Data processing methods and systems for ecological restoration

Abstract

This invention provides a data processing method and system for ecological restoration, relating to the field of data processing. The method includes: acquiring multidimensional soil characteristics and restoration strategies for multiple experimental soil samples, wherein the multidimensional soil characteristics include physical, chemical, biological, and pollution characteristics; determining multiple soil characteristic factor groups based on the multidimensional soil characteristics of the experimental soil; establishing multiple feature interpolation models based on the multiple soil characteristic factor groups; acquiring a sampling feature set of the soil to be restored; generating an interpolated feature set of the soil to be restored based on the sampling feature set of the soil to be restored using the multiple feature interpolation models; and generating a restoration strategy for the soil to be restored based on the multidimensional soil characteristics and restoration strategies of the multiple experimental soil samples and the interpolated feature set of the soil to be restored, for improving the effectiveness of ecological restoration.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to data processing methods and systems for ecological restoration. Background Technology

[0002] Ecological restoration refers to the process of repairing ecosystems damaged by human activities or natural disasters through artificial intervention or natural recovery, restoring their structure, function, and ecosystem service capacity, and ultimately achieving ecosystem sustainability and self-sustaining capacity. Its core objective is to rebuild the health, stability, and diversity of the ecosystem, enabling it to continue providing key ecosystem services such as clean air, water, soil conservation, and habitats for humankind. Soil remediation is the foundation of ecological restoration. Soil provides nutrients and water for plants, supporting vegetation growth; it also provides habitats for microorganisms and animals. If soil pollution or degradation is not remediated, vegetation is difficult to restore, and ecosystem functions (such as carbon cycling and soil and water conservation) cannot be rebuilt.

[0003] Existing soil remediation methods often prioritize short-term results while neglecting long-term sustainability. These techniques typically aim to rapidly improve soil conditions by adding chemicals or performing physical treatments, but this approach can lead to imbalances in the soil's ecological environment and introduce new sources of pollution. Furthermore, these methods often lack a comprehensive assessment of the soil ecosystem, failing to ensure that the remediated soil can coexist harmoniously with the surrounding environment. Moreover, many current remediation technologies are generic, failing to adequately consider the specific circumstances of soil pollution and remediation objectives. This one-size-fits-all approach may not only result in ineffective remediation but also waste significant resources and time.

[0004] Therefore, there is a need to provide data processing methods and systems for ecological restoration to improve the effectiveness of ecological restoration. Summary of the Invention

[0005] This invention provides a data processing method for ecological restoration, comprising: acquiring multidimensional soil characteristics and remediation strategies from multiple experimental soil samples, wherein the multidimensional soil characteristics include physical characteristics, chemical characteristics, biological characteristics, and pollution characteristics; physical characteristics include at least particle size distribution, porosity, and water content; chemical characteristics include at least pH value, organic matter content, and heavy metal content; biological characteristics include at least microbial biomass, enzyme activity, and respiration intensity; pollution characteristics include at least pollutant type and pollution degree; and the remediation strategy includes at least remediation technology and technical parameters, wherein the remediation technology is at least one of physical remediation, chemical remediation, and biological remediation; determining multiple soil characteristic factor groups based on the multidimensional soil characteristics of the experimental soil; establishing multiple feature interpolation models based on the multiple soil characteristic factor groups; acquiring a sampling feature set of the soil to be remediated; generating an interpolated feature set of the soil to be remediated based on the sampling feature set of the soil to be remediated using the multiple feature interpolation models; and generating a remediation strategy for the soil to be remediated based on the multidimensional soil characteristics and remediation strategies from multiple experimental soil samples and the interpolated feature set of the soil to be remediated.

[0006] Furthermore, based on the multidimensional soil characteristics of the experimental soils, multiple groups of soil characteristic factors are determined, including: for each experimental soil sample, generating a remediation strategy feature vector based on the remediation strategy of the experimental soil; for any two experimental soil samples, calculating the vector distance between the remediation strategy feature vectors of the two experimental soil samples; for each soil characteristic factor, based on the multidimensional soil characteristics of multiple experimental soil samples, calculating the feature difference between any two experimental soil samples corresponding to the soil characteristic factor, and calculating the strategy influence value of the soil characteristic factor based on the vector distance between the remediation strategy feature vectors of any two experimental soil samples and the feature difference between any two experimental soil samples corresponding to the soil characteristic factor; based on the strategy influence value of each soil characteristic factor, selecting multiple target soil characteristic factors; and based on the multiple target soil characteristic factors, determining multiple groups of soil characteristic factors.

[0007] Furthermore, based on multiple target soil characteristic factors, multiple soil characteristic factor groups are determined, including: for any two target soil characteristic factors, based on the multidimensional soil characteristics of multiple experimental soils, determining the plane variation correlation coefficient and vertical variation correlation coefficient corresponding to the two target soil characteristic factors; and based on the plane variation correlation coefficient and vertical variation correlation coefficient corresponding to any two target soil characteristic factors, determining multiple soil characteristic factor groups.

[0008] Further, based on the plane correlation coefficient and vertical correlation coefficient corresponding to any two target soil characteristic factors, multiple soil characteristic factor groups are determined, including: initializing a first optimization parameter, wherein the first optimization parameter includes at least the number of particles, the maximum number of iterations, and the number of soil characteristic factors included in the soil characteristic factor group; clustering the multiple target soil characteristic factors based on the plane variation correlation coefficient and vertical variation correlation coefficient corresponding to any two target soil characteristic factors to generate clustering results; initializing a first particle swarm based on the clustering results; establishing a first fitness function, wherein the first fitness function is related to the number of soil characteristic factor groups included in the particles and the plane variation correlation coefficient and vertical variation correlation coefficient corresponding to any two target soil characteristic factors included in each soil characteristic factor group; and iteratively optimizing the first particle swarm based on the first fitness function to determine multiple soil characteristic factor groups.

[0009] Furthermore, based on multiple soil characteristic factor groups, multiple feature interpolation models are established, including: for each soil characteristic factor group, an initial feature interpolation model corresponding to the soil characteristic factor group is established, wherein the feature interpolation model includes an interpolation encoder and an interpolation decoder. The interpolation encoder is used to extract the spatial features of the feature set of the first sampling density corresponding to the soil characteristic factor group, and the interpolation decoder is used to generate the feature set of the second sampling density corresponding to the soil characteristic factor group based on the spatial features output by the interpolation encoder. The first sampling density is less than the second sampling density. Based on the multidimensional soil features of multiple experimental soils, multiple training samples corresponding to the soil characteristic factor group are generated. Based on the multiple training samples corresponding to the soil characteristic factor group, the initial feature interpolation model corresponding to the soil characteristic factor group is trained to generate the feature interpolation model corresponding to the soil characteristic factor group.

[0010] Furthermore, based on multiple training samples corresponding to the soil feature factor group, an initial feature interpolation model corresponding to the soil feature factor group is trained to generate a feature interpolation model corresponding to the soil feature factor group. This includes: inputting the training samples corresponding to the soil feature factor group into the initial feature interpolation model, which outputs predicted values; calculating the loss function value based on the predicted values ​​output by the initial feature interpolation model and the true values ​​of the training samples corresponding to the soil feature factor group using a loss function, wherein the loss function includes a vertical consistency loss term with adaptive weights and a planar consistency loss term with adaptive weights; and adjusting the parameters of the initial feature interpolation model using an optimization algorithm to minimize the loss function value.

[0011] Furthermore, based on the sampled feature set of the soil to be remediated, an interpolated feature set of the soil to be remediated is generated using multiple feature interpolation models. This includes: generating a feature set with a first sampling density for each soil feature factor group of the soil to be remediated, based on multiple soil feature factor groups and the sampled feature set of the soil to be remediated; generating a feature set with a second sampling density for each soil feature factor group of the soil to be remediated, based on the feature set with the first sampling density for each soil feature factor group of the soil to be remediated, using multiple feature interpolation models; and generating an interpolated feature set of the soil to be remediated based on the feature set with the second sampling density for each soil feature factor group of the soil to be remediated.

[0012] Further, based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated, an interpolation feature set of the soil to be remediated is generated, including: calculating global interpolation consistency based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated; determining whether to perform interpolation fusion based on the global interpolation consistency; if so, initializing the second optimization parameter; initializing the second particle swarm based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated; establishing the second fitness function based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated; iteratively optimizing the second particle swarm based on the second fitness function to determine the interpolation feature set of the soil to be remediated.

[0013] Furthermore, based on the multidimensional soil characteristics and remediation strategies of multiple experimental soil samples and the interpolation feature set of the soil to be remediated, a remediation strategy for the soil to be remediated is generated, including: determining the target experimental soil based on the multidimensional soil characteristics of multiple experimental soil samples and the interpolation feature set of the soil to be remediated; and generating a remediation strategy for the soil to be remediated based on the remediation strategy of the target experimental soil.

[0014] This invention provides a data processing system for ecological restoration, comprising: a data acquisition module for acquiring multidimensional soil characteristics and remediation strategies from multiple experimental soil samples, wherein the multidimensional soil characteristics include physical characteristics, chemical characteristics, biological characteristics, and pollution characteristics; physical characteristics include at least particle size distribution, porosity, and water content; chemical characteristics include at least pH value, organic matter content, and heavy metal content; biological characteristics include at least microbial biomass, enzyme activity, and respiration intensity; pollution characteristics include at least pollutant type and pollution degree; and remediation strategies include at least remediation techniques and technical parameters, wherein the remediation technique is at least one of physical remediation, chemical remediation, and biological remediation; a factor grouping module for determining multiple soil characteristic factor groups based on the multidimensional soil characteristics of the experimental soils; a model building module for establishing multiple feature interpolation models based on the multiple soil characteristic factor groups; a feature interpolation module for acquiring a sampling feature set of the soil to be remediated and generating an interpolated feature set of the soil to be remediated based on the sampling feature set of the soil to be remediated through multiple feature interpolation models; and a strategy generation module for generating a remediation strategy for the soil to be remediated based on the multidimensional soil characteristics and remediation strategies of the multiple experimental soil samples and the interpolated feature set of the soil to be remediated.

[0015] Compared with existing technologies, the data processing method and system for ecological restoration provided by this invention have at least the following beneficial effects:

[0016] Integrating physical, chemical, biological, and pollution characteristics avoids the limitations of single indicators and ensures that remediation strategies comprehensively match actual soil conditions. By establishing interpolation models through grouping, data dimensionality complexity is reduced, model computation speed and interpolation accuracy are improved, and resource consumption is reduced. Models built using experimental soil data can generate a complete feature set of the soil to be remediated based on limited sampling points, solving the problems of high on-site testing costs and insufficient coverage. Based on the similarity matching between historical remediation cases and the characteristics of the soil to be remediated, validated strategies are directly transferred or integrated, reducing trial-and-error risks and improving the feasibility of the solutions. A closed loop is formed from data acquisition to strategy generation, reducing human intervention, shortening the remediation solution development cycle, and adapting to the needs of large-scale ecological restoration scenarios. Attached Figure Description

[0017] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0018] Figure 1 This is a flowchart illustrating a data processing method for ecological restoration according to some embodiments of this specification;

[0019] Figure 2 This is a schematic diagram of the structure of a feature interpolation model according to some embodiments of this specification;

[0020] Figure 3 This is a flowchart illustrating the generation of an interpolation feature set for the soil to be remediated, as shown in some embodiments of this specification.

[0021] Figure 4 This is a schematic diagram of the feature set of the first sampling density before interpolation, as shown in some embodiments of this specification;

[0022] Figure 5 This is a schematic diagram of the feature set of the second sampling density after interpolation, as shown in some embodiments of this specification;

[0023] Figure 6 This is a schematic diagram of a data processing system for ecological restoration, as shown in some embodiments of this specification. Detailed Implementation

[0024] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0025] Figure 1 This is a flowchart illustrating a data processing method for ecological restoration according to some embodiments of this specification, such as... Figure 1 As shown, the data processing methods used for ecological restoration can include at least S1-S6.

[0026] S1. Obtain multidimensional soil characteristics and remediation strategies from multiple experimental soil samples.

[0027] Multidimensional soil characteristics include physical, chemical, biological, and pollution characteristics.

[0028] Specifically, physical characteristics reflect the soil's structure, texture, and spatial distribution, directly affecting the transport of water and nutrients, and may include at least:

[0029] Particle size distribution: the proportion of sand, powder, and clay (determined by sieving or laser particle size analyzer);

[0030] Porosity: Total porosity, ventilation porosity (measured by ring cutter method or CT scan technology);

[0031] Bulk density: The dry weight of soil per unit volume (reflects soil compaction and is related to root growth).

[0032] Moisture content: natural moisture content, field water holding capacity (measured by drying method or time domain reflectometer);

[0033] Soil structure: Aggregate stability (assessed by wet sieving or ultrasonic dispersion).

[0034] Chemical characteristics can describe the composition, form, and reactivity of chemical elements in soil, determining nutrient supply capacity and pollution risk, and include at least:

[0035] pH value: Soil acidity and alkalinity (determined by potentiometric method);

[0036] Organic matter content: humus, carbohydrates, etc. (determined by potassium dichromate oxidation method);

[0037] Nutrient content: Total and available forms of nitrogen (N), phosphorus (P), and potassium (K) (determined by spectrophotometry or atomic absorption spectrometry);

[0038] Cation exchange capacity (CEC): The soil’s ability to adsorb nutrient ions (determined by the ammonium acetate exchange method).

[0039] Heavy metal content: cadmium (Cd), lead (Pb), arsenic (As), etc. (determined by atomic fluorescence spectrometry);

[0040] Heavy metal forms: exchangeable, carbonate-bound, iron-manganese oxide-bound, etc. (analyzed using Tessier's five-step continuous extraction method).

[0041] Biological characteristics can reflect the activity and diversity of microorganisms, animals, and plant roots in the soil, driving material cycling and energy flow, and include at least:

[0042] Microbial biomass: the number of bacteria, fungi, and actinomycetes (determined by plate count or phospholipid fatty acid analysis).

[0043] Enzyme activity: urease, phosphatase, dehydrogenase, etc. (determined by colorimetric or fluorescence methods);

[0044] Respiratory intensity: basal respiration and induced respiration (measured by alkaline absorption method or infrared gas analyzer).

[0045] Biodiversity: Microbial community structure (analysis of 16S rRNA or ITS genes using high-throughput sequencing technology).

[0046] Soil animals such as nematodes and earthworms: species, quantity and functional group composition (separated by hand picking or Tullgren funnel method).

[0047] Pollution characteristics can be used to describe the state of pollutants in soil and include at least:

[0048] Types of pollutants: heavy metals (such as Cd and Pb), organic pollutants (such as polycyclic aromatic hydrocarbons (PAHs) and pesticides), radioactive substances, etc.

[0049] Pollution level: Single-factor pollution index, Nemerow comprehensive pollution index (calculated based on national standards or background values).

[0050] The remediation strategy for experimental soil can be determined by researchers (e.g., experts) based on their remediation experience and experimental evaluation results, taking into account the multidimensional soil characteristics of the experimental soil. The remediation strategy can include at least the following: remediation techniques (e.g., physical remediation (e.g., topsoil replacement, soil replacement, electrokinetic remediation, etc.), chemical remediation, bioremediation (e.g., phytoremediation, microbial remediation, etc.)); and technical parameters (e.g., for soil replacement, determining the depth and volume of soil replacement; for electrokinetic remediation, determining the electrode arrangement, voltage, and energizing time; in chemical leaching, determining the type, concentration, and number of leaching solutions; for phytoremediation, selecting suitable plant varieties, planting density, and planting season; for microbial remediation, determining the dosage and culture conditions of highly efficient degrading bacteria).

[0051] As an example only, let's assume the experimental soil has the following characteristics:

[0052] Physical properties: High bulk density (1.6 g / cm³), low porosity (35%);

[0053] Chemistry: pH=4.5, total Cd=5 mg / kg (exchangeable form accounts for 60%);

[0054] Biological: Low microbial biomass carbon (200 mg / kg) and weak respiration intensity;

[0055] Pollution: Primarily Cd pollution.

[0056] The repair strategy can be:

[0057] Physical improvement: Adding straw or biochar reduces bulk density and increases porosity;

[0058] Chemical adjustment: Applying lime raises the pH to 6.5, promoting Cd precipitation;

[0059] Bioaugmentation: Inoculate with Cd-resistant microorganisms (such as Bacillus) and plant hyperaccumulating plants (such as Sedum aizoon).

[0060] S2. Based on the multidimensional soil characteristics of the experimental soil, multiple soil characteristic factor groups were determined.

[0061] Specifically, it includes:

[0062] For each experimental soil sample, a remediation strategy feature vector is generated based on the remediation strategy of the experimental soil.

[0063] For any two experimental soil samples, calculate the vector distance between the feature vectors of the remediation strategies of the two experimental soil samples;

[0064] For each soil characteristic factor, based on the multidimensional soil characteristics of multiple experimental soils, the characteristic difference between any two experimental soils corresponding to the soil characteristic factor is calculated. Based on the vector distance of the remediation strategy characteristic vectors of any two experimental soils and the characteristic difference between any two experimental soils corresponding to the soil characteristic factor, the strategy influence value of the soil characteristic factor is calculated.

[0065] Based on the strategy impact value of each soil characteristic factor, multiple target soil characteristic factors are screened.

[0066] Based on multiple target soil characteristic factors, multiple groups of soil characteristic factors were identified.

[0067] Specifically, the remediation strategy feature vector is a combination of remediation technologies and their key parameters of experimental soil that are transformed into quantifiable numerical vectors. This ensures that different remediation strategies (such as chemical, biological, and physical technologies) can be compared in the same vector space. The parameter encoding reflects the strength, type, and synergistic effect of the remediation technologies and satisfies mathematical operability (such as Euclidean distance and cosine similarity).

[0068] Assign a vector dimension to the remediation strategy and fix the vector length (e.g., 10 dimensions). If a soil type does not use a certain technology, the corresponding dimension value is 0 or a default parameter. Continuous parameters in the remediation strategy directly use measured values ​​(e.g., lime addition = 2.5 t / ha). Categorical parameters need to be converted to numerical values ​​(e.g., plant species: Sedum aizoon = 1, Centipede grass = 2). Normalize parameters of different dimensions (e.g., Min-Max normalization or Z-score normalization) to avoid large numerical differences affecting distance calculations. For example, the lime addition range is [0, 5 t / ha], and the lime addition in the remediation strategy of a certain experimental soil is 2.5 t / ha, so after normalization, it = 0.5; the plant planting density range is [0, 20 plants / m²], and the plant planting density in the remediation strategy of a certain experimental soil is 10 plants / m², so after normalization, it = 0.5. Fill the normalized parameters in dimensional order to generate a vector. Fill the dimensions of unused technologies with 0 or the default value.

[0069] As an example only, the remediation strategy for a certain experimental soil is: only lime (3 t / ha) is added, the remediation period is 30 days, the normalized lime addition amount is 0.6, and the normalized period after 30 days is 0.6. Then the corresponding remediation strategy feature vector is: [0.6, 0, 0, 0, 0, 0, 0, 0, 0, 0.6].

[0070] For example, the soil remediation strategy for a certain experimental soil is: lime addition (2 t / ha) + planting of Sedum aizoon (15 plants / m²) + electrokinetic remediation (1.2 V / cm), with a cycle of 60 days. The corresponding remediation strategy feature vector is: lime = 0.4, plant density = 0.75, voltage = 0.6, cycle = 1.0, vector: [0.4, 0, 0.75, 0, 0.6, 0, 0, 0, 0, 1.0].

[0071] The Euclidean distance or Manhattan distance between the feature vectors of the remediation strategies of two experimental soil samples can be calculated and used as the vector distance between the feature vectors of the remediation strategies of the two experimental soil samples.

[0072] Soil characteristic factors are measurable attributes of soil with clear ecological or environmental significance. They characterize soil state (such as pH, nutrient content, and pollutant concentration) and reflect soil function (such as microbial activity, enzyme catalytic capacity, and water retention capacity). Specific values ​​of soil characteristic factors are obtained through experiments or instrumental measurements.

[0073] By way of example only, soil characteristic factors may include:

[0074] 1. Physical characteristic factors, such as particle size distribution: the proportion of sand (2-0.05 mm), silt (0.05-0.002 mm), and clay (<0.002 mm); porosity: the proportion of pore volume to total volume in soil; bulk density: the dry weight of soil per unit volume (g / cm³); moisture content: the mass percentage of water in soil, etc.

[0075] 2. Chemical characteristic factors, such as pH value, organic matter content: the mass ratio of organic carbon in the soil, nutrient content: the concentration of available nutrients such as nitrogen (N), phosphorus (P), and potassium (K);

[0076] 3. Biological characteristics, such as microbial biomass carbon: the carbon content of microorganisms in the soil (mg / kg), respiration intensity: the CO2 release rate of soil microorganisms (mg CO2 / (kg·h)), etc.

[0077] 4. Pollutant characteristic factors, such as heavy metal concentrations: the content of pollutants such as cadmium (Cd), lead (Pb), and arsenic (As), and the residues of organic pollutants (such as total polycyclic aromatic hydrocarbons and total petroleum hydrocarbons) in the soil.

[0078] The multidimensional soil characteristics of the experimental soil can include the specific values ​​of each soil characteristic factor from multiple sampling points of the experimental soil.

[0079] For any given experimental soil sample, the mean and standard deviation of the soil characteristic factors can be calculated based on the specific values ​​of the soil characteristic factors at multiple sampling points. For any two experimental soil samples, the characteristic difference between the two samples can be calculated based on the mean and standard deviation of the soil characteristic factors corresponding to the two samples. For example, the absolute difference between the mean and the absolute difference between the standard deviations of the soil characteristic factors corresponding to the two samples can be calculated, and then a weighted sum of these absolute differences can be obtained to obtain the characteristic difference between the two samples.

[0080] Substitute the vector distance of the remediation strategy feature vectors of any two experimental soil samples and the feature difference of the soil feature factors corresponding to any two experimental soil samples into the correlation coefficient (e.g., Pearson correlation coefficient, Spearman rank correlation coefficient, etc.) calculation formula to obtain the correlation coefficient between the vector distance and the feature difference. Take the absolute value of the correlation coefficient between the vector distance and the feature difference as the strategy influence value of the soil feature factors.

[0081] Soil characteristic factors whose policy influence value is greater than a threshold (e.g., 0.6) can be used as target soil characteristic factors.

[0082] In some embodiments, multiple groups of soil characteristic factors are determined based on multiple target soil characteristic factors, including:

[0083] For any two target soil characteristic factors, based on the multidimensional soil characteristics of multiple experimental soil samples, determine the plane variation correlation coefficient and vertical variation correlation coefficient corresponding to the two target soil characteristic factors;

[0084] Based on the correlation coefficients of planar and vertical variations of any two target soil characteristic factors, multiple groups of soil characteristic factors are determined.

[0085] Specifically, in soil science research, there are often complex interactions among target soil characteristic factors (such as organic matter content, water content, pH, nitrogen, phosphorus, and potassium content). These relationships are not only reflected in the differences in distribution in horizontal space (planar variation), but also in the hierarchical changes in vertical profiles (vertical variation).

[0086] For each experimental soil sample, multiple sets of planar and vertical data were obtained based on the multidimensional soil characteristics of multiple samples. Each set of planar data included the values ​​of each target soil characteristic factor from multiple sampling points located on the same plane. The vertical data included the values ​​of each target soil characteristic factor from multiple sampling points located at different heights on the same horizontal coordinate. For each set of planar data, the values ​​of two target soil characteristic factors at each sampling point included in the planar data were substituted into the correlation coefficient calculation formula (e.g., Pearson correlation coefficient, Spearman rank correlation coefficient, etc.) to calculate the correlation coefficient between the two target soil characteristic factors and the corresponding planar data. The average of the correlation coefficients for each planar data point of all experimental soils corresponding to the two target soil characteristic factors was calculated to obtain the planar variation correlation coefficient between the two target soil characteristic factors. The calculation of the vertical variation correlation coefficient is similar to that of the planar variation correlation coefficient and will not be elaborated here.

[0087] In some embodiments, multiple groups of soil characteristic factors are determined based on the plane correlation coefficient and vertical correlation coefficient corresponding to any two target soil characteristic factors, including:

[0088] Initialize the first optimization parameters, which include at least the number of particles (e.g., 20), the maximum number of iterations (e.g., 150), and the number constraints of the soil characteristic factors included in the soil characteristic factor group. For example, the minimum number constraint is greater than or equal to 2, and the maximum number constraint is less than 5. The first optimization parameters may also include other parameters, such as inertia weight: controlling the inheritance ratio of particle velocity to balance global search and local search capabilities; learning factor: controlling the step size of the particle moving towards the individual optimal position and the global optimal position; velocity limit: the maximum value of particle velocity to prevent the particle from moving too fast and skipping the optimal solution, etc.

[0089] Based on the correlation coefficients of planar and vertical variations of any two target soil characteristic factors, multiple target soil characteristic factors are clustered to generate clustering results. Specifically, for any two target soil characteristic factors, the correlation coefficients of planar and vertical variations of the target soil characteristic factors are weighted and summed to calculate the comprehensive correlation coefficient of the two target soil characteristic factors. Then, a clustering algorithm (e.g., hierarchical clustering or K-means clustering) is used to cluster multiple target soil characteristic factors according to the comprehensive correlation coefficient of any two target soil characteristic factors to generate clustering results.

[0090] Based on the clustering results, the first particle swarm is initialized. Specifically, for each cluster, it can be determined whether the number of target soil feature factors included in the cluster meets the maximum number constraint. If so, the cluster is divided into two clusters, and the number of target soil feature factors included in the cluster meets the maximum number constraint again, until the number of target soil feature factors included in each cluster meets the maximum number constraint. The processed clusters are used as soil feature factor groups to form the first particle. By adding perturbation to the first particle, multiple other particles are formed, and the initialization of the first particle swarm is completed. For example, randomly swapping some target soil feature factors in the soil feature factor groups in the first particle, or increasing or decreasing the target soil feature factors in one or more soil feature factor groups in the first particle.

[0091] A first fitness function is established, which is related to the number of soil characteristic factor groups included in the particle and the plane variation correlation coefficient and vertical variation correlation coefficient corresponding to any two target soil characteristic factors included in each soil characteristic factor group. Specifically, the more soil characteristic factor groups the particle includes, the smaller the mean of the plane variation correlation coefficient and the mean of the vertical variation correlation coefficient corresponding to any two target soil characteristic factors included in each soil characteristic factor group, the smaller the first fitness value of the particle.

[0092] The first particle swarm is iteratively optimized based on the first fitness function to determine multiple soil characteristic factor groups. Specifically, in each iteration, the velocity and position of the particles are updated until the maximum number of iterations is reached or the first fitness value changes less than the threshold for 10 consecutive iterations. The multiple soil characteristic factor groups included by the particle with the largest first fitness value are then taken as multiple soil characteristic factor groups.

[0093] Specifically, in grouping soil characteristic factors, existing methods (such as thresholding or simple clustering) may overlook the following issues:

[0094] 1. Optimal number of groups: Manually setting the number of groups may lead to undergrouping (association factors are split) or overgrouping (weak association factors are merged).

[0095] 2. Balancing multidimensional correlations: There may be contradictions between planar and vertical correlation coefficients (such as strong planar correlation but weak vertical correlation), which need to be comprehensively weighed.

[0096] 3. Global optimization requirements: The relationships between soil characteristic factors are complex, and local optimal solutions (such as the results of a single clustering algorithm) may deviate from the true grouping.

[0097] The innovative method for determining multiple soil characteristic factor groups, by integrating planar correlation coefficients (reflecting horizontal spatial distribution patterns) and vertical correlation coefficients (reflecting profile layer variation patterns), ensures that the grouping results simultaneously conform to the natural variation characteristics of soil in both horizontal and vertical directions, avoiding biases caused by single-dimensional analysis. By recursively splitting over-limit clusters (e.g., splitting into two when the number of factors is ≥5), it ensures that each soil characteristic factor group meets the minimum (≥2) and maximum (<5) constraints, avoiding "single factor groups" (meaningless) or "over-merged groups" (masking inherent differences), thus improving the interpretability and practicality of the grouping results. The first particle is generated based on the constraint-corrected clusters, ensuring the initial solution has local rationality and providing a high-quality starting point for subsequent optimization. Perturbations are added to the baseline particle through random factor swapping, factor addition / reduction, etc., generating multiple differentiated particles, expanding the search space, reducing the risk of getting trapped in local optima, and improving global optimization capabilities. The particle swarm generated by the constraint correction and perturbation strategies is closer to the optimal solution distribution region than random initialization, significantly reducing the number of subsequent iterations (e.g., from 500 to 150), accelerating convergence. The grouping results based on the comprehensive correlation coefficient are consistent with the objectives of the subsequent fitness function (penalizing redundant grouping and strengthening intra-group homogeneity), forming a collaborative closed loop of "clustering-optimization" and further improving the scientific nature of the final grouping scheme.

[0098] S3. Based on multiple soil characteristic factor groups, establish multiple feature interpolation models.

[0099] Specifically, it includes:

[0100] For each soil feature factor group, an initial feature interpolation model corresponding to the soil feature factor group is established. The feature interpolation model includes an interpolation encoder and an interpolation decoder. The interpolation encoder is used to extract the spatial features of the feature set of the first sampling density corresponding to the soil feature factor group. The interpolation decoder is used to generate the feature set of the second sampling density corresponding to the soil feature factor group based on the spatial features output by the interpolation encoder. The first sampling density is less than the second sampling density.

[0101] Based on the multidimensional soil characteristics of multiple experimental soil samples, multiple training samples corresponding to soil characteristic factor groups are generated.

[0102] Based on multiple training samples corresponding to the soil characteristic factor group, the initial feature interpolation model corresponding to the soil characteristic factor group is trained to generate the feature interpolation model corresponding to the soil characteristic factor group.

[0103] Specifically, Figure 2 These are schematic diagrams illustrating the structure of the feature interpolation model according to some embodiments of this specification, such as... Figure 2As shown, the interpolation encoder may include an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a third convolutional layer, a spatial attention module, and an output layer. The input layer aligns the spatial range and number of channels of the low-sampling-density feature set (i.e., the feature set with the first sampling density). The first convolutional layer extracts local spatial features (e.g., factor gradients within a 3×3 neighborhood). The first pooling layer reduces spatial resolution (e.g., output size: 50×50×64) to enhance translation invariance. The second convolutional layer captures medium-scale spatial patterns. The second pooling layer further reduces dimensionality (e.g., output size: 25×25×128) to focus on global trends. The third convolutional layer extracts high-order features (e.g., nonlinear interactions between factors, such as the synergistic effect of pH and organic matter). The spatial attention module dynamically weights important spatial regions (e.g., assigning higher weights to highly variable regions). The spatial attention module may include a CBAM (Convolutional Block Attention Module). The output layer includes global average pooling and a fully connected layer to generate a fixed-length feature vector (e.g., 256 dimensions) as the decoder input.

[0104] The decoder may include an input layer, a first fully connected layer, a first deconvolutional layer, a first skip connection, a second deconvolutional layer, a second skip connection, a third deconvolutional layer, and an output layer. The input layer aligns with the encoder's output dimension. The first fully connected layer restores the feature vectors to a feature map, matching the output size of the encoder's second pooling layer. The first deconvolutional layer upsamples (e.g., output size: 50×50×128) to restore spatial resolution. The first skip connection concatenates with the encoder's second convolutional layer output. The second deconvolutional layer further upsamples (e.g., output size: 100×100×64). The second skip connection concatenates with the encoder's first convolutional layer output. The third deconvolutional layer upsamples to the target resolution (e.g., 100×100×3) to generate a high-density feature set (i.e., a feature set with the second sampling density). The output layer normalizes and linearly maps to the actual value range.

[0105] In some embodiments, based on multiple training samples corresponding to soil characteristic factor groups, an initial feature interpolation model corresponding to the soil characteristic factor groups is trained to generate a feature interpolation model corresponding to the soil characteristic factor groups, including:

[0106] The training samples corresponding to the soil feature factor group are input into the initial feature interpolation model corresponding to the soil feature factor group, and the initial feature interpolation model outputs the predicted value.

[0107] The loss function is calculated by comparing the predicted values ​​output by the initial feature interpolation model with the true values ​​of the training samples corresponding to the soil feature factor groups. The loss function includes a vertical consistency loss term with adaptive weights and a planar consistency loss term with adaptive weights.

[0108] The parameters of the initial feature interpolation model are adjusted by an optimization algorithm (such as gradient descent) to minimize the loss function value. In each iteration, the feature interpolation model generates a predicted value based on the current parameters, calculates the loss function value, and updates the parameters based on the gradient information of the loss function.

[0109] Evaluate the performance of the feature interpolation model on unseen data using validation set data. Calculate the loss function value on the validation set to determine whether the feature interpolation model is overfitting (performing well on the training set but poorly on the validation set) or underfitting (performing poorly on both the training and validation sets).

[0110] Specifically, a fixed-weight loss function may not adapt to the intensity of variation in different regions, leading to oversmoothing or loss of detail. For each training sample, multiple horizontal and vertical planes can be sampled. For each horizontal plane, the spatial features of the feature set corresponding to the first sampling density of the soil feature factor group are calculated. The numerical variance of each target soil feature factor in the soil feature factor group is calculated at that horizontal plane. The mean of the numerical variance of each target soil feature factor in the soil feature factor group at that horizontal plane is obtained. The larger the mean variance of the horizontal plane, the greater the weight of the horizontal plane. The method for calculating the weight of the vertical plane is similar to that for the horizontal plane, and will not be repeated here.

[0111] For each horizontal plane, based on the feature set of the second sampling density corresponding to the predicted soil characteristic factor group, the numerical variance of each target soil characteristic factor in that horizontal plane is calculated, called the predicted numerical variance. Based on the feature set of the second sampling density corresponding to the true soil characteristic factor group, the numerical variance of each target soil characteristic factor in that horizontal plane is calculated, called the true numerical variance. The absolute value of the difference between the predicted and true numerical variances of each target soil characteristic factor in that horizontal plane is calculated, and the absolute values ​​of the differences for each target soil characteristic factor in that horizontal plane are summed to obtain the sum of the absolute values ​​of the differences for the horizontal plane. Similarly, the sum of the absolute values ​​of the differences for the vertical planes is calculated. The weighted sum of the absolute values ​​of the differences for each horizontal plane is then calculated to obtain the value of the adaptive weighted plane consistency loss term. The weighted sum of the absolute values ​​of the differences for each vertical plane is then calculated to obtain the value of the adaptive weighted vertical consistency loss term.

[0112] Understandably, by assigning horizontal / vertical weights based on the mean variance of real data, the model retains details in areas of dramatic soil feature variation and forces smoothing in stable areas, avoiding over-smoothing or loss of detail caused by fixed weights. By calculating the sum of the absolute values ​​of the variance differences between predicted and true values ​​in the horizontal / vertical directions, the model's ability to capture spatial distribution patterns of soil features is directly quantified, guiding the model to learn spatial variation patterns that better reflect reality. The vertical consistency loss term with adaptive weights strengthens the logical connection between soil features between layers, while the planar consistency loss term ensures the continuity of spatial features within the same layer. The combination of these two significantly improves the spatial rationality of the 3D interpolation results.

[0113] S4. Obtain the sampling feature set of the soil to be remediated.

[0114] Specifically, the sampling feature set of the soil to be remediated can be the values ​​of multiple target soil characteristic factors at multiple sampling points of the soil to be remediated, collected according to a first sampling density, such as... Figure 4 As shown.

[0115] S5. Based on the sampled feature set of the soil to be remediated, generate the interpolated feature set of the soil to be remediated by multiple feature interpolation models.

[0116] Figure 3 This is a flowchart illustrating the generation of an interpolation feature set for the soil to be remediated, as shown in some embodiments of this specification. Figure 3 As shown, in some embodiments, S5 specifically includes:

[0117] Based on multiple soil characteristic factor groups and the sampling feature set of the soil to be remediated, a feature set of the first sampling density corresponding to each soil characteristic factor group of the soil to be remediated is generated.

[0118] Based on the feature set of the first sampling density of each soil feature factor group corresponding to the soil to be repaired, a feature set of the second sampling density of the soil to be repaired is generated by multiple feature interpolation models.

[0119] Based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated, an interpolated feature set of the soil to be remediated is generated.

[0120] like Figure 3 As shown, in some embodiments, based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated, an interpolated feature set of the soil to be remediated is generated, including:

[0121] Global interpolation consistency is calculated based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated.

[0122] Based on global interpolation consistency, determine whether to perform interpolation fusion. If not, calculate the mean of the interpolation results for each soil feature factor group that includes the target soil feature factor, and use it as the interpolation result of the target soil feature factor to generate the interpolation feature set of the soil to be remediated.

[0123] If so, initialize the second optimization parameters, which may include at least the number of particles (e.g., 20), the maximum number of iterations (e.g., 150), and the inertia weight;

[0124] Based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated, the second particle swarm is initialized. The second particle swarm includes multiple second particles, and the second particles can include the interpolation result of the second sampling density corresponding to each target soil characteristic factor, that is, the specific value of each target soil characteristic factor at multiple sampling points corresponding to the second sampling density.

[0125] A second fitness function is established based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated.

[0126] The second particle swarm is iteratively optimized based on the second fitness function to determine the interpolation feature set of the soil to be remediated.

[0127] Specifically, for each target soil characteristic factor, the interpolation result corresponding to each soil characteristic factor group including the target soil characteristic factor can be determined from the feature set of the second sampling density corresponding to the soil characteristic factor group including the target soil characteristic factor. The Euclidean distance between any two interpolation results of the target soil characteristic factor including the target soil characteristic factor can be calculated. The sum of the Euclidean distances between any two interpolation results of the target soil characteristic factor including the target soil characteristic factor can be obtained.

[0128] The sum of interpolation distances corresponding to each target soil feature factor is calculated to obtain the global interpolation distance sum. Based on the global interpolation distance sum, the global interpolation consistency is calculated. The larger the global interpolation distance sum, the smaller the global interpolation consistency. If the global interpolation consistency is less than a threshold (e.g., 0.5), interpolation fusion is determined to be performed.

[0129] The perturbation within an allowable error range can be added to the interpolation results of the soil characteristic factor group corresponding to the target soil characteristic factor, and the second particle swarm can be initialized.

[0130] For a target soil characteristic factor, the interpolation result of the second sampling density of the target soil characteristic factor in the second particle and the Euclidean distance with the interpolation result of the soil characteristic factor group including the target soil characteristic factor can be calculated. The interpolation distance of the target soil characteristic factor is obtained by summing the interpolation result of the second sampling density of the target soil characteristic factor in the second particle and the Euclidean distance of each interpolation result with the soil characteristic factor group including the target soil characteristic factor.

[0131] The interpolation distances of each target soil characteristic factor are summed to calculate the global interpolation distance corresponding to the second particle. The larger the global interpolation distance of the second particle, the smaller the second fitness value of the second particle.

[0132] The second particle swarm is iteratively optimized based on the second fitness function until the maximum number of iterations is reached or the second fitness value changes less than a threshold for 10 consecutive iterations. The interpolation results of the target soil characteristic factors included in the second particle with the largest second fitness value are used to determine the interpolation feature set of the soil to be remediated. Figure 5 As shown.

[0133] S6. Based on the multidimensional soil characteristics and remediation strategies of multiple experimental soil samples and the interpolation feature set of the soil to be remediated, a remediation strategy for the soil to be remediated is generated.

[0134] Specifically, it includes:

[0135] Based on the multidimensional soil characteristics of multiple experimental soil samples and the interpolation feature set of the soil to be remediated, the target experimental soil is determined.

[0136] Based on the remediation strategy of the target experimental soil, a remediation strategy for the soil to be remediated is generated.

[0137] Specifically, for each experimental soil sample, a feature set corresponding to the second sampling density can be determined based on the multidimensional soil characteristics and multiple target soil feature factors. The similarity between this feature set and the interpolated feature set of the soil to be remediated is calculated. Experimental soil samples with a similarity greater than a similarity threshold (e.g., 0.7) are designated as target experimental soils. If an experimental soil sample is highly similar to the soil to be remediated (e.g., similarity greater than 0.9), its remediation strategy is directly adopted. A weighted fusion approach can be used to fuse the remediation strategies of the target experimental soil samples to generate a remediation strategy for the soil to be remediated. Alternatively, a decision tree or neural network model based on an experimental soil database can be constructed to predict the optimal strategy for the soil to be remediated based on the remediation strategies of the target experimental soil samples, reducing reliance on manual rules.

[0138] Figure 6 These are schematic diagrams of modules for a data processing system for ecological restoration, as shown in some embodiments of this specification. Figure 6As shown, a data processing system for ecological restoration may include a data acquisition module, a factor grouping module, a model building module, a feature interpolation module, and a strategy generation module.

[0139] The data acquisition module is used to acquire multidimensional soil characteristics and remediation strategies from multiple experimental soil samples. The multidimensional soil characteristics include physical, chemical, biological, and pollution characteristics.

[0140] The factor grouping module is used to determine multiple soil characteristic factor groups based on the multidimensional soil characteristics of the experimental soil.

[0141] The model building module is used to build multiple feature interpolation models based on multiple soil feature factor groups;

[0142] The feature interpolation module is used to obtain the sampling feature set of the soil to be remediated, and generate the interpolated feature set of the soil to be remediated based on the sampling feature set of the soil to be remediated through multiple feature interpolation models.

[0143] The strategy generation module is used to generate a remediation strategy for the soil to be remediated based on the multidimensional soil characteristics and remediation strategies of multiple experimental soil samples and the interpolation feature set of the soil to be remediated.

[0144] For more details on data processing systems used for ecological restoration, please refer to [link / reference]. Figure 1 The data processing method shown is used for ecological restoration.

[0145] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A data processing method for ecological restoration, characterized in that, include: The study aims to obtain multidimensional soil characteristics and remediation strategies from multiple experimental soil samples. The multidimensional soil characteristics include physical characteristics, chemical characteristics, biological characteristics, and pollution characteristics. The physical characteristics include at least particle size distribution, porosity, and water content. The chemical characteristics include at least pH value, organic matter content, and heavy metal content. The biological characteristics include at least microbial biomass, enzyme activity, and respiration intensity. The pollution characteristics include at least pollutant type and pollution degree. The remediation strategies include at least remediation technologies and technical parameters. The remediation technology is at least one of physical remediation, chemical remediation, and biological remediation. Based on the multidimensional soil characteristics of the experimental soil, multiple soil characteristic factor groups were identified. Multiple feature interpolation models were established based on multiple soil feature factor groups; Obtain the sampling feature set of the soil to be remediated; Based on the sampled feature set of the soil to be remediated, an interpolated feature set of the soil to be remediated is generated by using multiple feature interpolation models. Based on the multidimensional soil characteristics and remediation strategies of multiple experimental soil samples and the interpolation feature set of the soil to be remediated, a remediation strategy for the soil to be remediated is generated. Based on the multidimensional soil characteristics of the experimental soil, several soil characteristic factor groups were identified, including: For each experimental soil sample, a remediation strategy feature vector is generated based on the remediation strategy of the experimental soil. For any two experimental soil samples, calculate the vector distance between the feature vectors of the remediation strategies of the two experimental soil samples; For each soil characteristic factor, based on the multidimensional soil characteristics of multiple experimental soils, the characteristic difference between any two experimental soils corresponding to the soil characteristic factor is calculated. Based on the vector distance of the remediation strategy characteristic vectors of any two experimental soils and the characteristic difference between any two experimental soils corresponding to the soil characteristic factor, the strategy influence value of the soil characteristic factor is calculated. Based on the strategy impact value of each soil characteristic factor, multiple target soil characteristic factors are screened. Based on multiple target soil characteristic factors, multiple soil characteristic factor groups were determined. Based on multiple target soil characteristic factors, multiple groups of soil characteristic factors were identified, including: For any two target soil characteristic factors, the plane variation correlation coefficient and vertical variation correlation coefficient corresponding to the two target soil characteristic factors are determined based on the multidimensional soil characteristics of multiple experimental soil samples. Based on the correlation coefficients of planar and vertical variations of any two target soil characteristic factors, multiple groups of soil characteristic factors are determined. Based on multiple soil characteristic factor groups, several feature interpolation models were established, including: For each soil feature factor group, an initial feature interpolation model corresponding to the soil feature factor group is established. The feature interpolation model includes an interpolation encoder and an interpolation decoder. The interpolation encoder is used to extract the spatial features of the feature set with a first sampling density corresponding to the soil feature factor group. The interpolation decoder is used to generate a feature set with a second sampling density corresponding to the soil feature factor group based on the spatial features output by the interpolation encoder. The first sampling density is less than the second sampling density. Based on the multidimensional soil characteristics of multiple experimental soil samples, multiple training samples corresponding to soil characteristic factor groups are generated. Based on multiple training samples corresponding to the soil characteristic factor group, the initial feature interpolation model corresponding to the soil characteristic factor group is trained to generate the feature interpolation model corresponding to the soil characteristic factor group.

2. The data processing method for ecological restoration according to claim 1, characterized in that, Based on the correlation coefficients of planar and vertical variations corresponding to any two target soil characteristic factors, multiple groups of soil characteristic factors are determined, including: Initialize the first optimization parameters, wherein the first optimization parameters include at least the number of particles, the maximum number of iterations, and the number of soil characteristic factors included in the soil characteristic factor group; Based on the correlation coefficients of the plane variation and the vertical variation of any two target soil characteristic factors, multiple target soil characteristic factors are clustered to generate clustering results; Based on the clustering results, initialize the first particle swarm; A first fitness function is established, wherein the first fitness function is related to the number of soil characteristic factor groups included in the particle and the plane variation correlation coefficient and vertical variation correlation coefficient corresponding to any two target soil characteristic factors included in each soil characteristic factor group; The first particle swarm is iteratively optimized based on the first fitness function to determine multiple soil characteristic factor groups.

3. The data processing method for ecological restoration according to claim 1, characterized in that, Based on multiple training samples corresponding to soil characteristic factor groups, an initial feature interpolation model corresponding to the soil characteristic factor groups is trained to generate a feature interpolation model corresponding to the soil characteristic factor groups, including: The training samples corresponding to the soil feature factor group are input into the initial feature interpolation model corresponding to the soil feature factor group, and the initial feature interpolation model outputs the predicted value. The loss function is calculated by comparing the predicted values ​​output by the initial feature interpolation model with the true values ​​of the training samples corresponding to the soil feature factor groups. The loss function includes a vertical consistency loss term with adaptive weights and a planar consistency loss term with adaptive weights. The parameters of the initial feature interpolation model are adjusted by optimizing the algorithm to minimize the loss function value.

4. The data processing method for ecological restoration according to claim 3, characterized in that, Based on the sampled feature set of the soil to be remediated, an interpolated feature set is generated using multiple feature interpolation models, including: Based on multiple soil characteristic factor groups and the sampling feature set of the soil to be remediated, a feature set of the first sampling density corresponding to each soil characteristic factor group of the soil to be remediated is generated. Based on the feature set of the first sampling density of each soil feature factor group corresponding to the soil to be repaired, a feature set of the second sampling density of the soil to be repaired is generated by multiple feature interpolation models. Based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated, an interpolated feature set of the soil to be remediated is generated.

5. The data processing method for ecological restoration according to claim 4, characterized in that, Based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated, an interpolated feature set of the soil to be remediated is generated, including: Global interpolation consistency is calculated based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated. Determine whether to perform interpolation fusion based on global interpolation consistency; If so, initialize the second optimization parameters; The second particle swarm is initialized based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be repaired. A second fitness function is established based on the feature set of the second sampling density corresponding to each soil characteristic factor group of the soil to be remediated. The second particle swarm is iteratively optimized based on the second fitness function to determine the interpolation feature set of the soil to be remediated.

6. The data processing method for ecological restoration according to claim 1 or 2, characterized in that, Based on the multidimensional soil characteristics and remediation strategies of multiple experimental soil samples, as well as the interpolation feature set of the soil to be remediated, a remediation strategy for the soil to be remediated is generated, including: Based on the multidimensional soil characteristics of multiple experimental soil samples and the interpolation feature set of the soil to be remediated, the target experimental soil is determined. Based on the remediation strategy of the target experimental soil, a remediation strategy for the soil to be remediated is generated.

7. A data processing system for ecological restoration, characterized in that, The data processing method for ecological restoration according to claim 1 includes: The data acquisition module is used to acquire multidimensional soil characteristics and remediation strategies for multiple experimental soil samples. The multidimensional soil characteristics include physical characteristics, chemical characteristics, biological characteristics, and pollution characteristics. The physical characteristics include at least particle size distribution, porosity, and water content. The chemical characteristics include at least pH value, organic matter content, and heavy metal content. The biological characteristics include at least microbial biomass, enzyme activity, and respiration intensity. The pollution characteristics include at least pollutant type and pollution degree. The remediation strategy includes at least remediation technology and technical parameters. The remediation technology is at least one of physical remediation, chemical remediation, and biological remediation. The factor grouping module is used to determine multiple soil characteristic factor groups based on the multidimensional soil characteristics of the experimental soil. The model building module is used to build multiple feature interpolation models based on multiple soil feature factor groups; The feature interpolation module is used to obtain the sampling feature set of the soil to be remediated, and generate the interpolated feature set of the soil to be remediated based on the sampling feature set of the soil to be remediated through multiple feature interpolation models. The strategy generation module is used to generate a remediation strategy for the soil to be remediated based on the multidimensional soil characteristics and remediation strategies of multiple experimental soil samples and the interpolation feature set of the soil to be remediated.

Images