Soil attribute intelligent prediction method and system based on multi-dimensional dynamic scale

Abstract

The invention discloses a soil attribute intelligent prediction method and system based on a multi-dimensional dynamic scale, and relates to the technical field of soil prediction.The method comprises the steps that remote sensing images, terrain factors, meteorological factors and soil actual measurement sample data are collected and stored in a multi-source raster data bin in a unified mode; preprocessing the multi-source data, and constructing raster data in a unified format; combining terrain difference, land utilization and climate fluctuation, dynamically dividing a spatial scale, a time scale, a terrain layering scale and a land utilization pattern scale, and constructing a multi-dimensional dynamic scale feature expression vector; carrying out soil attribute modeling and training based on a multi-task deep neural network model; and applying the model to a non-actually-measured area, and outputting a prediction result and a spatial distribution map layer. The system comprises a multi-source raster data construction and management module, a multi-dimensional dynamic scale feature construction module and a soil attribute modeling and prediction module. The method can be used for forestry management and protection, resource management, ecological monitoring, resource evaluation and other scenes.

Description

Technical Field

[0001] The present invention relates to the field of intelligent prediction technology, and in particular to a soil property intelligent prediction method and system based on multi-dimensional dynamic scale. Background Art

[0002] Accurate prediction of soil properties is of great significance to a variety of fields, including forest resource management, environmental protection, and land use planning. Traditional soil property prediction methods rely on ground sampling and measured data. However, due to the uneven spatial and temporal distribution of soil data and limitations in sample acquisition, these methods face significant challenges in large-scale regional prediction applications.

[0003] Although multi-source heterogeneous data, such as remote sensing imagery, topographic data, and meteorological data, provide a rich foundation for soil property prediction, effective integration, feature expression, and scale selection of these data remain significant challenges. Existing research primarily employs single-scale or static modeling approaches, lacking systematic modeling and mining of multidimensional dynamic scale features. This makes it difficult to capture the complex interactions between soil properties and environmental factors, limiting the accuracy and adaptability of model predictions. Summary of the Invention

[0004] In order to solve the above technical problems, a method and system for intelligent prediction of soil properties based on multidimensional dynamic scales are provided. This technical solution solves the problem proposed in the above background technology that most existing studies adopt single-scale or static-scale modeling methods, lack systematic modeling and mining of multidimensional dynamic scale characteristics, and find it difficult to capture the complex interactive relationship between soil properties and environmental factors, which limits the model prediction accuracy and adaptability.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is: A soil property intelligent prediction method based on multi-dimensional dynamic scaling, including: Acquire remote sensing image data, terrain factor data, meteorological factor data, and measured soil attribute sample data of the target area, and store the above data in a multi-source raster data warehouse; Based on the multi-source raster data warehouse, remote sensing image data, terrain factor data and meteorological factor data are preprocessed to construct multi-source raster data in a unified format; Based on multi-source raster data in a unified format, geographical indicators reflecting terrain differences, land use patterns, and climate fluctuation characteristics are extracted to obtain terrain complexity index, climate volatility index, land use pattern comprehensive index, and comprehensive stratification factor. The spatial scale, temporal scale, terrain stratification scale, and land use pattern scale are then divided, and corresponding scale labels are assigned to each raster unit in the target area. Perform weighted joint encoding of various scale labels and the spatial coordinate information of the grid to construct the corresponding multi-dimensional dynamic scale feature expression vector; The multi-dimensional dynamic scale feature expression vector is associated with the soil property sample data at its corresponding geographical location to build a soil property prediction model. The soil property prediction model is trained using the collected soil property samples; The trained soil property prediction model is applied to geographical units in the target area that have not been measured to generate soil property prediction results and output a spatial distribution layer.

[0006] In an optional embodiment, the step of acquiring remote sensing image data, terrain factor data, meteorological factor data, and measured soil attribute sample data of a target area and uniformly storing the data in a multi-source raster data warehouse specifically includes: Remote sensing satellites equipped with multispectral sensors and thermal infrared imaging equipment acquire remote sensing image data related to visible light images, near infrared images, and thermal infrared images of target areas at different spatial resolutions; Obtain terrain factor data related to elevation, slope and aspect of the target area through digital elevation model; Obtain multi-temporal meteorological factor data related to precipitation, temperature, humidity, and wind speed in the target area from national meteorological monitoring stations and regional automatic weather stations; Through field sampling at ground sampling points set up in the target area, soil measured attribute sample data related to soil organic matter, pH value, moisture content and particle composition at the sampling points are obtained based on laboratory testing and analysis methods; Remote sensing image data, terrain factor data, meteorological factor data and soil measured attribute sample data are uniformly stored in a multi-source raster data warehouse.

[0007] In an optional embodiment, the remote sensing image data, terrain factor data and meteorological factor data are preprocessed based on the multi-source raster data warehouse to construct multi-source raster data in a unified format, specifically including: Performing image denoising, band fusion and spatial registration processing on the remote sensing image data to obtain standard remote sensing image data; Performing format standardization, reference space unification and spatial gridding processing on the terrain factor data to obtain standard terrain factor data; Performing time series difference, spatial mapping and rasterization processing on the meteorological factor data to obtain standard meteorological factor data; The standard remote sensing image data, standard terrain factor data and standard meteorological factor data are sequentially processed for spatial resolution unification, coordinate reference system conversion and grid structure alignment, and the processed grid units are uniformly coded and data format standardized to construct multi-source raster data in a unified format.

[0008] In an optional embodiment, the weighted joint encoding of various scale labels and spatial coordinate information of the grid to construct a corresponding multi-dimensional dynamic scale feature expression vector specifically includes: Based on multi-source raster data in a unified format, geographical indicators reflecting terrain differences, land use patterns and climate fluctuation characteristics are extracted to obtain terrain complexity index, climate volatility index, land use pattern comprehensive index and comprehensive stratification factor respectively; The slope value and aspect information of each grid cell are extracted from the standard terrain factor data. The slope aspect change rate is calculated by analyzing the amplitude of the slope aspect change within the local neighborhood of the grid cell. The slope value and slope aspect change rate of each grid cell are normalized respectively, and the normalized slope value and slope aspect change rate are weighted fused to obtain the terrain complexity index of each grid cell. : Where, is the target area, represents the slope of the grid, represents the slope change rate of the grid, is the total number of grids, is the grid number; Extract multi-temporal precipitation of each grid cell in the target area from the standard meteorological factor data and construct precipitation time series data; Based on this time series data, the standard deviation and mean of precipitation in each grid cell are calculated to further quantify the degree of fluctuation of precipitation in the temporal dimension. Constructing a climate volatility index by measuring the ratio of the standard deviation to the mean , the specific calculation formula is as follows: Where, represents the standard deviation of the precipitation time series, represents the mean of the precipitation time series; Extract land use type information from standard remote sensing image data, divide the target area into cultivated land, forest land, grassland, water body, and construction land through supervised classification algorithm, and rasterize and encode each land use type to obtain raster data of land use type; Based on the raster data of land use types and the spatial distribution of various land use types in the local neighborhood, the Shannon diversity index is calculated as an indicator of land use diversity. The formula is as follows: Where, Indicates the local neighborhood The proportion of land use types, is the total number of land use types; Using rasterized land use data, the boundary lengths and neighborhood grid areas between different land use types are calculated, and then the boundary density index is obtained. The land use diversity index and boundary density index are normalized and weighted to form a comprehensive index of land use pattern; From the standard terrain factor data, we extract elevation, slope, and aspect information, normalize the three factor data, and combine the normalized elevation, slope, and aspect values ​​to construct a comprehensive hierarchical factor that uniformly measures the complexity of the terrain structure. The calculation formula is as follows: Where, 、 are the elevation, slope and aspect values ​​of the normalized grid, ; According to the numerical range of the terrain complexity index, the target area is divided into multiple spatial scale levels; According to the size of climate volatility indicators, the target area is divided into time scale levels; Based on the comprehensive terrain stratification factor and land use pattern comprehensive index, the target area is divided into terrain stratification scale and land use pattern scale; Based on the division of spatial scale, temporal scale, terrain layer scale and land use pattern scale, each grid cell is assigned its corresponding spatial scale label, temporal scale label, terrain layer label and land use pattern label, and these labels are jointly encoded with their spatial location information to construct a multi-dimensional dynamic scale feature expression vector, which is expressed as follows: Where, and Represents the spatial coordinate position of the grid, Represents a grid The spatial scale level label, Represents a grid The time scale level label to which it belongs, Represents a grid The terrain layer labels are defined. Represents a grid Corresponding land use pattern stratification labels.

[0009] In an optional embodiment, the method of associating the multi-dimensional dynamic scale feature expression vector with the measured soil property sample data at the corresponding geographical location to construct a soil property prediction model specifically includes: Based on the measured soil sample data in the target area, the corresponding measured soil attribute values ​​are obtained, and the multi-dimensional dynamic scale feature expression vector at the position corresponding to its spatial coordinates is extracted to construct a training sample set. , where Indicates the Multi-dimensional dynamic scale expression vector of measured soil samples, Indicates the corresponding measured soil property value, Indicates the type of soil properties, Indicates the The first Class measured soil measured values, Number the measured soil samples; Based on the training sample set, a multi-task deep neural network model is used to jointly model multiple soil properties and construct a soil property prediction model; The constructed soil prediction model is trained. During the model training process, the prediction error of the current prediction model for the training sample set is analyzed based on the local clustering in spatial distribution. The scale division granularity corresponding to the error-significant area is analyzed to see if it matches. If it is judged to be mismatched, the division threshold and number of labels of the spatial scale level, temporal scale level and terrain stratification scale are further dynamically adjusted. On this basis, the mean square error loss of each soil property prediction task in the model is calculated separately, and an adjustable task weight factor is set to perform a weighted combination of the losses of each task to construct a unified multi-task prediction model loss function, as shown below: Where, Indicates the number of types of soil properties, It is represented as the number of training samples, Expressed as The first Class measured soil property values, is the soil property prediction value of the model, represents the set of model parameters, No. The loss weight corresponding to the soil-like attribute, is the regularization coefficient; Based on the loss function, an iterative optimization algorithm based on gradient descent is used to train the parameters of the multi-task deep neural network model, including the weights and bias terms of the shared representation layer and the output layers of each task; in each iteration, the gradient is calculated and the parameters are updated based on the weighted loss of the current model on the training samples; Combined with the prediction effect of the multi-dimensional dynamic scale feature expression vector in various soil property tasks, the task loss weight is dynamically adjusted to make the model focus on areas with significant scale changes; the training process continues until the loss decreases below the preset threshold.

[0010] In an optional embodiment, applying the trained soil property prediction model to geographical units in the target area that have not been measured to generate soil property prediction results and output a spatial distribution layer specifically includes: Locate the geographic coordinates of unsampled locations within the target area and obtain their spatial coordinate positions; Based on the spatial coordinate position, extracting a corresponding multi-dimensional dynamic scale feature expression vector including a spatial scale label, a temporal scale label and a terrain layer scale label from a multi-source raster data warehouse; Input the multi-dimensional dynamic scale feature expression vector into the trained soil property prediction model to obtain the corresponding soil property prediction value; All predicted values ​​are organized into continuous raster datasets according to spatial locations. The raster datasets are rendered using a geographic information system platform according to the color gradient mapping rules set for the attribute values ​​to generate a soil attribute spatial distribution layer with a color gradient.

[0011] Furthermore, a soil property intelligent prediction system based on multi-dimensional dynamic scaling is proposed to implement the above-mentioned prediction method, specifically including: A multi-source raster data construction and management module, which completes the acquisition, preprocessing and standardized fusion of remote sensing images, terrain factors and meteorological factors, constructs multi-source raster data in a unified format, and uniformly stores it in a multi-source data warehouse, while obtaining measured soil attribute sample data in the target area; A multidimensional dynamic scale feature construction module, which extracts geographical indicators reflecting terrain differences, land use patterns, and climate fluctuation characteristics, and then divides the spatial scale, temporal scale, terrain layer scale, and land use pattern scale, and fuses them to construct a multidimensional scale feature expression vector for each grid cell; The soil property modeling and prediction module uses multi-dimensional scale feature expression vectors and measured soil samples for supervised modeling, trains a deep neural network prediction model, and applies it to soil property prediction and layer output in unmeasured areas.

[0012] In an optional embodiment, the multi-source raster data construction and management module specifically includes: A remote sensing image processing unit, which is responsible for performing image enhancement and standardization processing on visible light, near infrared, and thermal infrared image data from satellites to generate standard remote sensing image data; A terrain factor processing unit, which extracts elevation, slope, and aspect information of the target area and performs standard rasterization processing to generate standard terrain factor data; A meteorological factor processing unit, which performs time series reconstruction and spatial interpolation on meteorological data related to precipitation, temperature, humidity, and wind speed from a meteorological station to generate standard meteorological factor data; A soil measured property management unit, which collects sample data of soil measured properties in a target area; The raster data unification and warehousing unit completes the spatial resolution unification, reference system conversion and raster structure alignment processing of various factor data to ensure the consistency of multi-source data in spatial dimension, and writes them into the multi-source raster data warehouse after unified encoding and formatting.

[0013] In an optional embodiment, the multi-dimensional dynamic scale feature construction module specifically includes: A geographic factor extraction unit, which is responsible for extracting key geographic indicators from remote sensing images, terrain factors, and meteorological factors to provide basic data for subsequent scale division; A spatial scale division unit, which divides the target area into spatial scale levels based on the terrain complexity index and generates spatial scale labels; A time scale division unit, which divides the time scale for each grid unit based on a multi-temporal meteorological factor time series analysis and generates a time scale label; A land use pattern stratification unit, wherein the land use pattern scale division unit calculates diversity and boundary density indicators based on land use type classification, spatial distribution and boundary information, divides the land use pattern scale levels, and labels the land use pattern scale labels for each regional grid unit; A terrain stratification scale division unit, which normalizes the three factor data of elevation, slope and aspect information, fuses the processed data, calculates the comprehensive terrain stratification factor, divides the terrain stratification scale, and generates a terrain stratification scale label; A multi-dimensional dynamic scale feature fusion unit is provided, which jointly encodes labels of spatial scale, temporal scale, terrain stratification scale and land use pattern scale to construct a multi-dimensional scale feature expression vector for each grid unit.

[0014] In an optional embodiment, the soil property modeling and prediction module specifically includes: A soil property modeling unit, wherein the soil property modeling unit constructs a soil property prediction model using training samples to achieve nonlinear relationship modeling between multidimensional scaling characteristics and soil properties; a soil property prediction unit, which applies the trained model to unmeasured geographic locations in the target area, performs inference based on the scale feature vector, and generates a prediction result; The prediction layer generation and visualization unit organizes the prediction results into a spatial raster layer, renders it using a geographic information platform, forms a soil property distribution map with a color gradient, and realizes the visualization expression of the prediction results.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This paper proposes a method and system for intelligent soil property prediction based on multidimensional dynamic scaling. It innovatively introduces a multidimensional dynamic scaling modeling mechanism that can dynamically assign scale labels based on the spatial heterogeneity of terrain, climate, and land use characteristics, effectively overcoming the limitations of traditional methods that use single or static scale modeling. By constructing a multidimensional scale feature expression vector and combining it with a multi-task deep neural network model for joint learning, the accuracy, stability, and generalization of soil property predictions are significantly improved. Furthermore, through adaptive error feedback and a dynamic scale adjustment mechanism, the scaling strategy can be proactively optimized based on model prediction errors, further enhancing the model's adaptability to complex environmental factors. Combined with spatial layer visualization output and a geographic information system platform, the system possesses excellent application practicality and promotional value, and can be widely used in the fields of precision forest resource management, land resource management, and ecological and environmental monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of a soil property intelligent prediction method based on multi-dimensional dynamic scaling proposed by the present invention; Figure 2 A flowchart of constructing a multi-dimensional dynamic scale expression vector in the present invention; Figure 3 A flowchart for obtaining predicted values ​​and generating spatial distribution layers in the present invention; Figure 4 This is a system framework diagram of a soil property intelligent prediction system based on multi-dimensional dynamic scaling proposed by the present invention. DETAILED DESCRIPTION

[0017] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0018] Reference Figure 1 - Figure 4 As shown, a soil property intelligent prediction method based on multi-dimensional dynamic scaling includes: Acquire remote sensing image data, terrain factor data, meteorological factor data, and soil property sample data of the target area, and store the above data in a multi-source raster data warehouse; Based on the multi-source raster data warehouse, remote sensing image data, terrain factor data and meteorological factor data are preprocessed to construct multi-source raster data in a unified format; Based on multi-source raster data in a unified format, geographical indicators reflecting terrain differences, land use patterns, and climate fluctuation characteristics are extracted to obtain terrain complexity index, climate volatility index, land use pattern comprehensive index, and comprehensive stratification factor. The spatial scale, temporal scale, terrain stratification scale, and land use pattern scale are then divided, and corresponding scale labels are assigned to each raster unit in the target area. Perform weighted joint encoding of various scale labels and the spatial coordinate information of the grid to construct the corresponding multi-dimensional dynamic scale feature expression vector; The multi-dimensional dynamic scale feature expression vector is associated with the soil property sample data at its corresponding geographical location to build a soil property prediction model. The soil property prediction model is trained using the collected soil property samples; The trained soil property prediction model is applied to geographical units in the target area that have not been measured to generate soil property prediction results and output a spatial distribution layer.

[0019] Furthermore, remote sensing image data, terrain factor data, meteorological factor data, and soil property sample data of the target area are obtained and stored in a multi-source raster data warehouse, specifically including: Remote sensing satellites equipped with multispectral sensors and thermal infrared imaging equipment acquire remote sensing image data related to visible light images, near infrared images, and thermal infrared images of target areas at different spatial resolutions; Obtain terrain factor data related to elevation, slope and aspect of the target area through digital elevation model; Obtain multi-temporal meteorological factor data related to precipitation, temperature, humidity, and wind speed in the target area from national meteorological monitoring stations and regional automatic weather stations; Through field sampling at ground sampling points set up in the target area, soil measured attribute sample data related to soil organic matter, pH value, moisture content and particle composition at the sampling points are obtained based on laboratory testing and analysis methods; Remote sensing image data, terrain factor data, meteorological factor data and soil measured attribute sample data are uniformly stored in a multi-source raster data warehouse.

[0020] Furthermore, based on the multi-source raster data warehouse, remote sensing image data, terrain factor data and meteorological factor data are preprocessed to construct multi-source raster data in a unified format, including: Perform image denoising, band fusion and spatial registration processing on remote sensing image data to obtain standard remote sensing image data; The terrain factor data is standardized in format, unified in reference space and processed into spatial grids to obtain standard terrain factor data; Perform time series difference, spatial mapping and rasterization processing on meteorological factor data to obtain standard meteorological factor data; Standard remote sensing image data, standard terrain factor data, and standard meteorological factor data are sequentially processed for spatial resolution unification, coordinate reference system conversion, and grid structure alignment. The processed grid units are then uniformly coded and the data format standardized to construct multi-source raster data in a unified format. Specifically, for the standard remote sensing image data, standard terrain factor data and standard meteorological factor data, an integrated intelligent alignment and fusion method for multi-source heterogeneous data is proposed. First, based on the principle of spatial consistency priority, an adaptive spatial resolution adjustment module is constructed: with 30 meters as the unified spatial granularity, the data resolution is automatically identified through a distributed discrimination mechanism, and the data with a resolution higher than 30 meters is spatially compressed using a neighborhood-preserving downsampling algorithm. A sliding window is constructed with a target grid size of 30 meters × 30 meters, and the texture gradient of the image unit is calculated in each window. The image unit value with significant or representative texture indicators is selected as the downsampling result; for data with a resolution lower than 30 meters, an improved bilinear interpolation method is introduced. Based on the traditional bilinear interpolation, this method integrates the gradient direction and change rate of the pixels around the interpolation center point, and dynamically adjusts the weight coefficient, so that the interpolation process can better retain the spatial trend and local change characteristics of the original data, improve spatial accuracy and suppress boundary blurring, thereby completing the consistent reconstruction of spatial scale while maintaining feature stability; Secondly, in the coordinate reference system, an automated geographic coordinate parsing and projection conversion module was designed to uniformly project all types of data into the EPSG:32650 coordinate system, achieving high-precision spatial alignment across geographic locations. Subsequently, a multi-source collaborative raster template generation mechanism was constructed to determine a unified spatial starting point, resolution, and raster dimension, and perform structural alignment and index mapping of multi-source data. This template not only has flexible configuration capabilities but also supports automatic marking of missing data areas and boundary completion. After completing the structural alignment, an efficient raster encoding strategy is used to perform hash-type unique encoding based on the raster row and column numbers and the coordinates of the spatial center point, providing a fast indexing mechanism for subsequent data access and spatial retrieval. At the same time, the format information such as the unit, data type, and missing value identifier of various data fields is standardized. For example, the temperature unit is unified as ℃, precipitation is unified as mm, and slope is unified as °, so as to construct a multi-source raster dataset with consistent structure and unified fields, providing consistent data support for subsequent feature extraction and model training.

[0021] Furthermore, various scale labels and the spatial coordinate information of the grid are weighted and jointly encoded to construct the corresponding multi-dimensional dynamic scale feature expression vector, which specifically includes: Based on multi-source raster data in a unified format, geographical indicators reflecting terrain differences, land use patterns and climate fluctuation characteristics are extracted to obtain terrain complexity index, climate volatility index, land use pattern comprehensive index and comprehensive stratification factor respectively; The slope value and aspect information of each grid cell are extracted from the standard terrain factor data. The slope aspect change rate is calculated by analyzing the amplitude of the slope aspect change within the local neighborhood of the grid cell. The slope value and slope aspect change rate of each grid cell are normalized respectively, and the normalized slope value and slope aspect change rate are weighted fused to obtain the terrain complexity index of each grid cell. : Where, is the target area, represents the slope of the grid, represents the slope change rate of the grid, is the total number of grids, is the grid number; Extract multi-temporal precipitation of each grid cell in the target area from the standard meteorological factor data and construct precipitation time series data; Based on this time series data, the standard deviation and mean of precipitation in each grid cell are calculated to further quantify the degree of fluctuation of precipitation in the temporal dimension. Constructing a climate volatility index by measuring the ratio of the standard deviation to the mean , the specific calculation formula is as follows: Where, represents the standard deviation of the precipitation time series, represents the mean of the precipitation time series; Extract land use type information from standard remote sensing image data, divide the target area into cultivated land, forest land, grassland, water body, and construction land through supervised classification algorithm, and rasterize and encode each land use type to obtain raster data of land use type; Specifically, classification features such as the target area's spectral reflectance, texture information, and vegetation index are extracted from standard remote sensing image data. A supervised classification model is constructed based on the training sample set to identify and classify different land features in the remote sensing image. The target area is divided into major land use types, such as cultivated land, forest land, grassland, water bodies, and construction land. The classification results are converted into raster-coded data on a pixel-by-pixel basis, with different types corresponding to different category numbers. For example, cultivated land is coded as 1, forest land is coded as 2, grassland is coded as 3, water bodies are coded as 4, and construction land is coded as 5. This forms a raster data layer of land use types, providing basic support for subsequent land use pattern analysis.

[0022] Based on the raster data of land use types and the spatial distribution of various land use types in the local neighborhood, the Shannon diversity index is calculated as an indicator of land use diversity. The formula is as follows: Where, Indicates the local neighborhood The proportion of land use types, is the total number of land use types; Using rasterized land use data, the boundary lengths and neighborhood grid areas between different land use types are calculated, and then the boundary density index is obtained. The land use diversity index and boundary density index are normalized and weighted to form a comprehensive index of land use pattern; Specifically, after completing land-use type extraction and raster encoding, a fixed-size (e.g., 5×5) neighborhood window is constructed centered on each grid cell. The proportion of each land-use type within the neighborhood is counted, and the Shannon diversity index is calculated to characterize the spatial heterogeneity of the local land-use pattern. To further comprehensively characterize the boundary variations of land-use types, the land-use boundary density index is calculated by combining the frequency of land-use changes between adjacent grid cells and the ratio of the total length of land-use type boundaries to the area of ​​the neighboring grid cells. Subsequently, these two indices are normalized and a unified land-use pattern composite index is constructed using a weighted overlay method to simultaneously reflect both the diversity of land features and the complexity of their boundaries.

[0023] From the standard terrain factor data, we extract elevation, slope, and aspect information, normalize the three factor data, and combine the normalized elevation, slope, and aspect values ​​to construct a comprehensive hierarchical factor that uniformly measures the complexity of the terrain structure. The calculation formula is as follows: Where, 、 are the elevation, slope and aspect values ​​of the normalized grid, ; According to the numerical range of the terrain complexity index, the target area is divided into multiple spatial scale levels. In terms of spatial scale level division, the terrain undulation degree of the grid area is graded based on the calculated terrain complexity index. Based on the numerical range of , it is divided into three types of spatial scale levels; for example, low complexity area (small scale area): ≤0.2; Medium complexity area (medium scale area): 0.2< ≤0.5; High complexity area (large scale area): >0.5; the classification threshold used can be dynamically adjusted according to the actual terrain complexity distribution in different regions to meet the needs of expressing different landform features; this division reflects the differences in regional terrain undulations. The higher the complexity, the coarser the spatial scale needs to be introduced to capture landform features; According to the size of the climate volatility index, the target area is divided into time scale levels; in terms of time scale level division, the climate volatility index based on precipitation time series Reflects the intensity of climate fluctuations, The time scale is divided into three categories based on the standard deviation ratio of ; for example, the stable zone (short period scale): ≤0.15; medium fluctuation range (medium cycle scale): 0.15< ≤0.35; High volatility area (long period scale): >0.35; the classification threshold can be adaptively adjusted according to the climate characteristics of the study area, thereby improving the generalization ability of the model in regions with different climate types; this division supports modeling with appropriate time scale granularity for regions with different climate stability; Based on the comprehensive terrain stratification factor and land use pattern comprehensive index, the target area is divided into terrain stratification scale and land use pattern scale; Specifically, in terms of terrain stratification scale division, based on the comprehensive terrain stratification factor , the area is divided into several terrain level layers using the natural breakpoint method; for example, the first level (lowland): ≤0.3; Level 2 (hilly layer): 0.3< ≤0.6; Level 3 (mountainous areas): >0.6; the classification threshold can be dynamically adjusted according to the distribution of terrain characteristics in different regions to improve the adaptability of terrain sensitivity analysis and local model construction.

[0024] At the same time, after obtaining the comprehensive index of land use pattern of each grid cell, the land use pattern of the entire region is divided into different levels based on the manually set classification criteria. Specifically, two classification thresholds are preset based on the actual complexity of the distribution of land use types in the study area and the statistical characteristics of the index values. and , the comprehensive index of land use pattern is divided into three levels: Level 1 (low heterogeneity-low boundary area): the comprehensive index is in the range Medium, indicating that the land use types are relatively concentrated, the boundaries of land features are clear, and the changes are gentle; Level 2 (medium heterogeneity-medium boundary area): the comprehensive index is in the range Medium, indicating that the land use type in the region is moderate and the boundary is relatively complex; Level 3 (high heterogeneity-high boundary area): the comprehensive index is between In the empirical study, it can be set to reflect the areas with diverse distribution of land features and frequent boundary intersections. =0.3, =0.6. That is, when the comprehensive land use pattern index of a grid cell is less than or equal to 0.3, it is classified as level 1; when the index is between 0.3 and 0.6, it is classified as level 2; and when the index is greater than 0.6, it is classified as level 3; Based on the division of spatial scale, temporal scale, terrain layer scale and land use pattern scale, each grid cell is assigned its corresponding spatial scale label, temporal scale label, terrain layer label and land use pattern label, and these labels are jointly encoded with their spatial location information to construct a multi-dimensional dynamic scale feature expression vector, which is expressed as follows: Where, and Represents the spatial coordinate position of the grid, Represents a grid The spatial scale level label, Represents a grid The time scale level label to which it belongs, Represents a grid The terrain layer labels are defined. Represents a grid Corresponding land use pattern stratification labels.

[0025] Specifically, after completing the division operations of spatial scale, temporal scale, terrain stratification scale and land use pattern scale, each grid cell in the target area is assigned various scale level labels in turn; in order to achieve the unified expression of scale labels and spatial positions, a multi-label spatial embedded joint encoding mechanism is proposed to integrate multi-dimensional scale information and spatial coordinate information to construct a multi-dimensional dynamic scale feature expression vector; the specific steps are as follows: first, obtain the label values ​​corresponding to each grid cell under the four scale divisions, including spatial scale level labels (such as 、 …), time scale level labels (e.g. 、 ...), terrain layer tags (such as 、 …) and land use pattern labels (e.g. 、 …); secondly, the position (x, y) of the grid in two-dimensional space is vectorized using sine and cosine position encoding; Then, based on the label type and hierarchical relationship, the four types of labels are jointly encoded using a multi-level weight combination method to form a scale label fusion vector with explicit structural constraints and semantic resolution capabilities; finally, by splicing the position encoding vector and the scale label fusion vector and constructing a multidimensional dynamic scale feature expression vector of uniform length through linear mapping, it is used to comprehensively characterize the spatiotemporal characteristic attributes of each raster unit at a multidimensional geographic scale; this expression method not only realizes the fusion of information at different scales, but also introduces spatial embedding and label semantic weighting mechanisms, providing subsequent deep learning models with interpretable and discriminative input feature representations, and improving the model's perception and expression capabilities of soil property differences in complex geographical backgrounds.

[0026] Furthermore, the multi-dimensional dynamic scale feature expression vector is associated with the measured soil property sample data at its corresponding geographical location to build a soil property prediction model, which specifically includes: Based on the measured soil sample data in the target area, the corresponding measured soil attribute values ​​are obtained, and the multi-dimensional dynamic scale feature expression vector at the position corresponding to its spatial coordinates is extracted to construct a training sample set. , where Indicates the Multi-dimensional dynamic scale expression vector of measured soil samples, Indicates the corresponding measured soil property value, Indicates the type of soil properties, Indicates the The first Class measured soil measured values, Number the measured soil samples; Based on the training sample set, a multi-task deep neural network model is used to jointly model multiple soil properties and construct a soil property prediction model; Specifically, a multi-task deep neural network model was constructed for joint modeling and collaborative prediction of multiple soil properties within a target area. The model takes the multidimensional dynamic-scale feature expression vector corresponding to each measured soil sample as input. First, a shared feature extraction layer extracts high-level semantic features that integrate multidimensional scales, such as space, time, topography, and land use pattern. This shared representation is then input into multiple task-specific output branches, each corresponding to a soil property prediction task. The model uses a joint loss function optimization strategy, assigning adjustable weights to the mean squared error loss of each task. In combination with a dynamic weight adjustment mechanism, the model focuses more on optimizing prediction performance in areas with significant scale changes during training. Through the combination of shared and dedicated structures, the model effectively enhances the collaborative modeling capabilities between soil properties, improves its adaptability and generalization capabilities to complex spatial nonstationarities, and exhibits good accuracy and stability.

[0027] The constructed soil prediction model is trained. During the model training process, the prediction error of the current prediction model for the training sample set is analyzed based on the local clustering in spatial distribution. The scale division granularity corresponding to the error-significant area is analyzed to see if it matches. If it is judged to be mismatched, the division threshold and number of labels of the spatial scale level, temporal scale level and terrain stratification scale are further dynamically adjusted. On this basis, the mean square error loss of each soil property prediction task in the model is calculated separately, and an adjustable task weight factor is set to perform a weighted combination of the losses of each task to construct a unified multi-task prediction model loss function, as shown below: Where, Indicates the number of types of soil properties, It is represented as the number of training samples, Expressed as The first Class measured soil property values, is the soil property prediction value of the model, represents the set of model parameters, No. The loss weight corresponding to the soil-like attribute, is the regularization coefficient; Based on this loss function, an iterative optimization algorithm based on gradient descent is used to train the parameters of the multi-task deep neural network model. The parameters include the weights and bias terms of the shared representation layer and the output layer of each task. In each iteration, the gradient is calculated and the parameters are updated based on the weighted loss of the current model on the training samples. Combined with the prediction effect of the multi-dimensional dynamic scale feature expression vector in various soil property tasks, the task loss weight is dynamically adjusted to make the model focus on areas with significant scale changes; the training process continues until the loss decreases below the preset threshold.

[0028] This method has been tested and verified in a variety of typical areas. The results show that its scale division strategy has good adaptability to different landform types and has wide regional applicability.

[0029] Furthermore, the trained soil property prediction model is applied to geographical units in the target area that have not been measured to generate soil property prediction results and output a spatial distribution layer, specifically including: Locate the geographic coordinates of unsampled locations within the target area and obtain their spatial coordinate positions; Based on the spatial coordinate position, the corresponding multi-dimensional dynamic scale feature expression vector including spatial scale label, temporal scale label and terrain layer scale label is extracted from the multi-source raster data warehouse; Input the multi-dimensional dynamic scale feature expression vector into the trained soil property prediction model to obtain the corresponding soil property prediction value; All predicted values ​​are organized into continuous raster datasets according to their spatial locations. The raster datasets are rendered using the geographic information system platform according to the color gradient mapping rules set for the attribute values ​​to generate a soil attribute spatial distribution layer with a color gradient.

[0030] Specifically, after completing the inference calculation of the multi-task soil property prediction model, for each type of soil property (such as pH, organic matter, nitrogen, phosphorus and potassium content, etc.), all prediction results are organized according to their corresponding grid spatial positions to construct a spatially continuous soil property prediction grid dataset. Specifically, the predicted value of each grid cell is mapped to a unified two-dimensional array structure using the spatial row and column numbers of the grid as indexes, thereby forming a standardized format raster data file; then, the generated soil attribute raster data is loaded into the geographic information system platform (GIS platform), and by setting the attribute value range and the corresponding color gradient mapping rules, a rendering layer is constructed using linear, segmented or quantile classification methods; for example, the pH value from 4.5 to 8.5 can be mapped to color transitions such as blue-green-yellow-red to intuitively reflect the changes in acidity and alkalinity; during the rendering process, in order to enhance the expressiveness of visualization and the interactivity of map information, auxiliary layers such as administrative boundaries and land use types are further superimposed, and legend descriptions, numerical annotations and scale control modules are enabled. The final output layer can be exported as a static map image, which can be used in multiple scenarios such as forest management zoning, quality assessment and precision fertilization; In addition, the present invention also supports the overlay analysis of multiple soil attribute layers, and uses spatial analysis tools to automatically extract thematic maps such as high-quality areas and low-nutrient areas, thereby realizing intelligent spatial decision-making support for resource management; for example, setting the screening rule of "organic matter content ≥ 20g / kg and pH in the range of 6.5-7.5" can generate a thematic layer of "high fertility soil area"; setting the condition of "available phosphorus <10mg / kg or pH <5.5", it can be identified as a "low-nutrient soil area".

[0031] Furthermore, a soil property intelligent prediction system based on multi-dimensional dynamic scaling is proposed to implement the above-mentioned prediction method, specifically including: Multi-source raster data construction and management module: The multi-source raster data construction and management module completes the collection, preprocessing and standardized fusion of remote sensing images, terrain factors and meteorological factors, constructs multi-source raster data in a unified format, and uniformly stores it in a multi-source data warehouse, while obtaining the measured soil attribute sample data of the target area; Multidimensional dynamic scale feature construction module: The multidimensional dynamic scale feature construction module extracts geographical indicators reflecting terrain differences, land use patterns and climate fluctuation characteristics, and then divides the spatial scale, temporal scale, terrain layer scale and land use pattern scale, and integrates them to construct a multidimensional scale feature expression vector for each grid unit; Soil property modeling and prediction module, the soil property modeling and prediction module uses multi-dimensional scale feature expression vectors and measured soil samples for supervised modeling, trains a deep neural network prediction model, and applies it to soil property prediction and layer output in unmeasured areas.

[0032] Furthermore, the multi-source raster data construction and management module specifically includes: Remote sensing image processing unit, which is responsible for image enhancement and standardization of visible light, near infrared and thermal infrared image data from satellites to generate standard remote sensing image data; The terrain factor processing unit extracts the elevation, slope and aspect information of the target area and performs standard rasterization processing to generate standard terrain factor data; Meteorological factor processing unit, which performs time series reconstruction and spatial interpolation on meteorological data related to precipitation, temperature, humidity, and wind speed from weather stations to generate standard meteorological factor data; The soil measured attribute management unit collects sample data of soil measured attributes in the target area; The raster data unification and warehousing unit completes the spatial resolution unification, reference system conversion and raster structure alignment processing of various factor data to ensure the consistency of multi-source data in spatial dimensions, and writes them into the multi-source raster data warehouse after unified encoding and formatting.

[0033] Furthermore, the multi-dimensional dynamic scale feature construction module specifically includes: Geographic factor extraction unit: The geographic factor extraction unit is responsible for extracting key geographic indicators from remote sensing images, terrain factors and meteorological factors to provide basic data for subsequent scale division; Spatial scale division unit, based on the terrain complexity index, divides the target area into spatial scale levels and generates spatial scale labels; Time scale division unit, based on the time series analysis of multi-temporal meteorological factors, divides the time scale for each grid unit and generates a time scale label; Land use pattern scale division unit: based on land use type classification, spatial distribution and boundary information, the land use pattern scale division unit calculates diversity and boundary density indicators, divides the land use pattern scale level, and labels the land use pattern scale label for each regional grid unit; Terrain stratification scale division unit: The terrain stratification scale division unit normalizes the three factor data of elevation, slope and aspect information, fuses the processed data, calculates the comprehensive terrain stratification factor, divides the terrain stratification scale, and generates terrain stratification scale labels; The multidimensional dynamic scale feature fusion unit jointly encodes the labels of spatial scale, temporal scale, terrain stratification scale and land use pattern scale to construct a multidimensional scale feature expression vector for each grid unit.

[0034] Furthermore, the soil property modeling and prediction module specifically includes: Soil property modeling unit, which uses training samples to build a soil property prediction model to achieve nonlinear relationship modeling between multi-dimensional scaling characteristics and soil properties; The soil property prediction unit applies the trained model to unmeasured geographic locations in the target area, performs inference based on scale feature vectors, and generates prediction results. The prediction layer generation and visualization unit organizes the prediction results into spatial raster layers, renders them using the geographic information platform, and forms a soil property distribution map with a color gradient to achieve visual expression of the prediction results.

[0035] In summary, the advantages of the present invention are as follows: by fusing multi-source heterogeneous data, constructing a multi-dimensional dynamic scale feature expression vector, and introducing a multi-task deep neural network joint modeling mechanism, the present invention achieves high-precision intelligent prediction of soil properties. It also introduces an adaptive scale feedback mechanism, which can dynamically adjust the scale division strategy, significantly improving the model's adaptability to complex terrain and spatial heterogeneity; at the same time, combined with visualization layer publishing and intelligent map extraction and analysis, it has good practicality and decision-making support value, and has the overall advantages of high data fusion depth, strong modeling accuracy, good adaptability and a wide range of application scenarios.

[0036] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A soil property intelligent prediction method based on multi-dimensional dynamic scale, characterized in that: include: Acquire remote sensing image data, terrain factor data, meteorological factor data, and measured soil attribute sample data of the target area, and store the above data in a multi-source raster data warehouse; Based on the multi-source raster data warehouse, remote sensing image data, terrain factor data and meteorological factor data are preprocessed to construct multi-source raster data in a unified format; Based on multi-source raster data in a unified format, geographical indicators reflecting terrain differences, land use patterns, and climate fluctuation characteristics are extracted to obtain terrain complexity index, climate volatility index, land use pattern comprehensive index, and comprehensive stratification factor. The spatial scale, temporal scale, terrain stratification scale, and land use pattern scale are then divided, and corresponding scale labels are assigned to each raster unit in the target area. Perform weighted joint encoding of various scale labels and the spatial coordinate information of the grid to construct the corresponding multi-dimensional dynamic scale feature expression vector; The multi-dimensional dynamic scale feature expression vector is associated with the soil property sample data at its corresponding geographical location to build a soil property prediction model. The soil property prediction model is trained using the collected soil property samples; The trained soil property prediction model is applied to geographical units in the target area that have not been measured to generate soil property prediction results and output a spatial distribution layer.

2. The soil property intelligent prediction method based on multidimensional dynamic scale according to claim 1 is characterized in that: The step of obtaining remote sensing image data, terrain factor data, meteorological factor data, and measured soil attribute sample data of the target area and uniformly storing the aforementioned data in a multi-source raster data warehouse specifically includes: Remote sensing satellites equipped with multispectral sensors and thermal infrared imaging equipment acquire remote sensing image data related to visible light images, near infrared images, and thermal infrared images of target areas at different spatial resolutions; Obtain terrain factor data related to elevation, slope and aspect of the target area through digital elevation model; Obtain multi-temporal meteorological factor data related to precipitation, temperature, humidity, and wind speed in the target area from national meteorological monitoring stations and regional automatic weather stations; Through field sampling at ground sampling points set up in the target area, soil measured attribute sample data related to soil organic matter, pH value, moisture content and particle composition at the sampling points are obtained based on laboratory testing and analysis methods; Remote sensing image data, terrain factor data, meteorological factor data and soil measured attribute sample data are uniformly stored in a multi-source raster data warehouse.

3. The soil property intelligent prediction method based on multidimensional dynamic scale according to claim 1 is characterized in that: The multi-source raster data warehouse is based on which remote sensing image data, terrain factor data and meteorological factor data are preprocessed to construct multi-source raster data in a unified format, specifically including: Performing image denoising, band fusion and spatial registration processing on the remote sensing image data to obtain standard remote sensing image data; Performing format standardization, reference space unification and spatial gridding processing on the terrain factor data to obtain standard terrain factor data; Performing time series difference, spatial mapping and rasterization processing on the meteorological factor data to obtain standard meteorological factor data; The standard remote sensing image data, standard terrain factor data and standard meteorological factor data are sequentially processed for spatial resolution unification, coordinate reference system conversion and grid structure alignment, and the processed grid units are uniformly coded and data format standardized to construct multi-source raster data in a unified format.

4. The soil property intelligent prediction method based on multi-dimensional dynamic scale according to claim 3 is characterized in that: The weighted joint encoding of various scale labels and the spatial coordinate information of the grid to construct the corresponding multi-dimensional dynamic scale feature expression vector specifically includes: Based on multi-source raster data in a unified format, geographical indicators reflecting terrain differences, land use patterns and climate fluctuation characteristics are extracted to obtain terrain complexity index, climate volatility index, land use pattern comprehensive index and comprehensive stratification factor respectively; The slope value and aspect information of each grid cell are extracted from the standard terrain factor data. The slope aspect change rate is calculated by analyzing the amplitude of the slope aspect change within the local neighborhood of the grid cell. The slope value and slope aspect change rate of each grid cell are normalized respectively, and the normalized slope value and slope aspect change rate are weighted fused to obtain the terrain complexity index of each grid cell. : Where, is the target area, represents the slope of the grid, represents the slope change rate of the grid, is the total number of grids, is the grid number; Extract multi-temporal precipitation of each grid cell in the target area from the standard meteorological factor data and construct precipitation time series data; Based on this time series data, the standard deviation and mean of precipitation in each grid cell are calculated to further quantify the degree of fluctuation of precipitation in the temporal dimension. Constructing a climate volatility index by measuring the ratio of the standard deviation to the mean , the specific calculation formula is as follows: Where, represents the standard deviation of the precipitation time series, represents the mean of the precipitation time series; Extract land use type information from standard remote sensing image data, divide the target area into cultivated land, forest land, grassland, water body, and construction land through supervised classification algorithm, and rasterize and encode each land use type to obtain raster data of land use type; Based on the raster data of land use types and the spatial distribution of various land use types in the local neighborhood, the Shannon diversity index is calculated as an indicator of land use diversity. The formula is as follows: Where, Indicates the local neighborhood The proportion of land use types, is the total number of land use types; Using rasterized land use data, the boundary lengths and neighborhood grid areas between different land use types are calculated, and then the boundary density index is obtained. The land use diversity index and boundary density index are normalized and weighted to form a comprehensive index of land use pattern; From the standard terrain factor data, we extract elevation, slope, and aspect information, normalize the three factor data, and combine the normalized elevation, slope, and aspect values ​​to construct a comprehensive terrain stratification factor that uniformly measures the complexity of the terrain structure. The calculation formula is as follows: Where, 、 are the elevation, slope and aspect values ​​of the normalized grid, ; According to the numerical range of the terrain complexity index, the target area is divided into multiple spatial scale levels; According to the size of climate volatility indicators, the target area is divided into time scale levels; Based on the comprehensive terrain stratification factor and land use pattern comprehensive index, the target area is divided into terrain stratification scale and land use pattern scale; Based on the division of spatial scale, temporal scale, terrain layer scale and land use pattern scale, each grid cell is assigned its corresponding spatial scale label, temporal scale label, terrain layer label and land use pattern label, and these labels are jointly encoded with their spatial location information to construct a multi-dimensional dynamic scale feature expression vector, which is expressed as follows: Where, and Represents the spatial coordinate position of the grid, Represents a grid The spatial scale level label, Represents a grid The time scale level label to which it belongs, Represents a grid The terrain layer labels are defined. Represents a grid Corresponding land use pattern stratification labels.

5. The soil property intelligent prediction method based on multi-dimensional dynamic scale according to claim 4 is characterized in that: The method of associating the multi-dimensional dynamic scale feature expression vector with the measured soil property sample data at its corresponding geographical location to construct a soil property prediction model specifically includes: Based on the measured soil sample data in the target area, the corresponding measured soil attribute values ​​are obtained, and the multi-dimensional dynamic scale feature expression vector at the position corresponding to its spatial coordinates is extracted to construct a training sample set. , where Indicates the Multi-dimensional dynamic scale expression vector of measured soil samples, Indicates the corresponding measured soil property value, Indicates the type of soil properties, Indicates the The first Class measured soil measured values, Number the measured soil samples; Based on the training sample set, a multi-task deep neural network model is used to jointly model multiple soil properties and construct a soil property prediction model; The constructed soil prediction model is trained. During the model training process, the prediction error of the current prediction model for the training sample set is analyzed based on the local clustering in spatial distribution. The scale division granularity corresponding to the error-significant area is analyzed to see if it matches. If it is judged to be mismatched, the division threshold and number of labels of the spatial scale level, temporal scale level and terrain stratification scale are further dynamically adjusted. On this basis, the mean square error loss of each soil property prediction task in the model is calculated separately, and an adjustable task weight factor is set to perform a weighted combination of the losses of each task to construct a unified multi-task prediction model loss function, as shown below: Where, Indicates the number of types of soil properties, It is represented as the number of training samples, Expressed as The first Class measured soil property values, is the soil property prediction value of the model, represents the set of model parameters, No. The loss weight corresponding to the soil-like attribute, is the regularization coefficient; Based on the loss function, an iterative optimization algorithm based on gradient descent is used to train the parameters of the multi-task deep neural network model, including the weights and bias terms of the shared representation layer and the output layers of each task; in each iteration, the gradient is calculated and the parameters are updated based on the weighted loss of the current model on the training samples; Combined with the prediction effect of the multi-dimensional dynamic scale feature expression vector in various soil property tasks, the task loss weight is dynamically adjusted to make the model focus on areas with significant scale changes; the training process continues until the loss decreases below the preset threshold.

6. The soil property intelligent prediction method based on multi-dimensional dynamic scale according to claim 5 is characterized in that: The method of applying the trained soil property prediction model to geographical units in the target area that have not been measured to generate soil property prediction results and output a spatial distribution layer specifically includes: Locate the geographic coordinates of unsampled locations within the target area and obtain their spatial coordinate positions; Based on the spatial coordinate position, extracting a corresponding multi-dimensional dynamic scale feature expression vector including a spatial scale label, a temporal scale label and a terrain layer scale label from a multi-source raster data warehouse; Input the multi-dimensional dynamic scale feature expression vector into the trained soil property prediction model to obtain the corresponding soil property prediction value; All predicted values ​​are organized into continuous raster datasets according to spatial locations. The raster datasets are rendered using a geographic information system platform according to the color gradient mapping rules set for the attribute values ​​to generate a soil attribute spatial distribution layer with a color gradient.

7. A soil property intelligent prediction system based on multidimensional dynamic scale, according to the soil property intelligent prediction method based on multidimensional dynamic scale according to claims 1-6, characterized in that: Specifically include: A multi-source raster data construction and management module, which completes the acquisition, preprocessing and standardized fusion of remote sensing images, terrain factors and meteorological factors, constructs multi-source raster data in a unified format, and uniformly stores it in a multi-source data warehouse, while obtaining measured soil attribute sample data in the target area; A multidimensional dynamic scale feature construction module, which extracts geographical indicators reflecting terrain differences, land use patterns, and climate fluctuation characteristics, and then divides the spatial scale, temporal scale, terrain layer scale, and land use pattern scale, and fuses them to construct a multidimensional scale feature expression vector for each grid cell; The soil property modeling and prediction module uses multi-dimensional scale feature expression vectors and measured soil samples for supervised modeling, trains a deep neural network prediction model, and applies it to soil property prediction and layer output in unmeasured areas.

8. The soil property intelligent prediction system based on multi-dimensional dynamic scale according to claim 7, characterized in that: The multi-source raster data construction and management module specifically includes: A remote sensing image processing unit, which is responsible for performing image enhancement and standardization processing on visible light, near infrared, and thermal infrared image data from satellites to generate standard remote sensing image data; A terrain factor processing unit, which extracts elevation, slope, and aspect information of the target area and performs standard rasterization processing to generate standard terrain factor data; A meteorological factor processing unit, which performs time series reconstruction and spatial interpolation on meteorological data related to precipitation, temperature, humidity, and wind speed from a meteorological station to generate standard meteorological factor data; A soil measured property management unit, which collects sample data of soil measured properties in a target area; The raster data unification and warehousing unit completes the spatial resolution unification, reference system conversion and raster structure alignment processing of various factor data to ensure the consistency of multi-source data in spatial dimension, and writes them into the multi-source raster data warehouse after unified encoding and formatting.

9. The soil property intelligent prediction system based on multi-dimensional dynamic scale according to claim 7, characterized in that: The multi-dimensional dynamic scale feature construction module specifically includes: A geographic factor extraction unit is responsible for extracting key geographic indicators from remote sensing images, terrain factors, and meteorological factors to provide basic data for subsequent scale division; A spatial scale division unit, which divides the target area into spatial scale levels based on the terrain complexity index and generates spatial scale labels; A time scale division unit, which divides the time scale for each grid unit based on a multi-temporal meteorological factor time series analysis and generates a time scale label; A land use pattern scale division unit, which calculates diversity and boundary density indicators based on land use type classification, spatial distribution and boundary information, divides land use pattern scale levels, and labels land use pattern scale labels for each regional grid unit; A terrain stratification scale division unit, which normalizes the three factor data of elevation, slope and aspect information, fuses the processed data, calculates the comprehensive terrain stratification factor, divides the terrain stratification scale, and generates a terrain stratification scale label; A multi-dimensional dynamic scale feature fusion unit is provided, which jointly encodes labels of spatial scale, temporal scale, terrain stratification scale and land use pattern scale to construct a multi-dimensional scale feature expression vector for each grid unit.

10. The soil property intelligent prediction system based on multi-dimensional dynamic scale according to claim 7, characterized in that: The soil property modeling and prediction module specifically includes: A soil property modeling unit, wherein the soil property modeling unit constructs a soil property prediction model using training samples to achieve nonlinear relationship modeling between multidimensional scaling characteristics and soil properties; a soil property prediction unit, which applies the trained model to unmeasured geographic locations in the target area, performs inference based on the scale feature vector, and generates a prediction result; The prediction layer generation and visualization unit organizes the prediction results into a spatial raster layer, renders it using a geographic information platform, forms a soil property distribution map with a color gradient, and realizes the visualization expression of the prediction results.

Images