A wetland slope bank precious tree species intelligent adaptation decision method and system based on multi-source data fusion

Abstract

The application provides a wetland slope bank precious tree species intelligent adaptation decision system based on multi-source data fusion, belongs to the wetland slope bank planting technical field, and comprises the following modules: multi-source data fusion, habitat accurate delineation, adaptability grade quantification, intelligent configuration, dynamic evaluation and optimization and the like. The application solves the problems of data isolation, poor adaptability, low-efficiency decision and no closed-loop optimization in the existing wetland slope bank precious tree species configuration, provides a wetland slope bank precious tree species intelligent adaptation decision system based on multi-source data fusion, realizes the whole-process intelligent decision of "habitat delineation-tree species matching-configuration optimization-effect evaluation-dynamic iteration", improves the survival rate of precious tree species and the stability of wetland ecological function, maximizes the ecological functions of soil and bank protection, water purification, biological conservation and carbon fixation and carbon sink, and provides a replicable and popularized intelligent technical scheme for various wetland slope bank ecological restoration.

Description

Technical Field

[0001] This invention belongs to the technical fields of wetland ecological restoration, vegetation ecological configuration, intelligent decision-making and multi-source data fusion, specifically involving an intelligent adaptation decision-making system for precious tree species on wetland slopes based on multi-source data fusion. Background Technology

[0002] As a transitional zone between water and land, wetland slopes are a key component of wetland ecosystems, possessing both ecological protection and landscape value. Precious tree species, due to their ecological, scientific research, and economic value, are the core carriers for wetland slope vegetation construction. However, their growth is highly specific to environmental conditions such as hydrology, soil, and climate, and they are sensitive to stress thresholds such as seasonal flooding, soil salinity, soil infertility, and slope wind force.

[0003] Currently, the selection of precious tree species for wetland slopes in China relies entirely on the experience and judgment of technical and construction personnel, resulting in three major technical challenges:

[0004] First, there is a serious data silo problem. Remote sensing, drones, ground IoT, soil sampling, meteorological and hydrological data, tree species baseline data, and historical operation and maintenance data are isolated from each other and have not been unified in time and space or fused with features, making it impossible to characterize the slope micro-habitat in a refined and quantitative manner.

[0005] Second, there is a lack of wetland slope-specific habitat evaluation and a quantitative system for the stress resistance of precious tree species. Adaptation models have not been built for special scenarios such as water level fluctuations, slope gradient, salinity stress, and seasonal flooding. Tree species and habitat matching is blind, resulting in low survival rates and difficulty in fully realizing ecological functions.

[0006] Third, the decision-making process is extensive and inefficient, relying on manual surveys and experience-based tree selection and configuration, which cannot support large-scale wetland slope vegetation planning. Furthermore, it lacks routine monitoring, effect evaluation, and dynamic optimization loops after planting, making it difficult to meet the needs of refined management for smart wetland construction and ecological restoration.

[0007] This situation has resulted in a double waste of resources: on the one hand, a large number of precious tree species are under protection pressure due to the lack of suitable habitats for propagation and utilization; on the other hand, the unique moist and fertile site conditions of wetland revetments and the public attention they attract have failed to be transformed into higher-level platforms for biodiversity conservation, distinctive landscape creation, and nature education.

[0008] Existing technologies primarily focus on monitoring single wetland indicators, configuring general wetland vegetation, and evaluating the flood tolerance of common tree species. They fail to address the unique challenges of wetland slopes where land and water intersect, lack multi-source heterogeneous data fusion technologies, and fail to establish dedicated intelligent adaptation models for precious tree species, thus lacking a closed-loop, end-to-end decision-making process. Therefore, there is an urgent need to propose an intelligent adaptation decision-making system for precious tree species on wetland slopes based on multi-source data fusion. This system should enable precise habitat characterization, intelligent tree species matching, multi-objective optimization of configuration, and dynamic iterative evaluation of effects, thereby improving the scientific rigor of wetland slope vegetation construction and ecosystem stability. Summary of the Invention

[0009] This invention aims to create an evaluation method and system that integrates multi-source data and intelligent analysis to solve the problems of isolated data, poor adaptability, inefficient decision-making, and lack of closed-loop optimization in the current configuration of precious tree species on wetland slopes. It provides an intelligent adaptation decision-making system for precious tree species on wetland slopes based on multi-source data fusion, realizing intelligent decision-making throughout the entire process of "habitat characterization - tree species matching - configuration optimization - effect evaluation - dynamic iteration". This improves the survival rate of precious tree species and the stability of wetland ecological functions, maximizes the ecological functions of slope soil stabilization, water purification, biological conservation, and carbon sequestration, and provides a replicable and scalable intelligent technical solution for the ecological restoration of various wetland slopes.

[0010] To achieve the above objectives, the specific technical solution of this invention is as follows:

[0011] A smart adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion includes:

[0012] S1. Obtain multi-source flood tolerance and plant growth assessment data of the target precious tree species, and use spatiotemporal alignment algorithm and feature-level fusion algorithm to fuse heterogeneous data to generate a standardized wetland slope habitat-tree species fusion dataset; the multi-source data includes native habitat distribution data, wetland spatial background data and hydrological and meteorological data;

[0013] S2. Based on the fused dataset, a quantitative evaluation index system for wetland slope habitats is constructed. The weight of each index is determined by the analytic hierarchy process (AHP), and the overall suitability of the habitat is calculated. The evaluation index system includes four dimensions: hydrological conditions, soil conditions, topographic conditions, and climatic conditions.

[0014] S3. Construct a model for assessing the suitability of precious tree species. The model takes the comprehensive suitability of the habitat and the indicators of each dimension, the physiological characteristics of the tree species and the ecological function parameters as inputs, and the survival rate, growth and ecological function efficiency of the precious tree species as outputs. The model is trained by machine learning algorithm to obtain a suitability scoring model, quantifies the suitability score of candidate tree species and target slope habitat, and classifies the suitability.

[0015] S4. Based on the adaptability score and combined with the wetland ecosystem function objectives, a multi-objective optimization algorithm is used to generate the optimal precious tree species configuration scheme; the configuration scheme includes tree species selection, spatial configuration, and planting density optimization;

[0016] S5. Regularly collect vegetation growth data and ecological function data after planting, and use the Kalman filter algorithm to fuse model predictions and actual observations to calculate the configuration effect evaluation index; when the configuration effect evaluation index is lower than the preset threshold, trigger the closed-loop optimization mechanism to dynamically adjust the tree species configuration scheme.

[0017] Furthermore, in step S1, a spatiotemporal alignment algorithm and a feature-level fusion algorithm are used to fuse heterogeneous data, specifically including:

[0018] For point data from sensors, ordinary kriging interpolation is used to spatially align the data to a unified spatial grid, and a system of ordinary kriging equations is constructed. Solving for Kriging weights , where the matrix The vector is composed of the variogram values ​​between known sampling points. The estimated value of the unknown point is composed of the variogram values ​​between known sampling points and unknown grid points. ;

[0019] To address the data redundancy arising from the fusion of multispectral remote sensing images and terrain features, principal component analysis (PCA) is employed for feature-level fusion and dimensionality reduction, and a multi-source feature data matrix is ​​calculated. covariance matrix Solve the characteristic equation Before selection The projection matrix is ​​composed of the eigenvectors corresponding to the largest eigenvalues. The dimensionality-reduced fused data is obtained. .

[0020] Furthermore, in step S2, the weights of each indicator are determined using the analytic hierarchy process (AHP) and the overall habitat suitability is calculated. Specifically, this includes:

[0021] Construct a judgment matrix The weight vector is calculated using the square root method, and the product of row elements is calculated. ,open The root is obtained Normalization yields the first Weight of each indicator ;

[0022] Calculate the largest eigenvalue And calculate the consistency ratio. ,when It passed the consistency test.

[0023] Standardize each habitat indicator as follows: Calculate the overall habitat suitability .

[0024] Furthermore, in step S3, a fitness scoring model is trained using a machine learning algorithm, specifically the XGBoost gradient boosting tree algorithm, and its training process includes:

[0025] Construct an objective function that includes a loss function and a regularization term. ;

[0026] Performing a second-order Taylor expansion on the loss function yields an approximate objective function. ,in and These are the first and second derivatives of the loss function, respectively.

[0027] Regroup the samples by leaf nodes and solve for the first... Optimal weights for leaf nodes ,in , ;

[0028] Calculate node split gain Iterate through all features and split points, and select... The largest point is used to grow a decision tree, and the final output is the fitness score. .

[0029] Furthermore, in step S4, a multi-objective optimization algorithm is used to generate the optimal configuration scheme for precious tree species. The multi-objective optimization algorithm is the NSGA-II algorithm, which specifically includes:

[0030] With the objective functions of maximizing survival rate, maximizing ecological function, and minimizing cost, a rapid non-dominated ordination of the population is performed, dividing the population into multiple non-dominated layers. ;

[0031] Within the same non-dominated layer, calculate the crowding distance for each individual. Individuals with lower non-dominant tiers and greater crowding distances are given priority for entering the next generation;

[0032] Offspring are generated using a simulated binary crossover operator, and the crossover formula is as follows: ,in For uniformly distributed random numbers and distribution index The expansion factor of the calculation;

[0033] Output the Pareto front non-dominated solution set as the optimal configuration scheme that includes planting location, row spacing, mixing ratio and planting density.

[0034] In step S5, the Kalman filter algorithm is used to fuse the model predictions and the measured observations, specifically including the following iterative steps:

[0035] State prediction: Based on the optimal estimate of the previous time step Predict the current state ;

[0036] Error covariance prediction: ,in The process noise covariance matrix;

[0037] Calculate the Kalman gain: ,in For the observation matrix, To measure the noise covariance matrix;

[0038] State update: Correcting predicted values ​​using observed residuals. ,in The measured observation vector;

[0039] Error covariance update: ;

[0040] Based on the updated optimal state estimate Calculate the configuration effect evaluation index ,when When it is below the preset threshold, As initial conditions, the model in steps S3 and S4 is re-inputted for rolling time-domain optimization.

[0041] A smart adaptation decision-making system for precious tree species on wetland slopes based on multi-source data fusion, used to implement the aforementioned smart adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion, includes:

[0042] The multi-source data acquisition and fusion module is used to acquire multi-source flood tolerance and plant growth assessment data of the target precious tree species. It uses spatiotemporal alignment algorithm and feature-level fusion algorithm to fuse heterogeneous data and generate a standardized wetland slope habitat-tree species fusion dataset.

[0043] The habitat precise characterization module is used to construct a quantitative evaluation index system for wetland slope habitats based on the fused dataset, and to determine the weight of each index using the analytic hierarchy process and calculate the overall habitat suitability.

[0044] The adaptability rating quantification module is used to construct an adaptability assessment model for precious tree species. It takes the comprehensive habitat suitability and indicators of each dimension, tree species physiological characteristics and ecological function parameters as inputs, and uses machine learning algorithms to train an adaptability scoring model to quantify the adaptability score of candidate tree species.

[0045] The intelligent configuration decision module is used to generate an optimal precious tree species configuration scheme that includes tree species selection, spatial configuration and planting density optimization based on the adaptability score and in combination with wetland ecosystem function objectives, using a multi-objective optimization algorithm.

[0046] The dynamic evaluation and optimization module is used to periodically collect vegetation growth data and ecological function data after planting. It uses the Kalman filter algorithm to fuse model predictions and actual observations, calculates the configuration effect evaluation index, and triggers a closed-loop optimization mechanism when the value is lower than a preset threshold.

[0047] The beneficial effects of the intelligent adaptation decision-making method and system for precious tree species on wetland slopes based on multi-source data fusion provided by this invention are as follows:

[0048] 1. Solve the problem of data silos: Integrate heterogeneous data from multiple sources such as remote sensing, ground monitoring, and baseline data to achieve integrated fusion of wetland slope habitat and precious tree species data, providing comprehensive and accurate data support for decision-making.

[0049] 2. Improve the accuracy of adaptability: Construct a habitat characterization index system and a special adaptation model for precious tree species for slope and bank scenarios, quantify the degree of adaptability between tree species and habitat, avoid blind configuration, and improve the survival rate of precious tree species (expected to increase by 20%-30%) and the efficiency of ecological function.

[0050] 3. Achieve intelligent and efficient decision-making: Through machine learning and multi-objective optimization algorithms, the optimal configuration scheme is automatically generated to replace traditional experience-based decision-making, greatly improving decision-making efficiency and supporting large-scale, dynamic wetland slope vegetation construction.

[0051] 4. Strengthen wetland ecological functions: By scientifically configuring precious tree species, we can enhance functions such as slope and bank stabilization, water purification, water conservation, and biodiversity maintenance, thereby contributing to the high-quality development of wetland ecosystems and providing replicable and scalable technical solutions for wetland protection and restoration. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the architecture of an intelligent adaptation decision-making system for precious tree species on wetland slopes based on multi-source data fusion provided by the present invention.

[0053] Figure 2 This is a flowchart illustrating an intelligent adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion, provided by the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific examples. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Example 1:

[0056] like Figure 2 As shown, it includes:

[0057] S1. Obtain multi-source flood tolerance and plant growth assessment data of the target precious tree species, and use spatiotemporal alignment algorithm and feature-level fusion algorithm to fuse heterogeneous data to generate a standardized wetland slope habitat-tree species fusion dataset; the multi-source data includes native habitat distribution data, wetland spatial background data and hydrological and meteorological data;

[0058] S2. Based on the fused dataset, a quantitative evaluation index system for wetland slope habitats is constructed. The weight of each index is determined by the analytic hierarchy process (AHP), and the overall suitability of the habitat is calculated. The evaluation index system includes four dimensions: hydrological conditions, soil conditions, topographic conditions, and climatic conditions.

[0059] S3. Construct a model for assessing the suitability of precious tree species. The model takes the comprehensive suitability of the habitat and the indicators of each dimension, the physiological characteristics of the tree species and the ecological function parameters as inputs, and the survival rate, growth and ecological function efficiency of the precious tree species as outputs. The model is trained by machine learning algorithm to obtain a suitability scoring model, quantifies the suitability score of candidate tree species and target slope habitat, and classifies the suitability.

[0060] S4. Based on the adaptability score and combined with the wetland ecosystem function objectives, a multi-objective optimization algorithm is used to generate the optimal precious tree species configuration scheme; the configuration scheme includes tree species selection, spatial configuration, and planting density optimization;

[0061] S5. Regularly collect vegetation growth data and ecological function data after planting, and use the Kalman filter algorithm to fuse model predictions and actual observations to calculate the configuration effect evaluation index; when the configuration effect evaluation index is lower than the preset threshold, trigger the closed-loop optimization mechanism to dynamically adjust the tree species configuration scheme.

[0062] Furthermore, in step S1, a spatiotemporal alignment algorithm and a feature-level fusion algorithm are used to fuse heterogeneous data, specifically including:

[0063] For point data from sensors, ordinary kriging interpolation is used to spatially align the data to a unified spatial grid, and a system of ordinary kriging equations is constructed. Solving for Kriging weights , where the matrix The vector is composed of the variogram values ​​between known sampling points. The estimated value of the unknown point is composed of the variogram values ​​between known sampling points and unknown grid points. ;

[0064] To address the data redundancy arising from the fusion of multispectral remote sensing images and terrain features, principal component analysis (PCA) is employed for feature-level fusion and dimensionality reduction, and a multi-source feature data matrix is ​​calculated. covariance matrix Solve the characteristic equation Before selection The projection matrix is ​​composed of the eigenvectors corresponding to the largest eigenvalues. The dimensionality-reduced fused data is obtained. .

[0065] Furthermore, in step S2, the weights of each indicator are determined using the analytic hierarchy process (AHP) and the overall habitat suitability is calculated. Specifically, this includes:

[0066] Construct a judgment matrix The weight vector is calculated using the square root method, and the product of row elements is calculated. ,open The root is obtained Normalization yields the first Weight of each indicator ;

[0067] Calculate the largest eigenvalue And calculate the consistency ratio. ,when It passed the consistency test.

[0068] Standardize each habitat indicator as follows: Calculate the overall habitat suitability .

[0069] Furthermore, in step S3, a fitness scoring model is trained using a machine learning algorithm, specifically the XGBoost gradient boosting tree algorithm, and its training process includes:

[0070] Construct an objective function that includes a loss function and a regularization term. ;

[0071] Performing a second-order Taylor expansion on the loss function yields an approximate objective function. ,in and These are the first and second derivatives of the loss function, respectively.

[0072] Regroup the samples by leaf nodes and solve for the first... Optimal weights for leaf nodes ,in , ;

[0073] Calculate node split gain Iterate through all features and split points, and select... The largest point is used to grow a decision tree, and the final output is the fitness score. .

[0074] Furthermore, in step S4, a multi-objective optimization algorithm is used to generate the optimal configuration scheme for precious tree species. The multi-objective optimization algorithm is the NSGA-II algorithm, which specifically includes:

[0075] With the objective functions of maximizing survival rate, maximizing ecological function, and minimizing cost, a rapid non-dominated ordination of the population is performed, dividing the population into multiple non-dominated layers. ;

[0076] Within the same non-dominated layer, calculate the crowding distance for each individual. Individuals with lower non-dominant tiers and greater crowding distances are given priority for entering the next generation;

[0077] Offspring are generated using a simulated binary crossover operator, and the crossover formula is as follows: ,in For uniformly distributed random numbers and distribution index The expansion factor of the calculation;

[0078] Output the Pareto front non-dominated solution set as the optimal configuration scheme that includes planting location, row spacing, mixing ratio and planting density.

[0079] In step S5, the Kalman filter algorithm is used to fuse the model predictions and the measured observations, specifically including the following iterative steps:

[0080] State prediction: Based on the optimal estimate of the previous time step Predict the current state ;

[0081] Error covariance prediction: ,in The process noise covariance matrix;

[0082] Calculate the Kalman gain: ,in For the observation matrix, To measure the noise covariance matrix;

[0083] State update: Correcting predicted values ​​using observed residuals. ,in The measured observation vector;

[0084] Error covariance update: ;

[0085] Based on the updated optimal state estimate Calculate the configuration effect evaluation index ,when When it is below the preset threshold, As initial conditions, the model in steps S3 and S4 is re-inputted for rolling time-domain optimization.

[0086] like Figure 1 As shown, an intelligent adaptation decision-making system for precious tree species on wetland slopes based on multi-source data fusion is used to implement the aforementioned intelligent adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion, including:

[0087] The multi-source data acquisition and fusion module is used to acquire multi-source flood tolerance and plant growth assessment data of the target precious tree species. It uses spatiotemporal alignment algorithm and feature-level fusion algorithm to fuse heterogeneous data and generate a standardized wetland slope habitat-tree species fusion dataset.

[0088] The habitat precise characterization module is used to construct a quantitative evaluation index system for wetland slope habitats based on the fused dataset, and to determine the weight of each index using the analytic hierarchy process and calculate the overall habitat suitability.

[0089] The adaptability rating quantification module is used to construct an adaptability assessment model for precious tree species. It takes the comprehensive habitat suitability and indicators of each dimension, tree species physiological characteristics and ecological function parameters as inputs, and uses machine learning algorithms to train an adaptability scoring model to quantify the adaptability score of candidate tree species.

[0090] The intelligent configuration decision module is used to generate an optimal precious tree species configuration scheme that includes tree species selection, spatial configuration and planting density optimization based on the adaptability score and in combination with wetland ecosystem function objectives, using a multi-objective optimization algorithm.

[0091] The dynamic evaluation and optimization module is used to periodically collect vegetation growth data and ecological function data after planting. It uses the Kalman filter algorithm to fuse model predictions and actual observations, calculates the configuration effect evaluation index, and triggers a closed-loop optimization mechanism when the value is lower than a preset threshold.

[0092] Example 2:

[0093] Taking the small wetland at the Jiangxi Provincial Forestry Science and Technology Experimental Center (25.36556266°N, 114.90521532°E) as an example, this study focuses on the configuration and planting of precious tree species along the wetland revetment. The small artificial lake, covering approximately 3.5 acres, has red soil with a pH of 7.0-7.6 and an average annual water level fluctuation of 0.8-1.5 meters. Its main ecological objectives are soil stabilization and bank protection, water quality improvement, and the creation of a beautiful landscape environment. The main precious tree species planted include *Senecio scandens*, *Dalbergia odorifera*, *Metasequoia glyptostroboides*, and *Cephalotaxus fortunei*.

[0094] 1. Multi-source data acquisition and fusion:

[0095] The slope DEM, slope aspect, slope length, and vegetation cover were obtained through Gaofen-2 satellite and UAV oblique photography; soil organic matter was sampled and tested in the field, with soil moisture content of 1.2%–1.8%, soil salinity of 25%–30%, and soil salinity of 0.3%–0.5%; IoT sensors were deployed to measure the annual average water level fluctuation of 1.5m, annual rainfall of 1500mm, and average summer wind speed on the slope of 2.5m / s; meteorological station data were collected, with an annual average temperature of 22℃ and annual sunshine duration of 1600h; parameters such as waterlogging tolerance, salinity tolerance, soil stabilization strength, and nitrogen and phosphorus absorption efficiency of various precious tree species were recorded; all data were uniformly normalized to a 1m×1m spatiotemporal grid, and feature-level fusion was performed using GCN+CNN to generate a standardized fused dataset.

[0096] 2. Precise depiction of wetland slope habitats:

[0097] A multi-dimensional evaluation system comprising 12 indicators covering hydrology, soil, topography, and climate was constructed. The Analytic Hierarchy Process (AHP) was used to determine the weights of each indicator, and the overall habitat suitability was calculated after normalization.

[0098] HIS = 0.25 × 0.82 + 0.20 × 0.78 + 0.15 × 0.83 + 0.20 × 0.92 = 0.6695, indicating that the slope as a whole is a suitable habitat for afforestation of precious tree species.

[0099] (3) Evaluation of the suitability of precious tree species:

[0100] Call the trained Random Forest (RF) model and output the SAS score and fitness level for each candidate tree species:

[0101] species Water resistance Salt and alkali resistance Soil stabilization and bank protection Nitrogen and phosphorus absorption rate SAS score Compatibility level Metasequoia 70d 0.5% 0.9MPa 15g / (m2a) 85 Highly adaptable Bald cypress 75d 0.6% 1.1MPa 18g / (m2a) 88 Highly adaptable Half Maple 70d 0.7% 1.0MPa 16g / (m2a) 82 Highly adaptable Rosewood 65d 0.4% 0.9MPa 12g / (m2a) 75 Moderately compatible

[0102] 4) Intelligent configuration decision output:

[0103] With soil stabilization and bank protection and water purification as the core objectives, the NSGA-II multi-objective optimization algorithm is used to generate the configuration scheme:

[0104] Tree species structure: 60% metasequoia, 30% bald cypress, 10% semi-maple schefflera as companion and supplementary planting, and scattered dalbergia odorifera in some areas;

[0105] Spatial layout: In areas with gentle slopes of 5° to 8°, bald cypress trees are planted at a spacing of 3m × 4m; in areas with steep slopes of 8° to 15°, pond cypress trees are planted in denser spacing to enhance soil stabilization and slope protection.

[0106] Community pattern: A layered mixed planting of trees, moisture-tolerant shrubs and aquatic herbs is adopted to construct a stable near-natural vegetation community;

[0107] The system automatically outputs GIS configuration drawings, tree species list, planting density parameters, and annual management and pest control technical solutions.

[0108] (5) Effect evaluation and dynamic optimization:

[0109] After afforestation is completed, drone remote sensing and ground quadrat monitoring are carried out quarterly to calculate the CEI configuration effect index. If the survival rate, water purification benefits, and vegetation coverage do not reach the preset threshold, the system automatically recalculates the habitat suitability and tree species compatibility score, and makes local adjustments to tree species replacement, planting density and management measures to achieve long-term closed-loop optimization.

[0110] (6) Implementation results:

[0111] Through the implementation of this patented technology, the planting scheme for precious tree species on the slopes of small wetlands at the Provincial Forestry Science and Technology Experimental Center has achieved intelligent and precise configuration. Monitoring data one year after implementation shows that the average survival rate of the four target precious tree species reached 89.7%, an increase of 32.5% compared to traditional planting methods; slope soil erosion decreased by 65%, and water purification capacity increased by 40%, forming a stable and functional wetland slope ecosystem. Simultaneously, it protected local precious tree species resources, achieving a win-win situation for both ecological benefits and species conservation.

[0112] While embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations fall within the scope and spirit of the invention as set forth in the claims. Furthermore, the invention described herein may have other embodiments and can be implemented or carried out in various ways.

Claims

1. A smart adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion, characterized in that, include: S1. Obtain multi-source flood tolerance and plant growth assessment data of the target precious tree species, and use spatiotemporal alignment algorithm and feature-level fusion algorithm to fuse heterogeneous data to generate a standardized wetland slope habitat-tree species fusion dataset; the multi-source data includes native habitat distribution data, wetland spatial background data and hydrological and meteorological data; S2. Based on the fused dataset, a quantitative evaluation index system for wetland slope habitats is constructed. The weight of each index is determined by the analytic hierarchy process (AHP), and the overall suitability of the habitat is calculated. The evaluation index system includes four dimensions: hydrological conditions, soil conditions, topographic conditions, and climatic conditions. S3. Construct a model for assessing the suitability of precious tree species. The model takes the comprehensive suitability of the habitat and the indicators of each dimension, the physiological characteristics of the tree species and the ecological function parameters as inputs, and the survival rate, growth and ecological function efficiency of the precious tree species as outputs. The model is trained by machine learning algorithm to obtain a suitability scoring model, quantifies the suitability score of candidate tree species and target slope habitat, and classifies the suitability. S4. Based on the adaptability score and combined with the wetland ecosystem function objectives, a multi-objective optimization algorithm is used to generate the optimal precious tree species configuration scheme; the configuration scheme includes tree species selection, spatial configuration, and planting density optimization; S5. Regularly collect vegetation growth data and ecological function data after planting, and use the Kalman filter algorithm to fuse model predictions and actual observations to calculate the configuration effect evaluation index; when the configuration effect evaluation index is lower than the preset threshold, trigger the closed-loop optimization mechanism to dynamically adjust the tree species configuration scheme.

2. The intelligent adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion according to claim 1, characterized in that, In step S1, a spatiotemporal alignment algorithm and a feature-level fusion algorithm are used to fuse heterogeneous data, specifically including: For point data from sensors, ordinary kriging interpolation is used to spatially align the data to a unified spatial grid, and a system of ordinary kriging equations is constructed. Solving for Kriging weights , where the matrix The vector is composed of the variogram values ​​between known sampling points. The estimated value of the unknown point is composed of the variogram values ​​between known sampling points and unknown grid points. ; To address the data redundancy arising from the fusion of multispectral remote sensing imagery and terrain features, principal component analysis (PCA) is employed for feature-level fusion and dimensionality reduction, calculating the multi-source feature data matrix. covariance matrix Solve the characteristic equation Before selection The projection matrix is ​​composed of the eigenvectors corresponding to the largest eigenvalues. The dimensionality-reduced fused data is obtained. .

3. The intelligent adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion according to claim 1, characterized in that, In step S2, the weights of each indicator are determined and the overall habitat suitability is calculated using the analytic hierarchy process (AHP), specifically including: Construct a judgment matrix The weight vector is calculated using the square root method, and the product of row elements is calculated. ,open The root is obtained Normalization yields the first Weight of each indicator ; Calculate the largest eigenvalue And calculate the consistency ratio. ,when It passed the consistency test. Standardize each habitat indicator as follows: Calculate the overall habitat suitability .

4. The intelligent adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion according to claim 1, characterized in that, In step S3, a fitness scoring model is trained using a machine learning algorithm, specifically the XGBoost gradient boosting tree algorithm. The training process includes: Construct an objective function that includes a loss function and a regularization term. ; Performing a second-order Taylor expansion on the loss function yields an approximate objective function. ,in and These are the first and second derivatives of the loss function, respectively. Regroup the samples by leaf nodes and solve for the first... Optimal weights for leaf nodes ,in , ; Calculate node split gain Iterate through all features and split points, and select... The largest point is used to grow a decision tree, and the final output is the fitness score. .

5. The intelligent adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion according to claim 1, characterized in that, In step S4, a multi-objective optimization algorithm is used to generate the optimal precious tree species configuration scheme. The multi-objective optimization algorithm is the NSGA-II algorithm, which specifically includes: With the objective functions of maximizing survival rate, maximizing ecological function, and minimizing cost, a rapid non-dominated ordination of the population is performed, dividing the population into multiple non-dominated layers. ; Within the same non-dominated layer, calculate the crowding distance for each individual. Individuals with lower non-dominant tiers and greater crowding distances are given priority for entering the next generation; Offspring are generated using a simulated binary crossover operator, and the crossover formula is as follows: ,in For uniformly distributed random numbers and distribution index The expansion factor of the calculation; Output the Pareto front non-dominated solution set as the optimal configuration scheme that includes planting location, row spacing, mixing ratio and planting density.

6. The intelligent adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion according to claim 1, characterized in that, In step S5, the Kalman filter algorithm is used to fuse the model predictions and the measured observations, specifically including the following iterative steps: State prediction: Based on the optimal estimate of the previous time step Predict the current state ; Error covariance prediction: ,in The process noise covariance matrix; Calculate the Kalman gain: ,in For the observation matrix, To measure the noise covariance matrix; State update: Correcting predicted values ​​using observed residuals. ,in The measured observation vector; Error covariance update: ; Based on the updated optimal state estimate Calculate the configuration effect evaluation index ,when When it is below the preset threshold, As initial conditions, the model in steps S3 and S4 is re-inputted for rolling time-domain optimization.

7. A smart adaptation decision-making system for precious tree species on wetland slopes based on multi-source data fusion, characterized in that, To implement the intelligent adaptation decision-making method for precious tree species on wetland slopes based on multi-source data fusion as described in claims 1 to 6, the method includes: The multi-source data acquisition and fusion module is used to acquire multi-source flood tolerance and plant growth assessment data of the target precious tree species. It uses spatiotemporal alignment algorithm and feature-level fusion algorithm to fuse heterogeneous data and generate a standardized wetland slope habitat-tree species fusion dataset. The habitat precise characterization module is used to construct a quantitative evaluation index system for wetland slope habitats based on the fused dataset, and to determine the weight of each index using the analytic hierarchy process and calculate the overall habitat suitability. The adaptability rating quantification module is used to construct an adaptability assessment model for precious tree species. It takes the comprehensive habitat suitability and indicators of each dimension, tree species physiological characteristics and ecological function parameters as inputs, and uses machine learning algorithms to train an adaptability scoring model to quantify the adaptability score of candidate tree species. The intelligent configuration decision module is used to generate an optimal precious tree species configuration scheme that includes tree species selection, spatial configuration and planting density optimization based on the adaptability score and in combination with wetland ecosystem function objectives, using a multi-objective optimization algorithm. The dynamic evaluation and optimization module is used to periodically collect vegetation growth data and ecological function data after planting. It uses the Kalman filter algorithm to fuse model predictions and actual observations, calculates the configuration effect evaluation index, and triggers a closed-loop optimization mechanism when the value is lower than a preset threshold.

Images