Cultivated land space layout optimization management system and method based on multi-source data fusion

Abstract

The invention discloses a cultivated land space layout optimization management system and method based on multi-source data fusion, and relates to the technical field of cultivated land space optimizing.The method comprises the steps that cultivated land pattern spot data, irrigation water source data and ecological protection elements in an evaluation area are obtained; basic characteristics of each cultivated land pattern spot are analyzed; collecting cultivated land change data in historical periods, and constructing a training data set containing stable cultivated land and variable cultivated land; analyzing the weight of each basic feature based on a stable fitting model; performing weighted summation based on each processed real-time basic feature and the weight of each basic feature, and calculating an optimized comprehensive score of each cultivated land pattern spot; determining a grading threshold of the optimized comprehensive score according to a threshold analysis method selected by an administrator; and dividing the cultivated land pattern spots according to the grading threshold, generating an optimization attribute list, and sending the optimization attribute list to an administrator. The uncertainty of subjective empowerment is avoided; and the accuracy of system analysis is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cultivated land space optimization, and in particular to a cultivated land space layout optimization management system and method based on multi-source data fusion. BACKGROUND

[0002] Cultivated land is the fundamental guarantee of national food security and important agricultural product supply, and is a strategic resource supporting sustainable economic and social development. Scientifically optimizing cultivated land space layout is an important link to coordinate the relationship between food security and economic and social development and ecological protection, promote the overall revitalization of rural areas, and is an inevitable requirement to implement the most strict cultivated land protection system, improve the comprehensive productivity of cultivated land, and adapt to the development of agricultural modernization.

[0003] Existing technologies rely on qualitative analysis and expert experience decision-making, and lack objective and quantitative decision-making support systems. When determining the importance of different characteristics, the weight is usually determined by using methods such as analytic hierarchy process that rely on expert subjective scoring. This method is easily affected by personal preferences and knowledge limitations of experts, resulting in a strong subjective color in weight distribution, and the conclusions obtained by different regions or different expert teams differ greatly, making it difficult to ensure the scientificity and universality of the planning scheme.

[0004] Therefore, the present application discloses a cultivated land space layout optimization management system and method based on multi-source data fusion to solve the above problems. SUMMARY

[0005] The purpose of the present application is to provide a cultivated land space layout optimization management system and method based on multi-source data fusion to solve the problems in the prior art.

[0006] To achieve the above purpose, the present application provides the following technical scheme: a cultivated land space layout optimization management method based on multi-source data fusion, which comprises the following steps: S1: obtaining cultivated land patch data, irrigation water source data and ecological protection elements in the evaluation area; analyzing the basic characteristics of each cultivated land patch; S2: collecting historical cultivated land change data, constructing a training data set containing stable cultivated land and changing cultivated land; analyzing the weight of each basic characteristic based on a stable fitting model; S3: based on the processed real-time basic characteristics and the weight of each basic characteristic, performing weighted summation to calculate the optimized comprehensive score of each cultivated land patch; S4: determining the classification threshold of the optimized comprehensive score according to the threshold analysis method selected by the administrator; dividing the cultivated land patches according to the classification threshold to generate an optimized attribute list and send it to the administrator.

[0007] According to the above scheme, in S1, the following contents are included: S101: Obtain cultivated land patch data in an evaluation area, the cultivated land patch data including geometric shapes, spatial positions, cultivated land areas, and cultivated land quality values of the cultivated land patches; obtain irrigation water source data and ecological protection elements; the irrigation water source elements include canal systems, river channels, water distribution outlets, and reservoirs; the irrigation water source data includes spatial positions of the irrigation water sources; and the ecological protection elements include ecological protection red lines; S102: Analyze basic characteristics of each cultivated land patch. For each cultivated land patch, all cultivated land patches with a cultivated land patch distance less than a search distance threshold are extracted to form a potential contiguous cultivated land patch set of each cultivated land patch; the cultivated land patch distance is equal to the shortest Euclidean distance between polygon boundaries of two cultivated land patches, and the cultivated land patch distance is equal to zero if the polygon boundaries of the two cultivated land patches intersect or contact; the cultivated land areas of all cultivated land patches in the contiguous cultivated land patch set are summed to obtain a contiguous expansion potential; and the contiguous expansion potentials of all cultivated land patches are normalized. The cultivated land quality values of all cultivated land patches in the evaluation area are extracted, the cultivated land quality values are normalized, and the difference between one and the normalized cultivated land quality value is taken as a quality improvement potential. The optimal irrigation distances of each cultivated land patch to the nearest irrigation water source are analyzed; the optimal irrigation distance is equal to the shortest Euclidean distance between the polygon boundary of the cultivated land patch and the irrigation water source; the optimal irrigation distances are normalized, and the difference between one and the normalized optimal irrigation distance is taken as an irrigation guarantee potential. For each cultivated land patch, the overlapping relationship with the ecological protection red line is determined by spatial overlay analysis; if the cultivated land patch overlaps with the ecological protection red line, the ecological restriction strength of the cultivated land patch is taken as zero; otherwise, the ecological restriction strength of the cultivated land patch is taken as one.

[0008] The present application comprehensively uses multi-source data such as cultivated land patches, irrigation water sources, and ecological protection elements, ensures the integrity and representativeness of feature extraction, and provides reliable input for subsequent analysis; the abstract spatial relationship is converted into a calculable feature by quantifying the contiguous expansion potential and the quality improvement potential, thereby enhancing the operability of the method; the feature values are normalized to eliminate the influence of dimensions, so that the comparison and fusion of different features are more fair and accurate.

[0009] According to the above scheme, in S2, the following content is included: S201: Extract cultivated land change data in a preset historical time window, identify cultivated land patches that have always remained as high-quality cultivated land in the preset historical time window as positive samples, and identify cultivated land patches that have been abandoned, converted into construction land, or severely degraded as negative samples; meanwhile, extract basic characteristic values of the cultivated land patches at the beginning of the preset time window to form a training data set, denoted as Dtrian = { (X i , Y i ) | i ∈ [1, I]}; wherein, I represents the total number of cultivated land patches in the training data set; X i represents the feature vector of the i cultivated land patch in the training data set; X i = (contig i , qual i , irrig i , eco i ) ; wherein contig i represents the contiguous expansion potential of the i cultivated land patch; qual i represents the quality improvement potential of the i cultivated land patch; irrig i represents the irrigation guarantee potential of the i cultivated land patch; eco i represents the ecological restriction intensity of the i cultivated land patch; Y i represents the label value of the i cultivated land patch, wherein the label value of the positive sample is 1; the label value of the negative sample is 0; S202: based on the training data set, a stable fitting model is constructed, and the specific formula of the stable fitting model is: Log (p / (1-p)) = β0+β1×contig i +β2×qual i +β3×irrig i +β4×eco i ; wherein, p represents the predicted label value of the cultivated land patch, β0represents the intercept term, β1, β2, β3and β4represent the fitting coefficients; the stable fitting model is trained using the maximum likelihood estimation method, and the final value of the fitting coefficient is generated by the optimization algorithm, so that the difference between the predicted label value and the true label value based on the feature vector of the stable fitting model is minimized; S203: the absolute value of the final value of the fitting coefficient is normalized, and the weight of each basic feature is analyzed, wherein the weight of each basic feature is equal to the ratio of the normalized value of the absolute value of the corresponding final value of the fitting coefficient to the sum of the normalized values of the absolute values of the final values of the fitting coefficients.

[0010] The present application trains the model by using historical cultivated land stability data, automatically determines the feature weight by the fitting coefficient, avoids the uncertainty of subjective weighting, improves the scientificity and repeatability of the weight, and allows the model to be retrained with the update of new data, so that the weight can adapt to the regional conditions and time changes, and has dynamic adjustment capability.

[0011] According to the above scheme, in S3, the following contents are included: S301: Obtain real-time basic features of each cultivated land patch, and perform forward transformation and extreme value standardization processing on each real-time basic feature of the cultivated land patch; S302: Based on the processed each real-time basic feature and the weight of each basic feature, weighted summation is performed, and the optimized comprehensive score of each cultivated land patch is calculated, and the optimized comprehensive score of the jth cultivated land patch is recorded as LOSS j ; LOSS j = W c *contig j + W q *qual j + W i *irrig j + W e *eco j ; Wherein W c , W q , W i and W e are the weights of the contiguous expansion potential, the quality improvement potential, the irrigation guarantee potential and the ecological restriction intensity; contig j , qual j , irrig j and eco j are the contiguous expansion potential, the quality improvement potential, the irrigation guarantee potential and the ecological restriction intensity of the jth cultivated land patch.

[0012] The present application simplifies the complex decision-making process by fusing multiple features into a single score through weighted summation; the weights come from historical data, but the score is applied to current data, so that the method can inherit historical rules and respond to current changes.

[0013] According to the above scheme, in S4, the following contents are included: S401: Statistics of the distribution of the optimized comprehensive score, and calculation of statistical quantities, including mean, median, quantile and sample variance; determine the first threshold and the second threshold according to the threshold value analysis method selected by the administrator; if the target area method is used, record the target total area used and the cumulative logic; if the expert decision method is used, record the expert weight and the decision basis; If the optimized comprehensive score is greater than or equal to the first threshold, the corresponding cultivated land patch is marked as a candidate for priority optimization area; if the optimized comprehensive score is less than the first threshold and greater than or equal to the second threshold, the corresponding cultivated land patch is marked as a general optimization area; if the optimized comprehensive score is less than the second threshold, the corresponding cultivated land patch is marked as a status maintenance area; if all the optimized comprehensive scores are less than the first threshold, a certain percentage selected by the quantile method is used as a backup; if all the optimized comprehensive scores are greater than or equal to the first threshold, the target area method is used to limit the total area of the candidate; S402: Extract the number, spatial position and cultivated land area of each marker as a priority optimization area candidate; integrate into an optimization attribute list and send to the administrator.

[0014] The application provides various threshold analysis methods (such as a target area method and an expert decision method), which are suitable for different management requirements, enhance the practicability and adaptability of the method, and ensure that the partition scheme is always feasible when the score distribution is abnormal (such as being higher or lower than the threshold value) by using an alternative method (such as a quantile method), thereby improving the robustness of the method and the accuracy of system analysis.

[0015] In another aspect of the application, a cultivated land spatial layout optimization management system based on multi-source data fusion is provided, which is applied to the above-mentioned cultivated land spatial layout optimization management method based on multi-source data fusion and comprises a basic feature analysis module, a feature weight analysis module, an optimization comprehensive score analysis module and a cultivated land division module. The basic feature analysis module is used to obtain cultivated land patch data, irrigation water source data and ecological protection elements in an evaluation area; and the basic features of each cultivated land patch are analyzed. The feature weight analysis module is used to collect cultivated land change data in a historical period, construct a training data set containing stable cultivated land and changed cultivated land; and the weights of the basic features are analyzed based on a stable fitting model. The optimization comprehensive score analysis module is used to perform weighted summation based on the processed real-time basic features and the weights of the basic features, and calculate the optimization comprehensive scores of the cultivated land patches. The cultivated land division module is used to determine the classification threshold of the optimization comprehensive scores according to the threshold analysis method selected by the administrator; and the cultivated land patches are divided according to the classification threshold, and an optimization attribute list is generated and sent to the administrator.

[0016] According to the above scheme, the basic feature analysis module comprises a cultivated land data acquisition unit and a feature analysis unit. The cultivated land data acquisition unit is used to obtain cultivated land patch data in an evaluation area, obtain an irrigation water source set and ecological protection elements. The feature analysis unit is used to analyze the basic features of each cultivated land patch; and the basic features include contiguous expansion potential, quality improvement potential, irrigation guarantee potential and ecological restriction intensity.

[0017] According to the above scheme, the feature weight analysis module comprises a sample acquisition unit, a model fitting unit and a weight analysis unit. The sample collection unit is configured to extract the cultivated land change data in a preset historical time window, identify a cultivated land patch that has always remained as high-quality cultivated land in the preset historical time window as a positive sample, and identify a cultivated land patch that has been abandoned, converted into construction land, or seriously degraded as a negative sample; meanwhile, the sample collection unit is configured to extract the basic feature values corresponding to the cultivated land patch at the beginning of the preset time window to form a training data set. The model fitting unit is configured to train the stable fitting model using the training data set, train the stable fitting model using a maximum likelihood estimation method, and generate a final value of a fitting coefficient through an optimization algorithm, so that the stable fitting model has a minimum difference between a predicted label value and an actual label value based on a feature vector. The weight analysis unit is configured to normalize the absolute value of the final value of the fitting coefficient and analyze the weight of each basic feature.

[0018] According to the above scheme, the optimization comprehensive score analysis module includes a real-time data processing unit and a score analysis unit. The real-time data processing unit is configured to obtain real-time basic features of each cultivated land patch, and perform positive transformation and extreme value standardization on each real-time basic feature of the cultivated land patch. The score analysis unit is configured to perform weighted summation based on the processed real-time basic features and the weight of each basic feature, and calculate the optimization comprehensive score of each cultivated land patch.

[0019] According to the above scheme, the cultivated land division module includes a cultivated land division unit and a list integration unit. The cultivated land division unit is configured to determine a first threshold value and a second threshold value according to an administrator-selected threshold analysis method, and divide the cultivated land patch based on the first threshold value and the second threshold value. The list integration unit is configured to extract the number, spatial position, and cultivated land area of each cultivated land patch marked as a candidate for a priority optimization area, and integrate the extracted information into an optimization attribute list and send the optimization attribute list to the administrator.

[0020] Compared with the prior art, the beneficial effects of the present application are: the present application integrates multiple source data such as cultivated land patches, irrigation water sources and ecological protection elements, ensures the integrity and representativeness of feature extraction, and provides reliable input for subsequent analysis; through the quantification of indicators such as contiguous expansion potential and quality improvement potential, the abstract spatial relationship is converted into a calculable feature, enhancing the operability of the method; the feature values are normalized, eliminating the influence of dimension, making the comparison and fusion of different features more fair and accurate; the present application uses historical cultivated land stability data to train the model, automatically determines the feature weight through the fitting coefficient, avoids the uncertainty of subjective weighting, and improves the scientificity and repeatability of the weight; the method allows the model to be retrained as new data is updated, so that the weight can adapt to regional conditions and time changes, and has dynamic adjustment capability; the present application fuses multiple features into a single score through weighted summation, simplifying the complex decision-making process; the weight comes from historical data, but the score is applied to current data, so that the method can inherit historical laws and respond to current changes; the present application provides multiple threshold analysis methods to adapt to different management needs, enhancing the practicality and adaptability of the method; when the score distribution is abnormal, the alternative method is used to ensure that the partition scheme is always feasible, improving the robustness of the method and the accuracy of system analysis. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. In the drawings: Figure 1 It is a flowchart of a cultivated land spatial layout optimization management method based on multi-source data fusion according to the present application; Figure 2 It is a structural schematic diagram of a cultivated land spatial layout optimization management system based on multi-source data fusion according to the present application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0023] Please refer to Figure 1 The present application provides a technical solution: a cultivated land spatial layout optimization management method based on multi-source data fusion, which comprises the following steps: S1: obtaining cultivated land patch data, irrigation water source data and ecological protection elements in the evaluation area; analyzing the basic characteristics of each cultivated land patch; S1 includes the following: S101: Obtain farmland patch data within the evaluation area, including the geometric shape, spatial location, farmland area, and farmland quality grade of the farmland patches; obtain irrigation water source data and ecological protection elements; irrigation water source elements include canals, rivers, water outlets, and reservoirs; irrigation water source data includes the spatial location of each irrigation water source; ecological protection elements include ecological protection red lines. S102: Analyze the basic characteristics of each cultivated land patch; For each cultivated land patch, extract all cultivated land patches whose spacing is less than the search distance threshold to form a potential contiguous cultivated land patch set for each cultivated land patch; the spacing between cultivated land patches is equal to the shortest Euclidean distance between the polygonal boundaries of two cultivated land patches, and the spacing between cultivated land patches is zero if the polygonal boundaries of two cultivated land patches intersect or contact; sum the cultivated land areas of all cultivated land patches in the contiguous cultivated land patch set, and record it as the contiguous expansion potential; normalize the contiguous expansion potential of all cultivated land patches. Extract the farmland quality grade values ​​of all farmland patches within the evaluation area, normalize the farmland quality grade values, and record the difference between the normalized farmland quality grade value and the normalized farmland quality grade value as the potential for quality improvement. Analyze the optimal irrigation distance from each cultivated plot to the nearest irrigation source; the optimal irrigation distance is equal to the shortest Euclidean distance between the polygonal boundary of the cultivated plot and the irrigation source; normalize the optimal irrigation distance, and record the difference between the normalized and the normalized optimal irrigation distance as the irrigation guarantee potential. For each cultivated land parcel, spatial overlay analysis is used to determine its overlap with the ecological protection red line. If the cultivated land parcel overlaps with the ecological protection red line, the ecological restriction intensity of the cultivated land parcel is recorded as zero; otherwise, the ecological restriction intensity of the cultivated land parcel is recorded as one.

[0024] S2: Collect historical data on changes in cultivated land and construct a training dataset containing both stable and changed cultivated land; analyze the weights of each basic feature based on a stable fitting model. S2 includes the following: S201: Extract farmland change data within a preset historical time window, identify farmland patches that have consistently remained high-quality farmland within the preset historical time window as positive samples, and farmland patches that have been abandoned, converted to construction land, or severely degraded as negative samples; simultaneously, extract the basic feature values ​​corresponding to the farmland patches at the beginning of the preset time window to form a training dataset, denoted as D. trian ={(X i Y i )|i∈[1,I]}; where I represents the total number of cultivated map patches in the training dataset; X iXi represents a feature vector of the i-th cultivated land patch in the training data set; X i = (contig i , qual i , irrig i , eco i ); wherein contig i represents the contiguously expanded potential of the i-th cultivated land patch; qual i represents the quality improvement potential of the i-th cultivated land patch; irrig i represents the irrigation guarantee potential of the i-th cultivated land patch; eco i represents the ecological restriction intensity of the i-th cultivated land patch; Y i represents the label value of the i-th cultivated land patch, wherein the label value of the positive sample is 1; the label value of the negative sample is 0; S202: constructing a stable fitting model based on the training data set, and the specific formula of the stable fitting model is: Log (p / (1-p)) = β0+ β1×contig i + β2×qual i + β3×irrig i + β4×eco i ; wherein, p represents the predicted label value of the cultivated land patch, β0represents the intercept term, β1, β2, β3and β4represent the fitting coefficients; the stable fitting model is trained using the maximum likelihood estimation method, and the final value of the fitting coefficient is generated by the optimization algorithm, so that the difference between the predicted label value and the true label value based on the feature vector of the stable fitting model is minimized; S203: normalizing the absolute value of the final value of the fitting coefficient, analyzing the weight of each basic feature, and the weight of each basic feature is equal to the ratio of the normalized value of the absolute value of the final value of the fitting coefficient to the sum of the normalized values of the absolute values of the final values of the fitting coefficients.

[0025] S3: based on the processed real-time basic features and the weights of the basic features, performing weighted summation to calculate the optimized comprehensive score of each cultivated land patch; In S3, the following contents are included: S301: obtaining the real-time basic features of each cultivated land patch, and performing positive direction and extreme value standardization on each real-time basic feature of the cultivated land patch; S302: based on the processed real-time basic features and the weights of the basic features, performing weighted summation to calculate the optimized comprehensive score of each cultivated land patch, and recording the optimized comprehensive score of the j-th cultivated land patch as LOSS j ; LOSS j = W c ×contig j+W q ×qual j +W i ×irrig j +W e ×eco j ; wherein W c , W q , W i and W e are the weights of the contig j , qual j , irrig j and eco j are the contig expansion potential, quality improvement potential, irrigation guarantee potential and ecological restriction intensity of the jth cultivated land patch, respectively.

[0026] S4: determining the grading threshold of the optimized comprehensive score according to the threshold analysis method selected by the administrator; dividing the cultivated land patches according to the grading threshold to generate an optimized attribute list and sending the list to the administrator.

[0027] In S4, the following contents are included: S401: counting the distribution of the optimized comprehensive score and calculating statistical quantities, the statistical quantities including mean, median, quantile and sample variance; determining the first threshold and the second threshold according to the threshold analysis method selected by the administrator; if the target area method is used, recording the target total area used and the cumulative logic; if the expert decision method is used, recording the expert weight and the decision basis; Embodiment 1: the threshold analysis method in this embodiment includes the target area method, the quantile method, the clustering division method and the expert decision method; wherein the target area method is that the administrator presets the target total area of the priority optimization area, and on the distribution of the optimized comprehensive score, the first threshold is taken as the optimized comprehensive score corresponding to the first cumulative area greater than or equal to the target total area from high to low, and the second threshold is equal to the optimized comprehensive score corresponding to the secondary target total area; The quantile method adopts the optimized comprehensive score corresponding to the quantile of the optimized comprehensive score as the threshold, and in this embodiment, the optimized comprehensive scores corresponding to 50% and 90% are adopted; The clustering division method is that the optimized comprehensive scores of all cultivated land patches are subjected to k-means clustering, the classification boundary is determined according to the cluster center and the cluster distance, finally the highest class is marked as the priority optimization area, the middle class is the general optimization area, and the lowest class is the maintenance area; The expert decision method is that the first threshold and the second threshold are determined manually by combining the planning target and the social and economic constraints through expert scoring; and the opinions and weight distribution of the expert group are recorded.

[0028] If the optimization comprehensive score is greater than or equal to the first threshold value, the corresponding cultivated land patch is marked as a priority optimization area candidate; if the optimization comprehensive score is less than the first threshold value and greater than or equal to the second threshold value, the corresponding cultivated land patch is marked as a general optimization area; if the optimization comprehensive score is less than the second threshold value, the corresponding cultivated land patch is marked as a status maintenance area; if all optimization comprehensive scores are less than the first threshold value, a certain percentage selected by the quantile method is used as a backup; if all optimization comprehensive scores are greater than or equal to the first threshold value, the target area method is used to limit the total area of the candidate. S402: Extract the number, spatial position and cultivated land area of each marked cultivated land patch as a priority optimization area candidate; integrate into an optimization attribute list and send to the administrator.

[0029] Please refer to Figure 2 The application provides a technical solution: a cultivated land spatial layout optimization management system based on multi-source data fusion, which comprises a basic feature analysis module, a feature weight analysis module, an optimization comprehensive score analysis module and a cultivated land division module. The basic feature analysis module is used to obtain cultivated land patch data, irrigation water source data and ecological protection elements in an evaluation area; and the basic features of each cultivated land patch are analyzed. The feature weight analysis module is used to collect historical cultivated land change data, construct a training data set containing stable cultivated land and changed cultivated land; and the weights of each basic feature are analyzed based on a stable fitting model. The optimization comprehensive score analysis module is used to perform weighted summation based on each processed real-time basic feature and the weight of each basic feature, and calculate the optimization comprehensive score of each cultivated land patch. The cultivated land division module is used to determine the classification threshold of the optimization comprehensive score according to the threshold value analysis method selected by the administrator; and the cultivated land patches are divided according to the classification threshold, and an optimization attribute list is generated and sent to the administrator.

[0030] The basic feature analysis module comprises a cultivated land data acquisition unit and a feature analysis unit. The cultivated land data acquisition unit is used to obtain cultivated land patch data in an evaluation area, obtain an irrigation water source set and ecological protection elements. The feature analysis unit is used to analyze the basic features of each cultivated land patch; the basic features include contiguous expansion potential, quality improvement potential, irrigation guarantee potential and ecological restriction intensity.

[0031] The feature weight analysis module comprises a sample collection unit, a model fitting unit and a weight analysis unit. The sample collection unit is configured to extract the cultivated land change data in a preset historical time window, identify a cultivated land patch that has always remained as high-quality cultivated land in the preset historical time window as a positive sample, and identify a cultivated land patch that has been abandoned, converted into construction land, or seriously degraded as a negative sample; meanwhile, the sample collection unit is configured to extract the basic feature values corresponding to the cultivated land patch at the beginning of the preset time window to form a training data set; The model fitting unit is configured to train the stable fitting model using the maximum likelihood estimation method, and generate a final value of the fitting coefficient through an optimization algorithm, so that the stable fitting model has a minimum difference between a predicted label value and an actual label value based on a feature vector. The weight analysis unit is configured to normalize the absolute value of the final value of the fitting coefficient and analyze the weight of each basic feature.

[0032] The optimization comprehensive score analysis module includes a real-time data processing unit and a score analysis unit. The real-time data processing unit is configured to obtain real-time basic features of each cultivated land patch, and perform positive transformation and extreme value standardization on each real-time basic feature of the cultivated land patch. The score analysis unit is configured to perform weighted summation based on the processed real-time basic features and the weight of each basic feature, and calculate an optimization comprehensive score of each cultivated land patch.

[0033] The cultivated land division module includes a cultivated land division unit and a list integration unit. The cultivated land division unit is configured to determine a first threshold value and a second threshold value according to an administrator-selected threshold analysis method, and divide the cultivated land patches based on the first threshold value and the second threshold value. The list integration unit is configured to extract the number, spatial position, and cultivated land area of each cultivated land patch marked as a candidate of a priority optimization area, and integrate the extracted information into an optimization attribute list and send the optimization attribute list to the administrator.

[0034] It should be noted that, in this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0035] It will be apparent to those skilled in the art that the application is not limited to the details of the above-exemplified embodiments and that the present application can be implemented in other particular forms without departing from the spirit or essential characteristics of the present application. The embodiments should therefore be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the above description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. No reference signs in the claims should be considered as limiting the scope of the claims with respect to the figures of the patent document.

Claims

1. A method for optimizing the spatial layout management of arable land based on multi-source data fusion, characterized in that, The method includes the following steps: S1: Obtain farmland patch data, irrigation water source data, and ecological protection elements within the evaluation area; analyze the basic characteristics of each farmland patch; S2: Collect historical data on changes in cultivated land and construct a training dataset containing both stable and changed cultivated land; analyze the weights of each basic feature based on a stable fitting model. S3: Based on the processed real-time basic features and the weights of each basic feature, perform weighted summation to calculate the optimized comprehensive score of each cultivated land patch; S4: Determine the grading threshold for the comprehensive optimization score based on the threshold analysis method selected by the administrator; divide the cultivated land patches according to the grading threshold, generate an optimization attribute list and send it to the administrator.

2. The method for optimizing and managing the spatial layout of cultivated land based on multi-source data fusion according to claim 1, characterized in that: S1 includes the following: S101: Obtain farmland patch data within the evaluation area, including the geometric shape, spatial location, farmland area, and farmland quality grade of the farmland patches; obtain irrigation water source data and ecological protection elements; the irrigation water source elements include canals, rivers, water outlets, and reservoirs; the irrigation water source data includes the spatial location of each irrigation water source; the ecological protection elements include ecological protection red lines; S102: Analyze the basic characteristics of each cultivated land patch; For each cultivated land patch, extract all cultivated land patches whose spacing is less than the search distance threshold to form a potential contiguous cultivated land patch set for each cultivated land patch; the spacing between cultivated land patches is equal to the shortest Euclidean distance between the polygonal boundaries of two cultivated land patches, and the spacing between cultivated land patches is zero if the polygonal boundaries of two cultivated land patches intersect or contact; sum the cultivated land areas of all cultivated land patches in the contiguous cultivated land patch set, and record it as the contiguous expansion potential; normalize the contiguous expansion potential of all cultivated land patches. Extract the farmland quality grade values ​​of all farmland patches within the evaluation area, normalize the farmland quality grade values, and record the difference between the normalized farmland quality grade value and the normalized farmland quality grade value as the potential for quality improvement. Analyze the optimal irrigation distance from each cultivated land patch to the nearest irrigation water source; the optimal irrigation distance is equal to the shortest Euclidean distance between the polygonal boundary of the cultivated land patch and the irrigation water source; normalize the optimal irrigation distance, and record the difference between the normalized and the normalized optimal irrigation distance as the irrigation guarantee potential. For each cultivated land parcel, spatial overlay analysis is used to determine its overlap with the ecological protection red line. If the cultivated land parcel overlaps with the ecological protection red line, the ecological restriction intensity of the cultivated land parcel is recorded as zero; otherwise, the ecological restriction intensity of the cultivated land parcel is recorded as one.

3. The method for optimizing and managing the spatial layout of cultivated land based on multi-source data fusion according to claim 2, characterized in that: In S2: Includes the following: S201: Extract farmland change data within a preset historical time window, identify farmland patches that have consistently remained high-quality farmland within the preset historical time window as positive samples, and farmland patches that have been abandoned, converted to construction land, or severely degraded as negative samples; simultaneously, extract the basic feature values ​​corresponding to the farmland patches at the beginning of the preset time window to form a training dataset, denoted as D. trian ={(X i Y i )|i∈[1,I]}; where I represents the total number of cultivated map patches in the training dataset; X i X represents the feature vector of the i-th cultivated patch in the training dataset; i =(contig) i qual i irrig i eco i ); where contig i This represents the contiguous expansion potential of the i-th cultivated plot; qual i Indicates the potential for quality improvement of the i-th cultivated plot; irrig i Eco represents the irrigation security potential of the i-th cultivated plot; i Y represents the ecological limitation intensity of the i-th cultivated land patch; i This represents the label value of the i-th cultivated patch, where the label value of a positive sample is 1 and the label value of a negative sample is 0. S202: Construct a stable fitting model based on the training dataset. The specific formula for the stable fitting model is: Log(p / (1-p))=β0+β1×contig i +β2×qual i +β3×irrig i +β4×eco i ; Where p represents the predicted label value of cultivated land patch, β0 represents the intercept term, and β1, β2, β3 and β4 represent the fitting coefficients; the maximum likelihood estimation method is used to train a stable fitting model, and the final value of the fitting coefficients is generated by the optimization algorithm to minimize the difference between the predicted label value and the true label value based on the feature vector of the stable fitting model. S203: Normalize the absolute values ​​of the final values ​​of the fitting coefficients, and analyze the weights of each basic feature. The weight of each basic feature is equal to the ratio of the normalized value of the absolute value of the corresponding final value of the fitting coefficient to the sum of the normalized values ​​of the absolute values ​​of the final values ​​of the fitting coefficients.

4. The method for optimizing and managing the spatial layout of cultivated land based on multi-source data fusion according to claim 3, characterized in that: S3 includes the following: S301: Obtain the real-time basic features of each cultivated land patch, and perform positiveization and extreme value standardization processing on each real-time basic feature of the cultivated land patch. S302: Based on the processed real-time basic features and their weights, perform a weighted summation to calculate the optimized comprehensive score for each cultivated land patch. The optimized comprehensive score for the j-th cultivated land patch is denoted as LOSS. j LOSS j =W c ×contig j +W q ×qual j +W i ×irrig j +W e ×eco j ; Among them W c W q W i and W e The weights for contiguous expansion potential, quality improvement potential, irrigation security potential, and ecological constraint intensity are respectively. j qual j irrig j and eco j These represent the contiguous expansion potential, quality improvement potential, irrigation guarantee potential, and ecological limitation intensity of the j-th cultivated land patch, respectively.

5. The method for optimizing and managing the spatial layout of cultivated land based on multi-source data fusion according to claim 4, characterized in that: S4 includes the following: S401: Statistically optimize the distribution of the comprehensive score and calculate various statistics, including the mean, median, quantiles and sample variance; determine the first threshold and the second threshold according to the threshold analysis method selected by the administrator; if the target area method is used, record the total target area and cumulative logic used; if the expert decision method is used, record the expert weights and decision basis. If the comprehensive optimization score is greater than or equal to the first threshold, the corresponding cultivated land patch is marked as a priority optimization candidate; if the comprehensive optimization score is less than the first threshold but greater than or equal to the second threshold, the corresponding cultivated land patch is marked as a general optimization area; if the comprehensive optimization score is less than the second threshold, the corresponding cultivated land patch is marked as a status quo maintenance area; if all comprehensive optimization scores are less than the first threshold, a certain percentage is selected as candidates using the quantile method; if all comprehensive optimization scores are greater than or equal to the first threshold, the total candidate area is limited using the target area method. S402: Extract the ID, spatial location, and cultivated land area of ​​each cultivated land patch marked as a candidate for priority optimization area; integrate them into an optimization attribute list and send it to the administrator.

6. A farmland spatial layout optimization management system based on multi-source data fusion, wherein the system is applied to the farmland spatial layout optimization management method based on multi-source data fusion as described in any one of claims 1-5, characterized in that, The system includes a basic feature analysis module, a feature weight analysis module, an optimized comprehensive scoring analysis module, and a farmland classification module; The basic feature analysis module is used to acquire farmland patch data, irrigation water source data, and ecological protection elements within the evaluation area; and to analyze the basic features of each farmland patch. The feature weight analysis module is used to collect historical data on changes in cultivated land, construct a training dataset containing both stable and changed cultivated land, and analyze the weights of each basic feature based on a stable fitting model. The optimized comprehensive scoring analysis module is used to perform weighted summation based on the processed real-time basic features and the weights of each basic feature to calculate the optimized comprehensive score of each cultivated land patch. The cultivated land division module is used to determine the grading threshold of the comprehensive optimization score based on the threshold analysis method selected by the administrator; the cultivated land patches are divided according to the grading threshold, and an optimized attribute list is generated and sent to the administrator.

7. A farmland spatial layout optimization management system based on multi-source data fusion according to claim 6, characterized in that: The basic feature analysis module includes a farmland data acquisition unit and a feature analysis unit; The cultivated land data acquisition unit is used to acquire cultivated land patch data within the evaluation area, and to acquire irrigation water source collection and ecological protection elements; The feature analysis unit is used to analyze the basic features of each cultivated land patch; the basic features include the potential for contiguous expansion, the potential for quality improvement, the potential for irrigation security, and the intensity of ecological constraints.

8. A farmland spatial layout optimization management system based on multi-source data fusion according to claim 6, characterized in that: The feature weight analysis module includes a sample acquisition unit, a model fitting unit, and a weight analysis unit. The sample collection unit is used to extract farmland change data within a preset historical time window, identify farmland patches that have always remained high-quality farmland within the preset historical time window as positive samples, and farmland patches that have been abandoned, converted to construction land, or severely degraded as negative samples; at the same time, it extracts the basic feature values ​​corresponding to the farmland patches at the beginning of the preset time window to form a training dataset. The model fitting unit is used to train the dataset to build a stable fitting model. The stable fitting model is trained using the maximum likelihood estimation method, and the final value of the fitting coefficients is generated through the optimization algorithm, so that the difference between the predicted label value and the true label value based on the feature vector of the stable fitting model is minimized. The weight analysis unit is used to normalize the absolute value of the final value of the fitting coefficients and analyze the weight of each basic feature.

9. A farmland spatial layout optimization management system based on multi-source data fusion according to claim 6, characterized in that: The optimized comprehensive scoring analysis module includes a real-time data processing unit and a scoring analysis unit; The real-time data processing unit is used to acquire the real-time basic features of each cultivated land patch and perform positiveization and extreme value standardization processing on each real-time basic feature of the cultivated land patch. The scoring analysis unit is used to perform weighted summation based on the processed real-time basic features and the weights of each basic feature to calculate the optimized comprehensive score of each cultivated land patch.

10. A farmland spatial layout optimization management system based on multi-source data fusion according to claim 6, characterized in that: The cultivated land division module includes a cultivated land division unit and a list integration unit; The cultivated land division unit is used to determine a first threshold and a second threshold according to the threshold analysis method selected by the administrator; and to divide cultivated land patches based on the first threshold and the second threshold. The list integration unit is used to extract the number, spatial location, and cultivated land area of ​​each cultivated land patch marked as a candidate for priority optimization area; and integrate them into an optimization attribute list and send it to the administrator.

Images