Fine-grained cultivated land protection potential optimization method, medium and system

Abstract

The invention provides a fine-grained cultivated land protection potential optimization method, medium and system, and belongs to the technical field of cultivated land data processing.The method comprises the steps that firstly, multi-source remote sensing, soil and climate data are obtained, and a basic feature matrix is extracted through multi-dimensional decomposition; then, multi-scale analysis of soil quality features is realized by utilizing a pyramid neural network model, and adaptive fusion is carried out through a feature integration module; and then constructing a contribution degree optimization equation set, comprehensively considering four aspects of benchmark evaluation, space correlation, time sequence evolution and weight balance, realizing accurate evaluation of the cultivated land protection potential, and generating an optimized protection partition scheme. On one hand, the pyramid neural network model overcomes the problem of insufficient feature extraction of a traditional method through multi-scale feature extraction and adaptive fusion; and on the other hand, through multi-dimensional information fusion and dynamic optimization, the contribution degree optimization equation set solves the problem that accurate evaluation of the cultivated land protection potential is difficult to realize in the prior art.

Description

Technical Field

[0001] The present invention belongs to the technical field of cultivated land data processing, and more particularly, relates to a fine-grained cultivated land protection potential optimization method, medium and system. Background Art

[0002] Cultivated land protection is an important foundation for ensuring national food security, and the evaluation of cultivated land protection potential is a key link in guiding cultivated land protection work. Traditional evaluation methods for cultivated land protection potential mainly rely on basic work such as remote sensing image interpretation, soil sampling analysis, and climate data statistics. By establishing an evaluation index system, methods such as the analytic hierarchy process and fuzzy comprehensive evaluation method are used for evaluation. These methods mainly construct evaluation models based on expert experience, obtain land use information by visually interpreting or automatically classifying remote sensing images, combine soil quality data obtained from on-site investigations and climate data observed at meteorological stations, calculate the weights of various evaluation indicators, and finally obtain the evaluation results of cultivated land protection potential.

[0003] However, the traditional evaluation methods have the following problems: First, the construction of the evaluation index system and the determination of weights mainly rely on expert experience, which is highly subjective and difficult to objectively reflect the differences in cultivated land protection potential in different regions; Second, the data acquisition method is single, and it is difficult to make full use of the rich information contained in multi-source remote sensing data, soil monitoring data, and climate observation data; Third, the evaluation method lacks a systematic analysis of the internal characteristics and external environment of cultivated land patches and cannot achieve multi-scale and multi-dimensional comprehensive evaluation; Finally, the evaluation results lack a dynamic update mechanism and are difficult to timely reflect the spatio-temporal change characteristics of cultivated land protection potential.

[0004] In the context of the rapid development of big data and artificial intelligence technologies, how to make full use of multi-source data, establish an objective and accurate evaluation method for cultivated land protection potential, and achieve fine-grained evaluation of cultivated land protection potential has become a technical problem that needs to be solved urgently. That is, there is a problem in the prior art that it is difficult to accurately evaluate the cultivated land protection potential. Summary of the Invention

[0005] In view of this, the present invention provides a fine-grained cultivated land protection potential optimization method, medium and system, which can solve the technical problem in the prior art that the fine-grained and accurate evaluation of cultivated land protection potential cannot be achieved.

[0006] The present invention is implemented as follows: In the first aspect of the present invention, a fine-grained cultivated land protection potential optimization method is provided. Remote sensing image data, soil monitoring data, and climate observation data of the target area are obtained as the basic data set; the basic data set is decomposed in multiple dimensions; cultivated land patches are extracted based on the land use type distribution matrix; the soil quality feature matrix is analyzed using the soil quality feature pyramid neural network model; the cultivated land patch condition index is evaluated according to the climate suitability matrix; a cultivated land protection potential evaluation model is constructed; an initial score is obtained by training using the neural network model; a cultivated land protection contribution matrix is established; an optimized score is obtained through optimization and adjustment; high, medium, and low protection potential areas are divided; spatial connectivity analysis is performed to identify priority protection areas and output them.

[0007] Among them, the remote sensing image data includes multispectral remote sensing image data, hyperspectral remote sensing image data, radar remote sensing image data, land cover classification data, and normalized difference vegetation index data.

[0008] Among them, the soil monitoring data includes soil organic matter content data, soil bulk density data, soil pH value data, soil nutrient content data, soil water content data, soil texture data, and soil erosion degree data.

[0009] Among them, the climate observation data includes annual average temperature data, annual precipitation data, sunshine hours data, accumulated temperature data, drought index data, precipitation seasonal distribution data, and extreme weather event data.

[0010] Among them, the basic data set is decomposed in multiple dimensions to obtain a land use type distribution matrix, a soil quality feature matrix, and a climate suitability matrix.

[0011] Among them, the cultivated land patches are extracted based on the land use type distribution matrix, and the area parameter, boundary length parameter, and fragmentation degree parameter of the cultivated land patches are calculated.

[0012] Among them, the soil quality feature pyramid neural network model includes five sequentially executed pyramid layers and a feature integration module.

[0013] Among them, the cultivated land protection contribution matrix is obtained by solving the contribution optimization equation set. The contribution optimization equation set includes a contribution benchmark equation, a spatial correlation equation, a temporal evolution equation, and a weight balance equation. The contribution benchmark equation is used to calculate the basic contribution of each evaluation index, the spatial correlation equation is used to evaluate the spatial coupling effect between evaluation indexes, and the temporal evolution equation is used to analyze the temporal variation characteristics of the contribution of evaluation indexes.

[0014] Among them, the pyramid layers include: the first pyramid layer is used to analyze the soil quality characteristics inside the cultivated land patches; the second pyramid layer is used to analyze the combination of the soil quality characteristics inside the cultivated land patches and the adjacent areas of the cultivated land patches; the third pyramid layer is used to analyze the combination of the soil quality characteristics inside the cultivated land patches, the adjacent areas of the cultivated land patches and the buffer zones around the cultivated land patches; the fourth pyramid layer is used to analyze the combination of the soil quality characteristics inside the cultivated land patches, the adjacent areas of the cultivated land patches, the buffer zones around the cultivated land patches and the watershed units of the cultivated land patches; the fifth pyramid layer is used to analyze the combination of the soil quality characteristics inside the cultivated land patches, the adjacent areas of the cultivated land patches, the buffer zones around the cultivated land patches, the watershed units of the cultivated land patches and the overall regional soil quality characteristics.

[0015] Among them, the feature integration module is used to determine whether the soil quality characteristic index output by the pyramid layer is greater than the soil quality characteristic credibility threshold. When it is greater than the soil quality characteristic credibility threshold, the soil quality characteristic index output by the current pyramid layer is used as the final soil quality characteristic index. When the feature integration module is less than the soil quality characteristic credibility threshold, it continues to perform the next-level pyramid analysis or uses the output of the last pyramid layer as the final soil quality characteristic index.

[0016] Among them, the light and heat resource condition index, the moisture condition index and the climate stress factor index of the cultivated land patches are evaluated according to the climate suitability matrix. The cultivated land protection contribution matrix is obtained by solving the contribution optimization equation set, and the contribution optimization equation set includes a contribution benchmark equation, a spatial correlation equation, a temporal evolution equation and a weight balance equation.

[0017] Among them, the inputs of the contribution benchmark equation include the normalized evaluation index values obtained from the cultivated land protection potential evaluation model, the index significance coefficients obtained from the historical monitoring data, the historical protection effect coefficients obtained from the historical protection records, and the regional development constraint coefficients obtained from the regional development plan. The inputs of the spatial correlation equation include the basic contribution matrix of the evaluation index, the spatial adjacency matrix obtained from the geographic information system, the terrain gradient coefficient obtained from the terrain data, the land use intensity index obtained from the remote sensing image data, and the landscape connectivity index obtained from the landscape pattern analysis. The inputs of the temporal evolution equation include the annual cultivated land protection effectiveness data obtained from the cultivated land protection monitoring system, the climate change trend coefficient obtained from the climate observation data, the social and economic development index obtained from the statistical department, and the policy regulation intensity index obtained from the policy documents.

[0018] The second aspect of the present invention provides a computer-readable storage medium, in which program instructions are stored. When the program instructions run on a computer, they are used to execute the above-mentioned fine-grained cultivated land protection potential optimization method.

[0019] The third aspect of the present invention provides a fine-grained cultivated land protection potential optimization system, which includes the above-mentioned computer-readable storage medium. The system can be any one of a computer, a server, and a single-chip microcomputer. The computer-readable storage medium is set inside the system, and a microprocessor for executing program instructions stored in the computer-readable storage medium is set inside the system.

[0020] Compared with the prior art, the present invention provides a fine-grained cultivated land protection potential optimization method, medium, and system. The fine-grained cultivated land protection potential optimization method proposed by the present invention realizes the accurate evaluation of cultivated land protection potential by constructing a soil quality feature pyramid neural network model and a cultivated land protection contribution degree optimization equation set. This method makes full use of multi-source remote sensing data, soil monitoring data, and climate observation data, and establishes an adaptive and dynamic evaluation framework through multi-dimensional decomposition and feature extraction.

[0021] The method of the present invention overcomes the problems existing in the traditional technology. First, by constructing a soil quality feature pyramid neural network model, automatic extraction of multi-scale features from local to global is realized, avoiding the influence of subjective experience. Second, by establishing a cultivated land protection contribution degree optimization equation set, effective fusion and comprehensive evaluation of multi-source data are realized, improving the objectivity and reliability of the evaluation results. Finally, by introducing a time series evolution equation and a spatial correlation equation, dynamic update and spatial optimization of the evaluation results are realized.

[0022] The present invention solves the technical problem of accurate evaluation of cultivated land protection potential, which is mainly reflected in the following aspects: First, the adaptive extraction of features is realized through the pyramid neural network model, improving the accuracy of feature expression. Second, the scientific fusion of multi-dimensional information is realized through the contribution degree optimization equation set, enhancing the reliability of the evaluation results. Third, the real-time update of the evaluation results is realized through the dynamic optimization mechanism, ensuring the timeliness of the evaluation results, and solving the problem in the prior art that it is difficult to achieve accurate evaluation of cultivated land protection potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0025] As Figure 1 shown, it is a flowchart of a fine-grained cultivated land protection potential optimization method provided by the first aspect of the present invention. This method includes the following steps: S01. Obtain remote sensing image data, soil monitoring data, and climate observation data of the target area as the basic data set. The remote sensing image data includes multi-spectral remote sensing image data, hyperspectral remote sensing image data, radar remote sensing image data, land cover classification data, and normalized difference vegetation index data. The soil monitoring data includes soil organic matter content data, soil bulk density data, soil pH value data, soil nutrient content data, soil water content data, soil texture data, and soil erosion degree data. The climate observation data includes annual average temperature data, annual precipitation data, sunshine hours data, accumulated temperature data, drought index data, precipitation seasonal distribution data, and extreme weather event data; S02. Perform multi-dimensional decomposition on the basic data set to obtain a land use type distribution matrix, a soil quality characteristic matrix, and a climate suitability matrix; S03. Extract cultivated land patches based on the land use type distribution matrix, and calculate the area parameter, boundary length parameter, and fragmentation degree parameter of the cultivated land patches; S04. Use the soil quality characteristic pyramid neural network model to analyze the soil quality characteristic matrix to obtain the soil fertility index, soil physical structure index, and soil chemical property index of the cultivated land patches. The soil quality characteristic pyramid neural network model includes five sequentially executed pyramid layers and a feature integration module: The first pyramid layer is used to analyze the soil quality characteristics inside the cultivated land patches; The second pyramid layer is used to analyze the combination of soil quality characteristics inside the cultivated land patches and the adjacent areas of the cultivated land patches; The third pyramid layer is used to analyze the combination of soil quality characteristics inside the cultivated land patches, the adjacent areas of the cultivated land patches, and the buffer zones around the cultivated land patches; The fourth pyramid layer is used to analyze the combination of soil quality characteristics inside the cultivated land patches, the adjacent areas of the cultivated land patches, the buffer zones around the cultivated land patches, and the watershed units of the cultivated land patches; The fifth pyramid layer is used to analyze the combination of soil quality characteristics inside the cultivated land patches, the adjacent areas of the cultivated land patches, the buffer zones around the cultivated land patches, the watershed units of the cultivated land patches, and the overall regional soil quality characteristics; The feature integration module is used to determine whether the soil quality characteristic index output by the pyramid layer is greater than the soil quality characteristic credibility threshold. When it is greater than the soil quality characteristic credibility threshold, the soil quality characteristic index output by the current pyramid layer is used as the final soil quality characteristic index. When it is less than the soil quality characteristic credibility threshold, continue the next-level pyramid analysis or use the output of the last pyramid layer as the final soil quality characteristic index; S05. Evaluate the light and heat resource condition index, water condition index, and climate stress factor index of the cultivated land patches according to the climate suitability matrix; S06. Construct an evaluation model for cultivated land protection potential, and input the area parameter, boundary length parameter, fragmentation degree parameter, soil fertility index, soil physical structure index, soil chemical property index, light and heat resource condition index, water condition index, and climate stress factor index of the cultivated land patches into the evaluation model for cultivated land protection potential; S07. Use a neural network model to train the evaluation model for cultivated land protection potential to obtain an initial score for cultivated land protection potential; S08. Establish a cultivated land protection contribution matrix based on the initial score for cultivated land protection potential, where the cultivated land protection contribution matrix is obtained by solving a contribution optimization equation set, and the contribution optimization equation set includes a contribution benchmark equation, a spatial correlation equation, a temporal evolution equation, and a weight balance equation; S09. Optimize and adjust the initial score for cultivated land protection potential according to the cultivated land protection contribution matrix to obtain an optimized score for cultivated land protection potential; S10. Divide the optimized score for cultivated land protection potential into high protection potential areas, medium protection potential areas, and low protection potential areas to generate a cultivated land protection zoning scheme matrix; S11. Conduct a spatial connectivity analysis on the cultivated land protection zoning scheme matrix to identify and output priority protection areas; The contribution benchmark equation is used to calculate the basic contribution of each evaluation index. The inputs include the normalized value of the evaluation index obtained from the evaluation model for cultivated land protection potential, the index significance coefficient obtained from historical monitoring data, the historical protection effect coefficient obtained from historical protection records, and the regional development constraint coefficient obtained from the regional development plan. The output is a matrix of the basic contribution of the evaluation index; The spatial correlation equation is used to evaluate the spatial coupling effect among evaluation indexes. The inputs include the matrix of the basic contribution of the evaluation index, the spatial adjacency matrix obtained from a geographic information system, the terrain gradient coefficient obtained from terrain data, the land use intensity index obtained from the remote sensing image data, and the landscape connectivity index obtained from landscape pattern analysis. The output is a matrix of spatial correlation contribution; The temporal evolution equation is used to analyze the temporal variation characteristics of the contribution of evaluation indexes. The inputs include the annual cultivated land protection effectiveness data obtained from the cultivated land protection monitoring system, the climate change trend coefficient obtained from the climate observation data, the social and economic development index obtained from the statistical department, and the policy regulation intensity index obtained from policy documents. The output is a matrix of temporal evolution contribution; The weight balance equation is used to comprehensively balance the final contribution of each evaluation index. The input includes the basic contribution matrix of the evaluation index, the spatial correlation contribution matrix, the time series evolution contribution matrix and the regional differentiation correction coefficient obtained from the regional differentiation assessment. The output is the final farmland protection contribution matrix.

[0026] The specific implementation methods of the above steps are described in detail below. The specific implementation method of step S01 is to obtain and preprocess remote sensing image data, soil monitoring data and climate observation data. The specific implementation first selects the data source. The remote sensing image data uses multi-source remote sensing data such as Landsat 8 and Sentinel 2, and the original image is processed by preprocessing methods such as geometric precision correction, radiation correction, and atmospheric correction; the soil monitoring data is obtained through field sampling and indoor analysis, and the layout of the sampling points adopts the stratified sampling method, with 1 to 3 sampling points set per square kilometer, and the sampling depth is 0 to 20 cm; the climate observation data comes from the measured data of the meteorological station and the meteorological satellite remote sensing data. Secondly, data quality control is carried out, and the 3 times standard deviation method is used to eliminate outliers, and the missing data is supplemented by the spatiotemporal interpolation method, and the interpolation method uses the Kriging interpolation method. Finally, data standardization is carried out, and the minimum and maximum standardization method is used to unify various types of data into the range of 0 to 1. The purpose of this step is to provide high-quality basic data support for subsequent analysis.

[0027] The specific implementation method of step S02 is to use the principal component analysis method to reduce the dimension and extract features of the basic data set. First, the basic data set is subjected to correlation analysis, the correlation coefficient matrix between the indicators is calculated, and the redundant indicators with correlation coefficients greater than 0.95 are eliminated. Secondly, the main eigenvectors are extracted using the principal component analysis method, and the principal components with cumulative contribution rates greater than 85% are selected as the feature matrix. Then, according to the load coefficients of each principal component, the original data is reconstructed into a land use type distribution matrix, a soil quality characteristic matrix, and a climate suitability matrix. Finally, the reconstructed feature matrix is ​​quality verified, and the data quality is ensured by calculating the reconstruction error, and the reconstruction error threshold is set to 0.15. The role of this step is to reduce the data dimension and extract key feature information.

[0028] The specific implementation of step S03 is to extract and analyze the characteristics of cultivated land patches based on the land use type distribution matrix. First, the regional growth algorithm is used to extract cultivated land patches, the growth threshold is set to 0.8, and the 8-neighborhood connectivity criterion is used for region merging. Secondly, the area parameters of each cultivated land patch are calculated using the grid counting method and converted to the actual area according to the spatial resolution. Then, the boundary tracking algorithm is used to calculate the boundary length parameters of the cultivated land patches, and the Freeman chain code method is used to extract boundary features. Finally, the fragmentation degree parameters are calculated using the landscape index method, including patch density index, boundary density index, and fractal dimension index, etc., and the fragmentation index is comprehensively calculated by the weighted average method, and the weight coefficients are determined by the analytic hierarchy process. The purpose of this step is to quantitatively describe the spatial characteristics of cultivated land patches.

[0029] The specific implementation of step S04 is to construct and apply a soil quality characteristic pyramid neural network model. First, the pyramid network structure is designed, and each layer uses a convolutional neural network to extract features. The size of the convolutional kernel increases from 3×3 to 11×11 layer by layer, and the max-pooling method is used in the pooling layer. Secondly, a feature integration module is designed, and the attention mechanism is used to adaptively fuse features at different levels, and the attention weights are optimized by the backpropagation algorithm. Then, the credibility threshold of soil quality characteristics is determined and set to 0.85 by the cross-validation method. Finally, the network model is trained, and the stochastic gradient descent method is used to optimize the network parameters. The initial value of the learning rate is set to 0.01, and the cosine annealing strategy is used for dynamic adjustment. The number of training epochs is set to 1000. The role of this step is to realize the multi-scale extraction and comprehensive analysis of soil quality characteristics.

[0030] The specific implementation of step S05 is to evaluate the climate condition index based on the climate suitability matrix. First, the light and heat resource condition index is calculated using the fuzzy comprehensive evaluation method. The annual average temperature, sunshine hours, cumulative temperature and other indicators are fuzzified, and the membership function uses the Gaussian function. Secondly, the moisture condition index is calculated using the drought index method, and a water balance model is constructed by combining the annual precipitation and the characteristics of precipitation seasonal distribution. Then, the climate stress factor index is calculated using the extreme value analysis method to evaluate the impact degree of extreme weather events on cultivated land. Finally, the entropy weight method is used to determine the weights of each index, and a comprehensive evaluation model is constructed. The purpose of this step is to evaluate the suitability of climate conditions for cultivated land.

[0031] The specific implementation manners of steps S06 and S07 are to construct and train a cultivated land protection potential evaluation model. First, a deep neural network is used to construct the evaluation model. The network structure includes an input layer, multiple hidden layers, and an output layer. The residual connection structure is adopted in the hidden layer to improve the model performance. Second, the batch normalization technique is used to standardize the input features, and the data augmentation technique is used to expand the training samples. Then, the cross-entropy loss function is used to measure the model performance, and the Adam optimizer is used for parameter optimization with a learning rate of 0.001. Finally, the early stopping method is adopted to prevent overfitting, and the training is stopped when the accuracy of the validation set no longer improves. The function of this step is to establish a prediction model for the cultivated land protection potential.

[0032] The specific implementation manners of steps S08 and S09 are to construct and solve the contribution degree optimization equations. First, a contribution degree benchmark equation is established, and multiple regression analysis is used to determine the basic contribution degrees of each index. Second, a spatial correlation equation is constructed, and the spatial autocorrelation analysis method is used to evaluate the spatial coupling effect among the indexes. Then, a temporal evolution equation is established, and the time series analysis method is used to study the dynamic change characteristics of the index contribution degrees. Finally, various contribution degrees are integrated through the weight balance equation, and the genetic algorithm is used to solve the optimization equations. The population size is set to 100, the number of evolutionary generations is 500, the crossover probability is 0.8, and the mutation probability is 0.1. The purpose of this step is to optimize the cultivated land protection potential score.

[0033] The specific implementation manners of steps S10 and S11 are to conduct cultivated land protection zoning and priority area identification. First, the natural breakpoint grading method is used to divide the optimized cultivated land protection potential score into three levels: high, medium, and low. The grading threshold is determined by the maximum inter-class variance method. Second, the spatial clustering analysis method is used to identify contiguous protection areas. The density clustering algorithm DBSCAN is adopted, with the core point neighborhood radius set to 500 meters and the minimum number of points set to 5. Then, the minimum spanning tree algorithm is used to analyze the connectivity among regions and construct a spatial connectivity network. Finally, the graph theory method is used to identify key nodes and determine the priority protection areas. The function of this step is to form a scientific and reasonable cultivated land protection spatial layout plan.

[0034] Among them, the soil quality feature pyramid neural network model realizes multi-scale feature analysis from local to global through 5 pyramid levels of different scales. Each layer of the pyramid expands the analysis scope on the basis of the previous layer, from a single cultivated land patch to the regional whole. This hierarchical analysis structure can fully capture the soil quality feature information at different spatial scales and avoid the limitations of traditional single-scale analysis. At the same time, through the setting of the credibility threshold, the feature synthesis module realizes the adaptive optimization of feature analysis, improving the calculation efficiency while ensuring the analysis quality.

[0035] Among them, the optimization equation system for the contribution degree of cultivated land protection constructs a complete analysis system including 4 interrelated equations. The basic contribution degree equation establishes a basic contribution degree evaluation framework through multiple evaluation indicators; the spatial correlation equation considers the spatial coupling relationship between indicators; the temporal evolution equation introduces dynamic analysis in the time dimension; and the weight balance equation realizes the comprehensive balance of multi-dimensional information. This analysis method combining multiple equations can comprehensively consider various influencing factors in the evaluation of cultivated land protection potential and provide more accurate and reliable evaluation results.

[0036] The pyramid neural network model of soil quality characteristics and the optimization equation system for the contribution degree of cultivated land protection, on the one hand, the pyramid model realizes the multi-scale adaptive analysis of soil quality characteristics and improves the accuracy and reliability of feature extraction; on the other hand, the equation system constructs a multi-dimensional and dynamic evaluation framework for cultivated land protection potential, providing a more scientific basis for cultivated land protection decision-making. This combination of technologies significantly improves the refinement level of cultivated land protection potential evaluation.

[0037] The overall implementation process of this method adopts a multi-level and multi-dimensional analysis framework, and through the combination of data-driven and model-driven methods, realizes the refined evaluation and optimization of cultivated land protection potential. In the data acquisition and preprocessing stage, multi-source remote sensing data fusion technology and geostatistical analysis methods are used to ensure the quality and reliability of basic data. In the feature extraction and analysis stage, deep learning and machine learning algorithms are used to realize the automatic extraction and multi-scale analysis of cultivated land features. In the evaluation model construction and optimization stage, intelligent optimization algorithms and spatial analysis methods are used to realize the scientific evaluation and spatial optimization of cultivated land protection potential.

[0038] The calculation of the output features of the convolutional layer of the pyramid neural network model of soil quality characteristics is expressed as follows: ; In the formula, is the output of the feature map of the th layer, where correspond to five pyramid layers respectively; is the input of the feature map of the th layer; is the weight of the convolutional kernel of the th layer; is the bias term; is the activation function, using the ReLU function; represents the convolution operation. When , represents the input soil quality characteristic matrix.

[0039] The calculation of the attention mechanism weight is expressed as follows: ; ; In the formula, is the attention weight of the th feature, ; is the attention score; is the hidden state; is the query vector; is the weight matrix, where is the hidden state weight matrix, is the query vector weight matrix; is the parameter vector, , is the number of features.

[0040] The contribution benchmark equation is expressed as follows: ; In the formula, is the basic contribution; is the normalized value of the th evaluation index; is the corresponding weight; is the index significance coefficient; is the historical protection effect coefficient; is the regional development constraint coefficient; is the regression coefficient; is the random error term; is the total number of evaluation indexes (including area parameters, boundary length parameters, fragmentation degree parameters, soil fertility index, soil physical structure index, soil chemical property index, light and heat resource condition index, water condition index, climate stress factor index).

[0041] The spatial correlation equation is expressed as follows: ; In the formula, is the spatial correlation contribution; is the element of the spatial weight matrix, indicating the spatial relationship between region and region ; is the current evaluation unit index; is the adjacent evaluation unit index, ; is the total number of adjacent evaluation units; is the spatial autocorrelation coefficient; is the terrain gradient coefficient; is the land use intensity index; is the landscape connectivity index; is the weight coefficient; is the random error term.

[0042] The time series evolution equation is expressed as follows: ; In the formula, is the contribution degree in the current period; is the contribution degree in the previous period; is the time decay coefficient; is the historical protection effectiveness index; is the climate change trend coefficient; is the social and economic development index; is the weight coefficient; is the random error term.

[0043] The weight balance equation is expressed as follows: ; In the formula, is the final contribution degree; is the comprehensive weight coefficient; is the regional differentiation correction coefficient; is the weight of the regional differentiation correction coefficient; is the random error term.

[0044] Description of the parameter acquisition method: 1. The parameters of the convolutional layer are optimized by the backpropagation algorithm. The initial value of the learning rate is 0.01, and the Adam optimizer is used; 2. The parameters of the attention mechanism are optimized by the gradient descent method, and the initial value is randomly initialized in the range of [-0.1, 0.1]; 3. In the basic contribution degree equation, the normalized value of the evaluation index is obtained by the min-max normalization method, the significance coefficient is calculated by the principal component analysis method, the historical protection effect coefficient is obtained by the time series analysis, and the regional development constraint coefficient is determined by the analytic hierarchy process; 4. In the spatial correlation equation, the spatial weight matrix is constructed by the inverse distance weighting method, the autocorrelation coefficient is calculated by the Moran's I index, the terrain gradient coefficient is calculated by the digital elevation model, and the land use intensity index and the landscape connectivity index are calculated by the landscape index method; 5. In the time series evolution equation, the value range of the time decay coefficient is [0.8, 0.95], the historical protection effectiveness index is calculated by the cumulative scoring method, the climate change trend coefficient is obtained by the Mann-Kendall test, and the social and economic development index is calculated by the entropy weight method; 6. In the weight balance equation, the comprehensive weight coefficient is determined by the combination of the analytic hierarchy process and the entropy weight method, and the regional differentiation correction coefficient is obtained by the cluster analysis; 7. Each error term Follow a normal distribution .

[0045] Explanation of the equation construction principle and significance: 1. The convolutional layer uses multi-scale convolutional kernels to extract features, introduces non-linearity through the ReLU activation function, and improves the model's expressive ability; 2. The attention mechanism adopts an additive attention model, normalizes the weights through the softmax function, and enhances the model's attention to important features; 3. The contribution benchmark equation considers three aspects: direct index impact, historical impact, and regional constraints, and adopts a linear weighted form for easy parameter estimation and interpretation; 4. The spatial correlation equation introduces a spatial lag term, considers spatial autocorrelation, and combines terrain, land use, and landscape features to reflect spatial heterogeneity; 5. The time series evolution equation adopts an autoregressive structure, introduces a time decay effect, and considers the influence of external factors such as climate and socioeconomic; 6. The weight balance equation realizes the fusion of multi-dimensional information through weighted combination, and introduces a correction coefficient to handle regional differences.

[0046] Among them, the derivation process of the convolutional layer feature extraction equation: First, based on the basic form of the traditional convolutional neural network , considering the multi-scale nature of soil features, a hierarchical structure is introduced to form ; To enhance the model's non-linear expressive ability, the ReLU activation function is introduced, and finally is obtained. This equation is optimized through the backpropagation algorithm, the loss function uses the mean squared error, and the optimization goal is to minimize the error between the predicted value and the true value. In this way, the model can adaptively extract soil feature information at different scales.

[0047] Among them, the derivation process of the attention mechanism weight calculation equation: Based on the traditional attention mechanism, first calculate the importance score of each feature , considering the correlation between features, introduce the query vector and the hidden state , and construct ; Then convert the score to a weight through the softmax function . Parameter optimization uses the gradient descent method and is achieved by minimizing the cross-entropy loss function. This design enables the model to automatically identify and emphasize important feature information.

[0048] Among them, the derivation process of the contribution benchmark equation: Starting from the basic relationship of the evaluation index , considering the influence of historical factors and regional restrictions, introduce the significance coefficient , historical effect coefficient and constraint coefficient , construct . Among them, the coefficient is obtained through multiple regression analysis, the significance coefficient is determined through principal component analysis, the historical effect coefficient is calculated based on time series analysis, and the constraint coefficient is determined through the analytic hierarchy process. This construction method can comprehensively reflect the comprehensive impact of evaluation indicators.

[0049] Among them, the derivation process of the spatial association equation: Based on the theory of spatial econometrics, introduce the spatial lag term , consider the influence of terrain, land use and landscape characteristics, and construct . The spatial weight matrix is constructed by the inverse distance weighting method, the autocorrelation coefficient is determined through spatial autocorrelation test, and each characteristic coefficient is estimated through a spatial regression model. This equation structure can effectively capture spatial dependence and heterogeneity.

[0050] Among them, the derivation process of the time series evolution equation: Adopt the autoregressive model framework , introduce external influencing factors, and construct . The time decay coefficient is determined through time series analysis, and the coefficients of each influencing factor are estimated through a panel data regression model. This equation can reflect the dynamic evolution characteristics of the contribution degree and the influence degree of external factors.

[0051] Among them, the derivation process of the weight balance equation: Based on the multi-criteria decision-making theory, adopt the weighted summation form , consider regional differences, introduce a correction term, and obtain . The weight coefficient is determined by combining the analytic hierarchy process and the entropy weight method, and the correction coefficient is obtained through cluster analysis. This structure can achieve the effective integration of multi-dimensional information and the reasonable adjustment of regional differences.

[0052] The second aspect of the present invention provides a computer-readable storage medium, in which program instructions are stored, and when the program instructions run on a computer, they are used to execute the above-mentioned fine-grained cultivated land protection potential optimization method.

[0053] The third aspect of the present invention provides a fine-grained cultivated land protection potential optimization system, including the above-mentioned computer-readable storage medium, the system is any one of a computer, a server, and a single-chip microcomputer, the computer-readable storage medium is arranged in the system, and a microprocessor for executing the program instructions stored in the computer-readable storage medium is arranged in the system.

[0054] Specifically, the principle of the present invention is: the technical principle of the present invention is based on deep learning and optimization theory. In terms of feature extraction, a deep neural network with a pyramid structure is used to achieve multi-scale feature extraction by expanding the receptive field layer by layer. Each layer contains convolution kernels of different sizes, which can capture feature information of different scales. By introducing the attention mechanism and residual structure, the model's ability to learn important features is enhanced, and the problem of deep network training difficulties is solved. The feature synthesis module realizes the effective fusion of features at different levels through an adaptive weight mechanism.

[0055] In terms of evaluation optimization, the constructed contribution optimization equation group includes four aspects: benchmark evaluation, spatial correlation, time series evolution and weight balance. The benchmark equation takes into account the direct impact of the evaluation indicators and establishes the relationship between the indicators and the contribution through multivariate regression analysis; the spatial correlation equation introduces spatial autocorrelation and heterogeneity analysis, and describes the spatial dependence between indicators through the spatial weight matrix; the time series evolution equation adopts an autoregressive structure to realize the dynamic evolution analysis of the contribution; the weight balance equation realizes the scientific integration of multi-dimensional information through multi-criteria decision-making theory.

[0056] The reason why the scheme of the present invention can solve the problem of accurate evaluation of cultivated land protection potential is mainly because: on the one hand, the pyramid neural network model overcomes the problem of insufficient feature extraction of traditional methods through multi-scale feature extraction and adaptive fusion; on the other hand, the contribution optimization equation group solves the problem of strong subjectivity and untimely update of evaluation results of traditional methods through multi-dimensional information fusion and dynamic optimization. The entire technical scheme forms a complete analysis and evaluation system, with clear logical relationships and mutual support between modules, and jointly achieves the goal of accurate evaluation of cultivated land protection potential.

[0057] A specific embodiment 1 of the present invention is provided below, and the specific implementation method of each step in this embodiment 1 is described in detail as follows.

[0058] The specific implementation of step S01 is data acquisition and preprocessing. First, the data source is selected. The remote sensing image data uses the Landsat 8 panchromatic band and multispectral band data, with spatial resolutions of 15 meters and 30 meters respectively. The hyperspectral data uses the Sentinel 2 data with a spatial resolution of 10 meters. The radar data uses the Sentinel 1 data. The remote sensing image is geometrically precisely corrected using the control point method. The number of control points is not less than 20, and the registration accuracy is better than 0.5 pixels. The radiation correction uses the radiation calibration equation: , where is a digital quantized value, is the radiance value, and are gain coefficient and offset respectively. Atmospheric correction uses the radiation transfer equation: , where is the radiance received by the sensor, is the radiance of the target itself, is the atmospheric transmittance, is the path radiance of the atmosphere. Soil monitoring data collection uses the stratified sampling method, with the sampling point layout density being 1 to 3 per square kilometer and the sampling depth being 0 to 20 cm.

[0059] The specific implementation of step S02 is multi-dimensional data decomposition. First, perform a correlation analysis on the basic data set and calculate the Pearson correlation coefficient: , where and are the observed values of two variables, and are the means. Screen and combine the indicators with a correlation coefficient greater than 0.95. Then, use the principal component analysis method for feature extraction and calculate the covariance matrix: , where is the sample matrix, is the sample mean. Solve the characteristic equation: , and obtain the eigenvalues and eigenvectors. Select the principal components with a cumulative contribution rate greater than 85% to reconstruct the feature matrix.

[0060] The specific implementation of step S03 is the extraction of cultivated land patch characteristics. First, use the region growing algorithm to extract cultivated land patches. The seed point selection criterion is gray value similarity, and the growth criterion is: , where is the pixel gray value, is the regional mean, is the standard deviation, is the growth threshold, with a value of 2.0. Use the 8-neighborhood connectivity criterion for region merging. Calculate the patch area parameter: , where is the actual area of the pixel. The boundary length is calculated using the Freeman chain code method: , where is the chain code unit length. Calculate the fragmentation index: , where is the area, is the perimeter.

[0061] The specific implementation of step S04 is the analysis of soil quality characteristics. The convolutional feature extraction for each pyramid layer uses the aforementioned convolutional equation: . The size of the convolutional kernel increases layer by layer from 3×3 in the first layer to 11×11 in the fifth layer. In the feature integration module, the attention weight calculation uses the aforementioned equations: and The credibility threshold of soil quality characteristics is determined to be 0.85 through cross-validation. Calculation of the credibility of the current layer characteristics: , where is the true value, is the predicted value.

[0062] The specific implementation of step S05 is climate condition assessment. The light and heat resource condition index adopts the fuzzy comprehensive evaluation method to construct an evaluation matrix: , where is the membership degree of the th evaluation object to the th index. The membership degree function adopts the Gaussian function: , where is the central value, is the standard deviation. The moisture condition index is calculated based on the drought index: , where is the precipitation, is the potential evaporation. The climate stress factor index is calculated through extreme value analysis: , where is the weight coefficient.

[0063] The specific implementation of steps S06 and S07 is evaluation model construction and training. The deep neural network structure includes an input layer with the number of nodes equal to the feature dimension, and the hidden layer adopts a residual structure: , where is the residual mapping. Batch normalization adopts: , where is the batch mean, is the batch variance, is the smoothing term. The cross-entropy loss function: , where is the true label, is the predicted probability.

[0064] The specific implementation of steps S08 and S09 is contribution degree evaluation and optimization. The contribution degree benchmark equation, spatial correlation equation, temporal evolution equation, and weight balance equation adopt the aforementioned formulas. The system of equations is solved using the genetic algorithm. The chromosome encoding adopts real number encoding, the population size is 100, and the number of evolutionary generations is 500. The selection operator adopts the roulette wheel method, and the crossover operator adopts arithmetic crossover: , where is the crossover coefficient, with a value of 0.8. The mutation operator adopts Gaussian mutation: , where is the mutation step size, with a value of 0.1. The fitness function is the reciprocal of the mean square error: .

[0065] The specific implementation manners of steps S10 and S11 are partition scheme generation and optimization. The natural break classification method is used to determine the partition threshold: , where is the th level, is the sample value, is the within-class mean. The DBSCAN algorithm is used for spatial clustering, and the core point determination criterion: , where is the point set within the neighborhood of point , takes the value of 500 meters, takes the value of 5. The minimum spanning tree algorithm is used to analyze the regional connectivity: , where is the edge weight.

[0066] Furthermore, the specific implementation manner of step S04 in this embodiment is to construct and apply a soil quality feature pyramid neural network model. The soil quality feature pyramid neural network model includes 5 sequentially executed pyramid layers and 1 feature integration module. Each pyramid layer adopts a deep convolutional neural network structure and is constructed according to the multi-scale feature analysis idea from local to global.

[0067] The first pyramid layer is used to extract the soil quality features inside the cultivated land patches. The network structure includes 4 convolutional layers and 3 pooling layers. Among them, the first and second convolutional layers use 3×3 convolutional kernels, the padding method is SAME, the stride is 1, and the output channels are 64 and 128 respectively; the third and fourth convolutional layers use 5×5 convolutional kernels, the padding method is SAME, the stride is 1, and the output channels are 256 and 512 respectively; the pooling layers all use 2×2 max pooling with a stride of 2. The input data of the first pyramid layer are the soil organic matter content data, soil bulk density data, soil pH value data, soil nutrient content data, soil water content data, soil texture data, and soil erosion degree data within the range of the cultivated land patches, and are input through the input layer with a feature map size of 64×64. This layer uses a batch normalization layer and a ReLU activation function to achieve the standardization processing and non-linear mapping of features.

[0068] The second pyramid layer is used to extract the combination of soil quality characteristics inside the cultivated land patches and in the adjacent areas of the cultivated land patches. The network structure consists of 4 convolutional layers and 3 pooling layers. Among them, the first and second convolutional layers use a 5×5 convolutional kernel, the padding method is SAME, the stride is 1, and the number of output channels is 128 and 256 respectively; the third and fourth convolutional layers use a 7×7 convolutional kernel, the padding method is SAME, the stride is 1, and the number of output channels is 512 and 1024 respectively; all the pooling layers use 2×2 max pooling with a stride of 2. The input data of the second pyramid layer includes the output feature map of the first pyramid layer and the soil quality data of the adjacent area, and the size of the feature map is enlarged to 128×128. This layer introduces an attention mechanism to highlight key feature information by calculating the importance weights of the features.

[0069] The third pyramid layer is used to extract the combination of soil quality characteristics inside the cultivated land patches, in the adjacent areas of the cultivated land patches, and in the buffer zones around the cultivated land patches. The network structure consists of 4 convolutional layers and 3 pooling layers. Among them, the first and second convolutional layers use a 7×7 convolutional kernel, the padding method is SAME, the stride is 1, and the number of output channels is 256 and 512 respectively; the third and fourth convolutional layers use a 9×9 convolutional kernel, the padding method is SAME, the stride is 1, and the number of output channels is 1024 and 2048 respectively; all the pooling layers use 2×2 max pooling with a stride of 2. The input data of the third pyramid layer includes the output feature map of the second pyramid layer and the soil quality data of the buffer zone, and the size of the feature map is enlarged to 256×256. This layer adopts a residual connection structure to alleviate the problem of difficult training of deep neural networks.

[0070] The fourth pyramid layer is used to extract the combination of soil quality characteristics inside the cultivated land patches, in the adjacent areas of the cultivated land patches, in the buffer zones around the cultivated land patches, and in the watershed units of the cultivated land patches. The network structure consists of 4 convolutional layers and 3 pooling layers. Among them, the first and second convolutional layers use a 9×9 convolutional kernel, the padding method is SAME, the stride is 1, and the number of output channels is 512 and 1024 respectively; the third and fourth convolutional layers use an 11×11 convolutional kernel, the padding method is SAME, the stride is 1, and the number of output channels is 2048 and 4096 respectively; all the pooling layers use 2×2 max pooling with a stride of 2. The input data of the fourth pyramid layer includes the output feature map of the third pyramid layer and the soil quality data of the watershed unit, and the size of the feature map is enlarged to 512×512. This layer introduces a channel attention mechanism to enhance the information interaction between feature channels.

[0071] The fifth pyramid layer is used to extract the combination of soil quality characteristics within cultivated land patches, adjacent areas of cultivated land patches, buffer zones around cultivated land patches, watershed units of cultivated land patches, and the overall region. The network structure includes 4 convolutional layers and 3 pooling layers. Among them, the first and second convolutional layers use an 11×11 convolutional kernel, the padding method is SAME, the stride is 1, and the output channels are 1024 and 2048 respectively; the third and fourth convolutional layers use a 13×13 convolutional kernel, the padding method is SAME, the stride is 1, and the output channels are 4096 and 8192 respectively; all the pooling layers use 2×2 max pooling with a stride of 2. The input data of the fifth pyramid layer includes the output feature map of the fourth pyramid layer and the soil quality data of the overall region, and the size of the feature map is expanded to 1024×1024. This layer uses a global average pooling layer to achieve the extraction of global information of features.

[0072] The feature integration module includes a feature fusion unit and a judgment unit. The feature fusion unit uses an adaptive weighting method to fuse the features output by different pyramid layers, and the weight coefficients are optimized through the backpropagation algorithm. The judgment unit judges whether the output of the current pyramid layer meets the accuracy requirements by calculating the feature credibility. When the feature credibility is greater than the threshold of 0.85, the output of the current layer is used as the final feature; when the feature credibility is less than the threshold, the next layer is continued to be analyzed or the output of the last layer is used as the final feature. The calculation of feature credibility uses the cross-validation method and is determined by comparing the prediction results with the true values.

[0073] The innovation points of this pyramid neural network model are as follows: adopting a multi-scale feature extraction strategy to achieve a progressive analysis from local to global; introducing an attention mechanism and a residual structure to improve the accuracy of feature extraction; realizing adaptive feature selection through the feature integration module to ensure the reliability of the analysis results. The model is trained using a stochastic gradient descent optimizer with an initial learning rate of 0.01, and the cosine annealing strategy is used for dynamic adjustment, and the number of training epochs is 1000. In each epoch, the mini-batch stochastic gradient descent method with a batch size of 32 is used to update the network parameters through the backpropagation algorithm. To prevent overfitting, the L2 regularization method with a weight decay coefficient of 0.0001 is used, and at the same time, a Dropout layer with a random dropout rate of 0.5 is used. This model can effectively extract multi-scale soil quality features and provide reliable feature support for the evaluation of the potential for cultivated land protection.

[0074] The above method realizes the refined evaluation of the potential for cultivated land protection through a multi-level and multi-dimensional analysis framework. At the data processing level, multi-source data fusion and feature extraction technologies are adopted to ensure the reliability of the basic data. At the model construction level, through deep learning and optimization algorithms, automatic feature extraction and optimized training of the model are achieved. At the evaluation optimization level, a complete contribution evaluation system is constructed to achieve the comprehensive balance of multi-dimensional information. The innovation of this method lies in establishing an adaptive and dynamic evaluation framework, which improves the accuracy and reliability of the evaluation of the potential for cultivated land protection.

[0075] To better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below:

[0076] The research area is located between 35 degrees north latitude and 36 degrees north latitude, and between 115 degrees east longitude and 116 degrees east longitude, with a total area of approximately 8,000 square kilometers, of which the cultivated land area is approximately 4,500 square kilometers. The terrain of this area is complex, including plains, hills and mountains, with an altitude ranging from 10 to 800 meters. The climate type belongs to the warm temperate semi-humid monsoon climate, with an average annual temperature of 14.2 degrees Celsius and an annual precipitation of 850 millimeters.

[0077] First, the research team established a multi-source data acquisition system. Remote sensing data acquisition includes: panchromatic images (15-meter resolution) and multispectral images (30-meter resolution) obtained by Landsat 8 on May 1, 2024, multi-temporal multispectral data (10-meter resolution) obtained by Sentinel 2 on April 28 and May 3, 2024, and SAR data obtained by Sentinel 1 on May 2, 2024. The soil monitoring data acquisition designed a systematic stratified sampling scheme: 100 standard plots of 1 square kilometer were established in the research area, 25 sampling points were set in each standard plot, and the sampling depth was divided into two layers of 0-20 cm and 20-40 cm, with a total of 5,000 samples. The climate data is sourced from the continuous observation records of 25 meteorological stations in the research area from 2019 to 2024, and 50 automatic weather stations were deployed for intensive observation.

[0078] The analysis results of the first batch of soil monitoring data are shown in Table 1 below: Table 1 Analysis results of the first batch of soil monitoring data

[0079] The specific implementation process of the soil quality characteristic pyramid neural network model is shown in Table 2 below: Table 2 Standardization results of the input data of the first pyramid layer:

[0080] Specific parameter settings and training effects of each layer of the pyramid network: The first layer uses 64 3×3 convolutional kernels, the activation function is ReLU, the pooling uses 2×2 max pooling, and the parameters of the batch normalization layer are = 0.02, = 0.015; The second layer uses 128 5×5 convolutional kernels, and other parameter settings are the same; The third layer uses 256 7×7 convolutional kernels; The fourth layer uses 512 9×9 convolutional kernels; The fifth layer uses 1024 11×11 convolutional kernels. The feature extraction effect of each layer is evaluated through credibility: The average credibility of the first layer is 0.832, the second layer is 0.856, the third layer is 0.878, the fourth layer is 0.891, and the fifth layer is 0.902.

[0081] Calculation results of the attention weights of the feature integration module: = 0.15 (local features), = 0.20 (proximal features), = 0.25 (buffer features), = 0.22 (watershed features), = 0.18 (regional features). The weight optimization uses the Adam algorithm, the initial learning rate is set to 0.01, and it decays to 0.8 times the original every 200 rounds.

[0082] Spatial distribution characteristics of the final evaluation results: The high protection potential area is about 1800 square kilometers (accounting for 40%), mainly distributed in the plain area, the average value of the soil quality characteristic index is 0.856, and the average value of the climate suitability index is 0.823; The medium protection potential area is about 2025 square kilometers (accounting for 45%), mainly distributed in the hilly area, the average value of the soil quality characteristic index is 0.685, and the average value of the climate suitability index is 0.742; The low protection potential area is about 675 square kilometers (accounting for 15%), mainly distributed in the mountainous area, the average value of the soil quality characteristic index is 0.523, and the average value of the climate suitability index is 0.612.

[0083] Compared with the traditional evaluation method, the technical solution adopted in this embodiment has significant advantages: first, the traditional method uses the expert scoring method to determine the weight, which is highly subjective and the evaluation results fluctuate greatly, while this solution automatically extracts features through the pyramid neural network, which is highly objective and the evaluation results are stable; second, the traditional method can only process limited evaluation indicators, while this solution can simultaneously process multi-source, multi-dimensional, and multi-scale data, and the information utilization is more sufficient; third, the traditional method lacks spatial correlation analysis, while this solution fully considers spatial dependence and heterogeneity through multi-level spatial analysis; finally, the evaluation results of the traditional method are static and difficult to update, while this solution can update the evaluation results in real time through a dynamic optimization mechanism. Compared with the expert evaluation results, the evaluation accuracy of this solution has increased by 23.5%, the spatial precision has increased by 35.8%, and the time response speed has increased by 68.2%.

[0084] This example fully verifies the feasibility and superiority of this technical solution in the refined evaluation of cultivated land protection potential, and provides a scientific basis for cultivated land protection decision-making. Adaptive feature extraction is achieved through the soil quality feature pyramid neural network model, and dynamic optimization of evaluation results is achieved through the contribution optimization equation group, which ultimately solves the technical problem of accurate evaluation of cultivated land protection potential.

[0085] It should be noted that the variables involved in the present invention are explained in detail as shown in Table 3 below.

[0086] Table 3 Variable explanation table

[0087] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A fine-grained cultivated land protection potential optimization method, characterized in that Including: Obtaining remote sensing image data, soil monitoring data, and climate observation data of the target area as the basic data set; performing multi-dimensional decomposition on the basic data set; extracting cultivated land patches based on the land use type distribution matrix; analyzing the soil quality characteristic matrix using the soil quality characteristic pyramid neural network model; evaluating the cultivated land patch condition index according to the climate suitability matrix; constructing a cultivated land protection potential evaluation model; training with a neural network model to obtain an initial score; establishing a cultivated land protection contribution matrix; optimizing and adjusting to obtain an optimized score; Dividing high, medium, and low protection potential areas; performing spatial connectivity analysis, identifying priority protection areas, and outputting.

2. The method according to claim 1, characterized in that, The remote sensing image data includes multi-spectral remote sensing image data, hyperspectral remote sensing image data, radar remote sensing image data, land cover classification data, and normalized difference vegetation index data.

3. The method according to claim 2, characterized in that, The soil monitoring data includes soil organic matter content data, soil bulk density data, soil pH value data, soil nutrient content data, soil water content data, soil texture data, and soil erosion degree data.

4. The method according to claim 3, characterized in that, The climate observation data includes annual average temperature data, annual precipitation data, sunshine hours data, accumulated temperature data, drought index data, precipitation seasonal distribution data, and extreme weather event data.

5. The method according to claim 4, wherein Performing multi-dimensional decomposition on the basic data set to obtain a land use type distribution matrix, a soil quality characteristic matrix, and a climate suitability matrix.

6. The method according to claim 5, characterized in that, Extracting cultivated land patches based on the land use type distribution matrix, and calculating the area parameter, boundary length parameter, and fragmentation degree parameter of the cultivated land patches.

7. The method according to claim 6, wherein The soil quality characteristic pyramid neural network model includes five sequentially executed pyramid layers and a feature integration module.

8. The method according to claim 7, wherein The cultivated land protection contribution matrix is obtained by solving a contribution optimization equation set, which includes a contribution benchmark equation, a spatial correlation equation, a temporal evolution equation, and a weight balance equation. The contribution benchmark equation is used to calculate the basic contribution of each evaluation index, the spatial correlation equation is used to evaluate the spatial coupling effect between evaluation indexes, and the temporal evolution equation is used to analyze the temporal change characteristics of the contribution of evaluation indexes.

9. A computer-readable storage medium, characterized in that, Program instructions are stored in the computer-readable storage medium. When the program instructions run on a computer, they are used to execute a fine-grained cultivated land protection potential optimization method according to any one of claims 1-8.

10. A fine-grained cultivated land protection potential optimization system, characterized in that, Including the computer-readable storage medium according to claim 9, the system is any one of a computer, a server, and a single-chip microcomputer. The computer-readable storage medium is set inside the system, and a microprocessor for executing the program instructions stored in the computer-readable storage medium is set inside the system.

Images