs a digital profile of soil characteristics through spatial analysis and multi-dimensional data fusion, providing a complete technology chain from data acquisition to decision support for precision agriculture.

[0068] The specific method for analyzing the total phosphorus content and spatial coordinates of the soil using a Gaussian model to obtain the differences in phosphorus content is as follows:

[0069] The phosphorus content of different grid blocks was extracted, and each grid block was analyzed from... Number them, among which The phosphorus content of two grid blocks with a distance of h is randomly selected. The spatial variation coefficient of phosphorus content is quantified using a semi-variogram function, and a scatter plot is generated in the three-dimensional soil grid. The expression for calculating the semi-variogram function is as follows:

[0070] ;

[0071] in, For the semi-mutation function, Let h be the number of sample point pairs (i,j) between all two grid blocks with a distance of h. Let be the phosphorus content of the soil in the i-th block. The phosphorus content of the soil in the j-th block;

[0072] A Gaussian model of soil phosphorus content is constructed, and the internal parameters are adjusted by the least squares method. The internal parameters are then fitted to the semivariogram function and matched to the scatter plot coordinates to generate the phosphorus content influence difference, which includes nugget value, sill value and range.

[0073] In this embodiment, the three-dimensional grid blocks of the agricultural planting area are numbered (e.g., Gx-yz, where x, y are planar coordinates and z is soil depth). Phosphorus content data is extracted from 30% of the grid blocks using a systematic sampling method (detection accuracy ±0.5 mg / kg). 1000 pairs of sample points (i, j) with a distance of h are randomly generated. The spatial coefficient of variation is calculated using a semi-variogram. The calculation results are plotted as a scatter plot with h as the horizontal axis and γ(h) as the vertical axis to show the spatial correlation characteristics of phosphorus content. A Gaussian model is constructed for curve fitting. The model expression is γ(h)=C0+C(1-exp(-h² / (2a²))), where C0 is the nugget value, C is the structural variance, and a is the range. The parameters are optimized iteratively using the least squares method (iteration number ≤ 100 times, convergence threshold 1e-6) to make the determination coefficient R² of the model prediction value and the semi-variogram calculation value ≥ 0.85. The final output consists of three core indicators that reflect the difference in phosphorus content: nugget value (C0), sill value (C0+C), and range (a), thus completing the transformation from discrete data to a spatial distribution model.

[0074] The quality evaluation module includes a data fusion unit, an interactive analysis unit, and a spatial prediction unit.

[0075] The data fusion unit includes:

[0076] Extract the nugget value of the first impact difference, calculate the product of the nugget value of the first impact difference and the corresponding sample coefficient of variation, generate a spatial correlation coefficient, determine the spatial relationship between the chemical data corresponding to the nugget value and the soil, generate a second impact difference, the method for determining the spatial relationship includes retaining the chemical data corresponding to the nugget value when the spatial correlation coefficient is greater than a set threshold, deleting the chemical data corresponding to the nugget value when the spatial correlation coefficient is less than a set threshold, and generating state auxiliary interaction data by mapping the second impact difference to the soil parameters one by one.

[0077] The data fusion unit first extracts the nugget value (C0) from the first impact difference, which characterizes the intensity of random variation in soil chemical data (such as sampling error). The spatial correlation coefficient (SRC) is calculated using the formula SRC = C0 × CV, where CV is the coefficient of variation (standard deviation / mean) of the corresponding parameter, reflecting the degree of data dispersion. If a significant spatial correlation is found between the chemical data and the soil (e.g., pH value SRC = 0.18), it is retained; if SRC < 0.18, it is discarded. If the heavy metal content (SRC=0.09) is considered to have weak spatial correlation, it will be excluded. The differences in the second influence after screening will be mapped one-to-one with soil physical and biological parameters to generate state-assisted interactive data, which will lay the foundation for subsequent analysis.

[0078] The interactive analysis unit includes:

[0079] Based on the state-assisted interaction data filtered and fused by the data fusion unit, a structural equation model is constructed. The second influence difference of the state-assisted interaction data is used as the dependent variable, and the chemical data corresponding to the nugget value of the second influence difference is used as the independent variable. A hypothetical causal relationship network is constructed. Path coefficients and model fit optimization are generated using maximum likelihood estimation. The path coefficients are used as edge weights, and the model fit optimization is used as a reference factor for drawing the hypothetical causal relationship network to generate a soil weight hypothetical path map.

[0080] Extracting a variable set from state-assisted interaction data: The second influence difference (five spatial influence difference indicators such as phosphorus content and pH value) is set as an endogenous latent variable (dependent variable), and the corresponding chemical data (effective parameters after nugget value screening) is set as an exogenous manifest variable (independent variable). A hypothetical causal network containing 12 observed variables and 3 mediating variables is constructed. Maximum likelihood estimation (MLE) is used to iteratively optimize the model parameters. The iteration convergence criterion is set as residual sum of squares <1e-6. Path coefficients (the strength of direct influence between variables) and model fit indices are calculated. The iteration stops when the goodness of fit reaches χ² / df <3, RMSEA <0.08, and CFI >0.9. Finally, using standardized path coefficients (β) as edge weights (range -1 to 1), and combining goodness-of-fit, a soil weighted hypothetical path diagram was drawn to visually demonstrate the causal chain from chemical data to mediating variables to the differences in secondary influences. In the causal deconstruction stage, SEM breaks through the limitations of traditional regression analysis, simultaneously estimating direct effects (e.g., the direct effect of pH on phosphorus availability β=0.32) and indirect effects (e.g., organic matter indirectly affects phosphorus availability by influencing pH, with a total effect β=0.47). The path strength quantification stage uses bootstrap sampling (n=5000) to test the significance of coefficients; paths with P<0.05 are retained, forming a simplified causal network.

[0081] The spatial prediction unit includes:

[0082] A non-stationary covariance function is constructed to obtain the nugget value of the second influence difference. A structural equation model is fitted, and the corresponding soil parameters are substituted into the structural equation model to predict the nugget value of the soil parameters, obtaining a spatially varying nugget value distribution map. The non-stationary covariance function is redefined, which is used to multiply the global stationary covariance function by the local variance scalar and connect it to the structural equation model result. The local standard deviation is set to be proportional to the square root of the nugget value, and a scaling constant is set to bring the local standard deviation and the overall variance level of the nugget value to a balance. The Kriging system equation is modified, and the covariance obtained by the non-stationary covariance function is used as the covariance of the Kriging system equation, and spatial prediction is performed to generate a soil spatial distribution prediction map. The soil spatial distribution prediction map includes a pH value distribution map, an organic matter distribution map, an available nitrogen distribution map, an available phosphorus distribution map, a available potassium distribution map, and a heavy metal distribution map.

[0083] By employing a technical approach of non-stationary covariance modeling, dynamic adjustment of nugget values, and multi-factor collaborative prediction, high-precision spatial distribution prediction of soil parameters is achieved. The average nugget value of the region is calculated from the nugget values ​​predicted by the result equation model. This embodiment breaks through the limitations of traditional geostatistical models by using non-stationary covariance to construct a digital map of the spatial distribution of soil parameters.

[0084] The dynamic optimization module includes a decision-making unit, an execution unit, and an evaluation unit;

[0085] The decision-making unit includes:

[0086] Analyze the soil parameter values ​​of the pixels in the predicted soil spatial distribution map, and assign the corresponding operation values ​​to the corresponding pixels in the output image according to the rules to generate a variable fertilization prescription map. The rule judgment includes a first rule, a second rule, and a third rule. When the soil parameter value meets the first rule, a first fertilization schedule is triggered. When the soil parameter value meets the second rule, a second fertilization schedule is triggered. When the soil parameter value meets the third rule, a third fertilization schedule is triggered.

[0087] In this embodiment, a multi-rule triggering mechanism is employed to achieve precise decision-making in soil nutrient management. First, soil parameter values ​​for each pixel are extracted from the soil spatial distribution prediction map (1m×1m resolution). A dataset containing six indicators, including pH (3-10), available phosphorus (0-50mg / kg), and available potassium (0-200mg / kg), is constructed. Based on a crop demand model, three fertilization rules are preset: Rule 1 (low fertility) triggers high-volume fertilization scheduling; when available phosphorus < 5mg / kg or available potassium < 50mg / kg, the operation amount is assigned as 1.2 times the difference between the target value and the measured value (compensation coefficient); Rule 2 (medium fertility) triggers conventional fertilization; when 5 ≤ available phosphorus < 15mg / kg and 50 ≤ available potassium < 120mg / kg, the operation amount is 1.0 times the difference; Rule 3 (high fertility) triggers reduced-volume fertilization; when available phosphorus ≥ 15mg / kg and available potassium ≥ 120mg / kg, the operation amount is 0.5 times or zero the difference. Using GIS spatial analysis tools, the operational values ​​are assigned to the corresponding pixels to generate an RGB pseudo-color variable fertilizer prescription map, where red represents high-dosage areas (>15kg / mu), yellow represents regular areas (8-15kg / mu), and green represents reduced-dosage areas (<8kg / mu).

[0088] By matching coordinates, the predicted image pixels are mapped one-to-one with the actual plot locations. The rule engine uses fuzzy logic reasoning and sets transitional ranges for parameter thresholds to avoid hard decision errors that are either / or. The calculation of the amount of fertilizer applied introduces a crop coefficient, so that the amount of fertilizer applied is dynamically matched with the crop's needs.

[0089] The execution unit includes:

[0090] According to the variable fertilization prescription map, the fertilization path is transmitted to the variable control agricultural machinery, the work field is set, the automatic navigation system is turned on, and the variable fertilization prescription map is queried based on the GPS coordinates obtained per unit time to obtain the target fertilization amount corresponding to the current GPS coordinate point, so as to accurately control the amount of fertilizer dispensed.

[0091] In this embodiment, the variable fertilizer prescription map is format-converted, and the 1m×1m pixel RGB pseudo-color image is parsed into vector data containing GPS coordinates (WGS84 coordinate system) and fertilizer application rate (kg / mu). This data is transmitted to the variable control agricultural machinery terminal via CAN bus (transmission delay ≤100ms). The boundary of the working field is set, and the automatic navigation system is activated (positioning accuracy 1-3cm, sampling frequency 10Hz). During agricultural machinery operation, the current GPS coordinates are acquired in real time (once every 0.1 seconds). The target fertilizer application rate of the corresponding pixel in the prescription map is queried through a spatial indexing algorithm. This is converted into a feeding command (0-100% opening) by the PID controller, which drives the variable frequency motor to control the speed of the fertilizer applicator's screw feeder, thereby achieving precise fertilizer application.

[0092] The evaluation unit includes:

[0093] After the execution unit completes its operation, it uses UAV multispectral remote sensing to monitor crop growth trends, check whether the crop response is uniform, and generate a yield distribution map. The yield map of this season is compared with the variable fertilization prescription map of the previous season. If the field has high fertilization but low yield, the current field area is marked as abnormal. If the field has normal fertilization but high yield, the execution unit adjusts the fertilization amount in the variable fertilization prescription map for the current field to reduce fertilizer input and generate the variable fertilization prescription map for the next season.

[0094] In this embodiment, the current season's yield map and the previous season's variable fertilization prescription map are spatially overlaid for analysis. The value of each grid cell is calculated, and a dual threshold rule is set to identify abnormal areas: when the response coefficient is <0.8 (high fertilizer, low yield), it is marked as an area to be improved; when the response coefficient is >1.2 (appropriate fertilizer, high yield), it is marked as an efficient area. Soil samples (0-20cm) are collected from abnormal areas to analyze their physicochemical properties. After excluding non-fertility limiting factors (such as heavy metal pollution), a machine learning model (random forest) is used to optimize fertilization parameters and generate the next season's variable fertilization prescription map.

[0095] 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.

[0096] 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 scope of the claims of the present invention.

Claims

1. An agricultural planting soil parameter analysis system, characterized by, The method comprises a data collection module, a data determination module, a quality evaluation module and a dynamic optimization module. The data collection module is configured to collect soil parameter and farming operation record data of each planting area. The data determination module is configured to analyze soil component difference values of each planting area to generate first influence differences. The quality evaluation module is configured to extract nugget values of the first influence differences, combine variation coefficients to generate spatial correlation coefficients, screen the first influence differences by using the spatial correlation coefficients to obtain second influence differences, integrate structural equation model results by using a non-stationary covariance function, dynamically fit nugget values of the second influence differences, and perform interpolation prediction by using an improved Kriging method to obtain a soil spatial distribution prediction map. The dynamic optimization module is configured to dynamically control soil content according to the soil spatial distribution prediction map. The data determination module comprises: A soil three-dimensional grid space is constructed for an agricultural planting area, the soil three-dimensional grid space includes x coordinates, y coordinates and z coordinates, the x coordinates and the y coordinates are used to frame the range of the agricultural planting area, and the z coordinates are used to represent current state values of the soil and standard deviations of different grid blocks are calculated respectively The average value of the phosphorus content in the chemical data is taken as a total soil phosphorus content value, and the total soil phosphorus content value and soil spatial coordinates are analyzed by using a Gaussian model to obtain phosphorus content influence differences, which are used to represent linear difference correlation of the phosphorus content and the soil spatial coordinates. The above operation is repeated to calculate influence differences of each component in the remaining chemical data, and the influence differences of each component are saved as first influence differences together with the phosphorus content influence differences, the remaining chemical data including soil pH value, nitrogen content, potassium content and heavy metal content. calculating the coefficient of variation for each sample of soil parameters The expression for calculating the coefficient of variation is: ； The first influence differences and variation coefficients are rendered into a soil three-dimensional grid as z coordinate values. The quality evaluation module comprises a data fusion unit, an interactive analysis unit and a spatial prediction unit. The data fusion unit comprises: The nugget values of the first influence differences are extracted, the product of the nugget values of the first influence differences and corresponding sample variation coefficients is calculated to generate spatial correlation coefficients, the spatial relationship degree of the chemical data corresponding to the nugget values and the soil is judged to generate second influence differences, the spatial relationship degree judgment method comprises retaining the chemical data corresponding to the nugget values when the spatial correlation coefficients are greater than a set threshold value, and deleting the chemical data corresponding to the nugget values when the spatial correlation coefficients are less than the set threshold value, and the second influence differences are generated as state auxiliary interaction data one by one corresponding to the soil parameters. The interactive analysis unit comprises: The state auxiliary interaction data filtered and fused by the data fusion unit are screened to construct a structural equation model, the second influence differences of the state auxiliary interaction data are taken as dependent variables, the chemical data corresponding to the nugget values of the second influence differences are taken as independent variables, a hypothetical causal relationship network is constructed, path coefficients and model fitting optimization degrees are generated by using maximum likelihood estimation, the path coefficients are taken as edge weights, the model fitting optimization degrees are taken as reference factors for drawing the hypothetical causal relationship network, and a soil weight hypothetical path diagram is generated; The spatial prediction unit comprises: The non-stationary covariance function is constructed, block gold values of the second influence difference are obtained, a structural equation model is fitted, corresponding soil parameters are brought into the structural equation model, block gold values of the soil parameters are predicted, a block gold value distribution map of spatial variation is obtained, the non-stationary covariance function is redefined, the non-stationary covariance function is used for multiplying a global stationary covariance function and a local variance scalar, and is connected with the structural equation model result, a local standard deviation is set to be proportional to a square root of the block gold value, and a scaling constant is set to balance the local standard deviation and the overall variance level of the block gold value, a Kriging system equation is modified, the covariance obtained by the non-stationary covariance function is taken as the covariance of the Kriging system equation, and spatial prediction is carried out to generate a soil spatial distribution prediction map, the soil spatial distribution prediction map including a pH value distribution map, an organic matter distribution map, an alkali-hydrolyzable nitrogen distribution map, an available phosphorus distribution map, an available potassium distribution map and a heavy metal distribution map.

2. An agricultural planting soil parameter analysis system as claimed in claim 1, characterized in that, The data acquisition module comprises: The agricultural planting area is divided according to a grid, and a soil sensor is arranged at the center point of each grid, and the soil sensor is used to collect soil parameters of each planting area in a unit time, the soil parameters including physical data, chemical data and biological data, the physical data including temperature, humidity, soil porosity, soil bulk density, soil particle, plough layer thickness and slope, the chemical data including soil pH value, conductivity, nitrogen content, phosphorus content, potassium content, heavy metal content and pesticide residue value, and the biological data including microbial carbon and nitrogen and urease activity, the soil sensor including a temperature and humidity sensor, a pH sensor, a conductivity sensor, an organic matter sensor and a heavy metal sensor; The historical planting records of the agricultural planting area are integrated with farm operation record data, the farm operation record data including planting number, tillage data, fertilizer data, irrigation data, pesticide data and harvest data, the tillage data including tillage depth and tillage method, the fertilizer data including fertilizer type, dosage, irrigation time and fertilization method, the irrigation data including irrigation method, irrigation time and irrigation amount, the pesticide data including pesticide type, use concentration and use amount, and the harvest data including harvest date and harvest quality; The conductivity is used to represent soil salt content. The soil porosity is used to represent the air permeability of soil.

3. An agricultural planting soil parameter analysis system as claimed in claim 2, characterized in that, The specific method for analyzing the total soil phosphorus content and soil spatial coordinates by using the Gaussian model is that: The phosphorus content of different grid blocks is extracted, and each grid block is subjected to are numbered, wherein The phosphorus content of two grid blocks with a distance of h is randomly selected, the spatial variation coefficient of the phosphorus content is quantified by using a semi-variation function, and a scatter plot is generated in the three-dimensional grid of the soil, and the calculation expression of the semi-variation function is: ； wherein, is the semi-variogram function, is the number of pairs of grid cell sample points (i,j) for all distances h, is the phosphorus content of the soil in the i-th block, is the phosphorus content of the soil in the j-th block; A Gaussian model of soil phosphorus content is constructed, internal parameters are adjusted by a least square method, the internal parameters are fitted into the semi-variogram function and then matched to coordinates in a scatter plot, a phosphorus content influence difference is generated, and the phosphorus content influence difference includes block gold value, base value and range.

4. The agricultural planting soil parameter analysis system of claim 1, wherein, The dynamic optimization module comprises a decision unit, an execution unit and an evaluation unit. The decision unit comprises: The soil parameter values of the pixel points of the soil spatial distribution prediction map are analyzed, and according to rule judgment, the corresponding operation amount is assigned to the corresponding pixel of the output image to generate a variable fertilization prescription map, the rule judgment includes a first rule, a second rule and a third rule, when the soil parameter value meets the first rule, the first fertilization scheduling is triggered, when the soil parameter value meets the second rule, the second fertilization scheduling is triggered, and when the soil parameter value meets the third rule, the third fertilization scheduling is triggered.

5. An agricultural planting soil parameter analysis system as claimed in claim 4, characterized in that, The execution unit comprises: According to the variable fertilization prescription map, the fertilization path is transmitted to the variable control agricultural machine, the operation field is set, the automatic navigation system is started, the GPS coordinates obtained in unit time are inquired from the variable fertilization prescription map, the target fertilization amount corresponding to the current GPS coordinate point is obtained, and the discharge amount is accurately controlled.

6. An agricultural planting soil parameter analysis system as claimed in claim 5, characterized in that, The evaluation unit comprises: After the execution unit completes the operation, the unmanned aerial vehicle multispectral remote sensing is used to monitor the crop growth trend, the crop response is checked, the yield distribution map is generated, the yield map of the current season is compared with the variable fertilization prescription map of the last season, when the field fertilization amount is high but the yield is low, the current field area is marked as abnormal, when the field fertilization amount is normal but the yield is high, the fertilization amount of the variable fertilization prescription map of the execution unit in the current field is adjusted, the fertilizer input is reduced, and the variable fertilization prescription map of the next season is generated.

Images