Intelligent evaluation method for rural construction land reduction based on multi-source spatial big data

Abstract

This invention discloses an intelligent assessment method for reducing rural construction land based on multi-source spatial big data, relating to the fields of land resource management and spatial information processing. The method involves collecting archival data and on-site status records of abandoned land parcels in rural areas from multiple sources, processing the discrepancies between the archival data and the status records, initially labeling the sources of deviation, and obtaining a list of inconsistencies in the abandoned land parcels. This intelligent assessment method for reducing rural construction land based on multi-source spatial big data improves the accuracy, intelligence level, and reliability of the assessment results, providing more scientific decision-making support for subsequent land remediation, land reclamation, spatial optimization, and implementation timing.

Description

Technical Field

[0001] This invention relates to the field of land resource management and spatial information processing technology, specifically to an intelligent assessment method for reducing rural construction land based on multi-source spatial big data. Background Technology

[0002] Reducing rural construction land use is an important means to promote the economical and intensive use of land, optimize rural spatial layout, and improve the efficiency of land resource allocation. It is also a crucial aspect of improving the rural living environment. In the work of reducing rural construction land use, it is usually necessary to comprehensively analyze the current use status, ownership information, historical changes, consolidation potential, and feasibility of the target plots to form a reduction assessment result. This assessment provides a basis for subsequent plot consolidation, land reclamation, layout optimization, and implementation scheduling. Therefore, the accuracy of the reduction assessment result largely depends on the completeness, authenticity, and consistency of the basic plot data.

[0003] In existing technologies, rural construction land reduction assessments are typically based on multi-source information, including land archive data, survey data, remote sensing interpretation data, and on-site verification records. Due to differences in collection time, update frequency, data standards, and expression formats among different data sources, inconsistencies easily arise between attributes such as land use, boundaries, ownership, and utilization status. This is particularly true in remote rural areas, where incomplete historical data, rapid changes in the current situation, and insufficient manual verification coverage lead to more common discrepancies between archive data and the actual situation, resulting in significant deviations in the underlying data maps used for assessment. In such cases, existing methods often rely on manual verification, single-field comparison, or simple rule screening to identify data anomalies and conduct reduction assessments based on these anomalies. While these methods can identify some obvious differences, they typically lack the ability to effectively identify hidden contradictions, conflicting relationships, and dynamic changes among multi-source data. When different attributes of the same land parcel conflict across multiple data sources, existing technologies often struggle to determine which information is closer to the true situation and to further clarify which attribute differences will have a key impact on the reduction assessment results. Furthermore, the assessment of rural construction land reduction does not rely solely on a single attribute; it requires a comprehensive analysis of multiple dimensions, including the current land use status, historical evolution, spatial distribution, remediation conditions, and potential implementation value. If the basic attribute data itself is biased or there are conflicts between different dimensions, it can easily lead to inaccurate extraction of assessment indicators, distorted assessment results, and improper prioritization of land parcels. For example, a parcel may be recorded as construction land in the archives, but its current status may have been abandoned for a long time, or different sources of data may record its use status, scope, and remediation conditions differently. Without effective identification and verification of the differences between multi-source data, it is difficult to accurately determine whether the parcel is suitable for reduction, thus affecting the analysis results of reduction potential and implementation priority. In addition, existing technologies typically lack a unified linkage analysis mechanism between data correction, assessment basis selection, and assessment priority determination during the assessment process. This makes it difficult to dynamically adjust the assessment basis based on multi-source data deviations and to form stable and reliable reduction assessment conclusions. This not only reduces the efficiency of the assessment, but also makes it easy for differences in human judgment to affect the consistency and objectivity of the assessment results, thereby hindering the scientific implementation of the work to reduce rural construction land. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent assessment method for reducing rural construction land based on multi-source spatial big data, thereby solving the problems existing in the prior art.

[0005] To achieve the above objectives, this invention provides the following technical solution: an intelligent assessment method for reducing rural construction land use based on multi-source spatial big data, comprising: S1, collecting archival data and on-site status records of abandoned land parcels in rural areas through multi-source information collection, processing the differences between archival data and on-site records, initially labeling the sources of deviation, and obtaining a list of inconsistencies in abandoned land parcels; S2, extracting the attribute features of the land parcels based on the list of inconsistencies, decomposing the attribute features into multiple dimensions using vector dimension partitioning technology, ranking the importance of each dimension, and determining the deviation distribution characteristics of the land parcel attributes; S3, if the deviation value of a certain dimension in the deviation distribution characteristics of the land parcel attributes exceeds a preset threshold, then obtaining historical data reference records for that dimension, and dynamically adjusting... S4. Analyze the updated deviation data to determine whether to generate a new correction direction vector; S5. Extract key correction reference points from the correction direction vector, filter out attributes related to the deviation based on the extraction logic, check the matching degree between the reference points and the current status records, and determine the specific content of the priority correction basis; S6. Based on the specific content of the priority correction basis, use correction scope definition technology to clarify the applicable land parcel attribute range, and combine the weight calculation method to assign correction priority to different attributes to obtain a correction guidance list for abandoned land parcels; S7. Based on the correction guidance list for abandoned land parcels, use correction path association technology to bind the guidance to the land parcel attribute update path, and determine the final correction execution sequence by setting a dynamic refresh time interval based on the update frequency.

[0006] Preferably, step S1 includes acquiring archival data and on-site status records of abandoned land parcels in rural areas collected from multiple sources; extracting spatial coordinates and ownership information from the archival data and on-site status records; performing spatial topological overlay processing on the archival data and on-site status records based on the spatial coordinates and ownership information to obtain an initial matching result containing land parcel area and utilization type; processing the differences between land parcel area and utilization type in the initial matching result using data field comparison rules to obtain a field difference feature set; if the area difference in the field difference feature set is greater than a preset area threshold, it is determined to be a boundary change deviation; and the differences between the archival data and on-site status records of abandoned land parcels are initially marked through the deviation source classification results to obtain a list of inconsistencies of abandoned land parcels.

[0007] Preferably, step S2 includes acquiring remote sensing monitoring data and a list of inconsistencies for abandoned land parcels; parsing the list of inconsistencies and mapping it to the remote sensing monitoring data to extract land parcel attribute features; dividing the land parcel attribute features into vector dimensions to construct a dimension division matrix; calculating the Euclidean distance between the orthogonal components in the dimension division matrix and the standard land parcel feature template to quantify the attribute deviation value; calculating the deviation weight allocation coefficient based on the attribute deviation value; ranking the importance of each attribute dimension according to the deviation weight allocation coefficient; and determining the deviation distribution characteristics of the land parcel attributes.

[0008] Preferably, step S3 includes acquiring the deviation distribution characteristics of the land parcel attributes; if the extracted dimensional deviation value exceeds a preset threshold, then acquiring historical data reference records for the dimension; extracting spatial coordinate offsets based on the historical data reference records, classifying the spatial coordinate offsets to obtain first deviation data; performing weighted calculations on the first deviation data through a dynamic adjustment mechanism to obtain second deviation data, and obtaining deviation analysis results based on the second deviation data; if the dispersion value in the deviation analysis results is greater than a preset dispersion, then extracting correction angle values ​​based on the deviation analysis results, and generating a new correction direction vector using the correction angle values.

[0009] Preferably, step S4 includes acquiring the original business data stream and calculating the correction direction vector; extracting key correction reference points based on the cluster density distribution of the correction direction vector; analyzing the mutual information correlation of the key correction reference points and selecting highly correlated deviation attributes; retrieving the real-time status records corresponding to the highly correlated deviation attributes; comparing the key correction reference points with the real-time status records using a data consistency verification model to generate a matching degree index; constructing an attribute priority index sequence based on the matching degree index; parsing the first and second features of the attribute priority index sequence to determine the specific content of the priority correction basis.

[0010] Preferably, step S5 includes obtaining the correction basis text and constructing a semantic feature vector using a bidirectional long short-term memory network; performing correlation analysis between the semantic feature vector and the land parcel attributes to determine the applicable scope boundary; calculating the information entropy value of each attribute within the applicable scope boundary and establishing the basis weight value based on the information entropy value; calculating a weighted correction urgency score by combining the current status value of the abandoned land parcel attributes and the basis weight value; generating an attribute correction priority ranking sequence based on the weighted correction urgency score; and generating an abandoned land parcel correction guidance list based on the attribute correction priority ranking sequence.

[0011] Preferably, step S6 includes obtaining a list of correction guidelines for abandoned land parcels, matching the list of correction guidelines with the land parcel attribute data using a path association algorithm to obtain a bound attribute update path; analyzing the bound attribute update path to obtain an update frequency index, setting a dynamic refresh time interval based on the update frequency index, and determining the refresh sequence of the path.

[0012] Preferably, step S6 further includes optimizing the dynamic refresh time interval using a frequency adjustment mechanism to obtain an optimized refresh configuration if the refresh sequence of the path exceeds a threshold, and deeply binding the correction guidance list with the bound attribute update path according to the optimized refresh configuration to determine the final correction execution sequence.

[0013] Preferably, the process also includes S7: forming a corrected land parcel attribute dataset based on the final corrected execution sequence, extracting assessment indicators characterizing the reduction of rural construction land, conducting reduction potential analysis, suitability analysis, and priority analysis on each abandoned land parcel, and generating intelligent assessment results for the reduction of rural construction land. Specifically, this includes acquiring execution sequence data and integrating land parcel attributes to determine the initial correction path, processing the execution sequence using indicator extraction rules to obtain the input dataset, calculating the reduction potential value based on the input dataset, and determining the land use matching standard set when the reduction potential value meets the conditions.

[0014] Preferably, step S7 further includes evaluating suitability scores based on a set of land use matching standards and generating a preliminary ranking list based on the reduction potential value, determining the output structure based on the preliminary ranking list, and generating intelligent assessment results for the reduction of rural construction land.

[0015] As can be seen from the above technical solution, the present invention has the following beneficial effects: This intelligent assessment method for reducing rural construction land based on multi-source spatial big data effectively identifies inconsistencies in land parcel attributes by collecting and comparing archival data and on-site records of abandoned land parcels in remote rural areas. Furthermore, it combines attribute feature decomposition, deviation distribution analysis, historical data reference, correction direction judgment, correction basis screening, and correction priority ranking to systematically verify and dynamically correct the basic data affecting the accuracy of the reduction assessment, thereby improving the completeness, consistency, and reliability of land parcel attribute data. Further, by extracting reduction assessment indicators from the corrected land parcel attribute dataset, it conducts reduction potential analysis, suitability analysis, and priority analysis on abandoned land parcels. This reduces the bias, lag, and human error problems caused by reliance on manual verification, single-field comparison, or simple rule screening in existing technologies, enhancing the stability and objectivity of the assessment basis under multi-source heterogeneous data conditions. Ultimately, it improves the accuracy, intelligence level, and application reliability of rural construction land reduction assessment results, providing more scientific decision-making support for subsequent land remediation, relocation and reclamation, spatial optimization layout, and implementation timing. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall signal transmission of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: As Figure 1 As shown, this invention provides a technical solution: an intelligent assessment method for reducing rural construction land use based on multi-source spatial big data, comprising: S1. Collect archival data and on-site status records of abandoned land parcels in rural areas through multi-source information collection, process the differences between archival data and status records, classify the sources of deviation and make preliminary annotations to obtain a list of inconsistencies of abandoned land parcels. S2. Extract the attribute features of the abandoned land parcels from the list of inconsistencies, decompose the attribute features into multiple dimensions using vector dimension partitioning technology, rank the importance of each dimension, and determine the deviation distribution characteristics of the land parcel attributes. S3. If the deviation value of a certain dimension in the deviation distribution characteristics of the land parcel attributes exceeds the preset threshold, then historical data reference records are obtained for that dimension, the analysis results of the deviation data are updated through a dynamic adjustment mechanism, and it is determined whether a new correction direction vector is generated. S4. Extract key correction reference points by correcting the direction vector, filter out the attributes related to the deviation based on the extraction logic, check the matching degree between the reference points and the current status record, and determine the specific content of the priority correction basis. S5. Based on the specific content of the priority correction basis, use the correction scope definition technology to clarify the applicable land parcel attribute range, and combine the correction priority ranking of different attributes according to the weight calculation method to obtain the correction guidance list of abandoned land parcels. S6. Based on the list of correction guidelines for abandoned land parcels, the correction path association technology is used to bind the guidelines with the land parcel attribute update path. By setting a dynamic refresh time interval based on the update frequency, the final correction execution sequence is determined. S7. Based on the final corrected execution sequence, a corrected land parcel attribute dataset is formed. Evaluation indicators representing the reduction of rural construction land are extracted. Reduction potential analysis, suitability analysis, and priority analysis are performed on each abandoned land parcel to generate intelligent evaluation results of rural construction land reduction.

[0019] In this embodiment, in step S1, archival data and on-site status records of abandoned land parcels in rural areas are collected through multi-source information collection. The archival data may include land registration data, cadastral survey results, historical approval information, village planning data, construction land ledgers, remote sensing interpretation data, and historical remediation records, etc.; the on-site status records may include on-site survey information, UAV aerial survey results, photos collected by mobile terminals, video patrol records, GNSS positioning data, and information reported by grassroots units, etc.

[0020] Furthermore, data from different sources undergoes unified coordinate registration, time alignment, field standardization, and object association processing to identify discrepancies between archival data and current records in areas such as plot boundaries, area, use, utilization status, building condition, vacancy level, and demolition status. It should be noted that these discrepancies are not simple data conflicts, but rather reflect deviations caused by factors such as data source, collection time, spatial precision, and business terminology. Therefore, this step further categorizes and initially labels the discrepancies according to their sources, forming a list of inconsistencies in abandoned plots, providing structured input for subsequent deviation analysis and correction.

[0021] In some embodiments, step S2 can extract the attribute features of the land parcels based on the aforementioned list of inconsistencies, and decompose the attribute features into multiple analytical dimensions using vector dimension partitioning technology. Specifically, attribute features may include spatial geometric features, ownership management features, land use features, current building status features, location and transportation features, infrastructure features, ecological constraint features, and policy constraint features. As a preferred embodiment, the aforementioned attribute features can be encoded into multi-dimensional attribute vectors, and divided into several dimensions according to preset rules or feature clustering results. Then, by combining factors such as historical correction samples, frequency of deviation occurrence, degree of impact on assessment results, and data source reliability, the importance of each dimension can be ranked. Thus, through vectorization and dimensionalization processing, the originally scattered, heterogeneous, and difficult-to-comparate attribute information can be transformed into calculable and quantifiable deviation distribution features, thereby facilitating the identification of key deviation dimensions that have a significant impact on the reduction assessment.

[0022] In this embodiment, in step S3, if the deviation value of a certain dimension in the deviation distribution characteristics of the land parcel attributes exceeds a preset threshold, it indicates that the attribute conflict corresponding to that dimension has reached a level requiring focused correction. Based on the aforementioned embodiments, to avoid making correction judgments solely based on data at the current point in time, this application further retrieves historical data reference records for that dimension. Historical data reference records may include previous remote sensing images, past survey results, historical version ledgers, correction logs, land parcel transfer records, demolition and reclamation records, and village construction change records, etc.

[0023] Furthermore, the analysis results of the current deviation data are updated through a dynamic adjustment mechanism. This mechanism comprehensively considers time decay factors, data source reliability, spatial neighborhood consistency, and historical change trends to reassess the deviation results and determine whether to generate a new correction direction vector. The correction direction vector characterizes the target change direction, correction magnitude, and priority correction tendency of a certain attribute during the correction process. Therefore, this step effectively establishes the technical foundation for the transition from static deviation identification to dynamic deviation correction.

[0024] In addition, in step S4, the system extracts key correction reference points based on the correction direction vector and filters out attributes related to the deviation according to a preset extraction logic to determine the specific content of the priority correction basis. It should be noted that key correction reference points can be spatial location points, boundary inflection points, key points of building outlines, historical time nodes, change evidence nodes, or ownership confirmation nodes, etc., which are essentially evidence units that strongly support the correction judgment. In another specific implementation, the system can match and verify the key correction reference points with the current status record to analyze the degree of consistency between them and the current land parcel status. When the matching degree between the reference point and the current status record is high, the corresponding data source or attribute item can be used as the priority correction basis; otherwise, its weight can be reduced or a pending review mark can be retained. As can be seen from the above, by setting the correction direction vector and key correction reference points, the correction of land parcel attributes can be based on a traceable chain of evidence, rather than simply relying on a single data source for overlay updates.

[0025] In this embodiment, in step S5, based on the specific content of the priority correction basis, the applicable land parcel attribute range is clarified using correction scope definition technology, and the correction priority is assigned to different attributes according to the weight calculation method to obtain a correction guidance list for abandoned land parcels. The correction scope definition technology is used to determine which attributes should be corrected simultaneously, which attributes are only used as auxiliary references, and which attributes need to be temporarily deferred and enter the subsequent review process. As a preferred embodiment, when a boundary attribute is confirmed to need correction, its associated area attribute, adjacent relationship attribute, and building coverage attribute can be corrected simultaneously; while when there is a dispute over ownership attributes, only the status flag can be updated without directly overwriting the original file fields.

[0026] Furthermore, based on the weighting calculation method, weight values ​​can be generated by combining factors such as the authority of the data source, its recentity, spatial precision, consistency level, and its impact on the reduction assessment results. These weights are then ranked according to their priority for correction. Therefore, this step provides clear boundaries and priorities for different types of attribute corrections, which helps improve the targeting and stability of the correction process.

[0027] Furthermore, in step S6, based on the correction guidance list, correction path association technology is used to bind each guidance to the land parcel attribute update path, and a dynamic refresh time interval is set according to the update frequency to determine the final correction execution sequence. In this embodiment, the correction path association technology can assign different correction items to automatic update paths, semi-automatic verification paths, or manual review paths according to attribute type, sufficiency of evidence, and complexity of business rules. For example, attributes with sufficient evidence and clear rules can directly enter the automatic update path; attributes with minor conflicts but which can be determined through auxiliary evidence can enter the semi-automatic verification path; attributes involving ownership disputes, significant changes in use, or policy restrictions can enter the manual review path. At the same time, this application also sets corresponding refresh cycles according to the data update frequency of different bases. For example, remote sensing images can be refreshed monthly or quarterly, on-site inspection records can be refreshed weekly, and archive ledgers can be refreshed when triggered by business changes. Thus, this step achieves an organic association between correction content, correction method, and correction timing.

[0028] In this embodiment, in step S7, a revised land parcel attribute dataset is formed based on the final revised execution sequence, and assessment indicators characterizing the reduction of rural construction land are extracted. Reduction potential analysis, suitability analysis, and priority analysis are performed on each abandoned land parcel, ultimately generating an intelligent assessment result for the reduction of rural construction land. Specifically, the reduction potential analysis is mainly used to identify whether a land parcel has space for idleness, inefficiency, abandonment, reclamation, or redevelopment; the suitability analysis is mainly used to determine whether a land parcel is suitable for inclusion in the reduction implementation scope in terms of location conditions, ecological constraints, infrastructure accessibility, planning control requirements, and subsequent remediation conditions; and the priority analysis is used to comprehensively assess the implementation benefits, implementation costs, social impact, and policy matching degree of different land parcels to determine the order of implementation for each land parcel. Therefore, after completing the basic attribute revision, this application further directly applies the revised results to the reduction decision analysis, enabling the assessment results to more realistically reflect the actual state and remediation value of abandoned rural land parcels.

[0029] In summary, this embodiment identifies differences in multi-source data, performs fractal analysis on deviations, dynamically corrects abnormal attributes, and conducts reduction assessments based on the correction results, thereby forming an intelligent assessment technology solution suitable for rural construction land reduction scenarios.

[0030] Example 2: S1 includes acquiring archival data and on-site status records of abandoned land parcels in rural areas from multi-source information collection, extracting spatial coordinates and ownership information from the archival data and on-site status records; performing spatial topological overlay processing on the archival data and on-site status records based on the spatial coordinates and ownership information to obtain an initial matching result containing land parcel area and utilization type; using data field comparison rules to process the differences between land parcel area and utilization type in the initial matching result to obtain a field difference feature set; if the area difference in the field difference feature set is greater than a preset area threshold, it is determined to be a boundary change deviation, and the differences between the archival data and on-site status records of abandoned land parcels are initially marked through the deviation source classification results to obtain a list of inconsistencies of abandoned land parcels.

[0031] In this embodiment, archival data and on-site status records of abandoned rural land parcels collected from multiple sources are first acquired, and spatial coordinates and ownership information are extracted from them. Specifically, the source attributes of each type of data entering the processing flow are verified, identifying the collection time, collection subject, collection area, data type, field composition, and record completeness, thereby distinguishing between historical management data and on-site collected data. Subsequently, the archival data is organized, and the following are extracted sequentially: parcel number, land parcel number, subject information, area record, utilization type record, boundary coordinate record, map outline record, and historical change record. The on-site status records are organized, and the following are extracted sequentially: on-site collection number, location information, on-site subject description, on-site area measurement results, on-site utilization status description, boundary trajectory record, and image supporting information.

[0032] As a preferred implementation, before formally extracting spatial coordinates and ownership information, synonymous fields in the two types of data are first uniformly named to merge fields with the same business meaning but different names into the same identification item; then missing fields are marked as missing items so that subsequent processing can identify whether the difference comes from missing records or changes in status; then obviously abnormal coordinate values, area values ​​and main information are removed and reviewed to prevent erroneous records from entering the subsequent matching stage.

[0033] In this embodiment, the extraction of spatial coordinates does not involve extracting only a single location point. Instead, it prioritizes extracting the coordinates of boundary nodes that can fully represent the spatial extent of the land parcel, followed by extracting the coordinates of the land parcel's center location to supplement the identification of the overall location of the land parcel. Similarly, the extraction of ownership information does not only involve extracting the entity name, but also simultaneously extracting the entity name, entity number, historical successor entities, related farmer information, and corresponding information from management ledgers. This is used to identify the continuous relationship of the same land parcel across different data sources.

[0034] Furthermore, the determination of spatial coordinates prioritizes data with high accuracy, recent time, and complete coverage, while the determination of ownership information prioritizes data with a unique subject, clear source, and continuous historical development. Therefore, the implementation process of this technical solution is not simply extracting two fields, but rather first organizing the data, then identifying valid fields, and then determining the priority order based on source quality, ultimately forming the spatial and subject foundations required for subsequent matching.

[0035] In some embodiments, after extracting spatial coordinates and ownership information, spatial topological overlay processing is performed on the archival data and the actual site records based on the spatial coordinates and ownership information to obtain an initial matching result including land parcel area and utilization type. Specifically, the archival data and the actual site records are first unified to the same spatial reference frame, so that the two types of land parcel information are expressed under the same coordinate reference, avoiding overall offset due to different spatial references. Subsequently, the positions of reference objects with stable location characteristics within the village area are compared. These reference objects include road edges, water body boundaries, ditch edges, long-standing building outlines, and public facility boundaries. These reference objects are used to identify whether there is overall translation, local offset, or boundary distortion between the two types of layers. If there is an overall offset, the position of the entire layer of land parcels is corrected first; if there is a local offset, boundary alignment processing is carried out according to regional blocks. Based on the aforementioned embodiments, it is clear that if spatial reference unification is not completed first, any subsequent area comparison and boundary determination may be distorted due to the offset of the underlying layer. Therefore, this processing is a prerequisite step for forming a true initial matching result.

[0036] In this embodiment, after the spatial benchmark is unified, the archived land parcels are used as benchmark objects. Candidate land parcels that overlap, contain, are adjacent to, or have boundary contact with the archival land parcels are searched in the on-site status records to form a candidate matching set. Subsequently, the extracted ownership information is introduced into the candidate matching set. Candidate objects that are spatially close but have no obvious connection between the subjects are excluded, while candidate objects that overlap spatially and have the same subject, a clear subject inheritance relationship, or whose connection has been confirmed by grassroots verification are retained. Thus, it can be seen that the method of this application does not simply determine the matching relationship based on spatial distance, but simultaneously incorporates spatial correspondence and subject correspondence into the judgment.

[0037] As a preferred embodiment, after retaining valid candidate objects, spatial topological overlay analysis is performed on each group of candidate objects. The overlapping areas, non-overlapping areas, archive-specific areas, and archive-specific areas are statistically analyzed between the archive plots and existing plots, and their intersection, containment, and adjacency states are recorded. Then, based on the overlay results, the areas of the archive plots, existing plots, and their overlapping areas are extracted. Simultaneously, the archive utilization type and the existing utilization type are extracted, thus forming the initial matching result. As can be seen from the above embodiment, the initial matching result is generated in the following order: first, spatially corresponding objects are determined; then, the subject association status is checked; then, area relationships are statistically analyzed; and finally, the utilization type comparison content is retained, thereby ensuring that the initial matching result has both spatial and business significance.

[0038] In this embodiment, after obtaining the initial matching results, the differences between the land parcel area and the utilization type are processed using data field comparison rules to obtain a set of field difference features. In specific implementation, the area field is first compared item by item, and the archived area and the current area are read in sequence. The area difference between the two is calculated, and the direction of the difference is recorded, that is, the current area is greater than the archived area or the current area is less than the archived area.

[0039] Furthermore, after determining the area difference, the source of the difference is identified by combining it with the overlapping area. If the overlapping area is large but the area difference is small, it is primarily determined to be a boundary offset or a general measurement error; if the overlapping area is small but the area difference is large, it is primarily determined to be a significant boundary change, patch splitting, patch merging, or change in the occupied area. In addition, it should be noted that the area difference alone is not sufficient to fully reveal the boundary change process. Therefore, it is also necessary to examine the boundary outline segment by segment, comparing the boundary turning point positions, edge direction, and outline shape changes in turn to identify whether the boundary change manifests as overall expansion, local intrusion, local contraction, or boundary redrawing.

[0040] In this embodiment, after processing the area field, a rule-based comparison is performed on the utilization type field. Specifically, a unified correspondence is first established between archival utilization types and actual utilization types. Types with the same business meaning but different expressions are grouped into the same category, and then reviewed item by item according to the unified category. If the archival utilization type and the actual utilization type point to the same business status, they are considered to be of the same type; if they belong to the same higher-level category but have different subcategories, they are considered to be of different levels; if they reflect significantly different land use statuses, they are considered to be in conflict of utilization types. It should be noted that the data field comparison rules are not set temporarily, but are formed item by item based on the local land management classification standards, on-site survey standards, historical ledger expressions, and manual review conclusions. The purpose is to eliminate superficial conflicts caused only by differences in name or standards. Therefore, by processing area differences, reviewing boundary contours, and reviewing utilization types, the key differences in the initial matching results can be clearly broken down item by item. Subsequently, the results of area difference, boundary contour change, utilization type comparison, ownership consistency, and spatial overlay are summarized to form a field difference feature set. Each item in this feature set corresponds to a specific land parcel and clearly records the difference field, difference size, difference direction, difference location, difference type, and associated basis.

[0041] In this embodiment, if the area difference in the field difference feature set is greater than a preset area threshold, it is determined to be a boundary change deviation. Specifically, the area difference in the field difference feature set is read one by one and then compared with the area threshold. The area threshold is used to distinguish between normal errors and substantial boundary changes. Its determination process includes the following: First, within the current area, samples of plots that have undergone manual verification and have clear conclusions are selected. The archived area and the on-site verified area in the samples are compared item by item to identify the actual distribution range of normal measurement errors. Then, combined with the current area's map accuracy requirements, plot size characteristics, village layout density, historical survey accuracy, and on-site data collection accuracy, a reasonable upper limit for the area error in this area is determined. Afterwards, sample plots that significantly exceed this reasonable upper limit and have been manually verified as boundary changes are used as references to extract the area boundary that can distinguish between ordinary errors and boundary changes, thus forming the area threshold. Therefore, this area threshold is formed comprehensively based on the regional sample verification results, survey accuracy requirements, and actual map features.

[0042] In this embodiment, when the area difference exceeds the area threshold, it indicates that the difference has exceeded the fluctuation range corresponding to normal measurement error, input error, and slight offset of the map patch, and should be identified as a boundary change deviation. Furthermore, if the area difference does not exceed the area threshold, but the boundary outline shows obvious expansion, obvious contraction, local encroachment, boundary breakage, or traces of map patch splitting and merging, a supplementary review is conducted on the plot simultaneously to prevent the omission of substantial boundary changes based solely on the size of the area difference.

[0043] In another specific embodiment of this application, it can be seen that for fragmented plots, irregularly shaped plots, or plots where boundary changes occur concentrated in localized areas, the boundary contour change status plays a more prominent supporting role in the deviation determination. Therefore, even if the area difference value has not reached the limit, further review is still required in conjunction with the contour change status. As can be seen from the above, the area difference value is first formed, then an area threshold is formed based on the sample and accuracy requirements, then the threshold judgment is performed, and the result is supplemented and corrected based on the boundary contour status, ultimately forming a conclusion on the boundary change deviation.

[0044] In this embodiment, after the boundary change deviation identification is completed, the differences between the archive data and the actual status records of abandoned plots are initially marked based on the deviation source classification results, resulting in a list of inconsistencies. In specific implementation, the cause of each difference is first identified layer by layer. If multiple plots in the same area have approximately the same distance offset along the same direction, it is preferentially identified as a coordinate reference deviation; if the spatial location of the plots is consistent but the subject name, subject number, or subject inheritance relationship is inconsistent, it is preferentially identified as an ownership association deviation; if the area difference exceeds the area threshold and the boundary contour comparison shows outward expansion, inward contraction, splitting, merging, or encroachment characteristics, it is identified as a boundary change deviation; if the boundary is basically stable but the archive utilization type and the on-site utilization type are inconsistent for a long time, it is identified as a utilization status lag deviation; if the missing key fields make it impossible to complete an effective judgment, it is identified as an information missing deviation. It can be seen that the method of this application does not treat all differences uniformly, but first classifies the differences into clear source categories, and then forms a targeted basis for subsequent analysis based on the source categories.

[0045] In this embodiment, after identifying the sources of deviation, preliminary annotations are generated for each discrepancy. Specifically, the following information is recorded sequentially: plot number, discrepancy field name, location of the discrepancy, discrepancy value, discrepancy category, judgment basis, associated data source, collection time, and processing order. As a preferred embodiment, the processing order is determined by first examining the impact of the discrepancy on plot boundary identification, area calculation, abandonment status assessment, and subsequent reduction assessment; then examining whether the source of the discrepancy is clear, whether the evidence is sufficient, and whether the verification path is direct. Differences with high impact, clear sources, and clear verification paths are prioritized, while differences with lower impact or those requiring further evidence are prioritized. Subsequently, all preliminary annotations for the same plot are compiled to form a list of inconsistencies. This list of inconsistencies is not simply a summary of problems, but rather clearly indicates where the discrepancy exists, its type, its extent, the supporting evidence, and which aspects should be prioritized for processing for each plot.

[0046] Example 3: S2 includes acquiring remote sensing monitoring data and a list of inconsistencies for abandoned land parcels; parsing the list of inconsistencies and mapping it to the remote sensing monitoring data to extract land parcel attribute features; dividing the land parcel attribute features into vector dimensions to construct a dimension division matrix; calculating the Euclidean distance between the orthogonal components in the dimension division matrix and the standard land parcel feature template to quantify the attribute deviation value; calculating the deviation weight allocation coefficient based on the attribute deviation value; ranking the importance of each attribute dimension according to the deviation weight allocation coefficient to determine the deviation distribution characteristics of the land parcel attributes.

[0047] In this embodiment, remote sensing monitoring data and a list of inconsistencies for abandoned land parcels are first acquired. The list of inconsistencies is then parsed and mapped to the remote sensing monitoring data to extract land parcel attribute features. Specifically, remote sensing monitoring data corresponding to abandoned land parcels is collected around the target area, and the acquired data undergoes temporal, coverage integrity, and spatial accuracy filtering. Temporal filtering involves comparing the acquisition time of the remote sensing monitoring data with the formation time of the actual site record, prioritizing data with shorter time intervals to avoid discrepancies between the actual site conditions and the image conditions due to large differences in imaging time points. Coverage integrity filtering examines whether the image completely covers the target land parcel and its surrounding areas to prevent partial omissions from affecting attribute identification. Spatial accuracy filtering compares the image resolution with the land parcel scale to confirm that the image has the ability to identify land parcel boundaries, building outlines, bare land distribution, vegetation cover, and road proximity relationships.

[0048] As a preferred implementation, after screening, the remote sensing monitoring data undergoes geometric correction, image stitching, brightness equalization, and boundary cropping. Geometric correction corrects positional offsets in the remote sensing images by using existing spatial benchmarks. Image stitching connects multiple remote sensing images covering adjacent areas according to a unified spatial relationship, ensuring image continuity. Brightness equalization eliminates brightness and color differences between images, making similar land features appear more consistent across different images. Boundary cropping separates local image regions corresponding to the target land area from the overall remote sensing image based on the spatial boundaries of abandoned land parcels. Therefore, before subsequent analysis, the remote sensing monitoring data undergoes basic processing related to spatial location, temporal state, and image quality, ensuring a reliable image foundation for subsequent attribute extraction.

[0049] In this embodiment, after acquiring remote sensing monitoring data, the inconsistency list is then parsed. Specifically, each discrepancy record in the inconsistency list is read sequentially, and the corresponding land parcel number, discrepancy field, discrepancy category, discrepancy location, discrepancy degree, and basis for discrepancy are identified. The discrepancy field indicates which of the following occurs: area, use type, boundary shape, ownership status, or idle status; the discrepancy category indicates whether the discrepancy is due to boundary changes, use status conflicts, information lag, or other sources of anomalies; the discrepancy location indicates whether the discrepancy occurs in the entire land parcel or a local area; the discrepancy degree indicates the level of the discrepancy in terms of value or status; and the basis for discrepancy indicates what data supports the discrepancy. It should be noted that the parsing of the inconsistency list is not merely a matter of reading fields, but rather restoring each discrepancy record to a structured semantic unit that "corresponds to which land parcel, occurs under which attribute, falls within which part of the land parcel, and affects what kind of status judgment."

[0050] In this embodiment, after the inconsistency list is parsed, it is mapped to remote sensing monitoring data to extract land parcel attribute features. Specifically, the spatial range of the target land parcel is first determined based on the land parcel number, and then this spatial range is projected onto the preprocessed remote sensing monitoring data to lock the image area corresponding to the land parcel. Subsequently, the corresponding image recognition content is selected based on the difference fields and difference categories. If the difference field involves the type of use, then the surface cover status, building distribution status, bare land distribution status, vegetation growth status, and road connectivity status are identified sequentially within the land parcel image area to determine the main use attributes of the land parcel from a remote sensing perspective. If the difference field involves boundary changes, then the boundary direction, edge continuity status, boundary inflection point location, and adjacent feature boundary status are identified segment by segment along the edge of the land parcel to determine whether there are outward expansion, contraction, encroachment, or reconstruction phenomena of the boundary. If the difference field involves vacancy status, then the building damage status, courtyard abandonment status, weed coverage status, bare surface status, and long-term inactivity traces are identified to determine whether the land parcel has a tendency to be abandoned. If the difference field involves changes in the internal structure of the land parcel, then the building footprint ratio, vacant area ratio, internal road accessibility status, and paved area distribution status are identified to determine whether the internal spatial organization of the land parcel has changed significantly.

[0051] As a preferred embodiment, after image recognition is completed, the recognition results are converted into unified attribute items, forming an attribute feature set corresponding to each land parcel. This attribute feature set includes at least spatial morphological features, land cover features, land use status features, structural status features, and ecological characterization features. As can be seen from the above, the implementation process of this technical solution involves first locating the land parcel image area, then selecting recognition content according to the difference type, and finally consolidating the image recognition results into unified attribute feature items, thereby completing the extraction of land parcel attribute features.

[0052] In this embodiment, after extracting the attribute features of the land parcels, the attribute features are divided into vector dimensions to construct a dimension partitioning matrix. Specifically, the extracted attribute features of each land parcel are first uniformly sorted to ensure that similar attributes across different parcels are in the same position. This uniform sorting process involves first establishing an attribute directory, then arranging the attributes in the order of spatial morphology, utilization status, structural status, ecological representation, and change status, providing a unified basis for comparing the attribute positions of different land parcels in the future.

[0053] Subsequently, the attributes were categorized according to their business implications. Specifically, attributes reflecting the geometric outline, boundary complexity, area variation, and morphological regularity of the land parcel were assigned to the spatial dimension; attributes reflecting building proportion, vacancy rate, utilization intensity, and land cover status were assigned to the utilization dimension; attributes reflecting vegetation cover, bare land exposure, and ecological disturbance status were assigned to the ecological dimension; and attributes reflecting historical changes, abrupt changes, and temporal continuity were assigned to the change dimension. It should be noted that this dimensional division is not an arbitrary combination, but rather based on the needs of land parcel status identification and deviation quantification, grouping attributes with similar meanings, similar directions of influence, and common characterizing capabilities for the same type of status into the same dimension. Thus, after dimensional division, the originally scattered attribute characteristics were organized into several attribute groups with clear business orientations.

[0054] In this embodiment, after completing the attribute classification, a dimensional partitioning matrix is ​​further constructed. Specifically, land parcels are used as the vertical arrangement basis, and attribute dimensions are used as the horizontal arrangement basis. The attribute values ​​of each land parcel in each dimension are sequentially filled into the corresponding positions, thus forming a multidimensional data table with a unified row and column structure. For cases where multiple attributes are contained within the same dimension, they are first arranged according to a pre-defined attribute order, and then the attribute values ​​are combined to form the attribute sequence corresponding to that dimension.

[0055] As a preferred implementation, to avoid imbalances in subsequent comparisons due to different attribute value ranges, a consistency processing is performed on the attribute values ​​before constructing the dimensionality partitioning matrix. This consistency processing involves uniformly converting the value ranges of each attribute based on its actual distribution range across all samples, ensuring that different attributes reflect their relative strengths on the same comparative basis. Therefore, the formation process of the dimensionality partitioning matrix includes not only attribute classification but also unified sorting, structural organization, and numerical consistency processing, thereby guaranteeing that the matrix possesses the conditions for subsequent bias quantification.

[0056] In this embodiment, after constructing the dimensional partitioning matrix, the Euclidean distance between the orthogonal components in the dimensional partitioning matrix and the standard plot feature template is calculated to quantify the attribute deviation value. Specifically, the standard plot feature template is first determined. The formation process of the standard plot feature template involves selecting standard plot samples with clear status, accurate boundaries, clear utilization relationships, and manual verification within the target area or nearby areas. Then, typical manifestations of various attribute characteristics are extracted from these standard plot samples and arranged in an order consistent with the dimensional partitioning matrix to form a standard reference structure. It should be noted that the standard plot feature template does not originate from a single plot, but rather from the comprehensive characteristics of a group of verified plots. Its purpose is to reflect the overall characteristic level that each dimension attribute should present under normal conditions. The determination criteria for the standard plot samples mainly include the accuracy of boundaries, the clarity of utilization status, the consistency of on-site verification, the continuity of historical data, and the stability of image representation.

[0057] In this embodiment, after determining the standard land parcel feature template, each dimension in the dimensional partitioning matrix is ​​independently decomposed to form corresponding orthogonal components. In this embodiment, orthogonal components refer to treating attribute groups of different dimensions as distinct independent analysis objects. This ensures that spatial, utilization, ecological, and change dimensions are calculated separately during deviation quantification, avoiding the mixing of attributes from different dimensions that could affect the deviation identification results.

[0058] In practice, the process begins by extracting all attribute values ​​for a specific land parcel within a given dimension from the dimensionality matrix. Then, corresponding attribute values ​​for the same dimension are extracted from the standard land parcel feature template. The two sets of attribute values ​​are then compared item by item at their corresponding positions to identify the degree of difference between each attribute and the standard template. Subsequently, a comprehensive distance calculation is performed on the differences in each attribute within that dimension to generate the attribute deviation value for that dimension. This distance calculation involves first determining the difference between the target land parcel and the standard template for each attribute, then uniformly converting the overall dispersion of these differences, and finally generating a numerical result that reflects the overall deviation of that dimension. Therefore, the attribute deviation value is not a single attribute difference, but a comprehensive representation of multiple attribute differences within the same dimension. A larger value indicates a more significant deviation of the land parcel from the standard state in that dimension.

[0059] In this embodiment, the deviation weight allocation coefficient is calculated based on the attribute deviation value. Specifically, the attribute deviation values ​​for each dimension of the same land parcel are first summarized to identify the proportion of each dimension's deviation in the overall deviation. Then, the weight of each dimension's deviation is calculated based on the actual impact of each dimension on the identification of abandoned land parcel status. The process of determining the deviation weight allocation coefficient involves first determining the relative magnitude of each dimension's deviation value; dimensions with larger deviation values ​​have a stronger foundation for influence in the overall analysis. Secondly, the indicative role of this dimension in identifying abnormal land parcel status in historical verification samples is examined; dimensions with significant indicative roles have a correspondingly higher weight. Thirdly, it is determined whether the dimension attribute is directly related to boundary stability, idle status, degree of utilization imbalance, and land parcel reduction potential; the stronger the direct correlation, the higher the weight.

[0060] As a preferred implementation, the weight results of each dimension are then uniformly balanced to ensure that the weight results of all dimensions are on the same comparative basis. Therefore, the formation process of the deviation weight allocation coefficients does not simply rely on the magnitude of the deviation values ​​for direct sorting, but simultaneously considers the degree of deviation, the identification function, and the business impact, thus making the weight results more consistent with the needs of land parcel status analysis.

[0061] In this embodiment, the importance of each attribute dimension is ranked according to the deviation weight allocation coefficient to determine the deviation distribution characteristics of the land parcel attributes. Specifically, the deviation weight allocation coefficients of the same land parcel on each dimension are first arranged from high to low to obtain the dimension importance order of the land parcel. A dimension with a high weight indicates that the dimension plays a dominant role in the formation of the current land parcel's deviation; a dimension with a low weight indicates that although the dimension has a deviation, its supporting effect on the overall abnormal state of the current land parcel is relatively weak.

[0062] Subsequently, by combining the actual magnitudes of the deviation values ​​across each dimension, the concentration and dispersion of deviations across different dimensions are identified. If the deviations are mainly concentrated in the spatial and utilization dimensions, it indicates that the anomalies of the land parcel are primarily manifested in boundary changes and utilization imbalances; if the deviations are mainly concentrated in the ecological and change dimensions, it indicates that the anomalies of the land parcel are more characterized by abnormal surface disturbance and historical change traces. Thus, the so-called deviation distribution characteristics of land parcel attributes not only reflect which dimension has the largest deviation, but also how the deviations are distributed across multiple attributes, identifying which are dominant and which are secondary. This result can be directly used for subsequent identification of correction directions and selection of correction criteria.

[0063] Furthermore, it should be noted that, in another specific embodiment of this application, when the number of regional samples is large and the types of land parcels are diverse, the dimensional ranking results of similar land parcels can be grouped and summarized to form a regional-level deviation distribution reference structure. The implementation process involves first grouping land parcels according to their land use background, spatial location characteristics, village type, and historical change frequency, and then performing centralized statistics on the dimensions with high deviation weights in each group of land parcels.

[0064] Example 4: S3 includes obtaining the deviation distribution characteristics of land parcel attributes. If the extracted dimensional deviation value exceeds a preset threshold, historical data reference records are obtained for the dimension. Spatial coordinate offsets are extracted based on the historical data reference records, and the spatial coordinate offsets are classified to obtain first deviation data. Second deviation data is obtained by weighting the first deviation data through a dynamic adjustment mechanism, and deviation analysis results are obtained based on the second deviation data. If the dispersion value in the deviation analysis results is greater than a preset dispersion, a correction angle value is extracted based on the deviation analysis results, and a new correction direction vector is generated using the correction angle value.

[0065] In this embodiment, the deviation distribution characteristics of land parcel attributes are first obtained. If the extracted dimensional deviation value exceeds a preset threshold, historical data reference records are obtained for that dimension. Specifically, the deviation distribution results of the same land parcel across each attribute dimension are read item by item, and the deviation values ​​corresponding to the spatial dimension, utilization dimension, ecological dimension, and change dimension are extracted sequentially and arranged according to a pre-set dimensional order. Subsequently, the deviation value of each dimension is compared item by item with the preset threshold. It should be noted that the preset threshold is used to distinguish between general deviations and significant deviations that require further tracking and analysis. Its determination process is not arbitrary. Instead, sample plots that have been manually verified and have clear conclusions are first selected in the target area. Then, the distribution of deviation values ​​in each dimension of the sample plots is statistically reviewed to identify the normal fluctuation range. After that, the upper limit of the normal fluctuation range is further corrected by combining the influence of this dimension on the plot status identification, boundary judgment, utilization change judgment, and reduction assessment results, thereby forming a threshold limit that can reflect the degree of abnormal deviation. In some embodiments, the threshold also needs to be adjusted for regional adaptation by combining the fragmentation of regional plots, remote sensing accuracy, completeness of historical data, and deviation quantification caliber in the previous steps.

[0066] In this embodiment, when the deviation value of a certain dimension exceeds the aforementioned threshold, it indicates that the attribute deviation corresponding to that dimension has exceeded the normal fluctuation range, suggesting a possible continuous anomaly or significant deviation. Therefore, historical data reference records are obtained for that dimension. Specifically, based on the target plot's plot number, spatial location range, and dimensional attribute category, the historical database is first searched for all corresponding archive records, remote sensing monitoring records, survey and verification records, correction records, and related change records. Subsequently, the retrieved historical records are sorted chronologically to form a historical data sequence reflecting the plot's change process.

[0067] As a preferred implementation, after forming the historical data sequence, the historical data content directly related to the current threshold is selected based on the specific dimension that exceeds the threshold. For example, when the threshold dimension reflects spatial morphological anomalies, priority is given to retaining previous boundary mapping results, boundary patch change records, and historical image edge recognition results; when the threshold dimension reflects abnormal utilization status, priority is given to retaining previous utilization type records, idle status verification records, building status change records, and remote sensing interpretation results; when the threshold dimension reflects abnormal change traces, priority is given to retaining multi-temporal image records, historical change node records, and previous manual review results.

[0068] In this embodiment, spatial coordinate offsets are extracted from historical data reference records, and the spatial coordinate offsets are classified to obtain the first deviation data. Specifically, spatial coordinate information corresponding to the target plot is extracted from the historical data reference records point by point. This spatial coordinate information not only refers to the center position of the plot but also includes the position of boundary nodes, the boundary outline range, and the spatial relationship with adjacent features. Subsequently, the spatial coordinate information of each time point is arranged in chronological order, and based on the coordinate changes between adjacent time points, the spatial position changes of the plot in different periods are identified segment by segment.

[0069] Specifically, the changes in the boundary node positions at adjacent time points are first compared to identify phenomena such as boundary expansion, boundary contraction, local encroachment, local retreat, or boundary redrawing. Then, the changes in the center positions of plots at adjacent time points are compared to identify the overall offset direction and magnitude. Finally, the changes in the plot outline shape are considered to identify whether overall movement and local deformation occur simultaneously. It should be noted that the spatial coordinate offset is not a single numerical value, but rather a spatial change representation composed of the offset direction, offset magnitude, and offset location.

[0070] In this embodiment, after extracting the spatial coordinate offset, the spatial coordinate offset is classified to obtain the first deviation data. Specifically, the spatial coordinate offset is first classified according to the offset direction, distinguishing between unidirectional offset, multidirectional offset, and directionally unstable offset. Unidirectional offset is used to characterize the situation where the land parcel generally changes in the same direction over consecutive time points; multidirectional offset is used to characterize the situation where the land parcel exhibits two or more obvious directional changes at different time periods; and directionally unstable offset is used to characterize the situation where the offset direction of the land parcel changes frequently over consecutive time periods without a clear dominant direction.

[0071] Subsequently, the spatial coordinate offsets are classified according to their magnitude, into slight offset, moderate offset, and significant offset. The classification of offset magnitude is mainly determined by combining the boundary change magnitude in historical samples, the allowable range of surveying accuracy, and the characteristics of the land parcel scale. Then, they are classified according to time series change characteristics, into persistent offset, phased offset, and sudden offset. Persistent offset is used to characterize situations where multiple consecutive time points maintain similar change directions and trends; phased offset is used to characterize situations where changes are concentrated within a certain period but relatively stable in other periods; and sudden offset is used to characterize situations where significant changes occur at a single point or within a short period. As a preferred embodiment, after completing the classification of direction, magnitude, and change characteristics, the classification results corresponding to the same land parcel in the same dimension are aggregated to form structured first deviation data. This first deviation data includes at least the offset direction category, offset magnitude category, offset persistence characteristic, and corresponding time interval.

[0072] In this embodiment, a second deviation data is obtained by weighting the first deviation data through a dynamic adjustment mechanism, and a deviation analysis result is obtained from the second deviation data. Specifically, a weight allocation order is first established for the data at each time point in the first deviation data. This weight allocation is not an even distribution, but rather differentiated based on the reference value of each time point data for judging the current deviation.

[0073] Specifically, the first deviation data that is closer to the current time point reflects the true state of the current plot more directly, and therefore has a high weight; the first deviation data with clear data sources, standardized collection processes, and high spatial accuracy has a high degree of reliability, and therefore has a high weight; the first deviation data that is directly related to the current threshold dimension has a strong supporting role, and therefore has a high weight.

[0074] In this embodiment, after establishing the weight allocation order, the first deviation data is weighted and integrated. Specifically, the first deviation data is read sequentially by time, and then each item is added together with its corresponding weight, ensuring that the deviation information at each time point reflects different degrees of influence in the overall calculation. As a preferred embodiment, during the weighting process, data that significantly deviates from the overall trend needs to undergo anomaly screening. The anomaly screening process involves first identifying whether the offset direction or magnitude of a certain time point is significantly inconsistent with its preceding and following time points; if the change at that point cannot be supported by other historical records, image features, or survey records, it is determined that the point may be affected by acquisition errors, annotation errors, or local anomalies, and its influence in the weighted integration is then reduced.

[0075] After the above processing, the second deviation data is generated. It should be noted that the second deviation data differs from the first deviation data. The first deviation data mainly reflects the classification results of historical offset information, while the second deviation data reflects the overall deviation status after adjustments for time factors, reliability, and correlation. Subsequently, deviation analysis is conducted on the second deviation data. The specific content of the deviation analysis includes identifying whether the deviation is continuously expanding, whether it is stabilizing, whether it occurs concentratedly in a local time period, and whether the dominant direction of the deviation is clear. As can be seen, this scheme first establishes a weighted basis based on time, source, and correlation, then dynamically integrates the classified first deviation data, and judges the deviation change trend based on the integration results, ultimately forming the deviation analysis results.

[0076] In this embodiment, if the dispersion value in the deviation analysis result is greater than the preset dispersion value, a correction angle value is extracted based on the deviation analysis result, and a new correction direction vector is generated using the correction angle value. Specifically, the dispersion of the time-series deviation state reflected by the second deviation data is first analyzed. The dispersion is used to characterize the consistency or dispersion of deviation performance at different time points. Specifically, the analysis first confirms whether the deviation directions at each time point tend to be consistent. If the deviation directions at most time points are similar, it indicates low directional dispersion; if multiple significantly different deviation directions appear at different time points, it indicates high directional dispersion. Then, it confirms whether the changes in the deviation amplitude at each time point are concentrated. If the deviation amplitude fluctuates around a similar level, it indicates low amplitude dispersion; if the deviation amplitude differs significantly between different time points, it indicates high amplitude dispersion. Finally, the directional dispersion state and the amplitude dispersion state are combined to form a dispersion value, which is used to characterize the stability of the deviation in this dimension over time.

[0077] It should be noted that the purpose of the preset dispersion is to distinguish between stable and unstable deviations. The determination process involves first selecting plots with confirmed stable trends and plots with confirmed abnormal fluctuation trends from historical samples, then extracting their time-series deviation performance, confirming the differences in the degree of directional dispersion and amplitude dispersion between the two, and then using the boundary that can distinguish between stable samples and abnormal fluctuation samples as the preset dispersion. In some embodiments, it is also necessary to adapt and adjust the dispersion boundary by taking into account regional plot morphology differences, data time intervals, and the completeness of historical records.

[0078] In this embodiment, when the dispersion value is greater than the preset dispersion, it indicates that the deviation in this dimension exhibits significant dispersion characteristics over time, suggesting the possible coexistence of multiple deviation directions or drastic fluctuations in deviation amplitude. Therefore, it is necessary to re-extract the correction direction. Specifically, the dominant deviation direction is first identified based on the deviation analysis results. This identification process prioritizes directions with high frequency of occurrence, large deviation amplitudes, and stronger correlation to recent states among deviation directions at multiple time points as the primary correction reference direction. Subsequently, this dominant deviation direction is compared with a pre-set standard reference direction to identify its relative deflection degree, thereby obtaining the correction angle value. The correction angle value reflects the directional adjustment that the dominant deviation direction of the target plot should make relative to the standard state in this dimension. Then, this correction angle value is combined with the current deviation amplitude state to form a new correction direction vector. This new correction direction vector at least reflects the correction direction, correction tendency, and correction strength, and is used in subsequent steps to extract key correction reference points and determine priority correction criteria. Therefore, this scheme first determines the dominant direction through time series analysis, then determines its deflection relationship with the standard reference direction, and finally forms a corrected direction vector with clear directional significance.

[0079] In addition, it should be noted that, as can be seen in another specific embodiment of this application, when the dispersion value does not exceed the preset dispersion, it indicates that the current dimensional deviation remains relatively stable in time sequence and the deviation direction does not show obvious disorder.

[0080] Example 5: S4 includes acquiring the original business data stream and calculating the correction direction vector; extracting key correction reference points based on the cluster density distribution of the correction direction vector; analyzing the mutual information correlation of the key correction reference points and filtering out highly correlated deviation attributes; retrieving the real-time status records corresponding to the highly correlated deviation attributes; comparing the key correction reference points with the real-time status records using a data consistency verification model to generate a matching degree index; constructing an attribute priority index sequence based on the matching degree index; parsing the first and second features of the attribute priority index sequence to determine the specific content of the priority correction basis.

[0081] In this embodiment, the original business data stream is first acquired and the correction direction vector is calculated. Key correction reference points are then extracted based on the cluster density distribution of the correction direction vector. Specifically, the original business data stream related to the target plot is retrieved from the business system. This original business data stream includes archive registration data stream, remote sensing interpretation data stream, field acquisition data stream, historical correction record data stream, and dynamically updated data stream. Subsequently, the data from different sources are sorted temporally and aligned spatially to form a unified data sequence within the same plot. As is known from the aforementioned embodiment, the correction direction vector has already been generated in S3; therefore, this correction direction vector is directly read and associated with the original business data stream.

[0082] Furthermore, cluster analysis is performed on the correction direction vectors. The specific implementation process is as follows: First, the correction direction vectors corresponding to multiple plots or different time points of the same plot are collected and organized. Then, they are grouped according to directional similarity and amplitude proximity, so that vectors with consistent or similar directions are grouped into the same category. Subsequently, the number of vectors in each category is counted, and their spatial concentration is identified, thus forming a cluster density distribution result. As a preferred embodiment, high-density clustered areas are regarded as the main correction trend concentration areas, and representative spatial locations within these areas are selected as key correction reference points. The selection process for these reference points prioritizes spatial locations with high consistency in correction direction, stable data sources, continuous historical changes, and located in areas of concentrated deviation. Therefore, this solution first aggregates business data streams, then clusters and groups the correction direction vectors, then identifies areas of concentrated density, and extracts key correction reference points that represent the main correction trend from these areas.

[0083] In this embodiment, the mutual information correlation degree of key correction reference points is analyzed to screen out highly correlated attributes of deviations. Specifically, multi-dimensional attribute data corresponding to the key correction reference points are first extracted, including spatial attributes, utilization attributes, structural attributes, ecological attributes, and change attributes. Then, the above attributes are correlated with the identified deviation results, and the degree of correlation between each attribute and the deviation is determined item by item. The mutual information correlation degree reflects the strength of the impact of a change in a certain attribute on the deviation result. Its implementation process is as follows: first, the synchronous occurrence of attribute changes and deviation changes in historical data is statistically analyzed; then, it is identified whether the attribute change persists before and after the deviation occurs; finally, the closeness of the correlation between the attribute change and the deviation is comprehensively judged.

[0084] In some embodiments, the correlation degree of all attributes is sorted, and attributes with high correlation are selected as highly correlated attributes of deviation. It should be noted that highly correlated attributes of deviation are not simply determined based on a single matching result, but are formed based on a comprehensive analysis of multi-time-series data and multi-sample statistical results. Therefore, this scheme first extracts multi-dimensional attribute data around key reference points, then identifies the correlation between attributes and deviations through historical data comparison, and finally selects the set of attributes that have a major impact on the formation of deviations.

[0085] In this embodiment, real-time status records corresponding to highly correlated deviation attributes are retrieved. A data consistency verification model is used to compare key correction reference points with the real-time status records to generate a matching degree index. Specifically, based on the filtered highly correlated deviation attributes, the corresponding real-time status records are retrieved item by item. These real-time status records include the latest remote sensing imagery, on-site acquisition records, dynamic monitoring data, and the latest business reporting information. Subsequently, the historical attribute statuses corresponding to the key correction reference points are compared item by item with the real-time status records.

[0086] Furthermore, a data consistency verification model is used to determine the degree of consistency between historical reference states and the current state. The specific implementation process is as follows: First, spatial location consistency is verified to determine whether the positions of key reference points in real-time data have significantly shifted; then, attribute values ​​are verified to determine whether the current attribute values ​​maintain a continuous relationship with historical reference values; next, attribute states are verified to determine whether there have been abrupt changes in utilization, structural, and ecological states. Subsequently, the consistency results of each step are integrated to form a matching degree index. The matching degree index reflects the degree of consistency between the key corrected reference point in the current state and the historical reference state; a higher value indicates stronger consistency, while a lower value indicates a more significant deviation. Therefore, this scheme first retrieves real-time data, then performs spatial and attribute consistency verification item by item, and finally integrates the verification results into a matching degree index.

[0087] In this embodiment, an attribute priority index sequence is constructed based on the matching degree index. The first and second features of the attribute priority index sequence are analyzed to determine the specific content of the priority correction basis. In specific implementation, the matching degree indices corresponding to all highly correlated deviation attributes are first summarized and sorted from low to high according to the matching degree. Attributes with low matching degree indicate that their current state differs greatly from the historical reference state, and therefore should be processed first in the correction process; attributes with high matching degree indicate that their current state is relatively stable and can be treated as secondary processing objects.

[0088] As a preferred implementation, after sorting, the attributes are arranged sequentially according to the sorting results to form an attribute priority index sequence. Then, the first attribute in this sequence is parsed. The parsing process for the first attribute is as follows: first, the dimension to which the attribute belongs and its corresponding deviation type are identified; then, combined with the spatial location of the key correction reference point and the correction direction vector, the specific correction content of the attribute in the current land parcel is determined. For example, when the first attribute is a boundary-related attribute, the boundary correction basis is determined first; when the first attribute is a utilization status attribute, the utilization status correction basis is determined first. Therefore, by parsing the first feature of the priority index sequence, the correction basis that should be prioritized for the current land parcel can be clearly identified.

[0089] Example 6: S5 includes obtaining the correction basis text, constructing a semantic feature vector using a bidirectional long short-term memory network; performing correlation analysis between the semantic feature vector and the land parcel attributes to determine the applicable scope boundary; calculating the information entropy value of each attribute within the applicable scope boundary, and establishing the basis weight value based on the information entropy value; calculating a weighted correction urgency score by combining the current status value of the abandoned land parcel attributes and the basis weight value; generating an attribute correction priority ranking sequence based on the weighted correction urgency score; and generating an abandoned land parcel correction guidance list based on the attribute correction priority ranking sequence.

[0090] In this embodiment, the text used as the basis for correction is obtained, and a semantic feature vector is constructed using a bidirectional long short-term memory network. Specifically, the priority correction criteria determined in the preceding steps are first extracted and organized into standardized text data. This text data includes attribute description information, deviation explanation information, and correction suggestion information. Subsequently, the text undergoes word segmentation, semantic segmentation, and invalid information removal to ensure a clear and parsable text structure.

[0091] Furthermore, the processed text is input into a bidirectional long short-term memory network model to encode the semantic context of the text, thus fully expressing the dependencies between words in the text. As a preferred embodiment, this model generates a corresponding semantic feature vector for each corrected text segment. This vector can represent the semantic information contained in the text and its importance. Therefore, this scheme first completes the text normalization process, and then extracts semantic features through the model, thereby achieving a vectorized representation of the text.

[0092] In this embodiment, semantic feature vectors are correlated with land parcel attributes to determine the boundaries of the applicable scope. Specifically, the semantic feature vectors are first matched against existing attribute data of the land parcels to identify the attribute categories involved in the semantic content, such as boundary attributes, utilization status attributes, or structural attributes. Then, a correlation strength analysis is performed on each attribute category to determine the degree of matching between the attributes involved in the semantic features and the actual attributes of the land parcel. Furthermore, by combining the distribution of values ​​in each dimension of the semantic feature vector, it is identified which attributes play a major role in the current correction basis. Therefore, attributes with high correlation are identified as target attributes within the applicable scope, and the boundaries of the applicable scope are defined accordingly, thus clarifying which attribute sets are applicable to this correction basis.

[0093] In this embodiment, the information entropy value of each attribute within the applicable scope boundary is calculated, and the weight value is determined based on the information entropy value. Specifically, for each attribute within the applicable scope, its value distribution across different land parcels or at different time points is statistically analyzed. Then, the information entropy value of each attribute is calculated based on the dispersion of its values. It should be noted that the information entropy value reflects the degree of uncertainty of an attribute; the more dispersed the attribute's value changes, the larger its information entropy value, indicating that the attribute plays a more significant role in distinguishing the states of different land parcels. Based on the aforementioned embodiment, it is clear that the importance of an attribute in the correction analysis is closely related to its distinguishing ability; therefore, weights are assigned to each attribute based on its information entropy value.

[0094] As a preferred implementation, attributes with higher information entropy values ​​are assigned higher weights, and attributes with lower information entropy values ​​are assigned lower weights, thus forming weighted values. Therefore, this scheme first statistically analyzes the attribute distribution, then calculates the information entropy value, and finally converts the information entropy value into attribute weights.

[0095] In this embodiment, a weighted correction urgency score is calculated by combining the current attribute value of abandoned land parcels with weighted values. An attribute correction priority ranking sequence is generated based on the weighted correction urgency score, and an abandoned land parcel correction guidance list is generated based on the attribute correction priority ranking sequence. Specifically, the current attribute status value of each land parcel is first extracted and combined with the corresponding attribute's weighted value for calculation. During the calculation process, the attribute deviation degree and weight value are comprehensively processed, so that attributes with larger deviations and higher weights receive higher urgency scores. Furthermore, the urgency scores of all attributes are summarized and sorted according to their scores, thus forming an attribute correction priority ranking sequence. Therefore, attributes ranked higher indicate that their correction needs are more urgent in the current land parcel.

[0096] As a preferred embodiment, the correction basis for each attribute is extracted according to the priority sequence of attribute correction, and a correction guidance list for abandoned land parcels is formed in the sorted order. This list includes at least the attribute to be corrected, the source of the correction basis, the correction priority, and the corresponding processing order. It should be noted that, in another specific embodiment of this application, when the urgency scores of multiple attributes are close, the sorting results can be adjusted based on the correlation between the attributes to improve the rationality of the correction guidance.

[0097] Example 7: S6 includes obtaining a list of correction guidelines for abandoned land parcels, using a path association algorithm to match the correction guidelines list with the land parcel attribute data to obtain bound attribute update paths; analyzing the bound attribute update paths to obtain update frequency indicators, setting dynamic refresh time intervals based on update frequency indicators, and determining the refresh sequence of the paths; if the refresh sequence of the paths exceeds a threshold, using a frequency adjustment mechanism to optimize the dynamic refresh time interval to obtain an optimized refresh configuration, and deeply binding the correction guidelines list with the bound attribute update paths according to the optimized refresh configuration to determine the final correction execution sequence.

[0098] In this embodiment, a correction guidance list for abandoned land parcels is obtained, and a path association algorithm is used to match the correction guidance list with the land parcel attribute data to obtain the bound attribute update path. In specific implementation, the correction guidance list for abandoned land parcels generated in the previous steps is first retrieved, and the attributes to be corrected, the basis for correction, the order of correction, and the processing requirements in the list are analyzed item by item, so that each correction task corresponds to a clear attribute object and processing target.

[0099] Subsequently, the land parcel attribute data corresponding to the target abandoned land parcel is retrieved. This land parcel attribute data includes boundary attributes, area attributes, utilization status attributes, structural status attributes, ownership association attributes, and historical update records. Based on the aforementioned embodiments, the preceding steps have already generated an attribute correction priority ranking result. Therefore, this step first establishes the correspondence between correction tasks and attribute objects according to this ranking result, and then constructs the attribute update chain based on the dependencies, data transmission relationships, and processing order relationships between each attribute.

[0100] Furthermore, the execution process of the path association algorithm is as follows: first, the attribute category corresponding to a certain correction task is identified; then, the processing node of that attribute in the land parcel attribute data is located; and finally, the correction basis node, attribute verification node, status update node, and result confirmation node are connected in sequence to form the attribute update path corresponding to the correction task. Therefore, after this step, each correction item in the correction guidance list forms a one-to-one correspondence with the specific processing node in the land parcel attribute data, thus obtaining the bound attribute update path.

[0101] In this embodiment, the bound attribute update paths are analyzed to obtain update frequency indicators. Based on these indicators, dynamic refresh intervals are set to determine the refresh sequence of the paths. Specifically, each bound attribute update path is analyzed individually to extract the update time, number of updates, time intervals between adjacent updates, and the source record triggering the update for each path during historical processing. Then, the update activity of the same path at different time stages is statistically analyzed to form the corresponding update frequency indicator. The process for determining the update frequency indicator is as follows: first, the total number of updates occurring on the path within a preset time range is counted; then, the distribution of time intervals between each update is analyzed; finally, the update activity level of the path in the historical stage is determined by combining this with the source type triggering the update. Paths with more update frequency and shorter update intervals correspond to higher update frequency indicators; paths with fewer update frequency and longer update intervals correspond to lower update frequency indicators.

[0102] Subsequently, a dynamic refresh interval is set based on the update frequency index. Attribute update paths with high update frequency indices are assigned shorter refresh intervals, while those with low update frequency indices are assigned longer refresh intervals, ensuring that the refresh rhythm of different attribute paths aligns with their historical change rhythm. Furthermore, after determining the dynamic refresh intervals for all paths, they are arranged according to the chronological order of their corresponding refresh times, resulting in a refresh sequence. This refresh sequence clearly records the refresh time, refresh order, and time distribution of each attribute update path. Therefore, this step completes the transformation from attribute update path-based to time-based refresh scheduling.

[0103] In this embodiment, if the refresh sequence of a path exceeds a threshold, a frequency adjustment mechanism is used to optimize the dynamic refresh time interval to obtain an optimized refresh configuration. Specifically, the already formed path refresh sequence is first reviewed as a whole, and the number of attribute update paths entering the refresh state per unit time, the time density between adjacent refresh nodes, and the path concurrency within the same time period are statistically analyzed. Then, the above statistical results are compared with a preset threshold. The threshold is used to limit the upper limit of the number of paths allowed to enter refresh processing per unit time, and its determination process is based on system processing capacity, historical refresh load, the current total number of correction tasks, and business processing time limits.

[0104] When the number of paths entering the refresh sequence within a certain time period exceeds a certain threshold, it indicates that the current dynamic refresh time interval setting is too concentrated, resulting in a congested path refresh sequence. Based on the aforementioned embodiments, a frequency adjustment mechanism needs to be activated to redistribute the dynamic refresh time intervals for each path. The execution process of the frequency adjustment mechanism is as follows: first, identify the time periods with excessively dense tasks in the refresh sequence; then, extract the attribute update paths within those time periods; subsequently, rearrange the refresh positions of each path according to the path update frequency index, the urgency of the correction, and the dependencies between paths; extend the refresh time intervals for low-urgency paths; and keep the refresh time intervals for high-urgency paths within a shorter range. After the above adjustments, an optimized refresh configuration is formed. This optimized refresh configuration clearly records the adjusted refresh time intervals, refresh times, and new arrangement order for each attribute update path. Therefore, after this step, the time distribution of the path refresh sequence is more balanced, and the refresh load in subsequent execution processes remains within a preset range.

[0105] In this embodiment, the correction guidance list is deeply bound to the attribute update paths according to the optimized refresh configuration to determine the final correction execution sequence. Specifically, based on the optimized refresh configuration, each correction item in the correction guidance list is remapped to its corresponding attribute update path node, ensuring that each correction task simultaneously possesses the attribute object, execution order, refresh time, and processing basis. Subsequently, all attribute update paths within the same abandoned land parcel are uniformly arranged. First, the dependencies between paths are identified, then the parallel relationships and result propagation relationships between different paths are identified, and then they are sequentially organized according to correction priority, refresh time order, and path dependency order. For paths where the result of a previous attribute correction directly affects the judgment of a subsequent attribute correction, the path corresponding to the previous attribute is executed first, followed by the path corresponding to the subsequent attribute. For independent attribute update paths, they are processed sequentially according to the optimized refresh time order. Finally, all path nodes, time nodes, and execution nodes are integrated in a unified order to form the final correction execution sequence. This correction execution sequence clearly records the execution start point, execution order, path node, refresh time, attribute object, and completion conditions for each correction task. As can be seen from the above embodiments of this application, after deep binding processing, the correction guidance list no longer remains at the attribute sorting level, but is transformed into a correction execution result with a clear time schedule, clear execution path and clear sequential relationship.

[0106] Example 8: S7 includes acquiring execution sequence data and integrating land parcel attributes to determine the initial correction path; processing the execution sequence using index extraction rules to obtain the input dataset; calculating the reduction potential value based on the input dataset; determining the land use matching standard set when the reduction potential value meets the conditions; evaluating the suitability score based on the land use matching standard set and generating a preliminary priority ranking list based on the reduction potential value; determining the output structure based on the preliminary ranking list; and generating the intelligent assessment result of rural construction land reduction.

[0107] In this embodiment, execution sequence data is acquired and land parcel attributes are integrated to determine the initial correction path. The execution sequence is then processed using index extraction rules to obtain the input dataset. Specifically, the correction execution sequence formed in the preceding steps is retrieved first, and each execution item in the execution sequence is broken down to extract the execution node, execution order, corresponding attribute object, correction result, and completion status item by item.

[0108] Subsequently, the attribute data corresponding to the target plot is retrieved. This attribute data includes boundary attributes, area attributes, utilization status attributes, structural status attributes, and the attribute results corrected in previous steps. Based on the aforementioned embodiments, the execution sequence clearly defines the order of attribute corrections. Therefore, this step first maps the correction results in the execution sequence to the corresponding attributes one by one according to the execution order. Then, the change states of each attribute in different execution nodes are sequentially linked together, forming a path structure in which the plot attributes continuously evolve from the original state to the corrected state. This path structure is the initial correction path. Afterward, the execution sequence is processed using index extraction rules. First, nodes directly related to the reduction assessment in the execution sequence are identified. Then, the area change information, boundary change information, utilization change information, and structural change information corresponding to each node are extracted. Finally, these are organized according to a unified field structure to form the input dataset.

[0109] In this embodiment, a reduction potential value is calculated based on the input dataset, and a set of land use matching criteria is determined when the reduction potential value meets certain conditions. Specifically, core indicators reflecting the land reduction value of a plot are first extracted from the input dataset. These core indicators include the degree of inefficient land use, spatial redundancy, structural imbalance, and the degree of change in the corrected land use status. Subsequently, the indicators are normalized and organized according to a unified calculation order, ensuring that different types of indicators participate in the calculation on the same evaluation basis. Afterward, the indicators are comprehensively processed to obtain the reduction potential value. This reduction potential value directly reflects the strength of the target plot's value in entering the scope of construction land reduction under the current corrected state.

[0110] As can be seen from the foregoing embodiments, the reduction potential value is a prerequisite for subsequent suitability analysis. Therefore, after the reduction potential value is formed, it is compared with preset conditions. These preset conditions are determined based on regional reduction targets, land consolidation requirements, current utilization standards, and historical assessment results. When the reduction potential value meets these conditions, it indicates that the land parcel has the basis for entering the subsequent matching assessment. Subsequently, combining the land type attributes, spatial structure attributes, utilization status attributes, and consolidation direction of the land parcel, a set of land use matching standards corresponding to it is determined. This set of land use matching standards clearly records the type standards, layout standards, utilization standards, and structural standards corresponding to the target land parcel when participating in the reduction assessment. Thus, this step completes the transformation from input data to reduction potential judgment and the basis for standard matching.

[0111] In this embodiment, a preliminary ranking list is generated by assessing suitability scores based on a set of land use matching criteria and combining them with reduction potential values. Specifically, the current modified attribute status of the target plot is first compared item by item with the set of land use matching criteria, examining the plot's compliance with the corresponding criteria in terms of boundary shape, area configuration, utilization status, spatial layout, and structural relationships. Subsequently, the compliance results are summarized to form a suitability score. The suitability score directly reflects the degree of matching between the target plot, in its current modified state, and the predetermined reduction configuration requirements. Based on the aforementioned embodiment, the reduction potential value reflects whether the plot has reduction value, while the suitability score reflects whether the plot meets the post-reduction configuration requirements. Therefore, this step further combines the reduction potential value and the suitability score for joint analysis.

[0112] During the joint analysis, the reduction potential value and suitability score of each plot were first matched, and then a comprehensive evaluation result was formed according to a unified ranking rule. Afterwards, all target plots were arranged from highest to lowest according to the comprehensive evaluation result, generating a preliminary priority ranking list. It is evident that plots located at the top of the preliminary ranking list indicate high reduction value and a high degree of matching, placing them at the forefront of the overall reduction plan.

[0113] In this embodiment, the output structure is determined based on the preliminary sorting list to generate intelligent assessment results for the reduction of rural construction land. Specifically, the items in the preliminary sorting list are first structured, extracting the plot number, reduction potential value, suitability score, comprehensive ranking position, and corresponding key attribute results for each item. Then, the above content is uniformly arranged according to a preset output format, ensuring that the assessment information for each plot is presented completely under the same result structure. The process of determining the output structure is as follows: first, the order of the fields in the result is determined; then, the correspondence between the fields is determined; and finally, the sorting information, score information, and attribute information are integrated into a unified result item. Finally, all result items are summarized to generate intelligent assessment results for the reduction of rural construction land. The intelligent assessment results clearly record the reduction potential level, suitability evaluation results, priority position, and corresponding attribute basis of each abandoned plot.

[0114] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A rural construction land reduction intelligent evaluation method based on multi-source spatial big data, characterized in that, include: S1. Collect archival data and on-site status records of abandoned land parcels in rural areas through multi-source information collection, process the differences between archival data and status records, classify the sources of deviation and make preliminary annotations to obtain a list of inconsistencies of abandoned land parcels. S2. Extract the attribute features of the abandoned land parcels from the list of inconsistencies, decompose the attribute features into multiple dimensions using vector dimension partitioning technology, rank the importance of each dimension, and determine the deviation distribution characteristics of the land parcel attributes. S3. If the deviation value of a certain dimension in the deviation distribution characteristics of the land parcel attributes exceeds the preset threshold, then historical data reference records are obtained for that dimension, the analysis results of the deviation data are updated through a dynamic adjustment mechanism, and it is determined whether a new correction direction vector is generated. S4. Extract key correction reference points by correcting the direction vector, filter out the attributes related to the deviation based on the extraction logic, check the matching degree between the reference points and the current status record, and determine the specific content of the priority correction basis. S5. Based on the specific content of the priority correction basis, use the correction scope definition technology to clarify the applicable land parcel attribute range, and combine the correction priority ranking of different attributes according to the weight calculation method to obtain the correction guidance list of abandoned land parcels. S6. Based on the list of correction guidelines for abandoned land parcels, the correction path association technology is used to bind the guidelines with the land parcel attribute update path. By setting a dynamic refresh time interval based on the update frequency, the final correction execution sequence is determined. 2.The method of claim 1, wherein the method is characterized by: S1 includes: Acquire archival data and on-site status records of abandoned land parcels in rural areas from multi-source information collection, and extract spatial coordinates and ownership information from the archival data and on-site status records; Based on spatial coordinates and ownership information, spatial topological overlay processing is performed on archival data and on-site status records to obtain initial matching results including land parcel area and utilization type. The differences between land parcel area and land use type in the initial matching results are processed using data field comparison rules to obtain a set of field difference features. If the area difference in the field difference feature set is greater than the preset area threshold, it is determined to be a boundary change deviation. The difference between the archive data and the actual status record of the abandoned land plot is initially marked by the deviation source classification result, and a list of inconsistencies of the abandoned land plot is obtained.

3. The intelligent assessment method for reducing rural construction land based on multi-source spatial big data as described in claim 1, characterized in that: S2 includes: Obtain remote sensing monitoring data and inconsistency list of abandoned land parcels, parse the inconsistency list and map it to the remote sensing monitoring data to extract land parcel attribute features; The attribute features of land parcels are divided into vector dimensions to construct a dimension division matrix. The Euclidean distance between the orthogonal components in the dimension division matrix and the standard land parcel feature template is calculated to quantify the attribute deviation value. The deviation weight allocation coefficient is calculated based on the attribute deviation value. The importance of each attribute dimension is ranked according to the deviation weight allocation coefficient to determine the deviation distribution characteristics of the land parcel attributes.

4. The intelligent assessment method for reducing rural construction land based on multi-source spatial big data according to claim 1, characterized in that: S3 includes: Obtain the deviation distribution characteristics of land parcel attributes. If the extracted dimension deviation value exceeds a preset threshold, obtain historical data reference records for the dimension. Based on historical data and reference records, spatial coordinate offsets are extracted, and the spatial coordinate offsets are classified to obtain the first deviation data; The second deviation data is obtained by weighting the first deviation data through a dynamic adjustment mechanism, and the deviation analysis results are obtained based on the second deviation data. If the dispersion value in the deviation analysis result is greater than the preset dispersion, then the correction angle value is extracted based on the deviation analysis result, and a new correction direction vector is generated using the correction angle value.

5. The intelligent assessment method for reducing rural construction land based on multi-source spatial big data according to claim 1, characterized in that: S4 includes: Obtain the original business data stream and calculate the correction direction vector, then extract key correction reference points based on the cluster density distribution of the correction direction vector; Analyze the mutual information correlation of key correction reference points and filter out highly correlated attributes of deviations; Retrieve real-time status records corresponding to highly correlated deviation attributes, and use a data consistency verification model to compare key correction reference points with real-time status records to generate a matching degree index. Based on the matching degree index, construct an attribute priority index sequence, analyze the first and second features of the attribute priority index sequence, and determine the specific content of the priority correction basis.

6. The intelligent assessment method for reducing rural construction land based on multi-source spatial big data according to claim 1, characterized in that: S5 includes: Obtain the text on which the correction is based, and construct semantic feature vectors using a bidirectional long short-term memory network; The semantic feature vectors are correlated with land parcel attributes to determine the boundaries of the applicable scope; Calculate the information entropy value of each attribute within the boundary of the applicable scope, and determine the basis weight value based on the information entropy value; By combining the current status values ​​of abandoned land parcel attributes with weighted values, a weighted correction urgency score is calculated. Based on the weighted correction urgency score, an attribute correction priority ranking sequence is generated. Based on the attribute correction priority ranking sequence, an abandoned land parcel correction guidance list is generated.

7. The intelligent assessment method for reducing rural construction land based on multi-source spatial big data according to claim 1, characterized in that: S6 includes: Obtain a list of correction guidelines for abandoned land parcels, and use a path association algorithm to match the list of correction guidelines with the land parcel attribute data to obtain the bound attribute update path; Analyze the bound attribute update path to obtain the update frequency index, set the dynamic refresh time interval based on the update frequency index, and determine the refresh sequence of the path.

8. The intelligent assessment method for reducing rural construction land based on multi-source spatial big data according to claim 7, characterized in that: S6 further includes: If the refresh sequence of the path exceeds the threshold, the dynamic refresh time interval is optimized using the frequency adjustment mechanism to obtain an optimized refresh configuration. Based on the optimized refresh configuration, the correction guidance list is deeply bound to the bound attribute update path to determine the final correction execution sequence.

9. The intelligent assessment method for reducing rural construction land based on multi-source spatial big data according to claim 1, characterized in that, It also includes S7, which forms a revised land parcel attribute dataset based on the final revised execution sequence, extracts assessment indicators characterizing the reduction of rural construction land, conducts reduction potential analysis, suitability analysis, and priority analysis on each abandoned land parcel, and generates intelligent assessment results for the reduction of rural construction land, specifically including: The execution sequence data is obtained and the parcel attributes are integrated to determine the initial correction path. The execution sequence is then processed using index extraction rules to obtain the input dataset. Calculate the reduction potential value based on the input dataset, and determine the set of land use matching criteria when the reduction potential value meets the conditions.

10. The intelligent assessment method for reducing rural construction land based on multi-source spatial big data according to claim 9, characterized in that: The S7 also includes: Based on the suitability score of the land use matching standard set and combined with the reduction potential value, a preliminary ranking list is generated to determine the priority order. The output structure is determined according to the preliminary ranking list to generate the intelligent assessment result of rural construction land reduction.

Images