A Method and System for Recommending Landscape Vegetation Based on Ecological and Environmental Data

Abstract

This invention relates to the field of data processing, proposing a method and system for recommending garden vegetation based on ecological and environmental data. First, environmental monitoring signals are acquired using sensor networks. Based on geostatistical variation patterns, spatial reconstruction and spatiotemporal interpolation of the signals are performed to construct a digital environmental model. Second, a global ecological neighborhood benchmark is constructed to standardize the signals, and a resource pressure index is calculated and introduced into a dynamic competition model as a nonlinear gain environmental factor weight. Third, multiple ecological strategy control logics are pre-set, and a Gaussian probability mapping between the resource pressure index and the applicability of each strategy is established. The strategy outputs are weighted and integrated using posterior probability weights to obtain configuration instructions including initial maintenance measures. Finally, feedback signals after the implementation of the instructions are monitored. When deviations exceed limits, physical adjustment mechanisms such as water and fertilizer regulation or microhabitat improvement are activated, and the model parameters are adaptively corrected. This invention achieves precise decision-making and closed-loop dynamic control of garden vegetation configuration.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to a method and system for recommending garden vegetation based on ecological environment data. Background Technology

[0002] Recommendations for landscape vegetation configuration determine the sustainability and ecological service functions of the landscape. Faced with increasingly complex urban microclimate changes and the need for refined maintenance, existing technologies mostly use single-point discrete sampling or directly refer to climate zone (such as weather station) data as environmental input.

[0003] However, due to the high spatial heterogeneity of garden habitats, the weights of soil physicochemical properties and microclimate environmental factors often fluctuate dramatically on a small scale. Discrete monitoring points are unable to capture this continuous spatial variation pattern, resulting in a serious lack of spatial representativeness of environmental data.

[0004] Furthermore, there is a significant scale mismatch between meteorological data and plant growth requirements. Making decisions directly based on such data with spatiotemporal discontinuities can easily lead to severe niche mismatches in local micro-topography (such as depressions and wind gaps), resulting in large-scale vegetation death.

[0005] Meanwhile, the severity of environmental conditions has a nonlinear catalytic effect on the intensity of interspecies competition. As resource pressure increases, the competitive behavior of organisms vying for limited survival resources intensifies exponentially, resulting in extremely low accuracy in predicting survival rates under extreme climate or resource-scarce conditions.

[0006] Existing recommendation algorithms typically employ deterministic single-objective optimization logic when addressing the aforementioned problems, lacking adaptability to environmental uncertainties. In dynamically fluctuating environments, using a single strategy often faces the risk of overfitting, meaning it may perform well under specific conditions but quickly fail when the environment fluctuates. Furthermore, over time, the deviation between the static model and the real environment gradually accumulates, leading to the failure of maintenance decisions.

[0007] Overcoming the spatiotemporal fragmentation of environmental perception, constructing a dynamic model capable of quantifying the nonlinear gain of environmental stress, and establishing an intelligent decision-making system with multi-strategy adaptation and closed-loop feedback capabilities are key technical problems that urgently need to be solved in the field of landscape engineering. Summary of the Invention

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution.

[0009] A method for recommending garden vegetation based on ecological environment data includes: acquiring monitoring signals of the target area environment using a sensor network; performing spatial reconstruction and spatiotemporal interpolation of the signals based on geostatistical variation laws to construct a digital environmental model reflecting environmental heterogeneity; constructing a global ecological neighborhood benchmark to standardize the monitoring signals; calculating a resource pressure index characterizing the degree of stress posed by actual environmental conditions to vegetation survival; and introducing this index as a nonlinear gain environmental factor weight into a dynamic competition model for simulating survival competition among vegetation types; the dynamic competition model outputs the dynamic competition intensity; in the construction of the dynamic competition model: estimating the canopy radius of vegetation using the allometric growth equation and calculating the distance attenuation coefficient; and incorporating ecological niche overlap, including the resource pressure index... The dynamic competition intensity is obtained by multiplying the environmental stress nonlinear gain environmental factor weight and the spatial attenuation term based on the distance attenuation coefficient. Multiple ecological strategy control logics for different pressure gradients are pre-set, and a Gaussian probability mapping is established between the resource pressure index and the applicability of each strategy to generate posterior probability weights under the current environmental state. The outputs of each ecological strategy control logic are probabilistically weighted and integrated using these posterior probability weights to obtain vegetation configuration instructions that guide planting construction and include targeted initial maintenance measures. These initial maintenance measures are implemented, and the growth feedback signal after the instructions are implemented is monitored. When deviations exceed limits, physical regulation mechanisms, including water and fertilizer regulation or microhabitat improvement, are activated, and the model parameters are adaptively corrected based on the adjusted feedback signal.

[0010] Preferably, the spatial reconstruction and spatiotemporal interpolation completion of the signal based on geostatistical variation law includes: performing random pre-sampling in the target area, calculating the experimental semivariogram of soil environmental data and extracting the minimum range; setting the side length of the square grid to half the value of the minimum range according to the spatial sampling theorem, and deploying sensors at the center of the grid to collect time series data; for missing data, constructing a product sum spatiotemporal covariance model, and using known spatiotemporal point data to perform unbiased optimal estimation completion by solving the Kriging linear equation system.

[0011] Preferably, the standardization process and resource pressure index calculation include: acquiring environmental data within a specific radius and a specific historical period around the target area; calculating the reference mean and reference standard deviation of the environmental factor weights corresponding to each environmental monitoring signal to construct a global reference benchmark; subtracting the reference mean from the original monitoring signal and dividing by the reference standard deviation to obtain a standardized environmental vector; using the same reference mean and reference standard deviation to perform a standardized transformation on the suitable growth range of vegetation to obtain a standardized preference center value; and defining the resource pressure index as the arithmetic mean of the absolute deviations of all environmental factor weights from the global reference benchmark.

[0012] Preferably, before introducing the index as a nonlinear gain environmental factor weight into the dynamic competition model used to simulate survival competition among vegetation, the method further includes calculating the real-time fitness: using the analytic hierarchy process to determine the weights and calculate the weighted Euclidean distance between the standardized environmental vector and the preference center value, and then converting it into the real-time fitness.

[0013] Preferably, the pre-set multiple ecological strategy control logics for different pressure gradients include: establishing a robust survival strategy control logic that focuses on real-time adaptability and includes edge effect penalties for high-pressure environments; establishing a diversity symbiosis strategy control logic that focuses on competitive coexistence and includes diversity index constraints for low-pressure environments; and establishing an equilibrium recovery strategy control logic that focuses on economy and rapid coverage for transitional environments; each strategy control logic is configured with independent objective function weights.

[0014] Preferably, the step of establishing a Gaussian probability mapping between the resource pressure index and the applicability of each strategy to generate posterior probability weights under the current environmental state includes: setting the typical pressure mean and ecological tolerance parameter corresponding to each ecological strategy control logic; constructing a mapping relationship using a Gaussian kernel function, calculating the square of the difference between the current resource pressure index and the typical pressure mean of each ecological strategy control logic, calculating an index decay term based on the ratio of the square of the difference to the ecological tolerance parameter; and normalizing the index decay term in all ecological strategy control logics to obtain the posterior probability weights of each ecological strategy control logic under the current environmental state.

[0015] Preferably, the step of obtaining vegetation configuration instructions for guiding planting construction and including targeted initial maintenance measures includes: multiplying the independent scoring functions of each ecological strategy control logic with their corresponding posterior probability weights and then summing them to construct the final expected integrated objective function; using a genetic algorithm to globally optimize the expected integrated objective function and output the optimal vegetation configuration and maintenance instructions under the current environmental probability distribution.

[0016] Preferably, the activation of the physical regulation mechanism, including water and fertilizer regulation or microhabitat improvement, includes: triggering an overall diagnosis when the real-time adaptability of vegetation within the grid is lower than a preset safety threshold; or calculating the absolute deviation of the monitoring value of each environmental factor weight relative to the vegetation preference range, and generating a precise water and fertilizer regulation or habitat improvement instruction for a specific environmental factor weight when the deviation exceeds a dynamic threshold determined by the vegetation ecological amplitude.

[0017] Preferably, the adaptive correction of model parameters based on the adjusted feedback signal includes: collecting the environmental changes and vegetation health changes after physical adjustment, and obtaining the environmental factor weight influence coefficients through ridge regression fitting; using Bayesian principles, combined with prior strength and sample size, and using the environmental factor weight influence coefficients to weight and update the old environmental factor weights to obtain new environmental factor weights.

[0018] Secondly, a landscape vegetation recommendation system based on ecological environment data includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the landscape vegetation recommendation method based on ecological environment data described in any one of the claims is implemented.

[0019] The beneficial effects of this invention are:

[0020] 1. Overcoming the spatiotemporal limitations of environmental perception, this invention constructs a precise digital environmental foundation: By combining sensor networks with geostatistical variation patterns, it utilizes the Kriging spatiotemporal interpolation algorithm to spatially reconstruct discrete signals. This method not only fills the spatiotemporal blind spots of monitoring data but also ensures the maximization of spatial autocorrelation of environmental factor weights within the grid based on the spatial sampling theorem, thus physically solving the problem of insufficient spatial representativeness of environmental data. The resulting digital environmental model, reflecting environmental heterogeneity, provides a continuous, complete, and high-precision input source for subsequent refined configuration, significantly improving the solution's ability to match micro-topographic habitats.

[0021] 2. This invention reveals and quantifies the nonlinear mechanism of "environmental stress driving intensified competition," significantly improving prediction accuracy under extreme environments: Addressing the issue of existing competition models being detached from actual environmental pressures, this invention innovatively constructs a standardized processing system and calculates a resource pressure index characterizing the degree of stress posed by actual environmental conditions to vegetation survival. By introducing this index as a nonlinear gain environmental factor weight into the dynamic competition model, this invention mathematically simulates the ecological dynamic phenomenon that "the harsher the environment, the more intense the resource competition." This processing method enables the model to dynamically sense changes in environmental pressure and adjust competition resistance parameters accordingly, effectively avoiding the blind overestimation of vegetation survival rates by traditional linear models under high-pressure scenarios such as drought and barrenness, ensuring the high resilience of the configuration scheme.

[0022] 3. This invention achieves multi-strategy dynamic game theory based on probability weighting, solving the problem of lack of robustness of single strategies in non-stationary environments: It abandons rigid single optimization logic, pre-sets multiple ecological strategy control logics for different pressure gradients (such as robust survival and biodiversity symbiosis), and uses a Gaussian kernel function to establish a probabilistic mapping between the resource pressure index and the applicability of each strategy. By generating posterior probability weights and performing probability cascade fusion, the system can automatically adjust the "discourse power" of different strategies during environmental fluctuations, outputting a "Bayesian average" instruction that balances high-probability gains and bottom-line survival risks. This mechanism endows the system with a "risk hedging" wisdom similar to biological evolution, enabling it to maintain the long-term stability of the community structure in the face of climate change or environmental disturbances.

[0023] 4. A dual closed-loop feedback mechanism of "physical adjustment - parameter correction" was established, eradicating the persistent model degradation problem caused by "open-loop" decision-making: This invention extends the recommendation process from simple "scheme design" to "engineering control," directly using configuration instructions containing targeted initial maintenance measures to guide construction, and establishing a dual response logic based on growth feedback signals: On the one hand, when deviations exceed limits, physical adjustment mechanisms such as water and fertilizer regulation or microhabitat improvement are directly triggered to correct the actual environment; on the other hand, using feedback data before and after adjustment, the model parameters are adaptively corrected through a Bayesian update algorithm, enabling the system to have continuous learning capabilities. This closed-loop design ensures that the decision-making model can self-evolve as the environment evolves, significantly reducing the maintenance cost and frequency of manual intervention throughout the entire life cycle. Attached Figure Description

[0024] Figure 1 This is a flowchart of steps S1-S4 in a garden vegetation recommendation method based on ecological environment data according to an embodiment of the present invention. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0026] Reference Figure 1 A method for recommending garden vegetation based on ecological environment data includes steps S1-S4, as follows:

[0027] S1: Use sensor networks to acquire monitoring signals of the target area environment, perform spatial reconstruction and spatiotemporal interpolation of the signals based on geostatistical variation patterns, and construct a digital environmental model that reflects environmental heterogeneity.

[0028] The growth environment in different areas of a garden ecosystem varies greatly. The weights of soil nutrients, water, and microclimate environmental factors often fluctuate dramatically on a small scale. The sparse distribution of single-point discrete monitoring points makes it difficult to capture the continuous changes in the weights of these environmental factors, resulting in the acquired environmental monitoring signals failing to represent the characteristics of local microhabitats. If discrete signals collected based on sparse distribution are directly used for vegetation recommendation decisions, the recommended scheme is prone to local niche mismatch due to the bias of the input data.

[0029] Based on geostatistical variability patterns, grid-based calibration was performed. A portable soil analyzer equipped with a high-precision positioning module was used to conduct systematic random sampling in the target area, with the number of pre-sampling points set. Of the 55 that met the statistical significance requirement, The empirical values ​​can be adjusted by the implementer according to the specific implementation scenario, so that the sampling range covers different micro-topography and vegetation cover types in the area.

[0030] Simultaneously, the experimental semivariograms of the weights of key environmental factors such as soil organic matter and volumetric water content were calculated. Weighted least squares method was used to fit the experimental data to obtain spherical model parameters, and then the minimum range of each environmental factor weight was extracted. The minimum value among them is taken as the basis for grid division, and the weighted least squares method for fitting the semivariogram function to extract the range is a well-known technique.

[0031] The side length of the square grid is set according to the spatial sampling theorem. ,make This ensures that the spatial autocorrelation of environmental factor weights within the grid is maximized, enabling data collected by a single sensor to effectively represent the environmental characteristics within the grid.

[0032] After obtaining the determined grid area division results, sensors using low-power wide-area network communication protocols are deployed to construct a sensor network. Data is collected at a frequency of 1 hour, with a data acquisition planning period. For a period of 3 months, the original time series signals of the weights of each environmental factor within the planned collection period are obtained, denoted as . Among them, the collection frequency and collection planning period are empirical values ​​that can be adjusted by the implementer according to the specific implementation scenario.

[0033] Because the sensors are located in the garden for extended periods, potential equipment malfunctions or packet loss due to signal obstruction may occur, affecting the acquisition of raw time-series signals. There is often a lack of spatiotemporal dimensions, and direct use of such data can lead to data gaps.

[0034] Therefore, unbiased optimal estimation is performed on the missing points to achieve spatiotemporal interpolation completion of the signal. The unbiased optimal estimation of the missing points satisfies the expression:

[0035]

[0036] In the formula, Indicates the missing position and missing moments Estimated values ​​of environmental monitoring signals at the location; Indicates a known location and time The effective monitoring signal value at the location; This represents the total number of known valid sample points used for interpolation. It can be set according to the requirements of interpolation accuracy and computational efficiency, and is usually set to 12. It can be adjusted by the implementer according to the specific implementation scenario. Indicates the first The weighting coefficients for each known sample point are obtained by solving the Kriging equations using the Lagrange multiplier method from the product-sum spatiotemporal covariance model, and must satisfy the unbiased condition. And the estimated variance is minimized.

[0037] The original time series signal is calculated using an unbiased optimal estimation expression. The missing values ​​are filled in to obtain the completed time series signal.

[0038] S2: Construct a global ecological neighborhood benchmark to standardize the monitoring signals, calculate the resource pressure index that characterizes the degree of stress that actual environmental conditions pose to vegetation survival, and introduce this index as a nonlinear gain environmental factor weight into a dynamic competition model to simulate survival competition among vegetation.

[0039] Local monitoring data are typically expressed as absolute values ​​with specific physical units, while general vegetation demand data are often based on a broad range of climate zones. There is a significant scale mismatch between the two, and the absolute readings of a single environmental factor weight cannot directly reflect its relative superiority or inferiority in the current ecosystem, resulting in unclear physical meaning. To solve this problem, this specific implementation method first constructs a global reference environment database based on ecologically similar neighborhoods of the target area, rather than using generalized broad climate zone data. Based on this, a standardized processing system is established to ensure that monitoring signals and plant preferences are mathematically comparable in the same dimensionless space. The global reference benchmark constructed above and the generated standardized environmental vector set together constitute a digital environmental model.

[0040] Specifically, the radius of the target area is obtained from meteorological and soil databases. (Usually within 50km) and with altitude differences less than (Usually taken as 100m) All long-term stations in the past Historical data (e.g., 30 years); calculate the weight of each environmental factor. Reference mean and reference standard deviation A reference benchmark covering the potential ecological amplitude of the target area is constructed; these two global statistics are used to complete the interpolated real-time monitoring signal. Z-score normalization is performed to obtain the normalized environment vector. .

[0041] Standardized environment vector The calculation follows a standardized processing principle, which maps the originally heterogeneous environmental data to a relative niche space with universal physical meaning. The calculation satisfies the expression:

[0042]

[0043] In the formula, Represents a grid At any moment Regarding the weight of environmental factors Standardized environmental values; This represents the reconstructed signal value after spatiotemporal interpolation. and These are the historical reference mean and reference standard deviation of the environmental factor weights, respectively. Through this transformation, the environmental factor weights of different dimensions are uniformly converted into a dimensionless index that measures their deviation from historical ecological stability.

[0044] Based on the standardized environmental vectors described above, the degree of matching between specific vegetation species and the current environment can be further evaluated. (Calculation of vegetation...) In the grid At the moment Real-time adaptability The process is as follows:

[0045] First, the weights of each environmental factor are determined using the analytic hierarchy process (AHP). Then, the standardized environment vector is calculated. with vegetation Standardized preference center value Weighted Euclidean distance between :

[0046] in, Environmental factors The weights satisfy .

[0047] Subsequently, the weighted Euclidean distance was... Convert to real-time adaptability This transformation aims to map distance values, which characterize the degree of environmental deviation, to a positive index characterizing the degree of fit, with a value range located at... Within the interval, the calculation formula is:

[0048] This function ensures that the environment perfectly matches the vegetation preference. When ), the real-time adaptability is As the degree of environmental deviation increases, the adaptability decreases from... Towards Monotonically decreasing, which aligns with ecological intuition. This will serve as a key input parameter for subsequent configuration optimization and closed-loop control.

[0049] After standardization, to address the technical challenge of existing competition models' inability to quantify the catalytic effect of environmental severity on biological competition intensity, this embodiment introduces a resource pressure index to characterize the degree of stress posed by actual environmental conditions to vegetation survival; the resource pressure index is defined as follows: This is the arithmetic mean of the absolute deviations of the weights of all environmental factors. A larger value indicates that the current environment deviates further from its historically suitable average, the environmental stress is more severe, and the competition among vegetation for limited resources will be increasingly fierce; Resource Pressure Index The calculation formula is:

[0050]

[0051] In the formula, This represents the total number of sets of environmental factor weights involved in the assessment; It represents the absolute value of the standardized environmental vector; by aggregating the deviation of the weights of multidimensional environmental factors, this index dynamically quantifies the “environmental severity”, providing a quantitative basis for introducing nonlinear gain environmental factor weights into the competition model, and solving the logical defect that environmental resistance and competition resistance are independent in the traditional model.

[0052] Furthermore, a dynamic competition model incorporating a resource pressure index is constructed to address the problem of distorted survival rate predictions under extreme climates caused by neglecting environmental stress gains in traditional models; firstly, grid calculations are performed. and adjacent grids The cosine similarity of the vegetation preference vectors yields the niche overlap. The dynamic competition intensity calculation formula is optimized as follows:

[0053]

[0054] In the formula, Indicates time Adjacent grids For the grid The intensity of the applied dynamic competition; For environmental factor weights based on the nonlinear gain of environmental stress according to the resource pressure index, when As the value increases, the intensity of dynamic competition is mathematically amplified; The Euclidean distance is the center of the grid. This is the distance attenuation coefficient.

[0055] For the distance decay coefficient in the dynamic competition model To address the lack of biophysical basis for this approach, this embodiment quantifies vegetation canopy parameters to dynamically correlate the competition mechanism; and clearly defines the canopy radius. The system prioritizes reading from the local plant attribute database; if the database is missing, it employs the allometric growth equation. To make an estimate, among which The parameter is the expected diameter at ground level or the diameter at ground level of seedlings as specified in the candidate vegetation scheme. The values ​​are based on empirical parameters for species of the same genus in the "Handbook of Allometric Growth Equations for Major Tree Species," such as... In this context, the "Handbook of Allometric Growth Equations for Major Tree Species" is a reference to background technology or common knowledge, not referring to a specific, publicly published book. Rather, it represents a set of empirical parameters or technical data generally known and used by those skilled in the art in their research and practice; at this point, the attenuation coefficient is set as... The competition effect decreases to 10% when the spacing equals the minimum crown width, because ,Right now .

[0056] The dynamic competition intensity calculated above is not only used for theoretical simulation, but also directly serves as the weight of key environmental constraints in vegetation configuration optimization. This intensity value is embedded in the scoring function of each ecological strategy control logic to quantify the competitive inhibition effect between adjacent vegetation, thereby penalizing species combinations with high competition risk in the objective function. In the subsequent genetic algorithm optimization process, the system automatically avoids configuration schemes with severe niche overlap or excessive spatial canopy overlap, ensuring that the recommended results meet environmental adaptability requirements while also possessing community stability and construction feasibility. Simultaneously, this competition intensity also provides a basis for generating initial maintenance measures (such as plant spacing adjustment and buffer species configuration) and is used in the closed-loop feedback stage to diagnose model biases and correct competition parameters.

[0057] S3: Pre-set multiple ecological strategy control logics for different pressure gradients, establish a Gaussian probability mapping between the resource pressure index and the applicability of each strategy, and generate posterior probability weights under the current environmental state.

[0058] Environmental stress in ecosystems often exhibits non-stationary temporal fluctuations, and a single, fixed ecological strategy cannot simultaneously adapt to the full spectrum of changes from favorable habitats to extremely stressed habitats. If the traditional threshold switching method is used to make hard jumps between different strategies, it is very easy to cause violent oscillations in the decision-making system when the environmental stress is in the critical transition zone, resulting in a lack of stability and robustness in the final vegetation configuration scheme. Therefore, it is key to constructing a soft trade-off mechanism based on probability theory, which dynamically allocates the posterior probability weights of multiple strategies according to the environmental state, to achieve adaptive and robust decision-making.

[0059] Specifically, this embodiment pre-configures... An ecological strategy control logic targeting different environmental pressure gradients is denoted as... Among them, robust survival strategy control logic Suitable for high-stress environments (Ψ≥2.5), its objective function focuses on the real-time physiological adaptability and marginal tolerance of plants; diversity symbiosis strategy control logic. Suitable for low-stress environments (Ψ<0.8), it focuses on maximizing community species richness and niche diversity; equilibrium restoration strategy control logic. Suitable for transitional environments with moderate pressure (0.8≤Ψ<2.5), focusing on balancing construction cost and coverage speed; each strategy control logic has independent optimization weight parameters and constraints to correspond to its specific ecosystem service objectives.

[0060] Furthermore, to establish a nonlinear mapping relationship between the resource pressure index and the applicability of each strategy, this invention introduces a Gaussian kernel function to construct a probabilistic model. It is assumed that each ecological strategy performs optimally only within a specific pressure range, which follows a normal distribution. The Gaussian function is used to calculate the membership degree of the current environmental pressure to the "dominance domain" of each strategy, and this membership degree is transformed into a normalized posterior probability weight. This Gaussian probability mapping process realizes the mathematical transformation from the deterministic resource pressure index to the posterior probability weight of each strategy, thereby quantifying the uncertainty in decision-making.

[0061] Specifically, based on the resource pressure index calculated from the current grid, the posterior probability weights of each ecological strategy control logic are generated, satisfying the expression:

[0062]

[0063] In the formula, Indicates the first The posterior probability weights of each ecological strategy control logic in the current environmental state; The resource pressure index is calculated in the previous steps; Represents an exponential function with the natural constant as its base; Indicates the first The optimal typical stress mean for each strategy, based on experience, can be set to 3.0, 0.5 and 1.5 for survival, symbiosis and recovery strategies, respectively, representing the most suitable stress center position for each strategy.

[0064] Following the previous formula, Indicates the first The ecological tolerance parameter of each strategy is used to control the bell width of the Gaussian curve. It is usually set between 0.5 and 1.0. The larger the value, the stronger the tolerance of the strategy to environmental fluctuations. In this embodiment, it is uniformly set to 0.8. This represents the total number of preset ecological strategy control logics, in this embodiment... The denominator is the normalized environmental factor weights, ensuring that the sum of the posterior probability weights of all strategies is 1, thus forming a complete probability distribution space.

[0065] The posterior probability weights obtained through the above calculations It is not a static value, but rather an index that changes with real-time monitoring of resource pressure. The system dynamically slides according to changes; when environmental pressure gradually increases, the system automatically reduces the weight of the diversity symbiosis strategy and smoothly increases the weight of the robust survival strategy, and vice versa; this mechanism simulates the "hedging" evolutionary strategy of organisms in nature under environmental stress, ensuring that the final generated configuration instructions can still maximize the expected mathematical benefits in fuzzy transitional environments, and avoiding the ecological risks brought about by extreme single coordinates.

[0066] S4: Use posterior probability weights to perform probability weighted integration of the outputs of each strategy control logic to obtain vegetation configuration instructions that guide planting construction and include targeted initial maintenance measures.

[0067] When faced with highly uncertain and dynamically changing garden ecological environments, relying solely on a single ecological strategy model to generate configuration schemes often leads to overfitting, where the decision results perform well in specific environmental scenarios but fail completely in other scenarios. Furthermore, traditional deterministic optimization methods struggle to effectively integrate diverse evaluation perspectives of different strategies for the same environmental state, resulting in a lack of resilience against environmental disturbances in the final scheme. Therefore, constructing an expected integrated objective function that can integrate the advantages of multiple strategies is crucial for achieving robust vegetation configuration.

[0068] Specifically, this implementation utilizes the posterior probability weights generated in the preceding steps to linearly weight and assemble the independent scoring functions of each ecological strategy control logic, constructing the final expected integrated objective function. Mathematically, this function represents the maximization of the expected ecological benefits that the vegetation configuration scheme can obtain under the current environmental probability distribution. The expected integrated objective function... Satisfying the expression:

[0069]

[0070] In the formula, Indicates vegetation configuration scheme Expected integration fitness score; The vegetation configuration vector to be solved contains information on vegetation type and planting density at each grid point; This represents the total number of preset ecological strategy control logics; in this embodiment, it is set to 3. For the first The posterior probability weights of each strategy.

[0071] For the first Each ecological strategy control logic is aimed at the solution The calculated independent ratings are as follows:

[0072] In the formula, This is the normalized real-time adaptability utility term, whose value is the solution. Real-time adaptability of all grid vegetation The sum of Real-time adaptability The calculation method is shown in step S2.

[0073] The normalized competition cost penalty term is represented by the value of the scheme. All adjacent grid pairs (defined as spatial distance) Dynamic competition intensity between mesh pairs The sum multiplied by the penalty weight, i.e. ,in For the first The competitive inhibition penalty weight of each strategy Representation scheme The set of all adjacent grid pairs in the set. For grid pairs The intensity of dynamic competition between them.

[0074] These are additional constraint terms after normalization, the specific form of which depends on the strategy. And determined:

[0075] For robust survival strategies ( ): ,in For the plan The number of grid cells located at the edge of the planting area. This is the edge penalty coefficient, with a value of 0.05, which can be adjusted by the implementer according to the specific implementation scenario.

[0076] For diversity symbiosis strategies ( ): ,in For the plan Shannon diversity index, For species The proportion of the total crown area This represents the total number of species.

[0077] For the equilibrium recovery strategy ( ): ,in For the plan The estimated total cost of planting and initial maintenance, This is the cost penalty coefficient, with a value of 0.001, which can be adjusted by the implementer according to the specific implementation scenario.

[0078] For strategy The preset weights of each item satisfy the following conditions. It was determined using the Hierarchical Analysis of Experts (AHP).

[0079] The normalization process employs the range method, applying it to all candidate solutions in the current iteration. of , and The calculated value is mapped to the interval [0,1].

[0080] Wherein, the dynamic competition intensity The calculation formula is obtained from step S2. In the formula Niche overlap. As a resource pressure index, This is the distance attenuation coefficient. is the Euclidean distance from the grid center.

[0081] and The specific calculation method is as follows:

[0082] Economic Costs Calculation: A cost parameter database is pre-built, including the unit seedling cost for each plant variety. Construction costs per unit area for planting And the monthly cost per unit area for various initial maintenance measures. For a given scheme Its total cost is calculated using the following formula:

[0083]

[0084] in, This represents the index for traversing all grid cells. For grid Plant varieties planted For planting density, For grid area, To implement maintenance measures in accordance with the plan, For the maintenance cycle (e.g., 3 months); After normalization, it is used as a negative penalty term.

[0085] Shannon Diversity Index Calculation: First, statistical scheme The set of all plant species appearing in the text is denoted as set. For each species Calculate its relative abundance in the entire scheme. :

[0086]

[0087] in, Represents a grid species The planting density (0 if no plants are planted). For species The crown radius. Then the Shannon diversity index is:

[0088]

[0089] When the regulations are in effect When the time is right, the corresponding item takes the value of 0. After normalization, it is used as a positive incentive term.

[0090] Furthermore, a genetic algorithm is used to globally optimize the desired ensemble objective function; the vegetation species to be configured are encoded into chromosome gene sequences, in order to... As an individual's fitness function, the algorithm iteratively searches the solution space through selection, crossover, and mutation operations. During the iteration process, the algorithm automatically tends to retain "all-rounder" gene combinations that score highly under strategies with a high probability of applicability and can maintain a basic survival baseline under strategies with a low probability of applicability, thereby outputting the optimal vegetation configuration instruction. The instruction specifies the specific vegetation species and spatial layout to be planted in each grid unit within the target area.

[0091] Furthermore, to ensure the survival rate of vegetation in the early stages of transplantation, this invention also generates targeted initial maintenance measures based on the distribution characteristics of posterior probability weights; if the calculation results show the weight of high-stress survival strategies... If a dominant strategy is employed, the system automatically adds a stress-resistance maintenance tag for high-frequency micro-sprinkler irrigation and shade netting to the configuration instructions; if a low-pressure symbiotic strategy is dominant, a low-intervention tag for routine inspections and pruning is added; simultaneously, for dynamic competition intensity exceeding a preset threshold (e.g., For grid pairs, instructions such as "increase plant spacing" or "set up buffer vegetation strips" are added to the maintenance measures of the corresponding grids to alleviate competitive pressure. The resulting vegetation configuration instructions integrate maintenance intensity definitions that match the current level of environmental stress, forming an engineering output that can directly guide construction.

[0092] Implement initial maintenance measures and monitor growth feedback signals after the instructions are implemented. When deviations exceed limits, activate physical regulation mechanisms, including water and fertilizer regulation or microenvironment improvement, and adaptively correct model parameters based on the regulated feedback signals.

[0093] After implementing the vegetation configuration instructions that include targeted maintenance measures, the process enters the closed-loop control phase. This phase involves continuous monitoring of vegetation growth feedback signals, with particular attention to the consistency between predicted dynamic competition intensity and actual growth performance, providing a basis for subsequent model adjustments.

[0094] Traditional landscape vegetation recommendation systems often exhibit an "open-loop" characteristic, meaning that once a plan is output and implemented, the system becomes disconnected from the subsequent vegetation growth. However, the actual ecological environment is highly dynamic and time-varying, and plant organisms exhibit individual differences in their responses to the environment. Static prediction models struggle to accurately address real-time environmental fluctuations and growth deviations. Without a feedback-based correction mechanism, model prediction errors accumulate over time, ultimately leading to the failure of maintenance decisions and the degradation of ecological functions. Therefore, establishing a closed-loop system that couples physical regulation with model correction is crucial for maintaining long-term ecological stability.

[0095] Specifically, after implementing initial maintenance measures, a sensor network is used to continuously monitor real-time environmental signals and vegetation growth status feedback signals in the target area; a dual response triggering logic based on real-time adaptability and environmental deviation is constructed, firstly calculating the grid... Inner vegetation At the present moment Real-time adaptability When this value falls below a preset safety threshold of 0.5, it is determined that the current microhabitat threatens vegetation survival, and the overall diagnostic procedure is immediately initiated; then, the absolute deviation of the weighted monitoring values ​​of each environmental factor relative to the standardized preference range of the vegetation is calculated. This allows for the quantification of the specific sources of environmental stress.

[0096] Furthermore, to avoid resource waste or over-regulation due to a "one-size-fits-all" approach to maintenance, this embodiment sets a differentiated threshold that is dynamically linked to the vegetation ecological width. For broadly adaptable plants with wide ecological amplitudes, the dynamic threshold is set to 0.25 times the width of the preference interval, allowing them to self-regulate within a certain range; for sensitive plants with narrow ecological amplitudes, the dynamic threshold is strictly set to 0.05 times; when the absolute deviation of the weight of a certain environmental factor... Exceeding the corresponding dynamic threshold At that time, the system generates precise physical adjustment instructions for the weight of the environmental factor, driving irrigation, fertilization or supplemental lighting equipment to perform water and fertilizer regulation or micro-habitat improvement, bringing the environmental parameters back to the appropriate range.

[0097] While implementing the physical adjustment mechanism, the system enters an adaptive learning phase, aiming to correct prior knowledge using measured data; the system automatically records the changes in environmental factor weights before and after each physical adjustment. and the resulting changes in vegetation health Vegetation health is quantified by the difference in the Normalized Difference Vegetation Index (NDVI); when the accumulated effective sample size... When there are more than 30 groups, the data are considered statistically significant. At this point, the ridge regression algorithm is used to fit a linear response model of the impact of environmental changes on health. The specific regression model expression is as follows:

[0098]

[0099] In the formula, Indicates the first Changes in vegetation health within each grid; This is the intercept term, representing the basic growth rate excluding environmental factors; For the first The regression coefficient of each environmental factor weight represents the actual contribution rate of that environmental factor weight to vegetation health. For the first Within the first grid The changes in the weights of each environmental factor were analyzed; a 10-fold cross-validation method was used to determine the regularization parameter to eliminate the interference of multicollinearity on the regression results, thereby extracting the factors that passed the significance test. The key influencing environmental factors weight coefficients.

[0100] Subsequently, the Bayesian inference principle is used to adaptively adjust the model parameters by fusing the regression coefficients extracted from the measured data with the expert experience weights initially set in the model; firstly, significant regression coefficients are... Normalization is performed to obtain data-driven weights. Then, the corrected new weights are calculated using the Bayesian update formula. This allows the model weights to evolve dynamically as data accumulates; the Bayesian update formula satisfies the expression:

[0101]

[0102] In the formula, This represents the updated environmental factor weights. This represents the initial weights before the update (i.e., the expert experience weights determined based on the AHP analytic hierarchy process). This indicates the number of measured samples involved in the calculation, reflecting the strength of the evidence in the current data; This represents the weights driven by normalized data obtained from ridge regression analysis; This is a priori strength parameter used to control the degree to which the model depends on initial expert experience; in this embodiment, it is set as follows: This setting means that as the sample size increases... With the addition of data, the model will smoothly transition from "experience-driven" to "data-driven," continuously approaching the optimal response logic in real-world environments.

[0103] The system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the garden vegetation recommendation method based on ecological environment data according to the first aspect of the present invention.

[0104] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0105] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for recommending garden vegetation based on ecological environment data, characterized in that, include: By using sensor networks to acquire monitoring signals of the target area environment, and performing spatial reconstruction and spatiotemporal interpolation of the signals based on geostatistical variation patterns, a digital environmental model reflecting environmental heterogeneity is constructed. A global ecological neighborhood benchmark is constructed to standardize the monitoring signals. A resource pressure index, which characterizes the degree of stress posed by actual environmental conditions to vegetation survival, is calculated. This index is then used as a nonlinear gain environmental factor weight to introduce into a dynamic competition model for simulating survival competition among vegetation. The dynamic competition model outputs the dynamic competition intensity. In the construction of the dynamic competition model: the vegetation canopy radius is estimated using the allometric growth equation and the distance attenuation coefficient is calculated; the ecological niche overlap, the environmental stress nonlinear gain environmental factor weight including the resource pressure index, and the spatial attenuation term based on the distance attenuation coefficient are multiplied together to obtain the dynamic competition intensity. Pre-defined ecological strategy control logic for different pressure gradients, including: Establish robust survival strategy control logic that focuses on real-time adaptability and includes edge effect penalties for high-stress environments; Establish a diversity symbiosis strategy control logic that focuses on competition and coexistence and includes diversity index constraints for low-stress environments; Establish a balanced recovery strategy control logic that emphasizes economy and rapid coverage for use in transitional environments; each strategy control logic is configured with independent objective function weights; Establish a Gaussian probability mapping between the resource pressure index and the applicability of each strategy to generate posterior probability weights under the current environmental state; By using the posterior probability weights to perform a probabilistic weighted integration of the outputs of each ecological strategy control logic, a vegetation configuration instruction is obtained to guide planting operations and includes targeted initial maintenance measures, including: The independent scoring functions of each ecological strategy control logic are multiplied by their corresponding posterior probability weights and then summed to construct the final expected integrated objective function. A genetic algorithm is used to globally optimize the desired integrated objective function, and the optimal vegetation configuration and maintenance instructions under the current environmental probability distribution are output. Implement the initial maintenance measures and monitor the growth feedback signal after the instructions are implemented. When the deviation exceeds the limit, activate physical regulation mechanisms, including water and fertilizer regulation or microenvironment improvement, and adaptively correct the model parameters based on the regulated feedback signal.

2. The method for recommending garden vegetation based on ecological environment data according to claim 1, characterized in that, The spatial reconstruction and spatiotemporal interpolation of signals based on geostatistical variation patterns include: Random pre-sampling was performed in the target area to calculate the experimental semivariogram of soil environmental data and extract the minimum range. Based on the spatial sampling theorem, the side length of the square grid is set to half of the minimum range, and sensors are deployed at the center of the grid to collect time series data. For missing data, a product sum spatiotemporal covariance model is constructed, and unbiased optimal estimation is performed by solving the Kriging linear equation system using known spatiotemporal location data to complete the missing data.

3. The method for recommending garden vegetation based on ecological environment data according to claim 1, characterized in that, Standardization and resource stress index calculation, including: Obtain environmental data within a specific radius and historical period around the target area, and calculate the reference mean and reference standard deviation of the environmental factor weights corresponding to each environmental monitoring signal to construct a global reference benchmark; The original monitoring signal is subtracted from the reference mean and then divided by the reference standard deviation to obtain the standardized environmental vector; the suitable growth range of vegetation is standardized using the same reference mean and reference standard deviation to obtain the standardized preference center value. The resource stress index is defined as the arithmetic mean of the absolute deviations of the weights of all environmental factors relative to the global reference baseline.

4. The method for recommending garden vegetation based on ecological environment data according to claim 3, characterized in that, Before introducing nonlinear gain environmental factor weights into the dynamic competition model used to simulate survival competition among vegetation, a step is also included to calculate the real-time fitness: The weights are determined using the analytic hierarchy process (AHP) and the weighted Euclidean distance between the standardized environment vector and the preference center value is calculated, which is then converted into real-time fit.

5. The method for recommending garden vegetation based on ecological environment data according to claim 1, characterized in that, The process of establishing a Gaussian probability mapping between the resource pressure index and the applicability of each strategy to generate posterior probability weights under the current environmental state includes: Set the typical average pressure and ecological tolerance parameters corresponding to each ecological strategy control logic; A mapping relationship is constructed using the Gaussian kernel function. The square of the difference between the current resource pressure index and the typical pressure mean of each ecological strategy control logic is calculated. The index decay term is calculated based on the ratio of the square of the difference between the current resource pressure index and the typical pressure mean of each ecological strategy control logic to the ecological tolerance parameter. The exponential decay term is normalized in all ecological strategy control logics to obtain the posterior probability weights of each ecological strategy control logic under the current environmental state.

6. The method for recommending garden vegetation based on ecological environment data according to claim 1, characterized in that, The activation of physical regulation mechanisms, including water and fertilizer regulation or microenvironment improvement, includes: When the real-time vegetation adaptation rate within the grid is lower than the preset safety threshold, an overall diagnosis is triggered. Alternatively, the absolute deviation of the monitoring values ​​of each environmental factor weight relative to the vegetation preference range can be calculated. When the absolute deviation exceeds the dynamic threshold determined by the vegetation ecological amplitude, precise water and fertilizer regulation or habitat improvement instructions can be generated for specific environmental factor weights.

7. The method for recommending garden vegetation based on ecological environment data according to claim 1, characterized in that, The adaptive correction of model parameters based on the adjusted feedback signal includes: We collected the environmental changes and vegetation health changes after physical regulation, and obtained the environmental factor weighting coefficients by ridge regression fitting. Using Bayesian principles, combined with prior strength and sample size, the old environmental factor weights are updated by weighting them using the environmental factor weight influence coefficient, resulting in new environmental factor weights.

8. A garden vegetation recommendation system based on ecological environment data, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the garden vegetation recommendation method based on ecological environment data according to any one of claims 1-7.

Images