A method and system for identifying salinity gradients in saline soil

Abstract

This invention relates to the field of soil salinity identification technology, and discloses a method and system for identifying salinity gradients in saline soil. The system includes a multi-source data acquisition and preprocessing module, a salinity migration coupling analysis module, a gradient division and dynamic correction module, a feature fusion and intelligent interpretation module, and a visualization and decision execution module. By collecting multi-source surface environmental data, subsurface layered soil physicochemical data, and microclimate environmental data, and combining them with salinity migration coupling analysis, regional salinity spatiotemporal distribution benchmark data is generated, ensuring the comprehensiveness and accuracy of the salinity distribution benchmark and providing a reliable basis for gradient division. The regional salinity spatiotemporal distribution benchmark data is divided into grid levels using a preset salinity gradient grading threshold set, and gradient drift correction is performed in conjunction with real-time dynamic monitoring data, ensuring the rationality of the initial salinity gradient zoning and improving the dynamic calibration accuracy of the salinity gradient distribution.

Description

Technical Field

[0001] This invention relates to the field of soil salinity identification technology, specifically to a method and system for identifying salinity gradients in saline soil. Background Technology

[0002] Saline soil is a general term for saline soil, alkaline soil, and various salinized and alkalized soils. Saline soil refers to soil with a soluble salt content that is significantly harmful to crop growth. The salt content index varies depending on the salt composition. Alkaline soil refers to soil containing a large amount of exchangeable sodium that is harmful to plant growth and changes soil properties. Saline soil is mainly distributed in inland arid and semi-arid regions, and is also found in coastal areas.

[0003] Currently, due to the complex environment of saline soil areas and the fact that the water and salt transport process is affected by multiple factors, when carrying out salinity gradient identification of saline soil, the collected surface environment and underground stratification data often have spatiotemporal matching deviations, making it difficult to reflect the actual migration status of salt at different depths in real time, and failing to guarantee the accuracy of gradient division, which in turn leads to a lack of targeted improvement measures.

[0004] Therefore, a method and system for identifying the salinity gradient of saline soil are proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for identifying salinity gradients in saline soils. This solves the problem mentioned in the background technology that it is difficult to reflect the actual migration status of salinity at different depths in real time, and the gradient division cannot be guaranteed to be accurate, which in turn leads to a lack of targeted improvement measures.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method and system for identifying salinity gradients in saline soil, the method comprising the following steps: S1. Collect surface multi-source environmental data and subsurface layered soil physicochemical data of the target saline soil area; S2. Based on the surface multi-source environmental data and the underground layered soil physicochemical data, perform salt migration coupling analysis to generate regional salt spatiotemporal distribution benchmark data. S3. Based on the regional salinity spatiotemporal distribution benchmark data and the preset salinity gradient classification threshold set, perform gradient level division processing to generate preliminary salinity gradient partition data. S4. Based on the preliminary salinity gradient partitioning data and real-time dynamic monitoring data, perform salinity gradient drift correction processing to generate dynamically calibrated salinity gradient distribution data. S5. Integrate the surface multi-source environmental data, the underground layered soil physicochemical data, the preliminary salinity gradient zoning data, and the dynamically calibrated salinity gradient distribution data into salinity identification feature parameters for the target area. S6. Based on the salinity identification feature parameters, multi-scale feature fusion and gradient boundary calculation are performed through a pre-trained salinity gradient interpretation model to generate salinity gradient identification result data with clear boundaries. S7. Based on the high-precision salinity gradient identification result data, perform visualization map construction processing to generate salinity gradient spatial distribution map data; S8. Based on the salinity gradient spatial distribution map data, trigger the execution instructions for irrigation and salt leaching and biological improvement, and generate and issue soil improvement control instruction data.

[0007] Preferably, the step S1, which involves collecting surface multi-source environmental data and subsurface layered soil physicochemical data of the target saline soil area, includes the following steps: S11. Collect multi-source surface environmental data of the target area using satellite remote sensing equipment and UAV-borne spectrometer. The multi-source surface environmental data includes surface reflectance spectrum, surface temperature, vegetation cover index and surface roughness, and generate surface environmental characteristic data. S12. Collect subsurface physicochemical data of the target area through a distributed soil sensor array. The subsurface physicochemical data includes soil moisture content, electrical conductivity, pH value, sodium ion adsorption ratio and concentration of eight major ions at different depths, and generate a subsurface physicochemical dataset. S13. Collect microclimate environmental data of the target area through meteorological monitoring stations. The microclimate environmental data includes rainfall, evaporation, temperature and relative humidity, and generate microclimate characteristic data. S14. In response to the user's instruction to conduct a detailed regional salinity survey, obtain the boundary coordinates of the sub-region to be analyzed, and retrieve historical salinity background data and geological and hydrological data from the regional database based on the boundary coordinates.

[0008] Preferably, the salt migration coupling analysis in S2 includes the following steps: S21. Obtain the surface environmental feature data and the underground stratified physicochemical dataset; S22. Based on the surface reflectance spectrum and vegetation cover index in the surface environmental characteristic data, the surface soil salinity content is inverted to generate surface salinity inversion data. S23. Based on the electrical conductivity and water content vertical profile data in the underground stratified physicochemical dataset, construct a one-dimensional vertical salt diffusion model, calculate the salt flux and accumulation at different depths, and generate underground salt distribution simulation data. S24. Couple the surface salinity inversion data with the underground salinity distribution simulation data, introduce microclimate characteristic data as boundary conditions, perform joint simulation of water and salt transport, and generate regional salinity spatiotemporal distribution benchmark data.

[0009] Preferably, the gradient level division process in S3 includes the following steps: S31. Obtain the regional salinity spatiotemporal distribution benchmark data and the preset salinity gradient classification threshold set, wherein the salinity gradient classification threshold set includes salinity value ranges corresponding to light, moderate, heavy and extremely heavy levels. S32. Perform gridded slicing on the regional salinity spatiotemporal distribution benchmark data, and calculate the weighted average salinity of each grid cell; S33. Compare the weighted average salinity of each grid cell with the corresponding numerical range in the salinity gradient classification threshold set, assign the corresponding gradient level label based on the comparison result, and generate preliminary salinity gradient partition data. S34. Perform topological continuity verification on the preliminary salinity gradient partition data, eliminate isolated noise points, and generate smoothed preliminary salinity gradient partition data.

[0010] Preferably, the salt gradient drift correction process in S4 includes the following steps: S41. Identify key inflection points in the preliminary salinity gradient partitioning data and extract buffer data at the junctions of different gradient levels. S42. Acquire real-time dynamic monitoring data from sensors deployed in the buffer zone, wherein the real-time dynamic monitoring data includes high-frequency collected data on soil solution salinity concentration and groundwater level changes. S43. Based on the real-time dynamic monitoring data, calculate the displacement distance and direction of the salinity interface per unit time, and generate a gradient drift vector; S44. Based on the gradient drift vector, the boundary of the preliminary salinity gradient partition data is interpolated and corrected to generate dynamically calibrated salinity gradient distribution data.

[0011] Preferably, the salt content identification feature parameters of the target region in S5 include the following steps: S51. Normalize the surface environmental characteristic data, the underground stratified physicochemical dataset, the microclimate characteristic data, the preliminary salinity gradient zoning data, and the dynamically calibrated salinity gradient distribution data. S52. Map the normalized multi-source heterogeneous data to a unified geographic coordinate system and construct a multi-dimensional feature tensor indexed by spatial location. S53. The multidimensional feature tensor is weighted by feature importance to extract deep semantic features that characterize the source, migration path and accumulation state of salt, and generate salt identification feature parameters.

[0012] Preferably, the multi-scale feature fusion and gradient boundary calculation processing in S6 includes the following steps: S61. Obtain the salt content identification feature parameters; S62. Input the salt identification feature parameters into the pre-trained salt gradient interpretation model. The salt gradient interpretation model adopts an encoder and decoder architecture and extracts the spatial distribution features of salt under different receptive fields through a multi-scale dilated convolution module. S63. In the decoding stage, an attention mechanism is introduced to weight the pixel features of the gradient boundary region, enhance the model's ability to perceive subtle gradient changes, and generate a pixel-level salt gradient probability map. S64. Perform cluster analysis on the pixel-level salt gradient probability map, and perform boundary regularization in combination with the dynamically calibrated salt gradient distribution data to generate salt gradient identification result data with clear boundaries.

[0013] Preferably, the visualization map construction process in S7 includes the following steps: S71. Overlay the salt gradient identification result data with clear boundaries with the terrain elevation undulation data; S72. Configure differentiated rendering colors and transparency according to different salt gradient levels to generate two-dimensional planar distribution maps and three-dimensional perspective views; S73. Mark the measured data and improvement suggestion labels of key monitoring points on the two-dimensional planar distribution map and the three-dimensional perspective view to generate salinity gradient spatial distribution map data; S74. The salinity gradient spatial distribution map data is encapsulated into a standard geospatial data format, supporting interactive queries on GIS platforms and mobile devices.

[0014] Preferably, the generation and issuance of soil improvement control command data in step S8 includes the following steps: S81. Analyze the salinity gradient spatial distribution map data to identify high-salinity areas that need priority improvement. S82. Based on the salt composition and gradient level of the high-salinity area, match the preset improvement strategy library to generate an improvement plan that includes irrigation quota, amount of amendment agent and planting structure. S83. The improved scheme is converted into specific equipment control commands and sent to the intelligent irrigation system and the unmanned agricultural machinery execution terminal. S84. Monitor the changes in salinity gradient after the implementation of improvement measures to form a closed-loop control.

[0015] Preferably, the system includes: The multi-source data acquisition and preprocessing module uses remote sensing units and sensor arrays to collect environmental data of the target area, performs standardization and cleaning through edge computing units, and outputs surface environmental feature data and subsurface stratified physicochemical datasets through data fusion units. The salinity migration coupling analysis module receives the surface environmental characteristic data and the underground stratified physicochemical dataset, constructs a vertical diffusion model through the water-salt transport simulation unit, integrates multi-source information through the data assimilation unit, and outputs regional salinity spatiotemporal distribution benchmark data through the gridded unit. The gradient division and dynamic correction module receives the regional salinity spatiotemporal distribution benchmark data, performs grade division through the threshold comparison unit, captures boundary changes through the drift monitoring unit, and outputs dynamically calibrated salinity gradient distribution data through the interpolation correction unit. The feature fusion and intelligent interpretation module receives the dynamically calibrated salt gradient distribution data, constructs a multidimensional feature tensor through the feature engineering unit, performs inference calculations using a pre-trained salt gradient interpretation model, and outputs salt gradient identification result data with clear boundaries through the boundary refinement unit. The visualization and decision execution module receives the salinity gradient identification result data with clear boundaries, generates a spatial distribution map through the map rendering unit, generates an optimization scheme through the improvement strategy matching unit, and sends control commands to the soil improvement execution device through the command issuing unit.

[0016] Compared with the prior art, the present invention provides a method and system for identifying the salinity gradient of saline soil, which has the following beneficial effects: 1. In this invention, when identifying the salinity gradient of saline soil, multi-source environmental data of the surface, physical and chemical data of underground layered soil, and microclimate environmental data are collected. Combined with salt migration coupling analysis, regional salinity spatiotemporal distribution benchmark data are generated to ensure the comprehensiveness and accuracy of the salinity distribution benchmark and provide a reliable basis for gradient division.

[0017] 2. In this invention, when dividing the salinity gradient, the regional salinity spatiotemporal distribution benchmark data is divided into grid levels by a preset salinity gradient classification threshold set, and gradient drift correction is performed in combination with real-time dynamic monitoring data to ensure the rationality of the initial salinity gradient partitioning and improve the dynamic calibration accuracy of the salinity gradient distribution.

[0018] 3. In this invention, when generating salinity gradient identification results, salinity identification feature parameters are constructed by integrating multi-source data, multi-scale feature fusion and gradient boundary calculation are performed using a pre-trained salinity gradient interpretation model, and soil improvement control commands are triggered by combining visualization maps to ensure the clarity of the boundaries of salinity gradient identification results and realize closed-loop control from identification to improvement. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for identifying the salinity gradient of saline soil according to the present invention; Figure 2 This is a schematic diagram of the architecture of a salinity gradient identification system for saline soil according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] For specific implementation examples, please refer to: Figures 1-2 A method and system for identifying salinity gradients in saline soil, the method comprising the following steps: S1. Collect surface multi-source environmental data and subsurface layered soil physicochemical data of the target saline soil area; S2. Based on surface multi-source environmental data and underground layered soil physicochemical data, salt migration coupling analysis is performed to generate regional salt spatiotemporal distribution benchmark data. S3. Based on the regional salinity spatiotemporal distribution benchmark data and the preset salinity gradient classification threshold set, perform gradient level division processing to generate preliminary salinity gradient partition data. S4. Based on the preliminary salinity gradient zoning data and real-time dynamic monitoring data, perform salinity gradient drift correction processing to generate dynamically calibrated salinity gradient distribution data. S5. Integrate surface multi-source environmental data, subsurface layered soil physicochemical data, preliminary salinity gradient zoning data, and dynamically calibrated salinity gradient distribution data into salinity identification feature parameters for the target area. S6. Based on the salt identification feature parameters, multi-scale feature fusion and gradient boundary calculation are performed through a pre-trained salt gradient interpretation model to generate salt gradient identification result data with clear boundaries. S7. Based on the high-precision salinity gradient identification results, perform visualization map construction processing to generate salinity gradient spatial distribution map data; S8. Based on the spatial distribution map data of salinity gradient, trigger the execution instructions for irrigation and salt leaching and biological improvement, and generate and issue soil improvement control instruction data.

[0022] S1 collects surface multi-source environmental data and subsurface layered soil physicochemical data of the target saline soil area, including the following steps: S11. Collect multi-source surface environmental data of the target area using satellite remote sensing equipment and UAV-borne spectrometers. The multi-source surface environmental data includes surface reflectance spectrum, surface temperature, vegetation cover index and surface roughness, and generate surface environmental characteristic data. S12. Collect subsurface physicochemical data of the target area through a distributed soil sensor array. The subsurface physicochemical data includes soil moisture content, electrical conductivity, pH value, sodium ion adsorption ratio and concentration of eight major ions at different depths, and generate a subsurface physicochemical dataset. S13. Collect microclimate environmental data of the target area through meteorological monitoring stations. The microclimate environmental data includes rainfall, evaporation, temperature and relative humidity, and generate microclimate characteristic data. S14. In response to the user's instruction to conduct a detailed regional salinity survey, obtain the boundary coordinates of the sub-region to be analyzed, and retrieve historical salinity background data and geological and hydrological data from the regional database based on the boundary coordinates.

[0023] The coupled analysis of salt migration in S2 includes the following steps: S21. Obtain surface environmental characteristic data and subsurface stratified physicochemical dataset; S22. Based on the surface reflectance spectrum and vegetation cover index in the surface environmental characteristic data, the surface soil salinity content is inverted to generate surface salinity inversion data. Based on surface reflectance spectra and vegetation cover index from surface environmental characteristic data, a nonlinear mapping relationship between surface reflectance and soil electrical conductivity is established using a semi-empirical radiative transfer model calibrated with ground-measured hyperspectral data. By performing principal component dimensionality reduction on the multidimensional spectral feature vector, the band features most sensitive to salt stress are extracted, thereby calculating the distribution of soil salinity in the target area. ; in This is the inversion value of surface soil salinity. For a specific band The surface reflectance below, The vegetation cover index, This is a salt inversion function model. This is the random error term of the model; S23. Based on the electrical conductivity and water content vertical profile data in the underground stratified physicochemical dataset, a one-dimensional vertical salt diffusion model is constructed to calculate the salt flux and accumulation at different depths, and to generate underground salt distribution simulation data. Combining soil texture parameters and meteorological boundary conditions, this step constructs a one-dimensional vertical soil water and salt transport dynamics model; by numerically discretizing the Richards equation and the diffusion equation with additional source and sink terms, the model simulates the movement trajectory of water on the soil profile and the migration process of salt it carries; within each calculation time step, the model synchronously updates the salt mass balance of each soil layer node, thereby calculating the vertical salt flux at different soil depths and the total salt accumulation in a specific soil layer within a specific time period. The governing equation for vertical soil salt transport in one dimension and its difference form are as follows: ; ; in Soil volumetric water content, This refers to the salt concentration in the soil solution. For time variables, Vertical depth coordinates Soil moisture diffusivity, This represents the vertical soil moisture flux. For salt source and sink items, For depth At any time The vertical flux of salt. This represents the vertical spatial gradient of salt concentration. S24. Couple surface salinity inversion data with underground salinity distribution simulation data, introduce microclimate characteristic data as boundary conditions, conduct joint simulation of water and salt transport, and generate regional salinity spatiotemporal distribution benchmark data. Rainfall, potential evaporation, and temperature from microclimate characteristic data are introduced as upper boundary conditions of the model. Microclimate driving data are coupled with soil hydraulic property parameters and substituted into the unsaturated water flow field described by Richards equations for solution. On this basis, the trajectory of dissolved salt moving with water is tracked simultaneously to realize dynamic simulation of water-salt coupling, and finally generate regional salinity spatiotemporal distribution benchmark data covering spatiotemporal dimensions. ; in The rate of change of soil moisture over time. For vertical depth Directional differentiation operator, For soil water matrix potential, Let be the soil hydraulic conductivity, which is a function of the matrix potential.

[0024] The gradient level partitioning process in S3 includes the following steps: S31. Obtain the regional spatiotemporal distribution benchmark data and the preset salinity gradient classification threshold set, which includes the salinity value range corresponding to the light, moderate, heavy and extremely heavy levels. S32. Perform gridded slicing on the regional salinity spatiotemporal distribution benchmark data, and calculate the weighted average salinity of each grid cell; ; in For the first The weighted average salt content of each grid cell. To fall into the first The number of valid data points within each grid cell For the first Salt content of each data point For the first Spatial weighting coefficients for each data point; S33. Compare the weighted average salinity of each grid cell with the corresponding numerical range in the salinity gradient classification threshold set, assign the corresponding gradient level label based on the comparison result, and generate preliminary salinity gradient partition data. S34. Perform topological continuity verification on the preliminary salinity gradient partition data, eliminate isolated noise points, and generate smoothed preliminary salinity gradient partition data. The gradient level labels of all grid cells are analyzed, and the distribution of salinity levels in the neighborhood of each cell is statistically analyzed. When a cell is determined to be of high salinity level, but its neighboring cells are all of low salinity level, the cell is determined to be an anomalous isolated point in the topology. Its salinity level attribute will be automatically replaced with the mode level of the neighboring cells to correct the topological errors in the initial salinity gradient partitioning data. ; in After isolated noise correction, the coordinates are... The final salt gradient label of the grid cell at that location. Based on spatial neighborhood relationships, coordinates The majority vote result of the surrounding neighboring grid cells. coordinates The original, uncorrected initial salinity gradient labels are shown. This represents the absolute difference between monitoring indicators in high-salinity areas and those in low-salinity areas. This is the preset gradient difference threshold.

[0025] The salt gradient drift correction process in S4 includes the following steps: S41. Identify key inflection points in the preliminary salinity gradient partitioning data and extract buffer data at the boundaries of different gradient levels. Based on preliminary salinity gradient partitioning data, the spatial coordinates of abrupt changes between different salinity levels, i.e. salinity interface inflection points, are located using edge detection algorithms. Centered on these inflection points, the data are expanded outward along the direction perpendicular to the salinity gradient to extract continuous monitoring data of soil properties, hydrogeology, and microclimate within the strip area, and a high-precision buffer dataset for subsequent dynamic correction is constructed. ; in For the first The set of buffer zones at the boundaries of elasticity, For the first The coordinates of the center point of each key inflection point in the geographic coordinate system The distance is the radius of the buffer zone. For any spatial coordinate point within the buffer zone; S42. Acquire real-time dynamic monitoring data from sensors deployed within the buffer zone. The real-time dynamic monitoring data includes high-frequency collected data on soil solution salinity concentration and groundwater level changes. S43. Based on real-time dynamic monitoring data, calculate the displacement distance and direction of the salinity interface per unit time, and generate a gradient drift vector; Extract the soil solution salinity concentration change sequence of different time phases recorded by high-frequency monitoring equipment in the buffer data, and determine the movement lag time of the salinity interface through cross-correlation analysis; combined with the monitoring time interval, calculate the displacement rate and normal deflection angle of the salinity interface per unit time, thereby quantifying the dynamic migration trend of the salinity interface. ; ; in The velocity vector representing the salinity interface migration. This represents the displacement distance of the salinity interface between the two monitoring sessions. The time interval between two monitoring sessions. The azimuth angle of the salt interface migration. , The components of the boundary displacement in the east-west and north-south directions; S44. Based on the gradient drift vector, interpolate and correct the boundaries of the preliminary salinity gradient partition data to generate dynamically calibrated salinity gradient distribution data. After obtaining the velocity vector, it is applied as a constraint on the edge of the initial salinity gradient partitioning data as the boundary offset. The Kriging interpolation algorithm is used to estimate the smooth offset trajectory of the entire boundary segment using the known offset of the boundary node, and then the gradient partitioning boundary is redrawn to complete the geometric correction of the initial partitioning result. ; in Target position after boundary correction The new coordinate values, For the first Interpolation weights of each sample point to the target point No. The original coordinates of each sample boundary point The total number of boundary sample points involved in the interpolation.

[0026] The salt content identification feature parameters of the target region in S5 include the following steps: S51. Normalize the surface environmental characteristic data, subsurface stratified physicochemical dataset, microclimate characteristic data, preliminary salinity gradient zoning data and dynamically calibrated salinity gradient distribution data. Normalization formula: ; in These are the normalized feature values. The original input feature values, , These are the global maximum and minimum values ​​for this feature dimension; S52. Map the normalized multi-source heterogeneous data to a unified geographic coordinate system and construct a multi-dimensional feature tensor indexed by spatial location. S53. Weight the multidimensional feature tensor according to feature importance, extract deep semantic features that characterize the source, migration path and accumulation state of salt, and generate salt identification feature parameters. This step uses a feature importance evaluation algorithm to screen out the dominant factors, and uses a random forest to train and parse the feature tensor, calculating the information gain of each feature during the node splitting process. Based on the score, high weights are assigned to key features that are strongly correlated with salt source, migration path and cumulative state. Subsequently, the weights are mapped back to the original feature tensor by element-wise multiplication to achieve dynamic weighting of features and enhance the model's ability to perceive core semantics. The formula for weighted calculation of feature importance is as follows: ; in For the weighted feature tensor, For the original input feature tensor, The feature importance weight matrix is ​​generated during model training and satisfies... , This is the Hadamard product operator; The multi-scale feature fusion and gradient boundary calculation processing in S6 includes the following steps: S61. Obtain salt content identification feature parameters; S62. Input the salt identification feature parameters into the pre-trained salt gradient interpretation model. The salt gradient interpretation model adopts an encoder and decoder architecture and extracts the spatial distribution features of salt under different receptive fields through a multi-scale dilated convolution module. We construct a deep learning network backbone based on encoder-decoder, embed multi-scale dilated convolutional layers during encoder downsampling, and by setting different dilation rate parameters, the network can capture spatial context information of different scales from local details to large-scale global areas in salt distribution images in parallel without increasing the number of parameters. ; in To output feature map at position Pixel value at that location, Input the pixel value of the feature map at the corresponding position. void ratio, For the convolution kernel in the th Weight parameters for each position, This represents the effective size of the convolution kernel; S63. In the decoding stage, an attention mechanism is introduced to weight the pixel features of the gradient boundary region, enhance the model's ability to perceive subtle gradient changes, and generate a pixel-level salt gradient probability map. While extracting features during forward propagation of the network, a batch normalization layer is introduced to standardize the intermediate activation values. When calculating the gradient during backpropagation of the network, the Adam optimizer is used to dynamically adjust the update step size of the weights of each layer to ensure that the convergence speed and stability of the model parameters are adaptively balanced when facing complex and variable water and salt transport data. ; ; in for The gradient of the loss function at time step 1. For a moment The updated model parameter vector, The global learning rate, For model parameters gradient operator, , These are the estimated values ​​of the first and second moments of the gradient after bias correction. For a moment The loss function value, For a moment The model parameter vector, For a moment The model parameter vector, It is a very small constant to prevent numerically stable terms with a denominator of zero; The initial learning rate of the Adam optimizer is set to 10. -4 The batch size is set to 32, the number of iterations is no less than 100, the specific choice of the loss function and the early stopping mechanism used to prevent overfitting are also specified. S64. Perform cluster analysis on the pixel-level salt gradient probability map, and perform boundary regularization in combination with the dynamically calibrated salt gradient distribution data to generate salt gradient identification result data with clear boundaries. The pixel-level salt gradient probability map output by the deep learning model is spatially superimposed with the dynamically calibrated salt gradient distribution data; density clustering algorithm is used to classify discrete pixel clusters in the boundary transition zone to identify the real salt interface line; then morphological closure operation is applied to fill the gaps and smooth the jagged edges, and finally the salt gradient identification result data with complete topological structure and clear boundary is output. ; , ; in For boundary pixels and Euclidean distance between them , For pixels The horizontal and vertical coordinates in a geographic coordinate system , For pixels The horizontal and vertical coordinates in a geographic coordinate system The set of spatial data points to be processed. This is the distance threshold for cluster merging. For the first Each salt gradient boundary cluster.

[0027] The visualization graph construction process in S7 includes the following steps: S71. Overlay the salt gradient identification results with clear boundaries with the terrain elevation undulation data; S72. Configure differentiated rendering colors and transparency according to different salt gradient levels to generate two-dimensional planar distribution maps and three-dimensional perspective views; S73. Mark the measured data and improvement suggestion labels of key monitoring points on the two-dimensional planar distribution map and the three-dimensional perspective view to generate salinity gradient spatial distribution map data; S74. Encapsulate the spatial distribution map data of salinity gradient into a standard geospatial data format, supporting interactive queries on GIS platforms and mobile devices.

[0028] The generation and issuance of soil improvement control command data in S8 includes the following steps: S81. Analyze the spatial distribution map data of salinity gradient to identify high-salinity areas that need priority improvement. S82. Based on the salt composition and gradient level of the high-salinity area, match the preset improvement strategy library to generate an improvement plan that includes irrigation quota, amount of amendment agent and planting structure. S83. Transform the improvement plan into specific equipment control commands and send them to the intelligent irrigation system and unmanned agricultural machinery execution terminals; S84. Monitor the changes in salinity gradient after the implementation of improvement measures to form a closed-loop control.

[0029] The system includes: The multi-source data acquisition and preprocessing module uses remote sensing units and sensor arrays to collect environmental data of the target area, performs standardization and cleaning through edge computing units, and outputs surface environmental feature data and subsurface stratified physicochemical datasets through data fusion units. The salinity migration coupling analysis module receives surface environmental characteristic data and underground stratified physicochemical datasets. It constructs a vertical diffusion model through a water-salt transport simulation unit, integrates multi-source information using a data assimilation unit, and outputs regional salinity spatiotemporal distribution benchmark data through a gridded unit. The gradient division and dynamic correction module receives the regional salinity spatiotemporal distribution benchmark data, performs grade division through the threshold comparison unit, captures boundary changes through the drift monitoring unit, and outputs dynamically calibrated salinity gradient distribution data through the interpolation correction unit. The feature fusion and intelligent interpretation module receives dynamically calibrated salt gradient distribution data, constructs a multidimensional feature tensor through the feature engineering unit, performs inference calculations using a pre-trained salt gradient interpretation model, and outputs salt gradient recognition result data with clear boundaries through the boundary refinement unit. The visualization and decision execution module receives salinity gradient identification results with clear boundaries, generates spatial distribution maps through the map rendering unit, generates optimization schemes through the improvement strategy matching unit, and sends control commands to the soil improvement execution equipment through the command issuing unit.

[0030] The operation steps of a method and system for identifying salinity gradients in saline soil are as follows: Step 1: Multi-source data acquisition and preprocessing: Multi-source surface environmental data, including surface reflectance spectrum, surface temperature, vegetation cover index, and surface roughness, were collected using satellite remote sensing equipment and UAV-borne spectrometers. Subsurface stratified soil physicochemical data, including soil moisture content, electrical conductivity, pH value, sodium ion adsorption ratio, and concentrations of eight major ions at different depths, were collected using a distributed soil sensor array. Simultaneously, microclimate environmental data, including rainfall, evaporation, air temperature, and relative humidity, were collected from meteorological monitoring stations. In response to regional salinity detailed investigation commands, the boundary coordinates of the sub-regions to be analyzed and the corresponding historical salinity background data, geological and hydrological data were obtained. After standardization, cleaning, and fusion, surface environmental characteristic data, subsurface stratified physicochemical datasets, and microclimate characteristic data were output.

[0031] Step 2: Salinity Migration Coupling Analysis and Benchmark Generation Surface salinity inversion data is generated by inverting surface soil salinity content based on surface reflectance spectrum and vegetation cover index from surface environmental characteristic data. A one-dimensional vertical salinity diffusion model is constructed based on electrical conductivity and water content vertical profile data from underground stratified physicochemical dataset. Salt flux and accumulation at different depths are calculated to generate underground salinity distribution simulation data. Surface salinity inversion data and underground salinity distribution simulation data are coupled, and microclimate characteristic data are introduced as boundary conditions to conduct joint simulation of water and salt transport, generating regional salinity spatiotemporal distribution benchmark data.

[0032] Step 3: Gradient Level Division and Dynamic Calibration Based on a preset threshold set for salinity gradient grading, which includes salinity ranges of light, moderate, heavy, and extremely heavy levels, the regional salinity spatiotemporal distribution baseline data is processed into gridded slices, and the weighted average salinity of each grid cell is calculated. The weighted average salinity is compared with the threshold set and assigned a gradient level label. After topological continuity verification to eliminate isolated noise points, preliminary salinity gradient partition data is generated. Buffer data at the boundaries of different gradient levels are extracted. Combined with real-time dynamic monitoring data of soil solution salinity concentration and groundwater level changes collected at high frequencies within the buffer zone, the displacement distance and direction of the salinity interface per unit time are calculated to generate a gradient drift vector. The boundaries of the preliminary salinity gradient partition data are interpolated and corrected to generate dynamically calibrated salinity gradient distribution data.

[0033] Step 4: Feature Fusion and Intelligent Interpretation The surface environmental feature data, subsurface stratified physicochemical dataset, microclimate feature data, preliminary salinity gradient zoning data, and dynamically calibrated salinity gradient distribution data are uniformly normalized and mapped to a unified geographic coordinate system to construct a multidimensional feature tensor indexed by spatial location. Deep semantic features representing salinity sources, migration paths, and accumulation states are extracted through feature importance weighting to generate salinity identification feature parameters. These parameters are then input into a pre-trained salinity gradient interpretation model employing an encoder-decoder architecture. A multi-scale dilated convolution module extracts the spatial distribution features of salinity under different receptive fields. During the decoding stage, an attention mechanism is introduced to weight the pixel features of gradient boundary regions, generating pixel-level salinity gradient probability maps. After cluster analysis and boundary normalization, clearly defined salinity gradient identification results are generated.

[0034] Step 5: Visualization and Decision Implementation The data of clearly defined salinity gradient identification is overlaid with topographic elevation data. Different rendering colors and transparency are configured according to different salinity gradient levels to generate two-dimensional planar distribution maps and three-dimensional perspective views. After labeling the measured data of key monitoring points and improvement suggestion tags, the data is packaged into a standard geospatial data format to generate salinity gradient spatial distribution map data. The map data is analyzed to identify high-salinity areas that need priority improvement. The pre-set improvement strategy library is matched to generate improvement plans that include irrigation quotas, improvement agent dosages and planting structures. The plans are then converted into specific equipment control commands and sent to the intelligent irrigation system and unmanned agricultural machinery execution terminals. The changes in salinity gradient after improvement are monitored to form a closed-loop control.

[0035] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0036] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for identifying the salinity gradient of saline soil, characterized in that: The method includes the following steps: S1. Collect surface multi-source environmental data and subsurface layered soil physicochemical data of the target saline soil area; S2. Based on the surface multi-source environmental data and the underground layered soil physicochemical data, perform salt migration coupling analysis to generate regional salt spatiotemporal distribution benchmark data. S3. Based on the regional salinity spatiotemporal distribution benchmark data and the preset salinity gradient classification threshold set, perform gradient level division processing to generate preliminary salinity gradient partition data. S4. Based on the preliminary salinity gradient partitioning data and real-time dynamic monitoring data, perform salinity gradient drift correction processing to generate dynamically calibrated salinity gradient distribution data. S5. Integrate the surface multi-source environmental data, the underground layered soil physicochemical data, the preliminary salinity gradient zoning data, and the dynamically calibrated salinity gradient distribution data into salinity identification feature parameters for the target area. S6. Based on the salinity identification feature parameters, multi-scale feature fusion and gradient boundary calculation are performed through a pre-trained salinity gradient interpretation model to generate salinity gradient identification result data with clear boundaries. S7. Based on the high-precision salinity gradient identification result data, perform visualization map construction processing to generate salinity gradient spatial distribution map data; S8. Based on the salinity gradient spatial distribution map data, trigger the execution instructions for irrigation and salt leaching and biological improvement, and generate and issue soil improvement control instruction data.

2. The method for identifying salinity gradients in saline soil according to claim 1, characterized in that: The step S1 involves collecting surface multi-source environmental data and subsurface layered soil physicochemical data of the target saline soil area, including the following steps: S11. Collect multi-source surface environmental data of the target area using satellite remote sensing equipment and UAV-borne spectrometer. The multi-source surface environmental data includes surface reflectance spectrum, surface temperature, vegetation cover index and surface roughness, and generate surface environmental characteristic data. S12. Collect subsurface physicochemical data of the target area through a distributed soil sensor array. The subsurface physicochemical data includes soil moisture content, electrical conductivity, pH value, sodium ion adsorption ratio and concentration of eight major ions at different depths, and generate a subsurface physicochemical dataset. S13. Collect microclimate environmental data of the target area through meteorological monitoring stations. The microclimate environmental data includes rainfall, evaporation, temperature and relative humidity, and generate microclimate characteristic data. S14. In response to the user's instruction to conduct a detailed regional salinity survey, obtain the boundary coordinates of the sub-region to be analyzed, and retrieve historical salinity background data and geological and hydrological data from the regional database based on the boundary coordinates.

3. The method for identifying salinity gradients in saline soil according to claim 2, characterized in that: The salt migration coupling analysis in S2 includes the following steps: S21. Obtain the surface environmental feature data and the underground stratified physicochemical dataset; S22. Based on the surface reflectance spectrum and vegetation cover index in the surface environmental characteristic data, the surface soil salinity content is inverted to generate surface salinity inversion data. S23. Based on the electrical conductivity and water content vertical profile data in the underground stratified physicochemical dataset, construct a one-dimensional vertical salt diffusion model, calculate the salt flux and accumulation at different depths, and generate underground salt distribution simulation data. S24. Couple the surface salinity inversion data with the underground salinity distribution simulation data, introduce microclimate characteristic data as boundary conditions, perform joint simulation of water and salt transport, and generate regional salinity spatiotemporal distribution benchmark data.

4. The method for identifying salinity gradients in saline soil according to claim 3, characterized in that: The gradient level partitioning process in S3 includes the following steps: S31. Obtain the regional salinity spatiotemporal distribution benchmark data and the preset salinity gradient classification threshold set, wherein the salinity gradient classification threshold set includes salinity value ranges corresponding to light, moderate, heavy and extremely heavy levels. S32. Perform gridded slicing on the regional salinity spatiotemporal distribution benchmark data, and calculate the weighted average salinity of each grid cell; S33. Compare the weighted average salinity of each grid cell with the corresponding numerical range in the salinity gradient classification threshold set, assign the corresponding gradient level label based on the comparison result, and generate preliminary salinity gradient partition data. S34. Perform topological continuity verification on the preliminary salinity gradient partition data, eliminate isolated noise points, and generate smoothed preliminary salinity gradient partition data.

5. The method for identifying salinity gradients in saline soil according to claim 4, characterized in that: The salt gradient drift correction process in S4 includes the following steps: S41. Identify key inflection points in the preliminary salinity gradient partitioning data and extract buffer data at the boundaries of different gradient levels. S42. Acquire real-time dynamic monitoring data from sensors deployed in the buffer zone, wherein the real-time dynamic monitoring data includes high-frequency collected data on soil solution salinity concentration and groundwater level changes. S43. Based on the real-time dynamic monitoring data, calculate the displacement distance and direction of the salinity interface per unit time, and generate a gradient drift vector; S44. Based on the gradient drift vector, the boundary of the preliminary salinity gradient partition data is interpolated and corrected to generate dynamically calibrated salinity gradient distribution data.

6. The method for identifying salinity gradients in saline soil according to claim 5, characterized in that: The salt content identification feature parameters of the target region in S5 include the following steps: S51. Normalize the surface environmental characteristic data, the underground stratified physicochemical dataset, the microclimate characteristic data, the preliminary salinity gradient zoning data, and the dynamically calibrated salinity gradient distribution data. S52. Map the normalized multi-source heterogeneous data to a unified geographic coordinate system and construct a multi-dimensional feature tensor indexed by spatial location. S53. The multidimensional feature tensor is weighted by feature importance to extract deep semantic features that characterize the source, migration path and accumulation state of salt, and generate salt identification feature parameters.

7. The method for identifying salinity gradients in saline soil according to claim 6, characterized in that: The multi-scale feature fusion and gradient boundary calculation process in S6 includes the following steps: S61. Obtain the salt content identification feature parameters; S62. Input the salt identification feature parameters into the pre-trained salt gradient interpretation model. The salt gradient interpretation model adopts an encoder and decoder architecture and extracts the spatial distribution features of salt under different receptive fields through a multi-scale dilated convolution module. S63. In the decoding stage, an attention mechanism is introduced to weight the pixel features of the gradient boundary region, enhance the model's ability to perceive subtle gradient changes, and generate a pixel-level salt gradient probability map. S64. Perform cluster analysis on the pixel-level salt gradient probability map, and perform boundary regularization in combination with the dynamically calibrated salt gradient distribution data to generate salt gradient identification result data with clear boundaries.

8. The method for identifying salinity gradients in saline soil according to claim 7, characterized in that: The visualization map construction process in S7 includes the following steps: S71. Overlay the salt gradient identification result data with clear boundaries with the terrain elevation undulation data; S72. Configure differentiated rendering colors and transparency according to different salt gradient levels to generate two-dimensional planar distribution maps and three-dimensional perspective views; S73. Mark the measured data and improvement suggestion labels of key monitoring points on the two-dimensional planar distribution map and the three-dimensional perspective view to generate salinity gradient spatial distribution map data; S74. The salinity gradient spatial distribution map data is encapsulated into a standard geospatial data format, supporting interactive queries on GIS platforms and mobile devices.

9. The method for identifying salinity gradients in saline soil according to claim 8, characterized in that: The generation and distribution of soil improvement control command data in S8 includes the following steps: S81. Analyze the salinity gradient spatial distribution map data to identify high-salinity areas that need priority improvement. S82. Based on the salt composition and gradient level of the high-salinity area, match the preset improvement strategy library to generate an improvement plan that includes irrigation quota, amount of amendment agent and planting structure. S83. The improved scheme is converted into specific equipment control commands and sent to the intelligent irrigation system and the unmanned agricultural machinery execution terminal. S84. Monitor the changes in salinity gradient after the implementation of improvement measures to form a closed-loop control.

10. A salinity gradient identification system for saline soil, used to implement the salinity gradient identification method for saline soil according to any one of claims 1-9, characterized in that: The system includes: The multi-source data acquisition and preprocessing module uses remote sensing units and sensor arrays to collect environmental data of the target area, performs standardization and cleaning through edge computing units, and outputs surface environmental feature data and subsurface stratified physicochemical datasets through data fusion units. The salinity migration coupling analysis module receives the surface environmental characteristic data and the underground stratified physicochemical dataset, constructs a vertical diffusion model through the water-salt transport simulation unit, integrates multi-source information through the data assimilation unit, and outputs regional salinity spatiotemporal distribution benchmark data through the gridded unit. The gradient division and dynamic correction module receives the regional salinity spatiotemporal distribution benchmark data, performs grade division through the threshold comparison unit, captures boundary changes through the drift monitoring unit, and outputs dynamically calibrated salinity gradient distribution data through the interpolation correction unit. The feature fusion and intelligent interpretation module receives the dynamically calibrated salt gradient distribution data, constructs a multidimensional feature tensor through the feature engineering unit, performs inference calculations using a pre-trained salt gradient interpretation model, and outputs salt gradient identification result data with clear boundaries through the boundary refinement unit. The visualization and decision execution module receives the salinity gradient identification result data with clear boundaries, generates a spatial distribution map through the map rendering unit, generates an optimization scheme through the improvement strategy matching unit, and sends control commands to the soil improvement execution device through the command issuing unit.

Images