Mountainous farmland mechanization ecological risk judgment method, system and electronic device

Abstract

The application provides a mountainous farmland mechanization ecological risk judgment method and system and electronic equipment, and relates to the technical field of agriculture, which comprises the following steps: generating risk index data based on the constructed multi-dimensional index system, calculating the ecological security level and the weight of each index by using the trained radial basis function neural network; further dynamically integrating the weight into the PLUS model to correct the risk weighting of the conversion resistance and the expansion probability, and generating dynamic control parameters; simulating the land use pattern under multiple scenarios based on the parameter-driven model, and feeding back the results to the RBF network for reevaluation until the convergence condition is met, realizing the iterative optimization of the control parameters, and finally generating a comprehensive evaluation result based on the optimal pattern and the ecological level; thereby solving the technical problems of the existing mountainous farmland mechanization ecological risk evaluation, such as insufficient comprehensive evaluation and insufficient accuracy, and improving the comprehensiveness and accuracy of the mountainous farmland mechanization ecological risk judgment.

Description

Technical Field

[0001] This invention relates to the field of agricultural technology, and more specifically, to a method, system, and electronic device for assessing the ecological risks of mechanization in mountainous farmland. Background Technology

[0002] With the intensification of global climate change and the increase in human activity, farmland ecosystems face increasingly severe ecological risks and challenges. Promoting the mechanization of farmland has become an important path to improve agricultural productivity and ensure food security. However, how to scientifically assess and effectively manage the impact of agricultural mechanization on the ecological environment has become a key issue in the field of sustainable agricultural development. Existing research generally recognizes that the mechanization of farmland not only involves land consolidation and optimization of the suitability of mechanized operations, but also profoundly affects regional ecological patterns, biodiversity maintenance, and the stable operation of ecosystem services.

[0003] In the development of ecological risk assessment systems, mainstream methods largely rely on the intrinsic characteristics of arable land quality, supplemented by natural elements such as vegetation cover and topographic slope as auxiliary discrimination parameters, forming a relatively fixed assessment process. While this framework reflects local ecological conditions to some extent, it often overlooks the structural impacts of farmland spatial pattern evolution, such as field fragmentation and decreased landscape connectivity. It also fails to fully integrate factors such as location conditions, transportation accessibility, and the interconnectedness of surrounding ecological networks, resulting in insufficient characterization of response mechanisms to complex ecological processes. In other words, current technologies for ecological risk assessment of mechanization adaptation in mountainous farmland suffer from incomplete assessment dimensions and insufficient accuracy. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, and electronic device for identifying ecological risks of mechanization in mountainous farmland, in order to solve the technical problems of insufficient comprehensiveness and accuracy in the assessment dimensions of existing technologies.

[0005] In a first aspect, embodiments of the present invention provide a method for identifying ecological risks associated with the mechanization of mountain farmland. This method includes: generating ecological security level values ​​and corresponding weights for each risk indicator based on multidimensional risk indicator data output from a mountain farmland mechanization ecological risk indicator system using a trained radial basis function neural network; performing risk-weighted correction on the land conversion resistance coefficient and land use expansion probability in the PLUS model according to preset ecological risk response rules, combined with the aforementioned ecological security level values ​​and the weights of each risk indicator, to generate dynamic control parameters for regulating model operation; running the PLUS model based on the aforementioned dynamic control parameters to generate target land use patterns under multiple scenarios; inputting the aforementioned target land use patterns into the aforementioned radial basis function neural network to recalculate the corresponding ecological security level values, and iteratively updating the aforementioned dynamic control parameters based on the calculation results until preset ecological risk control convergence conditions are met; and generating a comprehensive evaluation result of the ecological risks associated with the mechanization of mountain farmland based on the finally determined target land use patterns and their corresponding ecological security level values.

[0006] In some optional implementations, the above method also includes: using UAVs to synchronously acquire initial mapping data of mountain farmland; the initial mapping data includes high-resolution remote sensing images, hyperspectral biochemical parameters, and digital elevation data; correcting, denoising, and formatting the initial mapping data to generate mountain farmland data; and constructing a mechanization-friendly ecological risk indicator system for mountain farmland based on the mountain farmland data.

[0007] In some optional implementations, the methods for constructing an ecological risk indicator system for the mechanization of mountain farmland based on the above-mentioned mountain farmland data include: extracting multiple geographical elements related to ecosystem stability from the above-mentioned mountain farmland data based on a multi-indicator synergistic PSR system, and screening key geographical elements that have potential impacts on the mechanization transformation of mountain farmland; standardizing the screened key geographical elements and establishing multi-dimensional risk indicator data.

[0008] In some optional implementations, the above method also includes: determining the top 20% of regions corresponding to high-sensitivity risk areas based on the weight ranking of the risk indicators output by the above radial basis function neural network; for the above-mentioned high-sensitivity risk areas, adopting a higher simulation spatial resolution than that of conventional areas when running the above-mentioned PLUS model, and introducing micro-topographic constraint rules and neighborhood expansion effect enhancement mechanisms during the land expansion process.

[0009] In some optional implementations, the above micro-topography constraint rules include: proportionally decreasing the probability of conversion of mechanized farmland to types other than woodland with increasing slope; the above neighborhood expansion effect enhancement mechanism includes: enhancing the neighborhood clustering tendency of corresponding land use types in ecologically sensitive areas.

[0010] In some optional implementations, the method for generating the base value of the land use expansion probability before performing risk-weighted correction on the above land use expansion probability includes: obtaining multiple spatial driving factors affecting the land use change trend based on key geographic elements in the above multidimensional risk indicator data; training an expansion probability prediction model through machine learning algorithms based on the mapping relationship between historical land use change samples and the above spatial driving factors; and using the above expansion probability prediction model to infer the entire grid cell to generate an initial land use expansion probability distribution map as the base value of the above land use expansion probability.

[0011] In some optional implementations, the PLUS model is run based on the above dynamic control parameters to generate target land use patterns under multiple scenarios, including: loading the above dynamic control parameters into the above PLUS model, and driving the above PLUS model to perform land use pattern evolution simulations under three scenarios: natural development, farmland protection, and ecological priority, based on the above parameters, to generate corresponding target land use patterns.

[0012] Secondly, embodiments of the present invention provide a system for identifying ecological risks in the mechanization of mountain farmland. This system includes: an ecological security assessment unit, used to generate ecological security level values ​​and corresponding weights for each risk indicator based on multi-dimensional risk indicator data output from a system of ecological risk indicators for the mechanization of mountain farmland, using a trained radial basis function neural network; a dynamic parameter control unit, used to perform risk-weighted correction on the land conversion resistance coefficient and land use expansion probability in the PLUS model according to preset ecological risk response rules, combined with the aforementioned ecological security level values ​​and the weights of each risk indicator, generating dynamic control parameters for controlling the model's operation; a land pattern simulation unit, used to run the PLUS model based on the aforementioned dynamic control parameters, generating target land use patterns under multiple scenarios; a parameter iteration update unit, used to input the aforementioned target land use patterns into the aforementioned radial basis function neural network, recalculate the corresponding ecological security level values, and iteratively update the aforementioned dynamic control parameters based on the calculation results until preset ecological risk control convergence conditions are met; and a comprehensive evaluation generation unit, used to generate a comprehensive evaluation result for the ecological risks of the mechanization of mountain farmland based on the finally determined target land use patterns and their corresponding ecological security level values.

[0013] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the steps of the method described in any of the first aspects above.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to perform the method described in any of the first aspects above.

[0015] This invention provides a method, system, and electronic device for assessing the ecological risk of mechanization adaptation in mountainous farmland. The method includes: calculating the ecological security level and the weights of each indicator based on risk indicator data generated from a constructed multi-dimensional indicator system using a trained radial basis function neural network; further, dynamically integrating the weights into a PLUS model to perform risk-weighted correction on conversion resistance and expansion probability, generating dynamic control parameters; simulating land use patterns under multiple scenarios based on these parameters, and feeding the results back to the RBF network for re-evaluation until convergence conditions are met, achieving iterative optimization of the control parameters; and finally generating a comprehensive evaluation result based on the optimal pattern and ecological level. This solves the technical problems of insufficient comprehensiveness and accuracy in existing ecological risk assessments for mechanization adaptation in mountainous farmland, improving the comprehensiveness and accuracy of ecological risk assessment for mechanization adaptation in mountainous farmland. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating an ecological risk assessment method for mechanization of mountain farmland provided in an embodiment of the present invention;

[0018] Figure 2 A flowchart illustrating another method for identifying ecological risks of mechanization in mountainous farmland provided in an embodiment of the present invention;

[0019] Figure 3 This is a schematic diagram of the structure of an ecological risk assessment system for mechanization of mountain farmland provided in an embodiment of the present invention;

[0020] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Current technologies still have significant shortcomings in identifying ecological risks associated with farmland mechanization. These shortcomings primarily stem from a lack of systematic approach in the assessment process. Data collection, indicator construction, and model prediction are often treated in isolation, lacking an integrated technical path from multi-source data fusion to dynamic risk assessment and scenario simulation. This makes it difficult to track the entire process of ecological changes in complex hilly and mountainous environments. Furthermore, existing methods have limited ability to depict spatial details, generally relying on macro-statistical scales and low-resolution data. They fail to effectively reflect key spatial characteristics affecting mechanization, such as field morphology, micro-topographical undulations, and local slope changes, resulting in spatially coarse and homogeneous risk identification that cannot accurately pinpoint highly sensitive areas. In addition, current mainstream land use simulation models mostly employ fixed transformation rules and static parameter settings, lacking the ability to respond and adjust in real time according to regional ecological conditions. The simulation process lacks flexibility and adaptability, making it difficult to provide differentiated and dynamic support for land use decisions under different ecological protection objectives.

[0023] Based on this, the present invention provides a method, system and electronic device for identifying ecological risks of mechanization in mountainous farmland, in order to solve at least one of the above-mentioned technical problems existing in the prior art.

[0024] To facilitate understanding of this embodiment, a detailed description of the ecological risk assessment method for mechanization of mountain farmland disclosed in this embodiment of the invention will be provided first. (See [link to relevant documentation]). Figure 1 The diagram shows a flowchart of a method for assessing the ecological risks of mechanization in mountainous farmland. This method can be executed by electronic equipment and mainly includes the following steps S102 to S110:

[0025] Step S102: Based on the multidimensional risk index data output by the mountain farmland mechanization ecological risk index system, the trained radial basis function neural network is used to generate the ecological security level value of the mountain farmland mechanization ecosystem and the weights corresponding to each risk index.

[0026] In this embodiment, step S102 may include: taking the multidimensional risk index data output by the mountain farmland mechanization ecological risk index system as input, processing it through a trained radial basis function neural network, and outputting the ecological security level value of the mountain farmland mechanization ecosystem and the weight of each risk index.

[0027] In one embodiment, prior to step S102, the process may further include acquiring mountain farmland data to construct an ecological risk indicator system for the mechanization of mountain farmland. For example: first, initial mapping data of mountain farmland is acquired simultaneously using a drone, including high-resolution remote sensing images, hyperspectral biochemical parameters, and digital elevation data; then, the initial mapping data is corrected, denoised, and formatted to generate mountain farmland data; finally, an ecological risk indicator system for the mechanization of mountain farmland is constructed based on the mountain farmland data.

[0028] In step S102 above, the ecological risk indicator system for the mechanization of mountain farmland can be constructed based on a multi-indicator synergistic PSR system. This multi-indicator synergistic PSR system refers to a multi-indicator synergistic "stress-state-response" ecological risk discrimination indicator system. This PSR system can be constructed based on factors influencing the ecological risk of mechanization of mountain farmland selected from three aspects: stress indicators, state indicators, and response indicators. For example, by screening 16 multi-dimensional indicators such as soil thickness, land reclamation rate (stress indicator), average slope, remote sensing ecological index (state indicator), irrigation and drainage guarantee rate, and mechanical power per unit of cultivated land (response indicator), the system aims to achieve a specific analysis of the unique characteristics of farmland ecosystems, thereby overcoming the limitations of traditional evaluation systems, realizing a comprehensive quantification of the ecological risk of mechanization, and providing a scientific and systematic indicator framework for risk assessment.

[0029] In one embodiment, the method of constructing an ecological risk indicator system for the mechanization of mountain farmland based on mountain farmland data may include: extracting multiple geographical elements related to ecosystem stability from mountain farmland data based on a multi-indicator synergistic PSR system, and screening key geographical elements that have potential impacts on the mechanization transformation of mountain farmland; standardizing the screened key geographical elements and establishing multi-dimensional risk indicator data.

[0030] In this embodiment, preferably, the ecosystem state assessment paradigm adopted by the multi-indicator synergistic PSR system can be referenced to extract multiple geographic elements related to ecosystem stability (such as topography, soil, hydrology, and human activities) from mountain farmland data; then, based on the impact mechanism of mechanization transformation of mountain farmland on ecological security, key geographic elements with significant disturbance effects on farmland mechanization are screened out; finally, the screened key geographic elements are standardized and organized into multi-dimensional risk indicator data with unified spatial resolution and dimensions.

[0031] In step S102 above, the radial basis function neural network can be an RBF neural network built based on radial basis functions (RBF), which can be used to determine the ecological risk level of mechanization of farmland. The RBF neural network can be a three-layer feedforward network (input layer, hidden layer, output layer). Its core is to use radial basis functions as activation functions for the hidden layer neurons, and to map the input space to a high-dimensional feature space through a distance metric (such as Euclidean distance), thereby solving nonlinear problems.

[0032] After the aforementioned RBF neural network is trained, it can be further used to measure and classify new input data. Specifically, this can include: after determining the indicators used to assess the ecological risk level of mechanization of mountain farmland, preprocessing the relevant data for each indicator by handling missing values, feature scaling, and data splitting; using the selected indicators as input variables for the radial basis function; for each indicator, using the K-means clustering algorithm to determine the center and width of the radial basis function; and for each input variable, using the radial basis function model to calculate its output value, i.e., the corresponding ecological security level value and the weights corresponding to each risk indicator.

[0033] Furthermore, based on the ecological security level value and the weights of each risk indicator generated in step S102 above, the PLUS model can be used to simulate land use patterns under multiple scenarios and predict the changes in the ecological risk of mechanization of mountain farmland under the corresponding scenarios.

[0034] Step S104: Based on the preset ecological risk response rules, and combined with the ecological security level value and the weight of each risk indicator, the land conversion resistance coefficient and land use expansion probability in the PLUS model are risk-weighted and corrected to generate dynamic control parameters for regulating the operation of the model.

[0035] The PLUS model (Patch-generating Land Use Simulation model) is a data-driven spatial simulation model specifically designed to predict and simulate future land use / land cover (LULC) change patterns under various scenarios. Developed from the traditional CA-Markov model, this model is particularly suitable for simulating land use pattern evolution in complex terrain areas (such as hilly and mountainous regions).

[0036] In another embodiment, before performing risk-weighted correction on the land expansion probability in step S104 above, the method for generating the basic value of the land expansion probability may include: obtaining multiple spatial driving factors affecting the land use change trend based on key geographic elements in multidimensional risk indicator data; training an expansion probability prediction model through machine learning algorithms based on the mapping relationship between historical land use change samples and spatial driving factors; and using the expansion probability prediction model to infer the entire grid cell to generate an initial land expansion probability distribution map as the basic value of the land expansion probability.

[0037] For example, multiple spatial driving factors are acquired, including natural factors, socio-economic factors, and distance factors. Natural factors may include average slope, average elevation, and soil organic matter content; socio-economic factors may include fertilizer application intensity, mechanical power per unit of cultivated land, and investment in the primary industry; and distance factors may include distance to farm roads, distance to ditches, and distance to villages. Then, based on the sample points of mechanized farmland expansion extracted from 2015 to 2022 and their corresponding driving factor values, a random forest algorithm is used to train an expansion probability prediction model. This model is then used to infer the land use expansion probability distribution map of the entire grid cell, generating a land use expansion probability distribution map with a resolution of 30m. The expansion probability distribution map is used as part of the dynamic control parameters to regulate the operation of the PLUS model.

[0038] Furthermore, the basic values ​​for generating land conversion resistance coefficients can be generated by constructing an initial conversion resistance matrix based on historical land use change patterns and geographical constraints, and integrating multi-source spatial factors to achieve a spatialized expression of the resistance values.

[0039] Specifically, firstly, based on long-term land use evolution data for the region (such as remote sensing interpretation results from 2000 to 2022), the conversion frequency between different land types can be statistically analyzed, and initial resistance values ​​can be set in conjunction with expert experience. For example, in mountainous environments, the actual conversion from forest land to mechanized farmland is less frequent and ecologically restrictive, so a high resistance value (e.g., 90) is assigned; the conversion from unused land to arable land is easier to promote through remediation projects, so a lower resistance value (e.g., 40) is assigned.

[0040] Subsequently, spatial factors such as slope, ecological protection red line, and soil erosion intensity are introduced to perform geographical calibration of the initial resistance. For example, in areas with a slope >25° or located in water source protection areas, the resistance to the conversion of farmland expansion direction is automatically increased; while on low-slope, unused land adjacent to roads, the development resistance is appropriately reduced.

[0041] The basic conversion resistance layer formed by this process can be used as the initial parameter for running the PLUS model, reflecting the ease or difficulty of conversion between various land uses. In subsequent step S104, this basic value will be dynamically weighted and corrected in combination with the ecological risk weight output by the RBF neural network. That is, the lower the ecological security level of a region, the higher the resistance of the relevant high-risk conversion path will be, thereby achieving an upgrade from "static rules" to "risk-response dynamic regulation".

[0042] Based on the aforementioned basic values ​​of land expansion probability and land conversion resistance coefficient, weighted corrections or reconstructions can be made by introducing an ecological risk feedback mechanism to form dynamic control parameters with dynamic adjustment capabilities.

[0043] In other words, in this embodiment, the aforementioned dynamic control parameters can refer to a set of technical parameters generated by combining the ecological security level value output by the radial basis function neural network and the weights of each risk indicator with the basic parameters required for the operation of the PLUS model, and performing risk weighting correction, adaptive adjustment or priority reconfiguration according to the preset ecological risk response rules. This set of parameters is used to regulate the process of model operation and is used to achieve ecological sensitivity guidance and dynamic intervention in the evolution of land use patterns under multiple scenarios.

[0044] Among them, the aforementioned basic parameters refer to the initial input parameters on which the PLUS model depends for normal operation. These parameters may include the land conversion resistance coefficient and the land use expansion probability. The land conversion resistance coefficient can be used to characterize the ease or difficulty of conversion between different land types. For example, the resistance to converting forest land to arable land is usually higher than that to converting grassland land to arable land. The land use expansion probability can reflect the possibility that a certain grid cell will be converted into a specific land use type (such as mechanized farmland) in the future. It is usually determined by historical evolution trends and spatial driving factors (such as slope, distance from roads, etc.).

[0045] These fundamental parameters in traditional models are mostly statically set or based on empirical assignments, lacking the ability to respond to changes in the regional ecological state. In this embodiment, however, these fundamental parameters, as initial inputs, can be further transformed into dynamic control parameters. Preferably, the specific transformation method may include:

[0046] For the land conversion resistance coefficient, the weighted adjustment is performed according to the spatial distribution characteristics of high-weight risk indicators (such as "average slope" and "soil erosion modulus") and a preset response rule. For example, the resistance to the conversion of forest land to cultivated land is automatically increased in areas with low ecological security levels. For the land use expansion probability, its original distribution map (i.e., the base value) is spatially superimposed with the ecological sensitivity weight map output by the RBF neural network to generate a modified expansion probability distribution modulated by ecological risk.

[0047] The final set of dynamic control parameters not only includes the resistance coefficient and expansion probability after risk weighting correction, but also includes constraints that are temporarily activated in response to ecological safety thresholds (such as micro-topography constraint rules and neighborhood effect enhancement factors). These parameters are encapsulated and stored in structured data formats (such as JSON, XML, or database tables) for the PLUS model to read, load, and execute in each iteration.

[0048] In another embodiment, the aforementioned dynamic control parameters may further include a spatial constraint threshold that dynamically expands and contracts according to the scenario. This spatial constraint threshold can be dynamically adjusted based on the ecological security level value. For example, when the ecological security level value is lower than the first critical value, the ecological protection red line range is automatically expanded by 10%–20%, and land use conversion in high-risk areas is prohibited; when the ecological security level value is higher than the second critical value, the land use conversion restrictions in low-risk areas are relaxed under natural development scenarios, and the upper limit of the allowed conversion area is positively adjusted with the ecological security level value, etc.

[0049] Therefore, dynamic control parameters do not completely replace basic parameters. Instead, they are enhanced control variables with spatiotemporal variability and feedback adjustment functions, generated through mathematical modeling and a rule engine, starting from the basic parameters and integrating real-time assessed ecological risk information. This mechanism achieves a leap from "static simulation" to a closed loop of "perception, response, and regulation," significantly improving the model's ecological rationality and decision support capabilities in complex mountainous environments.

[0050] Furthermore, the above method may also include: determining the top 20% of regions corresponding to high-sensitivity risk areas based on the weight ranking of risk indicators output by the radial basis function neural network; for high-sensitivity risk areas, adopting a higher simulation spatial resolution than that of conventional areas when running the PLUS model, and introducing micro-topographic constraint rules and neighborhood expansion effect enhancement mechanisms during land expansion.

[0051] Specifically, the method of determining high-sensitivity risk areas based on the weighted ranking of risk indicators can include: First, determining multiple unified risk indicators based on the aforementioned multi-indicator synergistic PSR system; second, calculating the weight of each spatial evaluation unit on each of the aforementioned indicators using an RBF neural network, and then weighted and integrated to obtain the comprehensive ecological risk score of that unit; finally, ranking the comprehensive ecological risk scores of all units across the entire region, and identifying the spatial areas corresponding to the top 20% of units with the highest scores as high-sensitivity risk areas. In short, all regions are evaluated using the same indicators, but differences in their own geographical attributes lead to different comprehensive scores, and the top 20% of regions are the high-sensitivity risk areas.

[0052] As a preferred example, the aforementioned micro-topographic constraint rules may include: proportionally decreasing the probability of mechanized farmland converting to types other than woodland with increasing slope; for example, for every 5° increase in slope, the probability of mechanized farmland converting to types other than woodland decreases by 15%. The aforementioned neighborhood expansion effect enhancement mechanism may include: enhancing the neighborhood clustering tendency of corresponding land use types in ecologically sensitive areas; for example, increasing the probability of a grid around woodland converting to woodland by 20%.

[0053] Based on the above embodiments, high-sensitivity risk areas can be identified by using the risk index weights output by the radial basis function neural network. Within these areas, higher spatial resolution simulations, the introduction of micro-topographic constraint rules, and the enhancement of neighborhood expansion effects can solve the problems of traditional land use simulations, such as coarse spatial characterization of ecologically sensitive areas, insufficient dynamic response capabilities, and difficulty in accurately reflecting the impact of local topography and ecological aggregation processes. This enables refined and differentiated simulation and regulation of key ecological risk areas during the mechanization of mountain farmland, improving the ecological rationality and prediction accuracy of the model in complex mountain environments. As a result, unreasonable development in high-risk areas can be avoided more effectively, ensuring the stability of the ecosystem.

[0054] Step S106: Run the PLUS model based on dynamic control parameters to generate target land use patterns under multiple scenarios.

[0055] In one embodiment, step S106 may include: loading dynamic control parameters into the PLUS model, and performing land use pattern evolution simulations under three scenarios—natural development, farmland protection, and ecological priority—based on the parameter-driven PLUS model to generate corresponding target land use patterns.

[0056] The specific implementation process may include: First, loading the dynamic control parameters generated in the aforementioned steps into the PLUS model. These dynamic control parameters include the ecological risk index weights output by the RBF neural network, the regional ecological security level value, and the spatially differentiated simulation constraints determined based on risk sensitivity analysis. These parameters serve as the core regulatory factors for model operation, replacing the fixed empirical transformation rules and static resistance coefficients in traditional models.

[0057] Subsequently, for three preset scenarios—natural development, farmland protection, and ecological priority—corresponding total land use demand and conversion priority rules were set, and the spatial expansion process of various land uses was simulated using dynamic control parameters. During the simulation, based on the development probability map, the model appropriately relaxed development restrictions in areas with high ecological security levels, allowing orderly expansion in low-risk areas; while in high-risk weight areas, it automatically increased conversion resistance to suppress land use behaviors that did not conform to ecological protection goals. Simultaneously, local resolution enhancement and micro-topographic constraint mechanisms were implemented for identified highly sensitive ecological areas to enhance the accuracy of depicting small-scale land change processes.

[0058] The entire simulation employs a multi-round iterative approach and a random seed perturbation generation strategy to ensure the stability and reliability of the results. The final output is the land use pattern for the target year under each scenario, providing spatial foundational data for subsequent ecological risk reassessment and optimization decisions. This method, by introducing a dynamic feedback mechanism, achieves a shift from "passive simulation" to "active regulation," making the simulation process both ecologically sound and policy-adaptable.

[0059] In its specific implementation, the PLUS model, serving as the core spatial simulation engine of the ecological risk assessment system for mechanized mountain farmland, leverages its two main functional modules: the Land Expansion Analysis Strategy Module (LEAS) and the Conversion and Allocation Rules Simulation Module (CARS). This enables a complete technical loop, from uncovering historical evolution patterns to generating future multi-scenario patterns. The model not only scientifically predicts the spatial expansion trends of mechanized farmland under different development paths but also, through deep coupling with the RBF neural network, endows the simulation process with ecological risk perception capabilities, significantly enhancing the ecological rationality and decision support value of the prediction results.

[0060] Step S108: Input the target land use pattern into the radial basis function neural network, recalculate the corresponding ecological security level value, and iteratively update the dynamic control parameters according to the calculation results until the preset ecological risk control convergence conditions are met.

[0061] In this embodiment, the preset ecological risk control convergence condition may include: the change in the ecological security level value output in two consecutive iterations is less than a preset threshold, or the cumulative number of iterations reaches a preset upper limit. Preferably, after inputting the target land use pattern into the radial basis function neural network and recalculating the corresponding ecological security level value, the dynamic control parameters are iteratively updated according to the degree of difference between the current calculation result and the previous result until the change in the two calculation results is less than the preset threshold, or the maximum number of iterations is reached, so as to meet the ecological risk control convergence condition.

[0062] Step S110: Based on the final determined target land use pattern and its corresponding ecological security level value, generate a comprehensive ecological risk assessment result for the mechanization of mountain farmland.

[0063] Based on the final determined target land use pattern and its corresponding ecological security level value, the results of the ecological risk classification of mountain farmland suitable for mechanization or the risk heat map can be generated by combining the preset comprehensive evaluation function, which can be used as the comprehensive ecological risk evaluation result.

[0064] The specific implementation process may include: after obtaining the final determined target land use pattern and its corresponding ecological security level value, firstly, uniform preprocessing of the entire spatial data is performed, including geometric correction, projection transformation, resolution matching, and edge alignment, to ensure that the remote sensing imagery, topographic data, and land use classification results are consistent in spatial location and raster structure, providing a basic guarantee for subsequent continuous field analysis. Subsequently, the target land use pattern is used as input data and substituted into the trained RBF neural network model to calculate its corresponding ecological security level value pixel by pixel. This model comprehensively considers multi-dimensional features such as slope, soil thickness, shape index, remote sensing ecological index (RSEI), net primary productivity (NPP), farmland connectivity, and accessibility of roads and water conservancy facilities, and outputs a quantitative score reflecting ecosystem stability through nonlinear mapping relationships, forming a continuously distributed spatial layer of ecological security level.

[0065] Based on this, a comprehensive evaluation function is constructed, with the ecological security level as the core variable. Combining spatial autocorrelation analysis and a local weight adjustment mechanism, a regional ecological risk index is generated. This index achieves a "risk-oriented" expression through inverse transformation—that is, the lower the ecological security level, the higher the corresponding risk index, thus more intuitively reflecting the spatial distribution characteristics of potential ecological threats. Simultaneously, spatial smoothing and boundary optimization algorithms are introduced to reduce noise interference and highlight the main risk trends.

[0066] Furthermore, the Jenks Breaks method was used to optimally divide the ecological risk index into five levels: low risk, lower risk, medium risk, higher risk, and high risk, with standardized scores of 20, 40, 60, 80, and 100 respectively, to quantitatively represent the degree of risk. The classification results were visualized using a Geographic Information System (GIS) platform, generating a risk level map and a risk heat map for the mechanization of mountain farmland, clearly showing the risk distribution pattern and spatial clustering trend in different regions.

[0067] Ultimately, the above results were integrated into a comprehensive assessment of the ecological risks of mechanization in mountainous farmland. This assessment includes both global risk distribution characteristics and supports the identification of local hotspots and scenario comparison analysis, providing an actionable basis for decision support in scientifically formulating differentiated remediation strategies, optimizing land use allocation, and preventing ecological degradation risks.

[0068] The above-mentioned method for identifying ecological risks of mechanization adaptation in mountainous farmland provided by this invention calculates the ecological security level and the weights of each indicator using a trained radial basis function neural network based on risk indicator data generated from a constructed multi-dimensional indicator system. Furthermore, the weights are dynamically integrated into the PLUS model to perform risk-weighted correction on conversion resistance and expansion probability, generating dynamic control parameters. Based on these parameters, the model simulates land use patterns under multiple scenarios, and the results are fed back to the RBF network for re-evaluation until convergence conditions are met, achieving iterative optimization of the control parameters. Finally, a comprehensive evaluation result is generated based on the optimal pattern and ecological level. This solves the technical problems of insufficient comprehensiveness and accuracy in existing ecological risk assessments of mechanization adaptation in mountainous farmland, improving the comprehensiveness and accuracy of ecological risk identification for mechanization adaptation in mountainous farmland.

[0069] To facilitate understanding, this invention also provides an application example of the method for identifying ecological risks of mechanization in mountainous farmland, see [link to relevant documentation]. Figure 2 The flowchart shown is another method for identifying the ecological risks of mechanization in mountainous farmland. This method mainly includes the following steps S202 to S208:

[0070] Step S202: Use UAVs to survey and acquire mountain data, and generate mountain spatial range vector data, hyperspectral image data, and digital elevation data based on the acquired data. Perform data preprocessing, constrain segmentation on the hyperspectral image data, and extract parameter information.

[0071] Specifically, firstly, drones equipped with high-definition cameras are used to conduct surveys in mountainous areas to acquire high-precision mountain data. Then, the data processing module receives and processes the mountain data from the drone surveying module, performs geometric correction on the high-resolution remote sensing imagery, and performs projection conversion based on the vector data projection space of farmland suitable for mechanization.

[0072] Real-time position and attitude data of the UAV are acquired via GPS or an inertial navigation system (IMU). This data is compared and corrected against a preset flight path to ensure the accuracy of position and heading information during the mapping process. Next, sensor calibration is performed, correcting parameters such as angle, focal length, and distortion for the onboard high-definition camera to ensure the data acquired by the sensor has good geometric and optical characteristics. After calibration, data denoising is performed. Finally, according to the requirements of the generation unit, the raw data is converted into a standard format suitable for processing and analysis. This includes converting the data to point cloud format, raster format, or other specific data structures so that the subsequent terrain analysis module can effectively use and parse this data. After the data format is completed, coordinate systems and units are unified to ensure that the data uses the same geographic coordinate system and consistent units; this avoids errors caused by coordinate transformation or unit conversion during data processing and analysis. Through the above data calibration, denoising, and formatting process, the data processing module can effectively optimize the raw mountain data acquired from the UAV, laying a solid foundation for generating high-quality mountain spatial extent vector data, land use data, and digital elevation data (DEM). Geographic Information System (GIS) software is used to process the collected data for further analysis and visualization of vector and raster data.

[0073] Step S204: Based on the PSR system of multi-indicator synergy, construct the ecological risk index system for the mechanization of mountain farmland to quantify the changes in the ecological security pattern of the mechanization of mountain farmland, and use the radial basis function (RBF) neural network to generate the ecological security level value of the mechanization of mountain farmland and the weight of each indicator.

[0074] From the perspective of quantity, quality, and ecology, a multi-indicator collaborative "pressure-state-response" (PSR) ecological risk discrimination index system is constructed to quantify the changes in the ecological security pattern of mechanization in mountain farmland. The radial basis function (RBF) neural network is used to measure the ecological security level of mechanization in mountain farmland, and then the weight of each indicator is determined by the least squares method.

[0075] Furthermore, the traditional PSR (Potential Risk Assessment) system for assessing the ecological risk of mechanization in mountain farmland can specifically include:

[0076] Evaluation indicators were constructed by selecting factors affecting the ecological risk of mechanization of mountain farmland from three aspects: pressure indicators, status indicators, and response indicators.

[0077] In terms of stress indicators, soil layer thickness, land reclamation rate, fertilizer application intensity, and pesticide application intensity can be selected as evaluation indicators.

[0078] The formula for calculating soil layer thickness is:

[0079] ;

[0080] T represents the average soil thickness of mechanized farmland within the evaluation unit, Ti represents the soil thickness of the i-th plot, and n represents the number of plots within the evaluation unit.

[0081] Land reclamation rate refers to the ratio of farmland area transformed for mechanization in mountainous areas to the total land area. The transformed farmland is extracted from the land use data, and the perfection of the transformed farmland can be calculated using the buffer analysis and overlay analysis tools in ArcGIS 10.8. The intensity of fertilizer application and pesticide use is obtained from on-site soil sampling and testing.

[0082] Regarding state indicators, the average slope and average elevation, shape index, patch density, spatial clustering, remote sensing ecological index, net primary productivity, and grain yield per unit area of ​​cultivated land were selected as evaluation indicators.

[0083] The formula for calculating the average slope is:

[0084] ;

[0085] S is the average slope of the mechanized farmland within the evaluation unit, Si is the slope of the i-th plot, and n is the number of plots within the evaluation unit.

[0086] The formula for calculating average elevation is:

[0087] ;

[0088] H represents the average slope of the mechanized farmland within the evaluation unit, Hi represents the slope of the i-th plot, and n represents the number of plots within the evaluation unit.

[0089] The formula for calculating the shape index is:

[0090] ;

[0091] SE represents the regularity of the shape of the mechanized farmland within the evaluation unit, Pi represents the perimeter of the i-th plot, and Ai represents the area of ​​the i-th plot.

[0092] The formula for calculating patch density is:

[0093] ;

[0094] PD represents the number of plots of mechanized farmland within the evaluation unit, NP represents the number of plots, and A represents the total cultivated land area of ​​the evaluation region.

[0095] The formula for calculating spatial clustering is:

[0096] ;

[0097] BA represents the spatial clustering degree of mechanized farmland plots within the evaluation unit, Pi represents the perimeter of the i-th plot, and n represents the number of plots within the evaluation unit.

[0098] The formula for calculating the remote sensing ecological index is:

[0099] ;

[0100] RSEA is the average value of the remote sensing ecological index of mechanized farmland within the evaluation unit, VI is the greenness index, WET is the humidity index, LST is the heat index, and SI is the dryness index.

[0101] Net primary productivity can be obtained from high-precision remote sensing image interpretation. The data is processed using the Toolbox in ArcGIS 10.8, and vegetation cover is determined using remote sensing analysis tools within the GIS. The calculation formula is as follows: NPP is the net primary productivity per unit of mechanized farmland, GPP is the total primary productivity, R is the energy loss from plant respiration, and A is the total cultivated land area of ​​the evaluation region.

[0102] The grain yield per unit area of ​​cultivated land refers to the ratio of the total grain output of mechanized farmland within the evaluation unit to the total farmland area of ​​the evaluation region. The total grain output of mechanized farmland within the evaluation unit is obtained from field measurements. Mechanized farmland is extracted from the land use data, and the perfection of mechanized farmland can be calculated using the buffer analysis and overlay analysis tools in ArcGIS 10.8.

[0103] In terms of response indicators, irrigation and drainage guarantee rate, field road network density, total power of agricultural machinery per unit of cultivated land, and fixed investment in the primary industry can be selected as evaluation indicators.

[0104] The irrigation and drainage guarantee rate refers to the ratio of the area of ​​cultivated land that can be directly irrigated to the total cultivated land area of ​​the evaluation unit. The irrigation and drainage guarantee rate can be calculated by extracting ditches from the land use data and using the buffer analysis and overlay analysis tools in ArcGIS 10.8.

[0105] Field road network density refers to the ratio of farm road accessibility within an evaluation unit to the total cultivated land area of ​​the evaluation unit. Farm road accessibility can be calculated using the buffer analysis tool in ArcGIS 10.8.

[0106] The total power of agricultural machinery per unit of cultivated land refers to the ratio of the total power of agricultural machinery in the evaluation unit to the total cultivated land area of ​​the evaluation unit. The total power of agricultural machinery in the evaluation unit is obtained based on on-site measurements.

[0107] The fixed investment amount in the primary industry refers to the agricultural-related fixed asset investment amount within the evaluation unit, obtained based on on-site measurements.

[0108] Further utilizing radial basis function (RBF) neural networks to determine the ecological risk level of mechanization in mountainous farmland can specifically include selecting an RBF neural network for ecological risk level determination. The learning process of the RBF neural network can include the following steps:

[0109] The RBF neural network is a three-layer feedforward network (input layer, hidden layer, and output layer). Its core is to use radial basis functions as activation functions for hidden layer neurons and to map the input space to a high-dimensional feature space through distance metrics (such as Euclidean distance), thereby solving nonlinear problems.

[0110] The input layer is mainly used to receive feature vectors. The hidden layer is the core of the RBF neural network, consisting of a set of radial basis functions (RBFs). Each hidden layer neuron has an associated center point and a radial basis function as its activation function. The hidden layer neuron calculates the distance between itself and its center point from the input data and passes this distance as input to the radial basis functions, which then calculate the neuron's output. The formula for calculating the radial basis function is: , Let x be an n-dimensional input vector, and ci be the ith basis function center value, which has the same dimension as the input vector. The normalization constant of the width of the i-th center point of the basis function determines the sensitive region of the neuron. Let x be the Euclidean distance from the input x to the center ci.

[0111] The output layer is typically a linear layer. The final output is obtained by weighted summing of the outputs of the hidden layers. The formula for calculating the weighted sum is as follows: , Let b be the weight from the i-th neuron in the hidden layer to the k-th neuron in the output layer. k This is the bias term for the k-th neuron in the output layer.

[0112] Training an RBF neural network involves two main steps: determining the hidden layer parameters (ci and ) and calculate the output layer weights ( Hidden layer parameters are often determined using K-Means clustering or by randomly selecting centers. K-Means clustering refers to clustering the training data... Perform K-Means clustering (number of clusters m = number of hidden layer nodes), the cluster centers c1, c2, ..., c1 are the RBF centers, and calculate the average distance of each cluster to determine the width. The calculation formula is: The output layer weights are calculated using the least squares method. With the hidden layer outputs fixed, the output layer is a system of linear equations. The formula for calculating the linear equation system is:

[0113] ;

[0114] Target output matrix The solution for weight w is: After training, the RBF neural network can be used to measure and classify new input data. The specific steps are as follows:

[0115] After determining the indicators used to assess the ecological risk level of mechanization in mountain farmland, the relevant data for each indicator were preprocessed, including handling missing values, feature scaling, and data splitting. The selected indicators were used as input variables for the radial basis function. For each indicator, the center and width of the radial basis function were determined using the K-means clustering algorithm. For each input variable, the output value was calculated using the radial basis function model.

[0116] Step S206: Simulate land use patterns under multiple scenarios based on the PLUS model, and predict the changes in ecological risks of mechanization of mountain farmland under the corresponding scenarios based on land use data under different scenarios.

[0117] Furthermore, based on the PLUS model, land use patterns under multiple scenarios are simulated to predict the changes in the ecological risks of mechanization of mountain farmland under the corresponding scenarios. Specifically, this can include setting three future land use patterns under the 2030 natural development scenario, farmland protection scenario, and ecological priority scenario, respectively. Based on the land use data under different scenarios, the ecosystem service value and land use ecological risks under the corresponding scenarios are simulated respectively.

[0118] First, the land expansion portions from two different periods of the evaluation unit are extracted, and the Random Forest algorithm is used in the LEAS module to mine the relationship between driving factors and land expansion. In the LEAS module, random sampling is selected as the sampling method, the sampling rate is set to 0.01 by default, and the number of decision trees is 20. Since the number of mTry selections cannot exceed the number of driving factors, mTry is set to 10, and the number of parallel threads is 3. Various land development probabilities are generated. The formula for analyzing the driving mechanism of land use change based on the Random Forest algorithm is as follows:

[0119] ;

[0120] Among them, X m i is the driving factor. The factor weights for training the random forest are determined by the selection of driving factors based on the ecological risk discrimination index system. Secondly, by continuously changing the constraints, transformation rules, and transformation probabilities, and based on the threshold reduction mechanism and random seed generation mechanism, the 2030 land use patches for multiple scenario evaluation units are simulated under the CARS module, and the simulation accuracy is verified using the Kappa coefficient. The core formula for the Kappa coefficient is:

[0121] ,

[0122] Where, p o p represents the observation consistency ratio (actual accuracy). e The expected consistency ratio (consistency probability under random conditions) is used. Finally, during the simulation, it is assumed that other conditions remain unchanged, and only the land use-related parts of the ecosystem service value accounting and risk probability are changed. The rate of change of ecosystem service value is used as the loss amount to compare the land use ecological risk values ​​under different scenarios.

[0123] Step S208: Based on the evaluation system and the weight of each indicator, obtain the comprehensive evaluation results of the ecological risk assessment of the mechanization of mountain farmland, and classify the ecological risk level of the evaluation results according to the pre-constructed ecological risk assessment level standard.

[0124] Based on the comprehensive ecological risk assessment results of the target farmland, the ecological risk of mechanization adaptation in mountainous farmland was divided into five levels according to the natural discontinuity method. To unify the indicator dimensions, a graded scoring method was adopted for value assignment, where low risk = 20 points, lower risk = 40 points, medium risk = 60 points, higher risk = 80 points, and high risk = 100 points. The regional division results were visualized using the geographic software ArcGIS 10.8, which can intuitively show the ecological risk of mechanization adaptation in the target area. Based on the visualization results, the ecological status of cultivated land in the study area before and after mechanization adaptation and its changing patterns are clarified, providing a basis for subsequent targeted development of pathways to improve the ecological security level of cultivated land.

[0125] As a preferred example, in the PLUS model simulation process, 12 driving factors affecting land use change can first be selected based on the PSR indicator system, divided into three categories: "natural factors," "socio-economic factors," and "distance factors," and then rasterized using ArcGIS (30m resolution).

[0126] Natural factors: average slope (DEM calculation), average elevation (DEM calculation), soil organic matter content (hyperspectral inversion); Socioeconomic factors: fertilizer application intensity (field sampling), mechanical power per unit of cultivated land (farmer survey), investment in primary industry (statistical data); Distance factors: distance to farm road, distance to ditch, distance to village (obtained through Euclidean distance analysis of vector data).

[0127] Then, the random forest algorithm was used to analyze the impact of driving factors on the expansion of mechanized farmland. The expansion areas of mechanized farmland from 2015 to 2022 were extracted (obtained by comparing two periods of remote sensing images). 1000 expansion sample points and 1000 non-expansion sample points were randomly selected. The 12 driving factor values ​​of the sample points were used as inputs, and "whether to expand" (1=expansion, 0=non-expansion) was used as outputs. The random forest model was trained (number of decision trees=20, mTry=10) and the importance weight of each factor was output.

[0128] For example, if the weight of "distance from farm road" is 0.2, it indicates that this factor has the greatest impact on the expansion of mechanized farmland; based on the factor weights, the probability of mechanized farmland development in each grid (30m×30m) within the evaluation unit is calculated:

[0129] ;

[0130] in For factor weights, Let m be the factor value of the i-th grid, forming a probability distribution map.

[0131] Based on the constraints of different scenarios and combined with the development probability map, the land use pattern in 2030 is simulated: First, conversion rules are set: "Land use conversion priority" is defined (e.g., in the ecological priority scenario, the conversion priority of forest land → cultivated land = 0, and the conversion priority of cultivated land → forest land = 0.8); Threshold reduction mechanism: Starting from a high development probability (e.g., 0.9), the threshold is reduced in steps of 0.05, and each time the grid with a probability higher than the threshold is converted into mechanized farmland until the scenario constraints are met (e.g., the cultivated land protection scenario requires 30% of the area); Random seed generation: To avoid the simulation results being monotonous, 5 random seeds (123, 456, 789, 1011, 1213) are set to generate 5 sets of simulation results, and the average value is taken as the final pattern; Accuracy verification: The Kappa coefficient is used to verify the simulation accuracy. The simulated pattern in 2022 is compared with the actual pattern. If Kappa ≥ 0.8, the simulation is reliable.

[0132] Land use pattern data for each scenario in 2030 (such as the distribution of mechanized farmland and forest land) are input into the trained RBF network to recalculate the ecological security level value for each scenario, thereby predicting risk changes.

[0133] Furthermore, in the combined application of radial basis function neural networks and the PLUS model, the process is not a simple linear flow from "evaluation" to "simulation." Instead, it involves deep coupling through four stages: dynamic parameter permeation, real-time scenario adjustment, spatial precision focusing, and bidirectional closed-loop iteration, forming an organic system where "risk drives simulation, and simulation feeds back into risk." The specific application process is as follows:

[0134] First, the Restricted Baseline Ecosystem Function (RBF) was used to quantitatively analyze the ecological risks of mechanization in mountainous farmland. By inputting ecological risk indicators such as slope, shape index, and spatial clustering, the RBF outputs two core results after training: one is the weight of each risk indicator (e.g., slope risk weight, remote sensing ecological index risk weight), reflecting the intensity of different factors' impact on ecological security; the other is a quantitative value of the regional ecological security level, characterizing the overall stability of the current ecosystem. This provides the foundational data for the "risk perception" of the PLUS model.

[0135] Next, the risk weights of RBF are directly injected into the land conversion rule system of the PLUS model. In the "Land Conversion and Allocation Rule Simulation Module" of the PLUS model, traditional parameters such as conversion probability and resistance coefficient no longer depend on fixed historical data, but are dynamically bound to the risk weights output by RBF.

[0136] When RBF identifies "slope > 25°" as a high-weight risk indicator, the PLUS model can automatically increase the resistance coefficient of "conversion of forest land to mechanized farmland" in the region (resistance = basic resistance × risk weight multiple), while reducing the probability of farmland expansion in the region (probability = baseline probability × (1 - risk weight)). If the weight of "remote sensing ecological index" is high, the conversion of ecological land to farmland will still be restricted in the "farmland protection scenario" to avoid exacerbating ecological risks due to short-term farmland expansion.

[0137] This parameter permeation transforms the PLUS model's transformation logic from a blind replication of spatial patterns to a risk-aware decision-making simulation. Simultaneously, the ecological security level value output by RBF adjusts the scenario constraint strength of the PLUS model in real time.

[0138] In the three scenarios of natural development, farmland protection, and ecological priority set by the PLUS model, the constraint boundaries of each scenario (such as the ecological red line range and the elastic space of farmland retention) are no longer statically preset, but dynamically expand and contract with the safety level value.

[0139] When the RBF calculates a safety level value for a region below the critical value, even under the "farmland protection scenario," the PLUS model will automatically trigger enhanced constraints under the "ecological priority scenario." For example, it might temporarily extend the ecological protection red line by 10%-20% and prohibit any mechanization modifications in high-risk areas (such as steep slopes and water source areas). Conversely, under the "natural development scenario," the land use conversion restrictions in low-risk areas (such as slopes <15°) will be appropriately relaxed, allowing for the orderly conversion of forest land to terraced fields. The conversion scale is positively correlated with the safety level value (for every 5-point increase in the safety value, the maximum conversion area increases by 5%). This adjustment mechanism transforms scenario simulation from a scripted performance to one that adjusts strategies based on real-time conditions, better reflecting the dynamic nature of mountain ecosystems.

[0140] At the spatial simulation level, leveraging the risk sensitivity analysis of RBF (Risk-Based Risk Analysis), the PLUS model achieves precise characterization through both "global simulation" and "focused sensitive areas." By ranking the risk indicators using RBF, "high-sensitivity risk areas" (such as steep slopes and soil-poor areas in the top 20% of weight) are selected, and simulation parameters are customized for these areas: In the "Land Expansion Analysis Strategy Module," the simulation resolution of high-sensitivity areas is increased from the conventional 30m to 10m, capturing small-scale patch transformations below 200㎡ (such as whether scattered farmland on steep slopes has been converted to forest); simultaneously, micro-topographical constraint rules are added (such as a 15% decrease in the probability of mechanized farmland conversion for every 5° increase in slope), and the neighborhood expansion effect of ecological land is strengthened (the probability of a grid around forest land converting to forest land increases by 20%). Non-sensitive areas maintain conventional parameters, balancing simulation efficiency and accuracy. This differentiated approach solves the problem of the traditional PLUS model "focusing on the big picture while neglecting the details" in large-scale simulations, making the details of land use changes in high-risk areas more closely reflect the actual ecological risk characteristics.

[0141] Finally, dynamic optimization of risk and simulation is achieved through bidirectional closed-loop iteration. After the PLUS model simulates multi-scenario land use patterns based on the parameters and constraints of the RBF, these pattern data (such as new arable land distribution and vegetation cover changes) are back-input into the RBF to recalculate the ecological security level and risk weights under each scenario. If the simulation results of a certain scenario show that the risk reduction is not as expected (e.g., the safety value only increases by 3 points under the ecological priority scenario, which is lower than the target of 5 points), new risk weights are extracted (e.g., the weight of "soil layer thickness risk" is found to have increased), and the transformation rules of the PLUS model are adjusted in reverse. The simulation is repeated and evaluated until the land use scheme with the optimal risk prevention and control is obtained.

[0142] This closed loop of "RBF assessment, PLUS simulation, RBF reassessment, and PLUS optimization" upgrades the two from "one-way input and output" to an intelligent system of "co-evolution," ultimately outputting not only land use patterns but also decision-making solutions that integrate ecological risk prevention and control objectives.

[0143] In summary, throughout the process, RBF acts as the central hub for risk perception and quantification, providing dynamic parameters, scenario thresholds, and spatial focus for the PLUS model. Meanwhile, the PLUS model, as the center for spatial simulation and solution generation, transforms risk constraints into concrete land use patterns and continuously optimizes risk control effectiveness through feedback iteration. The synergy between the two enables the full-chain penetration of "ecological risk" from abstract indicators to spatial decision-making, ensuring that the simulation and prediction of mechanization-friendly transformation of mountain farmland conforms to spatial patterns while strictly adhering to the bottom line of ecological security.

[0144] Based on the same inventive concept, this invention also provides an ecological risk assessment system for mechanization of mountain farmland, see [link to relevant documentation]. Figure 3 As shown, the system mainly includes the following parts:

[0145] Ecological security assessment unit 310 is used to generate the ecological security level value of the mountain farmland mechanization ecosystem and the corresponding weights of each risk indicator based on the multidimensional risk indicator data output by the mountain farmland mechanization ecological risk indicator system.

[0146] The dynamic parameter control unit 320 is used to perform risk-weighted correction on the land conversion resistance coefficient and land use expansion probability in the PLUS model according to the preset ecological risk response rules, combined with the ecological security level value and the weight of each risk indicator, and generate dynamic control parameters for controlling the operation of the model.

[0147] Land use pattern simulation unit 330 is used to run the PLUS model based on dynamic control parameters and generate target land use patterns under multiple scenarios.

[0148] The parameter iteration update unit 340 is used to input the target land use pattern into the radial basis function neural network, recalculate the corresponding ecological security level value, and iteratively update the dynamic control parameters according to the calculation results until the preset ecological risk control convergence conditions are met.

[0149] The comprehensive evaluation generation unit 350 is used to generate a comprehensive evaluation result of the ecological risk of mechanization of mountain farmland based on the final determined target land use pattern and its corresponding ecological security level value.

[0150] In one embodiment, the system may further include: a multidimensional risk data acquisition unit, used to acquire multidimensional risk indicator data output by the ecological risk indicator system for the mechanization of mountain farmland.

[0151] The ecological risk assessment system and method for adapting mountain farmland to mechanization proposed in this invention is a systematic study covering the entire process. It fully utilizes the efficient mapping capabilities of UAVs to acquire high-precision remote sensing images, extracting rich and diverse ground feature information. It adopts a PSR model to comprehensively consider current characteristics while also taking into account potential impacts, establishing an ecological risk assessment framework for adapting farmland to mechanization based on the protection of ecosystem service value. Furthermore, it conducts specific analysis on the particularities of farmland ecosystems, enabling a more scientific and comprehensive assessment of the ecological risk status of regional farmland adaptability to mechanization.

[0152] The four linkage mechanisms between the RBF and PLUS models are the core features of this invention. First, the risk weights output by the RBF are used as parameters for the dynamic transformation rules of the PLUS model, breaking the limitations of traditional PLUS models that rely on historical data or experience to set transformation rules. This makes land use change simulation more aligned with risk prevention and control needs, improving the ecological rationality of the simulation. Second, a feedback adjustment mechanism is constructed from ecological security level values ​​to PLUS scenario constraints, changing the static preset nature of multiple scenarios in the PLUS model. This allows scenario simulation to be dynamically adjusted based on real-time safety status, better adapting to the dynamic changes in mountain ecosystems. Third, RBF risk sensitivity analysis is used to optimize the accuracy of PLUS patch simulation, solving the problem of insufficient spatial detail capture in ecologically risk-sensitive areas by traditional PLUS models. By focusing on highly sensitive risk areas, the accuracy of simulation results is improved. Fourth, a closed-loop linkage of "risk prediction, scenario simulation, and risk reassessment" is formed, breaking the limitations of the unidirectional process of the RBF and PLUS models. This enables dynamic iteration of "risk assessment, scenario optimization, and effect verification," ensuring the output of the optimal land use plan for risk prevention and control.

[0153] Meanwhile, the RBF neural network, combined with geographic software information processing, enables an objective assessment of the ecological risks associated with farmland mechanization, overcoming the shortcomings of traditional methods such as insufficient training speed and cross-interference of weights in local response characteristics. Based on the PLUS model, risk prediction overcomes the limitations of traditional models, such as reliance on linear regression, morphological distortion, difficulty in adjusting static constraints, and low computational efficiency. Overall, this invention can save a significant amount of work in identifying ecological risks associated with farmland mechanization and can also propose targeted pathways to improve ecological security, providing scientific decision-making support for the mechanization transformation of mountain farmland.

[0154] Based on the same inventive concept, this invention also provides an electronic device, see [link to relevant documentation]. Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, a communication interface 430, and a bus 440. The memory 420 stores machine-readable instructions that can be executed by the processor 410. When the electronic device is running, the processor 410 communicates with the memory 420 through the bus 440. The processor 410 executes the machine-readable instructions to perform the steps of the method described above.

[0155] Specifically, the memory 420 and processor 410 can be general-purpose memory and processor, without any specific limitations. When the processor 410 runs the computer program stored in the memory 420, it can execute the above method.

[0156] Processor 410 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 410 or by instructions in software form. The processor 410 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 420, and processor 410 reads the information from memory 420 and, in conjunction with its hardware, completes the steps of the above method.

[0157] Corresponding to the above method, this embodiment of the invention also provides a computer-readable storage medium storing machine-executable instructions. When the computer-executable instructions are called and run by a processor, the computer-executable instructions cause the processor to perform the steps of the above method.

[0158] The system / device provided in this embodiment of the invention can be specific hardware on a device or software or firmware installed on the device. The implementation principle and technical effects of the system / device provided in this embodiment of the invention are the same as those in the foregoing method embodiments. For the sake of brevity, any parts not mentioned in the system / device embodiments can be referred to the corresponding content in the foregoing method embodiments. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can all be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0159] In the embodiments provided by this invention, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0160] For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0161] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0162] In addition, the functional units in the embodiments provided by the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0163] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0164] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0165] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0166] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0167] Finally, it should be noted that the above embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. All should be covered within the scope of protection of the present invention.

Claims

1. A method for assessing the ecological risk of mechanization in mountainous farmland, characterized in that, include: Based on the multidimensional risk index data output by the ecological risk index system for mechanization of mountain farmland, the ecological security level value of the mechanization-friendly ecosystem of mountain farmland and the weights corresponding to each risk index are generated using a trained radial basis function neural network. Based on the preset ecological risk response rules, and combined with the ecological security level value and the weight of each risk indicator, the land conversion resistance coefficient and land use expansion probability in the PLUS model are risk-weighted and corrected to generate dynamic control parameters for regulating the operation of the model. The method for generating dynamic control parameters includes: obtaining multiple spatial driving factors affecting land use change trends based on key geographic elements in the multidimensional risk index data; training an expansion probability prediction model using machine learning algorithms based on the mapping relationship between historical land use change samples and the spatial driving factors; using the expansion probability prediction model to infer the entire grid cell to generate an initial land use expansion probability distribution map as the base value of the land use expansion probability; and spatially superimposing the base value of the land use expansion probability with the ecological sensitivity weight map output by the radial basis function neural network to generate a modified expansion probability distribution modulated by ecological risk. Running the PLUS model based on the dynamic control parameters to generate target land use patterns under multiple scenarios includes: loading the dynamic control parameters into the PLUS model, and driving the PLUS model to perform land use pattern evolution simulations under three scenarios: natural development, farmland protection, and ecological priority, based on the parameters, to generate corresponding target land use patterns; wherein, during the operation of the PLUS model, the ecological security level value calculated by the radial basis function neural network adjusts the constraint strength of the PLUS model in real time under different scenarios. The target land use pattern is input into the radial basis function neural network to recalculate the corresponding ecological security level value, and the dynamic control parameters are iteratively updated according to the calculation results until the preset ecological risk control convergence condition is met. Based on the final determined target land use pattern and its corresponding ecological security level value, a comprehensive ecological risk assessment result for the mechanization of mountain farmland is generated.

2. The method according to claim 1, characterized in that, The method further includes: Initial mapping data of mountain farmland was acquired synchronously using drones; the initial mapping data included high-resolution remote sensing images, hyperspectral biochemical parameters, and digital elevation data. The initial mapping data is corrected, denoised, and formatted to generate mountain farmland data; Based on the data of mountain farmland, an ecological risk index system for mechanization of mountain farmland was constructed.

3. The method according to claim 2, characterized in that, The methods for constructing an ecological risk indicator system for mechanization of mountain farmland based on the aforementioned mountain farmland data include: Based on the PSR system with multi-indicator synergy, multiple geographical elements related to ecosystem stability are extracted from the mountain farmland data, and key geographical elements with potential impact on the mechanization transformation of mountain farmland are screened; the screened key geographical elements are standardized and multi-dimensional risk indicator data are established.

4. The method according to claim 1, characterized in that, The method further includes: Based on the weight ranking of the risk indicators output by the radial basis function neural network, the region corresponding to the top 20% of the weight rankings is determined as a high-sensitivity risk area; For the high-sensitivity risk areas, the PLUS model adopts a higher simulation spatial resolution than that of conventional areas, and introduces micro-topographic constraint rules and neighborhood expansion effect enhancement mechanisms during land expansion.

5. The method according to claim 4, characterized in that, The micro-topography constraint rules include: proportionally decreasing the probability of converting mechanized farmland into types other than woodland as the slope increases; The enhanced neighborhood expansion effect mechanism includes: enhancing the tendency of adjacent land types to cluster in ecologically sensitive areas.

6. A system for assessing the ecological risk of mechanization in mountainous farmland, characterized in that, include: The ecological security assessment unit is used to generate the ecological security level value of the mountain farmland mechanization ecosystem and the corresponding weights of each risk indicator based on the multidimensional risk indicator data output by the mountain farmland mechanization ecological risk indicator system. The dynamic parameter control unit is used to perform risk-weighted correction on the land conversion resistance coefficient and land use expansion probability in the PLUS model according to the preset ecological risk response rules, combined with the ecological security level value and the weight of each risk indicator, and generate dynamic control parameters for controlling the operation of the model. The method for generating dynamic control parameters includes: obtaining multiple spatial driving factors affecting land use change trends based on key geographic elements in the multidimensional risk index data; training an expansion probability prediction model using machine learning algorithms based on the mapping relationship between historical land use change samples and the spatial driving factors; using the expansion probability prediction model to infer the entire grid cell to generate an initial land use expansion probability distribution map as the base value of the land use expansion probability; and spatially superimposing the base value of the land use expansion probability with the ecological sensitivity weight map output by the radial basis function neural network to generate a modified expansion probability distribution modulated by ecological risk. The land use pattern simulation unit is used to run the PLUS model based on the dynamic control parameters to generate target land use patterns under multiple scenarios; it loads the dynamic control parameters into the PLUS model and drives the PLUS model to perform land use pattern evolution simulations under three scenarios: natural development, farmland protection, and ecological priority, based on the parameters, to generate corresponding target land use patterns; wherein, during the operation of the PLUS model, the ecological security level value calculated by the radial basis function neural network adjusts the constraint strength of the PLUS model in real time under different scenarios. The parameter iteration update unit is used to input the target land use pattern into the radial basis function neural network, recalculate the corresponding ecological security level value, and iteratively update the dynamic control parameters according to the calculation results until the preset ecological risk control convergence condition is met. The comprehensive evaluation generation unit is used to generate a comprehensive evaluation result of the ecological risk of mechanization of mountain farmland based on the final determined target land use pattern and its corresponding ecological security level value.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 5.

Images