Intelligent database construction method for national space planning based on multi-source heterogeneous data integration

Abstract

This invention relates to the fields of land spatial planning, geographic information data processing, and spatial database construction. Specifically, it is a method for intelligent land spatial planning database construction based on the integration of multi-source heterogeneous data. The method includes: collecting multi-source heterogeneous spatial data, image data, and textual standard data; extracting geometric boundaries, timestamps, coordinate system precision features, semantic ontology label features, and element attributes; and dividing the space into grids. It also involves constructing a planning ontology database, identifying semantically ambiguous areas, and binding geometric boundaries; performing geometric correction and temporal smoothing based on timestamps and coordinate system precision features; overlaying detection to generate hard and soft conflict features; and performing conflict resolution based on the authority level of the data source to form a logically closed-loop database and output a data confidence heatmap. This invention can detect and correct planning contradictions during the database entry stage, reducing subsequent verification and approval costs.

Description

Technical Field

[0001] This invention relates to the fields of land spatial planning, geographic information data processing and spatial database construction, specifically to an intelligent database construction method for land spatial planning based on the integration of multi-source heterogeneous data. Background Technology

[0002] With the improvement of the national spatial planning management system and the increasing demand for multi-departmental collaborative governance, the creation of a unified map database for planning preparation, review, updating and supervision has become an important foundational task in the field of spatial governance. In order to achieve unified management of planning results, control boundaries, thematic data and project site selection information, the integration, verification and database processing of multi-source heterogeneous data has become particularly important. In the process of building a national spatial planning database, multi-source spatial data, image data, and textual standardized data often come from different business departments such as natural resources, housing and construction, water resources, and forestry and grassland. There are problems such as inconsistent format and structure, inconsistent classification terminology, different time versions, and differences in coordinate benchmarks and surveying accuracy among different data. Especially in the scenario of planning conflict identification, the existing processing methods mostly rely on manual comparison or basic overlay analysis, which are easily affected by semantic label ambiguity, slight boundary offset, mixed use of historical versions, and inconsistent attribute rules. As a result, it is difficult to timely and accurately discover and deal with conflicts between permanent basic farmland, ecological protection red lines, urban development boundaries and construction projects. At the same time, traditional database construction results usually focus more on graphic summarization and lack a closed-loop record of the conflict resolution process, data authority level, and spatial confidence level, which is difficult to support subsequent approval verification and dynamic updates. Therefore, it is crucial to ensure the consistency, accuracy, and traceability of territorial spatial planning data by uniformly collecting, semantically mapping, geometrically correcting, temporally smoothing, detecting and resolving topological conflicts in existing multi-source heterogeneous planning data to form intelligent database results with logical closed loops and confidence expression capabilities. This will facilitate more efficient database entry review and risk warning in conjunction with planning control rules. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent database construction method for land spatial planning based on the integration of multi-source heterogeneous data, and to solve the following technical problems: It avoids pseudo-conflicts and misjudgments caused by differences in spatiotemporal benchmarks and semantic standards between data sources from different departments. It can also automatically detect and repair planning contradictions according to rules at different levels during the data entry stage, effectively reducing the frequency of data interaction and system processing load in subsequent cross-departmental system verification, multi-level approval and dynamic planning adjustment.

[0004] The objective of this invention can be achieved through the following technical solutions: The intelligent database construction method for land spatial planning based on multi-source heterogeneous data integration includes the following steps: Step 1: Collect multi-source heterogeneous spatial data and text specification data; extract geometric boundaries, timestamps, coordinate system precision features, and feature attributes from the multi-source heterogeneous spatial data; and extract semantic ontology label features from the text specification data; divide the multi-source heterogeneous spatial data into several spatial grids according to a preset grid size. Step 2: Construct a planning ontology library, map the semantic ontology label features to the planning ontology library, identify differences through semantic comparison, generate semantic ambiguity region features, and bind the semantic ambiguity region features to their corresponding geometric boundaries; Step 3: Based on the timestamp and the coordinate system precision characteristics, and with a preset standard reference plane coordinate system as a reference, perform geometric correction and temporal smoothing on the multi-source heterogeneous spatial data to generate spatiotemporally consistent spatial features. Step 4: Spatial superposition of the semantic ambiguity region features bound to geometric boundaries, the spatial features of aligned standard classification nodes, and the spatiotemporal consistency spatial features; and conflict detection by combining the spatial overlap area and feature attributes between different elements using a preset topology operator to generate hard conflict features and soft conflict features. Step 5: Obtain the preset data source authority level. The data source authority level is determined by comprehensively considering standardization, approval effectiveness, data update procedures, and surveying accuracy. Based on the data source authority level, perform conflict resolution processing on the hard conflict features and the soft conflict features, record the number of conflict resolutions within each spatial grid, generate a logical closed-loop database, and output a data confidence heatmap based on the logical closed-loop database.

[0005] Preferably, step one specifically includes: Collect multi-source heterogeneous spatial data, image data, and text specification data from different business departments through preset data interfaces; The format of the multi-source heterogeneous spatial data, image data and text standard data is parsed, and the corresponding data generation time is extracted as the timestamp. The spatial reference information of the multi-source heterogeneous spatial data is analyzed, the error parameters of the spatial reference information are extracted as the coordinate system accuracy features, the business fields are extracted as the feature attributes, and the vector outer contour is extracted as the geometric boundary. Natural language processing is performed on the text specification data to extract classification standard vocabulary as semantic ontology label features; The multi-source heterogeneous spatial data is divided into several spatial grids according to the preset grid size.

[0006] Preferably, step two specifically includes: Construct a planning ontology library, which includes preset standard classification nodes and the relationships between the nodes; Calculate the semantic similarity between the semantic ontology label features and the standard classification nodes; When the semantic similarity is higher than a preset similarity threshold, the semantic ontology label features are aligned to the corresponding standard classification node; When the semantic similarity is lower than or equal to the preset similarity threshold, the semantic ontology label feature is marked as the semantic fuzzy region feature, and an association mapping between the semantic fuzzy region feature and the corresponding geometric boundary is established.

[0007] Preferably, step three specifically includes: Based on the accuracy characteristics of the coordinate system, the position offset of the multi-source heterogeneous spatial data relative to the preset standard reference plane coordinate system is calculated; Based on the position offset, the node coordinates of the multi-source heterogeneous spatial data are translated and affine transformed to complete the geometric correction process. The multi-source heterogeneous spatial data after the geometric correction process is completed is sorted by time sequence according to the timestamp. Calculate the boundary change rate of spatial elements at adjacent time nodes, perform interpolation fitting on boundary nodes that exceed the preset change rate threshold, and retain the original coordinates of boundary nodes that do not exceed the preset change rate threshold to complete the temporal smoothing process and output the spatiotemporal consistent spatial features.

[0008] Preferably, step four specifically includes: The semantically ambiguous region features bound to geometric boundaries, the spatial features of aligned standard classification nodes, and the spatiotemporally consistent spatial features are geometrically intersected in the same coordinate space to generate an overlay layer. In the overlay layer, the preset topology operator is invoked to calculate the spatial overlap area between different elements; When the spatial overlap area is greater than 0 and the feature attributes include mutually exclusive red line attributes, the hard conflict feature is determined and generated. When the spatial overlap area is greater than 0 and the feature attribute is a non-mutually exclusive but functionally incompatible attribute, the soft conflict feature is determined and generated. When the spatial overlap area is greater than 0 and the element attributes are mutually compatible, it is determined to be a regular overlap and there is no conflict; when the spatial overlap area is equal to 0, there is no conflict; wherein, the mutually exclusive red line attribute corresponds to the normative control requirements that cannot coexist; the non-mutually exclusive but functionally incompatible attribute corresponds to the situation that is not absolutely prohibited but is difficult to coordinate functionally; the mutually compatible attribute corresponds to the situation that can coexist functionally.

[0009] Preferably, step five involves performing conflict resolution processing on the hard-conflict features and the soft-conflict features based on the authority level of the data source, specifically including: Read the authority level of the data source corresponding to each of the two conflicting elements that generate the hard conflict feature; Compare the authority levels of the two data sources, retain the geometric boundaries and attributes of the conflicting elements with the higher authority level, trim or delete the overlapping parts of the conflicting elements with the lower authority level, and accumulate the number of conflict resolutions. For the soft conflict feature, a preset attribute compatibility mapping table is queried to obtain the transition attribute corresponding to the high-level feature attribute, the attribute of the low-level feature is modified to the transition attribute, and the number of conflict resolutions is accumulated. The elements that have undergone the cropping or deletion and the attribute modification are stored in the database to form the logical closed-loop database.

[0010] Preferably, the output data confidence heatmap based on the logical closed-loop database specifically includes: Extract the number of conflict resolutions occurring within each spatial grid in the logical closed-loop database and the coordinate system precision features of the original data; Obtain preset conflict resolution frequency weights and accuracy weights. Then, perform a weighted summation of the conflict resolution frequency and the coordinate system accuracy feature based on the conflict resolution frequency weights and accuracy weights to calculate the comprehensive confidence score for each spatial grid. Specifically, the conflict resolution frequency is negatively correlated with the comprehensive confidence score, and the coordinate system accuracy feature is positively correlated with the comprehensive confidence score. The weighted calculation includes mapping the conflict resolution frequency to a basic conflict resolution score. The mapping formula is Map the coordinate system precision features to a basic precision score. The mapping formula is ; The overall confidence score is mapped to a preset color gradient band; The spatial grid is rendered based on the color gradient band, and the data confidence heatmap is generated and output.

[0011] Preferred options also include: Receive externally input error correction instructions for the logical closed-loop database, and extract error tag data from the error correction instructions; Using the erroneous label data as training samples, the association weights in the planning ontology are updated through backpropagation. Specifically, the planning ontology is mapped to a semantic network graph structure consisting of classification nodes and associations. The weight parameters of the network connections corresponding to the associations are used to calculate the loss error between the predicted classification probability and the target classification label based on the erroneous label data. The loss error is then propagated along the association path of the semantic network graph structure using the chain rule of differentiation and backpropagation, thereby updating the association weights. Steps two through five are re-executed using the updated planning ontology library to achieve adaptive iteration of the library building logic.

[0012] The beneficial effects of this invention are: 1) This invention constructs a planning ontology library for semantic mapping, and marks semantically ambiguous regions that cannot be stably matched as features of semantic ambiguity and binds them to geometric boundaries. This design solves the semantic ambiguity caused by inconsistent classification terms across departments, transforms textual uncertainty into computational objects with clear spatial coordinates, provides stable and accurate spatial input parameters for subsequent topological conflict detection, and significantly improves the semantic integration accuracy of multi-source data. 2) Based on the extracted timestamps and coordinate system precision features, this invention performs geometric correction and temporal smoothing on spatial data to generate spatiotemporally consistent spatial features. This mechanism effectively eliminates local boundary misalignment caused by differences in surveying benchmarks and abnormal boundary jumps caused by historical data updates, avoiding misjudging conventional update mapping errors as substantial planning conflicts, and laying a continuous and unified spatial foundation for the accurate overlay review of subsequent control boundaries. 3) This invention spatially superimposes the semantic ambiguity features of the bounded boundary with the spatiotemporal consistency features, generates hard and soft conflict features through topological operators, and performs closed-loop resolution based on the authority level of the data source; this scheme realizes refined hierarchical identification of spatial overlap relationships, and automatically completes the retention of high-level elements and the attribute transition repair of low-level elements, which greatly reduces the business processing load of cross-departmental data verification and multi-level manual approval verification. 4) This invention forms a logical closed-loop database by recording the number of conflict resolutions within a spatial grid, and combines the coordinate system accuracy characteristics to perform weighted calculations to output a data confidence heatmap; this breaks through the limitation of traditional database construction that only outputs graphical conclusions, comprehensively quantifies and explicitly expresses the hidden conflict reconstruction frequency and the reliability of the base data, intuitively identifies high-risk anomaly verification grids, and provides intuitive risk warnings for planning quality control and approval review; 5) This invention receives error correction instructions for the logical closed-loop database and extracts error label data, which are used as training samples to dynamically update the association weights in the planning ontology using the backpropagation algorithm; this feedback mechanism realizes the adaptive iteration of the database building logic, enabling the system to continuously learn and adapt to the ever-evolving local control standards and special planning classification standards, effectively reducing the repetitive semantic misjudgments of the system in new business contexts. Attached Figure Description

[0013] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating the intelligent database construction method for land spatial planning based on the integration of multi-source heterogeneous data provided in this application embodiment. Detailed Implementation

[0014] 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.

[0015] Please see Figure 1 The intelligent database construction method for land and space planning based on the integration of multi-source heterogeneous data includes the following steps: Step 1: Collect multi-source heterogeneous spatial data and textual specification data; extract geometric boundaries, timestamps, coordinate system precision features and element attributes from the multi-source heterogeneous spatial data; and extract semantic ontology label features from the textual specification data; divide the multi-source heterogeneous spatial data into several spatial grids according to the preset grid size. Step 2: Construct a planning ontology library, map semantic ontology label features to the planning ontology library, identify differences through semantic comparison, generate semantically ambiguous region features, and bind semantically ambiguous region features to their corresponding geometric boundaries; Step 3: Based on the timestamp and coordinate system precision characteristics, and with the preset standard reference plane coordinate system as a reference, perform geometric correction and temporal smoothing on the multi-source heterogeneous spatial data to generate spatiotemporally consistent spatial features. Step 4: Spatial overlay of the semantically ambiguous region features bound to geometric boundaries, the spatial features of aligned standard classification nodes, and the spatiotemporal consistency spatial features; using a preset topological operator, and combining the spatial overlap area between different elements with element attributes to perform conflict detection, generating hard conflict features and soft conflict features. Step 5: Obtain the preset data source authority level. The data source authority level is determined by comprehensively considering standardization, approval effectiveness, data update procedures, and surveying accuracy. Based on the data source authority level, perform conflict resolution processing on hard conflict features and soft conflict features, record the number of conflict resolutions in each spatial grid, generate a logical closed-loop database, and output a data confidence heatmap based on the logical closed-loop database.

[0016] This embodiment provides an intelligent database construction mechanism for territorial spatial planning based on the integration of multi-source heterogeneous data. Specifically, this embodiment takes an exemplary target planning area as the main scenario for carrying out an annual update of the territorial spatial planning map and simultaneously reviewing whether the site selection of a proposed specific functional project conflicts with permanent basic farmland, ecological protection red line and urban development boundary, and explains the entire database construction process. The data accessed in Step 1 is not limited to the current land use vector data from the natural resources department, the control planning layers from the housing and construction department, the river and lake management scope data from the water resources department, the forestry and grassland thematic data, remote sensing image data, and text materials such as current regulations, glossaries, and classification standards from various departments. The system extracts the geometric boundaries of land parcels or linear features from spatial data and simultaneously reads the formation time and coordinate reference accuracy information of the batch of data; The technical motivation for extracting these three types of information is that conflicts in land spatial planning are not only caused by overlapping graphics, but also by the mixing of data from different years and slight spatial offsets under different coordinate benchmarks. Such offsets may only manifest as boundary misalignment at the level of a single plot, but after overlay analysis, they may lead to issues that have an impact on approval, such as red line crossing and land use classification conflicts. At the same time, the system also extracts semantic tags reflecting land use classification, control level and functional use from normative texts, and retains element attributes such as land category number, approval status, restriction conditions and competent department from business fields; then the system divides the data of the entire domain according to the preset grid, and the grid serves as both a parallel processing unit and a spatial container for subsequent statistical conflict density and output confidence heatmap. In step two, the system establishes a unified planning ontology library. This ontology library is not a simple field lookup table, but rather organizes standard nodes such as cultivated land, permanent basic farmland, general agricultural land, ecological restoration areas, construction land, and transportation facility land, as well as their hierarchical, compatibility, and exclusion relationships. The system compares the semantic tags extracted from the text with the standard nodes in the ontology library one by one. If the standard node feature alignment threshold is met, classification and normalization are completed directly; the feature stable matching condition means that the semantic similarity is greater than the preset similarity threshold; if the feature stable matching condition cannot be met, it is marked as a semantically ambiguous area feature, and the semantic feature is bound to the corresponding land parcel boundary. The technical necessity of performing geometric boundary binding lies in accurately mapping the semantic uncertainty features to the actual two-dimensional space polygon, so that the unstructured text abnormal state is transformed into a computational object with clear spatial coordinate boundaries, thereby providing stable spatial input parameters for subsequent conflict detection algorithms based on topological operators. In step three, the system performs spatiotemporal consistency processing based on timestamps and coordinate system accuracy characteristics. On the one hand, it performs geometric correction on data from different reference surfaces, different surveying accuracies, or different historical versions, so that the data is projected into a unified standard coordinate frame. On the other hand, time series processing is performed on the boundary changes of the same element at different time points, and the boundaries of abnormal jumps are smoothed to avoid misjudging data update errors as real land changes. For example, if a river management line shows an abrupt broken line in adjacent years' data, it may actually be due to changes in mapping scale rather than actual river course changes. In this case, it is necessary to suppress unreasonable boundary jumps through temporal smoothing. The resulting spatiotemporal consistency spatial features can serve as a stable base map for subsequent conflict detection. In step four, the system spatially overlays the semantically ambiguous region features bound to geometric boundaries with the spatiotemporally consistent spatial features. The spatial overlay operation here does not only perform basic geometric Boolean intersection judgment, but also combines complex topological relationships to accurately identify whether the elements overlap in a planar manner, cross in a linear manner, contain at a boundary, or are adjacent at an edge, and combines the element attributes to identify the conflict nature. If the overlapping area involves mutually exclusive red lines, such as the overlap between permanent basic farmland and the site selection of new industrial warehouses, then a hard conflict characteristic is formed; If it is not absolutely prohibited but functionally difficult to coexist, such as residential land and high-intensity logistics distribution functions overlapping each other, then a soft conflict characteristic is formed; through this logical closed-loop processing, the system transforms fragmented data from various sources that are independent and compliant into a conflict spatial feature set after multi-source spatiotemporal correlation verification. In step five, the system introduces the authority level of the data source to resolve conflicts; generally, the authority level of the standard weighted data, the approved land survey results, and the approved control plan is higher than that of temporary business reporting data, third-party survey results, or information collected from the Internet. For hard conflicts, the system prioritizes preserving the boundaries and attributes of high-authority data and performs pruning or removal on low-authority elements; for soft conflicts, the system performs transitional processing or conditional preservation of low-authority data according to attribute compatibility rules. Each conflict resolution process records the grid to which it belongs, the elements involved, the reason for the resolution, and the processing actions, ultimately forming a logical closed-loop database. Then, based on the number of conflict resolutions and the original accuracy within each grid, a data confidence heatmap is generated, thereby indicating the risk distribution of data review for planning and approval departments. When a data source is missing a timestamp, its batch entry time can be used as a temporary sequence identifier, and an incomplete time source flag can be added to the database. When coordinate reference information is missing, the batch of data does not directly participate in high-precision boundary clipping, but only participates in semantic comparison and low confidence prompts. When the quality of the scanned text is poor and terminology cannot be extracted reliably, the text can be included in the manual verification queue to prevent erroneous terms from being propagated to the ontology. When the number of conflicts within the same grid is too high and exceeds the threshold of the automatic repair strategy, it can be switched to manual review mode to retain all conflict traces without forcing automatic overwriting. In the target planning area scenario, the system accesses the 2023 permanent basic farmland data from the natural resources department, the 2022 version of the control plan construction land data from the housing and construction department, the river and lake shoreline data from the water resources department, and the site selection data of the specific functional projects to be built in the area recently. During the initial compilation, the plot of land for the proposed specific function project was marked as warehousing and logistics land in the housing and construction data, but in a certain text specification data, the corresponding term was described as a comprehensive service supporting area. After ontology comparison, this term was identified as a semantically ambiguous area. Meanwhile, the boundary of the plot is slightly offset compared to the natural resource data; after correction and overlay, the system found that the site selection boundary of the project partially overlaps with the boundary of permanent basic farmland, forming a hard conflict, and a soft conflict with the planning of the adjacent urban residential area. The system retains the boundaries of high-level farmland, performs cropping on the low-level reported map, adjusts the use of adjacent areas to transitional supporting land, and marks the number of high conflict resolutions on the corresponding grid to generate a warm-colored confidence alert area. The purpose of this step is to further enhance the integration of multi-source data formats into logical integration, spatiotemporal integration, and compliance integration, so as to discover and correct planning contradictions at the data entry stage and reduce the frequency of data interaction and system processing load in subsequent cross-departmental system verification, multi-level approval, and dynamic planning adjustment stages.

[0017] In a preferred embodiment of the present invention, step one specifically includes: collecting multi-source heterogeneous spatial data, image data and text specification data from different business departments through a preset data interface; parsing the format of the multi-source heterogeneous spatial data, image data and text specification data, and extracting the corresponding data generation time as a timestamp; The spatial reference information of multi-source heterogeneous spatial data is analyzed, the error parameters of the spatial reference information are extracted as coordinate system accuracy features, the business fields are extracted as feature attributes, and the vector outer contour is extracted as geometric boundary. Natural language processing is performed on the textual data to extract classification standard words as semantic ontology label features; the multi-source heterogeneous spatial data is divided into several spatial grids according to the preset grid size.

[0018] This embodiment provides a detailed set of steps for data collection and preprocessing. Specifically, in the aforementioned task of building a database in the target planning area, simply completing the overall data access is not enough to support subsequent compliance judgments. This is because the data sources, format structures, and mapping logic of different departments have significant feature distribution dispersion. If time, accuracy, boundaries, attributes, and terms are not decomposed first, subsequent semantic alignment and conflict judgment will be affected by source noise. The system has multiple preset data interfaces, including an interface for the government information sharing platform, a direct connection interface for the database, a batch processing interface for the file exchange directory, and a tile grabbing interface for image services; for vector data provided by the natural resources department, its layer structure, spatial reference, and attribute table can be directly parsed. For compressed files submitted by the housing and construction department, the graphic files, appendices and instructions can be identified first; for scanned standard texts, text recognition can be performed first, and then the classification standard vocabulary can be extracted; for timestamp extraction, the generation time, surveying time or approval time of the data itself should be read first. If there are multiple time fields in the same data, the primary timestamp can be determined in the order of effective time taking precedence over edit and save time, and edit and save time taking precedence over file creation time. The reason for this is that the planning and judgment focus on the effective time point of data business, rather than just when the file is copied to the server. Regarding the accuracy characteristics of the coordinate system, the system not only reads the coordinate name, but also further extracts error parameters, projection zone number, control point source, surveying scale and other content that can reflect the reliability of the positioning; whether the boundaries in the land space planning can be directly compared does not depend on whether the data are all called a certain coordinate system, but on whether its actual surveying accuracy is sufficient to support red line level superposition. For example, data labeled with the same unified coordinate system may come from different sources: one may be from high-precision land rights measurement, and the other may be from small-scale thematic compilation. The reliability of the two when used for boundary clipping is obviously different. The system also extracts element attributes such as land category code, land use zoning, control level, and approval document number from attribute fields, and extracts the outer contour as the geometric boundary from vector graphics. For complex surfaces, multi-component elements, or plots with holes, the outer contour and inner ring can be stored separately to avoid misidentifying the restricted internal space as a usable area later. For textual data, the system uses natural language processing steps such as word segmentation, part-of-speech recognition, proper noun extraction, and rule filtering to extract classification standard words such as permanent basic farmland, ecological restoration area, urban concentrated construction area, and mixed land for transportation facilities as semantic ontology label features. For ease of explanation, a simplified sandbox example can be used: if the three terms "preserved forest land", "ecological conservation forest" and "public welfare forest protection area" appear in a text, the system can first split them into three candidate labels, and then retain the complete phrases instead of single words based on the context, so as to avoid mistaking "forest" as an independent category later. After the above extraction is completed, the system divides the space into grids according to the preset grid size. The grid size can be set according to business objectives. For example, a first preset grid size is used in the core urban area to improve the local conflict location capability, and a second preset grid size is used in mountainous areas or ecological buffer zones to balance processing efficiency. The first preset size is smaller than the second preset size. The engineering significance of grids lies in discretizing a continuous space into manageable units, making the data quality, collision count, and repair records within each unit traceable. If a department's data lacks complete spatial reference information, but its boundaries have a clear overlap with existing high-level base maps, it can be attached as a layer to be confirmed and participate in low-level analysis but not in direct cropping. If the vocabulary extraction results in the text specification are too scattered, such as only identifying generalized terms like construction control and protection, they will not be directly written into the ontology tag pool, but will instead enter the manual terminology verification stage. If a grid contains both vector data and image data, but the time span characteristics of both exceed the preset timeliness verification threshold (e.g., the image is significantly earlier than the latest redline delineation result), then the grid will be marked as having insufficient timeliness of image evidence and will only be used for reference and not as the main basis for boundary judgment. In the aforementioned target planning area, the natural resources department uploaded a land parcel database with spatial reference, the housing and construction department uploaded several control planning layer files and explanatory documents, and the water resources department provided the river management scope through the interface. In addition, the proposed site selection plan and feasibility study text of the specific functional project to be built in this area were also submitted. The system extracts the outer contour, reporting time, and usage fields of the project site from these data, as well as terms such as integrated logistics warehousing area and port supporting service area in the text, and then divides the entire new area into a unified grid. After segmentation, it was found that the proposed project mainly falls into two adjacent grids. The side closer to the river channel simultaneously superimposed a housing and construction layer with the first historical time stamp and a water conservancy layer with the second historical time stamp. The second historical time stamp is later than the first historical time stamp, which established a spatial index foundation for subsequent difference identification in advance. The purpose of this step is to organize the disorganized business data into spatiotemporal semantic units that can be uniformly compared, thereby enabling the feasibility and traceability of subsequent ontology mapping, geometric correction, and conflict resolution.

[0019] In a preferred embodiment of the present invention, step two specifically includes: constructing a planning ontology library, which contains preset standard classification nodes and the relationships between each node; calculating the semantic similarity between semantic ontology label features and standard classification nodes; and aligning the semantic ontology label features to the corresponding standard classification nodes when the semantic similarity is higher than a preset similarity threshold. When the semantic similarity is lower than or equal to the preset similarity threshold, the semantic ontology label features are marked as semantic fuzzy region features, and an association mapping between the semantic fuzzy region features and the corresponding geometric boundaries is established.

[0020] This embodiment provides an ontology mapping step to address the problem of inconsistent semantic standards; specifically, even after data collection and vocabulary extraction are completed, there may still be situations where the same type of space is described by different departments using different terms. If the original fields are directly entered into the database, it is easy to mistakenly identify essentially the same spaces as different categories in the subsequent overlay stage, or to mistakenly identify categories with strict boundary relationships as compatible ordinary land use. Therefore, this embodiment further introduces a planning ontology database and a semantic ambiguity area identification mechanism. The planning ontology database should at least include standard classification nodes, hierarchical relationships, synonymous relationships, mutual exclusion relationships, and functional compatibility relationships between nodes; standard classification nodes can correspond to land space survey classification, land use control classification, ecological protection classification, and special management classification. Association relationships are used to express knowledge such as the mutual exclusion between permanent basic farmland and ecological protection red lines in the farmland protection and management system and general construction land, and the low compatibility between logistics and warehousing land and residential functions; After receiving the semantic tags extracted from the text, the system compares them one by one with the standard nodes. To clarify the identification and separation logic between the standard classification nodes and the semantically ambiguous features, the following quantitative deduction example is introduced: If three tags are extracted, namely, port supporting service area, warehousing and logistics land, and basic farmland protection area, the warehousing and logistics land can be directly aligned to the logistics and warehousing land standard node. Basic farmland protection zones can be aligned to permanent basic farmland or its superior protection nodes; while port supporting service areas may simultaneously cover service facilities, roads, and warehousing auxiliary spaces, making it difficult to uniquely fall into a single node, and are thus marked as semantically ambiguous areas. Semantic similarity is used to quantitatively evaluate the distance between extracted words and standard node vectors in the business rule feature space. It reflects both literal similarity and whether the contextual roles are consistent. When the system determines that the distance is higher than the threshold, it means that the term is sufficient to stably represent a certain standard planning concept and can be safely merged. When the value is below or equal to the threshold, it indicates that the term is ambiguous, has different local interpretations, or is too broad in its expression. If it is forcibly classified, the ambiguous features may be transformed into rigid misjudgment spatial conflicts. Therefore, this embodiment does not rush to perform error normalization, but retains its uncertain state and establishes an association mapping with the corresponding geometric boundary. Through this mapping mechanism, even when the local semantic features have not reached the threshold for complete resolution, the system can still identify the classification attributes of the polygonal region as having analytical uncertainty in a unified coordinate framework. If a label is similar to multiple standard nodes, the system does not directly select the one with the highest score, but instead retains the candidate node list and incorporates the label into the semantic ambiguity area. If a certain tag cannot find any neighboring nodes in the ontology, it can be written into the term pool to be expanded first, and then manually reviewed to decide whether to add a standard node. If multiple contradictory labels correspond to the same geometric boundary, such as the same plot of land being both an ecological conservation area and a commercial service area, the system will prioritize retaining the contradictory state and will not resolve it in this step, so as to avoid losing the information required for subsequent conflict detection. Within this target planning area, the term "comprehensive service supporting area" appears in the project's submitted materials, and warehousing and logistics land appears in the housing and construction control plan. However, the new district's investment promotion documents describe the same area as a demonstration area for the integration of port, industry, and city. Among these, the warehousing and logistics land can be relatively stably aligned with the standard construction land node. The demonstration area for the integration of port, industry and city is more like a development concept. The spatial scope and functional mapping characteristics involved exceed the definition boundary of a single standard node, and it is not suitable to directly replace the normative classification. The comprehensive service support area may correspond to offices, roads, parking or supporting commercial facilities on different plots. Therefore, it is identified as a semantically ambiguous area and is bound to the boundary of the candidate plot of the proposed project. In this way, during subsequent overlay, the system can know that the plot not only has a spatial location, but also that the classification of its spatial polygons has computational uncertainty characteristics. The purpose of this step is to explicitly expose the inconsistent features of text semantics and convert them into analyzable spatial annotation features, thereby enabling subsequent conflict detection to focus on the features of risky plots in advance.

[0021] In a preferred embodiment of the present invention, step three specifically includes: calculating the position offset of the multi-source heterogeneous spatial data relative to the preset standard reference plane coordinate system based on the coordinate system accuracy characteristics; performing translation and affine transformation on the node coordinates of the multi-source heterogeneous spatial data according to the position offset to complete the geometric correction processing; and sorting the multi-source heterogeneous spatial data after geometric correction processing according to the timestamp in a time sequence. Calculate the boundary change rate of spatial elements at adjacent time nodes, perform interpolation fitting on boundary nodes that exceed the preset change rate threshold, and retain the original coordinates of boundary nodes that do not exceed the preset change rate threshold to complete the temporal smoothing process and output spatiotemporally consistent spatial features.

[0022] This embodiment provides a spatiotemporal consistency processing step for spatial misalignment and historical version jumps; specifically, after semantic alignment is completed in the previous stage, if multi-source graphics are still directly superimposed, it is easy to misjudge the boundary misalignment caused by differences in surveying benchmarks and inconsistent version update rhythms as substantial planning conflicts. Especially in comprehensive target planning areas that simultaneously involve farmland, shorelines, construction boundaries, and project site selection, a deviation of a few meters to tens of meters can affect the approval conclusion; therefore, this embodiment further introduces geometric correction and temporal smoothing. Geometric correction uses a preset standard reference plane coordinate system as a unified reference to correct the positional consistency of spatial data from different sources with this reference. For data with sufficient control points and high accuracy, translation combined with affine transformation can be used to correct overall deviations and local rotation and stretching problems. For layers with only overall translation errors, simpler node translations can be performed; for historical data whose boundaries are vectorized from paper maps, more attention should be paid to eliminating systematic deviations, rather than excessively stretching local nodes to avoid destroying the original boundary shape; its engineering significance lies in bringing the boundaries that should coincide back to the same spatial reference, so that the subsequent overlay results reflect the real planning relationship rather than cartographic differences. After completing the geometric correction, the system sorts the data of the same spatial element across multiple periods according to the timestamp; it then observes whether the boundary changes at adjacent time points conform to the actual land evolution patterns. Specifically, the system defines the boundary change rate as the position offset distance of the same boundary node in two adjacent periods divided by the time span of the two periods; Here is a specific example of quantitative deduction: Let the boundary nodes be... At any moment The coordinates are At adjacent times The coordinates are The system then calculates the Euclidean spatial displacement distance between the two and divides it by the time span. Let the boundary change rate be... The calculation formula is:

[0023] To obtain the accurate boundary rate of change ; Land space objects usually have a certain degree of continuity. For example, although village boundaries, farmland patches, and river management lines may change, they will not exhibit discontinuous boundary jumps that exceed the preset topological curvature threshold within a preset time period. If a node experiences an abnormal jump in adjacent periods, that is, the boundary change rate exceeds the preset change rate threshold, such as being greater than the preset upper limit of 0.1 meters / day, it often means a change in the data update scale, different collection accuracy, or human editing error, rather than a sudden shift of the actual land parcel. At this point, the system performs interpolation fitting on boundary nodes that exceed the preset rate of change threshold. Specifically, it can use linear interpolation to determine their smooth coordinates by combining the reasonable positions of the preceding and following time nodes, so that the boundary changes are closer to the interpretable spatiotemporal evolution. For nodes whose changes are within the normal range, the original coordinates are retained to avoid over-smoothing that masks the real changes. In 2021, 2022 and 2023, the eastern boundary of a certain plot of land was generally continuous in 2021 and 2022, but in 2023 the eastern boundary suddenly concave to form a sharp angle. If there are no records of land acquisition, river regulation, or planning adjustments in the area, the system can treat the sharp corner as an abnormal jump node and smooth it out; if another area has a boundary setback due to the approval of road construction, even if the change is significant, the true boundary will be preserved because there is corresponding time reference. If there is only single-period data for the same element, then no time-series smoothing is performed; only the geometric correction result is retained and the time-series evidence is marked as insufficient. If the span between adjacent periods is too large and multiple years of data are missing in between, then the smoothing strength should be reduced to avoid fabricating non-existent intermediate evolutions. If the coordinate precision of a certain layer is too low or the control points are insufficient, the system can limit the affine transformation range, perform only overall alignment, and list it as a low-confidence reference layer; if obvious self-intersection, polygon damage, or topological breakage still occurs after correction, geometric repair or manual review of the original layer will be prioritized instead of directly entering the overlay process. In the target planning area, the 2022 version of the control plan layer from the housing and construction department and the 2023 farmland layer from the natural resources department have boundary offsets near the candidate plots of the proposed project. The water conservancy shoreline data also uses base maps from different periods. After the system projected the three onto the standard benchmark, it was found that the storage plots that seemed to occupy a large area of ​​farmland were actually only a false overlap caused by coordinate offset. Meanwhile, a time-series inspection of a broken line boundary along the river revealed that it was relatively straight in both the previous and following years, with an abnormal bend only appearing in the middle period. Therefore, the bend node was smoothed out. The boundary obtained after the smoothing process is closer to the actual approved spatial pattern of the area. The purpose of this step is to eliminate spurious conflicts and changes caused by differences in data sources, so that subsequent topological analysis can be based on a unified, continuous and physically interpretable spatial foundation.

[0024] In a preferred embodiment of the present invention, step four specifically includes: performing a geometric intersection operation on the semantically ambiguous region features bound to geometric boundaries and the spatiotemporally consistent spatial features in the same coordinate space to generate an overlay layer; and in the overlay layer, calling a preset topological operator to calculate the spatial overlap area between different elements. When the spatial overlap area is greater than zero and the feature attributes include mutually exclusive red line attributes, a hard conflict feature is determined and generated; when the spatial overlap area is greater than zero and the feature attributes are non-mutually exclusive but functionally incompatible, a soft conflict feature is determined and generated; when the spatial overlap area is greater than 0 and the feature attributes are mutually compatible, it is determined to be a regular overlap and there is no conflict; when the spatial overlap area is equal to 0, there is no conflict. Among these, mutually exclusive red line attributes correspond to regulatory control requirements that cannot coexist; non-mutually exclusive but functionally incompatible attributes correspond to situations that are not absolutely prohibited but are difficult to coordinate functionally; and mutually compatible attributes correspond to situations that can coexist functionally.

[0025] This embodiment provides a topology detection step for spatial conflict type identification; specifically, after completing the spatiotemporal consistency processing, although the data is relatively neat, if there is no clear conflict classification, the system still cannot distinguish between conflicts that must be eliminated immediately and functional conflicts that need to be negotiated and optimized. In actual planning and management, these two types of problems are handled in completely different ways. Therefore, this embodiment further introduces overlay layers and topology determination mechanisms. The system first places the geometric boundaries bound to the semantically ambiguous region features, the spatial features of the aligned standard classification nodes, and the spatiotemporal consistency spatial features after correction and smoothing in the same coordinate space to perform geometric intersection, forming an overlay layer; The significance of overlay layers lies in directly projecting semantically uncertain land parcels and spatiotemporally unified standard or business layers into the same spatial relationship, thereby determining the final boundary range of ambiguous terms and which control lines they come into contact with; The topology operator is called to analyze whether different elements overlap, cross each other, contain each other, and whether the overlapping area is greater than zero. The overlapping area here is not just a numerical value, but represents whether the same surface space is repeatedly occupied under multiple management rules. When land parcels overlap and both parcels contain mutually exclusive red line attributes, the system determines that the conflict is a hard conflict. Mutually exclusive red line attributes usually correspond to regulatory control requirements that prevent coexistence, such as permanent basic farmland and newly added industrial construction land, core protected areas of ecological protection red lines and commercial development land, etc. Once a hard conflict enters the approval process, it often means that the project cannot be directly promoted; when there is an overlap in land parcels, but the two parties are not absolutely prohibited from each other, but their functions are difficult to coordinate, the system judges it as a soft conflict. For example, problems such as logistics distribution centers encroaching on neighboring residential land, school land being adjacent to high-noise traffic facilities, and river landscape control areas overlapping with high-intensity storage yards can still be solved through functional adjustments, boundary optimization, or condition control. If the overlapping area is zero, it means that there is no direct spatial conflict at least at the current boundary level, and the process can proceed to the regular warehousing process. In a typical topological overlap verification scenario: Suppose there are a first plot, a second plot, and a third plot, where the first plot is permanent basic farmland, the second plot is land for proposed warehousing and logistics, and the third plot is land for supporting living services; if the geometric intersection operation of the first plot and the second plot generates the first overlapping area, then the first overlapping area corresponds to a hard conflict. If the second and third plots form an overlapping area, but the two are not absolutely prohibited, and there is only a functional conflict such as noise, traffic, and fire protection, then the second overlapping area corresponds to a soft conflict; if the first and third plots only have boundary contact and do not actually overlap, then it does not constitute a direct conflict. If two elements only have point contact or line contact without area overlap, the system can record them as adjacency and not directly identify them as conflict, but may issue a risk warning under specific rules; if a plot of land is missing attributes and it is impossible to determine whether it has mutually exclusive red line attributes, the system will first classify it as pending conflict and will not rashly give a hard or soft conflict conclusion. If a semantically ambiguous region corresponds to multiple candidate categories, the system can simulate and superimpose them separately and output multiple conflict scenarios for comprehensive judgment during subsequent resolution. If a fragmented region with an area smaller than a preset threshold appears after geometric intersection, the system can combine the minimum cartographic unit threshold to determine whether it belongs to cartographic noise, thus avoiding a large number of meaningless conflicts caused by irregular topological redundant patches. In the target planning area scenario, the system overlays the candidate plot of the proposed project with permanent basic farmland, river management area, urban development boundary and residential area control plan layers; the results show that the first side of the candidate plot forms a stable overlapping area with the permanent basic farmland, which is a hard conflict. The second side of the candidate plot, which is close to the existing residential area, has some overlap and partial encroachment on the living service function area, which is considered a soft conflict; the river management area only touches the boundary in a local area and does not have any areal encroachment, so it is not considered a direct conflict for the time being; therefore, the system can focus its subsequent processing on farmland reduction and functional transition, rather than evenly distributing the review effort. The purpose of this step is to transform complex spatial overlap relationships into conflict objects that can be handled in a tiered manner, so that subsequent automatic repair strategies can take differentiated actions based on the nature of the conflict.

[0026] In a preferred embodiment of the present invention, step five involves performing conflict resolution processing on hard conflict features and soft conflict features based on the authority level of the data source. Specifically, this includes: reading the authority level of the data source corresponding to each of the two conflicting elements that generate the hard conflict feature; comparing the authority levels of the two data sources, retaining the geometric boundaries and attributes of the conflicting elements with higher authority levels, trimming or deleting the overlapping parts of the conflicting elements with lower authority levels, and accumulating the number of conflict resolutions. For soft conflict characteristics, the preset attribute compatibility mapping table is queried to obtain the transitional attribute corresponding to the attribute of the high-level feature. The attribute of the low-level feature is modified to the transitional attribute, and the number of conflict resolutions is accumulated. The features that have been clipped or deleted and whose attributes have been modified are stored in the database to form a logical closed-loop database.

[0027] This embodiment provides a conflict resolution step based on authority level; specifically, simply identifying hard and soft conflicts is not enough to complete intelligent database building, because if the system remains at the problem discovery stage, planning and management personnel still need to manually compare the sources and determine the objects to be retained, and the processing efficiency and consistency are difficult to guarantee. Especially during the annual update phase of the new area, more than ten types of conflicts may occur simultaneously within the same grid, so it is necessary to introduce executable resolution rules; The system pre-maintains a data source authority ranking system; this ranking is not simply sorted by department from highest to lowest, but is determined by a comprehensive consideration of standardization, approval effectiveness, data update procedures, and surveying accuracy. For example, permanent basic farmland confirmed through standardized procedures is usually of a high grade, and officially approved control detailed plans are also of a high grade, while site selection lines drawn by the project applicant, investment concept maps, or third-party assessment maps are of a relatively low grade. For hard conflicts, after reading the data source levels of both conflicting parties, the system prioritizes preserving the geometric boundaries and attributes of higher-level features and clips the overlapping parts of lower-level features. If a low-level element falls entirely within the prohibited space of a high-level element, the conflicting part or the entire candidate object can be directly deleted. The basis for this approach is that hard conflicts often involve regulatory boundaries, which should not be blurred through compromise. For soft conflicts, the system does not directly delete low-level objects, but instead queries the attribute compatibility mapping table to adjust low-level features to transitional attributes that are more compatible with high-level features. The transitional attributes here can be understood as functional categories that serve as a buffer between the conflicting parties; for example, if a high-level element is a residential area and a low-level element was originally a high-intensity logistics and warehousing area, the conflicting part can be adjusted to transitional attributes such as low-interference supporting facilities, green isolation belts, parking and municipal facilities. If the high-level element is an ecological landscape corridor and the low-level element was originally a general commercial service facility, it can be adjusted to a low-intensity function such as ecological service facilities or walkway service nodes. Each time a cropping, deletion, or attribute adjustment occurs, the system accumulates the number of conflict resolutions within that spatial grid and writes the relationship before and after the processing into the database, forming a closed-loop record from original data, conflict identification, repair actions to repair results; If there is a first hard conflict feature pair consisting of a first conflict element and a second conflict element in the first target grid, where the first conflict element comes from the project site selection sketch, then the first conflict element is retained and the second conflict element is trimmed. If there is a soft conflict CD in the same grid, where C is the approved residential area control plan and D is the supporting plan for the proposed project, then D will not be deleted. Instead, the attribute of D within the conflict range will be adjusted from the storage operation surface to the supporting green buffer zone. In this way, the number of conflict resolutions for the first target grid will increase by at least 2, and two processing records will be left in the database. If the conflicting parties are of the same level, the system may give priority to the one with the newer effective time; if the effective time is also the same, it will be marked as a conflict of the same level and await manual review, and will not be automatically overwritten; if the low-level feature is clipped to form a fragmented polygon with an area feature lower than the preset minimum cartographic unit threshold, and no longer meets the minimum cartographic unit requirement, it can be cleared or merged into a neighboring area with the same attribute. If a soft conflict cannot find a corresponding transition attribute, it will not be forced to modify it. Instead, the original attribute will be retained and a missing compatibility relationship marker will be added so that the mapping table can be manually supplemented. If a conflict contains both boundary occupancy and attribute incompatibility, the boundary will be treated as a hard conflict first, and then the remaining part will be adjusted as a soft conflict. In the scenario of the target planning area, the part of the candidate plot that encroaches on the permanent basic farmland on the first side is a hard conflict; after reading the data, the system confirmed that the farmland data comes from the high-level specification layer, and the boundary of the candidate plot comes from the plan submitted by the project unit. Therefore, the farmland boundary is retained and the overlapping area of ​​the candidate plot is cut off. Meanwhile, the second side of the candidate plot, which is close to the residential area, constitutes a soft conflict. Based on the attribute compatibility mapping table, the system adjusts the originally planned heavy warehousing operation area into a combination of parking and changing area and green isolation belt as a transitional attribute. All processing actions and times are written into the corresponding grid records to form a traceable closed-loop database. The purpose of this step is to further transform the discovery of conflicts into the repair of conflicts according to rules, so as to ensure that the database output results are not only queryable, but also have logical integrity that allows them to directly enter the subsequent planning review.

[0028] In a preferred embodiment of the present invention, the confidence heatmap of the output data based on the logical closed-loop database specifically includes: extracting the number of conflict resolutions occurring in each spatial grid in the logical closed-loop database and the coordinate system precision features of the original data; Obtain preset conflict resolution frequency weights and accuracy weights. Then, perform a weighted summation of the conflict resolution frequency and coordinate system accuracy features based on these weights to calculate the overall confidence score for each spatial grid. The conflict resolution frequency and overall confidence score are negatively correlated, while the coordinate system accuracy features and overall confidence score are positively correlated. The specific weighting calculation includes mapping the conflict resolution frequency to a basic resolution score. The mapping formula is Map the coordinate system precision features to a basic precision score. The mapping formula is ; The overall confidence score is mapped to a preset color gradient band; the spatial grid is rendered according to the color gradient band to generate and output a data confidence heatmap.

[0029] This embodiment provides a data confidence heatmap generation step for result visualization and quality assessment. Specifically, after the aforementioned automatic repair, although the database has formed a closed loop, the planning and management department still needs to know which spatial grids fall into the low confidence risk range and needs to trigger the algorithm review or manual inspection mechanism first. If only the repaired final layer is output, many intermediate risk features will be hidden. Therefore, this embodiment combines the number of conflict resolutions with the original data precision characteristics to generate a confidence feature distribution expression result for space management decision-making; The system extracts two types of core information from the logical closed-loop database according to the spatial grid: the first is the number of conflict resolutions that have occurred cumulatively during the database construction process of the grid, and the second is the coordinate system precision characteristics corresponding to the original data entering the grid. The former reflects the density of data contradiction distribution in the region. The more times the contradiction is resolved, the more frequently the region undergoes feature reconstruction and the worse the original consistency. The latter reflects the geometric reliability of the basic data layer. The higher the precision feature, the more robust and reliable the base spatial positioning still is, even if the algorithm performs the contradiction resolution judgment. By combining the two for composite calculation, the system can scientifically distinguish between situations where the analysis results are unreliable due to missing precision parameters of the underlying spatial reference system or positional offset exceeding the preset spatial tolerance parameters, and situations where, although the conflict resolution algorithm has undergone high-frequency iterations, the final repair results have strong spatial topological convergence and reliability due to the stable accuracy of the base map control points. Specifically, since the number of conflict resolutions and the precision characteristics of the coordinate system differ in terms of dimensions and direction, the system performs reverse standardization on the number of conflict resolutions before weighted summation, mapping a higher number of conflict resolutions to a lower base score, indicating a decrease in data credibility. At the same time, the coordinate system accuracy characteristics are positively standardized to convert high precision into higher base scores, thereby eliminating differences in dimensions and polarity. To clarify the calculation rules for data transformation, the forward standardization process here adopts extreme value normalization logic, that is, targeting the precision characteristics of the coordinate system. To obtain its maximum value in the current spatial grid analysis sample set. and minimum value The specific base accuracy score, let the base accuracy score be... The calculation formula is as follows:

[0030] Based on calculation rules Map it to be in Precision base score of the interval ; Similarly, reverse standardization addresses the number of conflict resolution attempts. To obtain its maximum value in the current spatial grid analysis sample set. and minimum value The specific basic score for resolution, let the basic score for resolution be... The calculation formula is as follows:

[0031] Based on calculation rules Map it to be in Basic score for interval resolution ; The system obtains the preset resolution count weight and accuracy weight, and generates a comprehensive confidence score for each spatial grid by multiplying the two basic scores after transformation by the corresponding weights and performing a weighted summation calculation. The score is then mapped to a preset color gradient band. Colors can be represented by a sequence from cool to warm, and confidence feature scores can be set from high to low, or vice versa, as long as the rules remain consistent in the feature rendering system. To clarify the technical path of multidimensional parameter dimensionality reduction mapping and thermal coloring, the following quantitative deduction example is introduced: if the number of grid first analysis resolution operations is small and the mapping data accuracy features are high, the score is high and the rendering mapping is a cool color tone; If the second analysis of the mesh has a high number of resolution operations and some of the overlay layers have low precision parameters, then the rendering will be mapped to a warm color tone. If the third analysis of the grid is triggered multiple times to resolve and reconstruct the logic, but the conflict determination mainly comes from the reporting defects of the same specific business layer, and its core control network and other high-level basic reference data have not been offset, then it can be mapped, calculated and displayed as an intermediate color value; such a visualization feature transformation can help the system and the operator to intuitively identify high-risk abnormal grids. In engineering applications, heat map features do not replace standard physical boundaries, but serve as an auxiliary basis for spatial quality control rules. This feature data can be linked to planning verification servers and control business flow terminals to serve as judgment parameters to prompt the system whether to call higher-level risk verification algorithms, whether to automatically trigger the reading of original conflict log data packets, or whether to issue manual verification instructions. If no conflict resolution calculation event occurs in a certain grid, but the data accuracy index itself is also low, the system will not directly assign the grid the highest comprehensive confidence score feature, but should maintain the neutral or confidence discount state judgment feature. If the characteristic value of the number of conflict resolutions triggered in a certain grid is high, but all judgment rules are based on a high-level standard base to achieve closed-loop correction, the system can attach an attribute identifier vector in the heat map log output indicating that the rules have been stably converged and repaired, so as to prevent downstream applications from directly intercepting it as abnormal and unusable area data. If a grid lacks identifiable precision calculation feature parameters, the system can default to configuring a penalty factor to suppress its confidence score or force it to be labeled as a precision parameter missing classification label. If multiple high-risk score grids cross the boundary of the same special construction control area in terms of topological relationship, the system will aggregate and output a confidence downgrade prompt based on the business unit, which will facilitate system-level review of spatial compliance attributes. In the target planning area scenario, among the two grids where the candidate area of ​​the proposed project is located, the first spatial grid is rendered as a warm-colored warning by the system because it involves hard conflict calculation events of permanent basic farmland and multiple boundary shearing update operations, and the original data in the database has general accuracy features. Although the second spatial grid triggered the soft conflict adjustment logic related to the residential area rules, the underlying control plan layer and the basic base plate data have high relative accuracy feature parameters, and the rule loop is clear after the transition attribute conversion of the function and purpose is completed. Therefore, the system generates its intermediate feature score and corresponding rendering color. When the verification command is sent to the relevant area for automated approval and verification, the system reads the risk characteristic index of the first spatial grid value which is higher than the preset risk judgment threshold. Then, according to the preset rules, it can automatically prioritize calling the underlying records of the conflict repair calculation trajectory log and related data. The purpose of this step is to explicitly parameterize the differences in features and risks hidden in the database construction calculation process, so that the database results not only carry the final state element conclusions of geometric compliance, but also include the confidence level feature distribution basis of spatial reliability, which facilitates the subsequent implementation of differentiated verification calculations and multi-level verification instructions by the system.

[0032] In a preferred embodiment of the present invention, the method further includes: receiving an externally input error correction instruction for the logical closed-loop database, and extracting error label data from the error correction instruction; using the error label data as training samples, and updating the association weights in the planning ontology library through a backpropagation algorithm; specifically, mapping the planning ontology library into a semantic network graph structure composed of classification nodes and associations, calculating the loss error between the predicted classification probability and the target classification label based on the error label data using the weight parameters of the network connections corresponding to the associations, and using the chain rule to propagate the loss error along the association path of the semantic network graph structure through a backpropagation algorithm, thereby updating the association weights; and re-executing steps two to five using the updated planning ontology library to achieve adaptive iteration of the library building logic.

[0033] This embodiment provides an adaptive iterative step for continuous error correction; specifically, the aforementioned scheme can already complete automatic database construction and conflict repair, but in the land and space planning business, classification standards, local interpretations and special rules will continue to evolve, and new functional expressions may also emerge during the development of new areas; If the ontology database remains unchanged for a long period of time, the system may repeatedly identify the same semantic ambiguity in a new round of database construction, reducing efficiency; therefore, this embodiment introduces an external error correction feedback mechanism on the basis of a closed-loop database. The system receives error correction instructions for the logical closed-loop database from planning units, reviewers, surveying and mapping quality inspectors, or relevant departments in charge of specific topics; Error correction instructions can be used to annotate error labels, error classifications, error boundary bindings, or error compatibility relationships. Among these, extracting error label data from error correction instructions is particularly important because it directly reflects which term relationships in the ontology have not yet been fully learned. For example, in a new round of special planning, a certain place has clearly defined the comprehensive support area behind the port as a specific logistics supporting function. If the system has previously classified it as a semantically ambiguous area, the correct relationship between this term and the standard node can be fed back to the system through manual error correction. The system uses the mislabeled data as training samples and updates the association weights in the planning ontology library through the backpropagation algorithm, making it easier for subsequent similar terms to be accurately aligned. The update process here can be completed using the backpropagation algorithm; in specific implementation, the system maps the planning ontology library into a semantic network graph structure composed of classification nodes and associations, where each association corresponds to the initial weight parameters of the network connection. After receiving erroneous label data, the system calculates the loss error between the predicted classification probability output by the current network and the manually corrected target classification label; The system uses the cross-entropy algorithm as the judgment criterion, takes the target classification label as the ideal real distribution reference, and the predicted classification probability as the actual output distribution result, and calculates the cross-entropy between the two to obtain the loss error. Specifically, let the distribution of target classification labels be... The predicted classification probability distribution is as follows: Then the loss error The calculation formula is:

[0034] This formula quantifies the degree of deviation in semantic association; By using the chain rule of differentiation and the backpropagation algorithm, the loss error is propagated back step by step along the association path of the semantic network. The gradient update amount of each association weight is calculated. Based on the product of the preset learning rate and the gradient update amount, the corresponding association weights are precisely adjusted step by step to complete the update iteration. Its function is to dynamically adjust the semantic association strength between each classification node in the ontology based on the error correction samples, so as to gradually reduce the error matching. If a term T1 was originally closely related to nodes N1 and N2, causing the system to repeatedly classify it into the fuzzy region; When external error correction continues to indicate that T1 should belong to N2, the system will increase the association weight of T1 to N2 and decrease its association weight to N1 after the update. After the update is completed, the system uses the updated planning ontology library to re-execute steps two through five, forming a new round of more stable database construction results, thereby achieving adaptive iteration of the database construction logic; in this way, the database is no longer a one-time result, but can self-converge along with business iterations; If external error correction instructions contradict each other, such as different departments giving different classification opinions on the same term, the system will not directly overwrite the old rules, but will first sort them according to the level of error correction source, standard basis and time, and include contradictory samples in the queue to be reviewed. If the number of error correction samples is too small to stably update a certain relationship, the system can adopt a conservative update strategy to avoid individual samples causing the ontology to shift. If after re-executing steps two to five, the new repair results cause abnormal changes to the boundaries of the existing high-level specifications, the system will automatically roll back to the previous stable version and output an update impact report for manual review. In the target planning area scenario, after the first round of database construction, the reviewers pointed out that a plot of land in the candidate area of ​​the proposed project, which was marked as a comprehensive service supporting area, should be prioritized as a transportation distribution supporting facility land rather than a generalized service node, according to the latest special guidelines of the new area. After receiving the error correction instruction, the system updates the ontology relationship using the relevant tags, context descriptions, and corresponding plots as training samples. After the update is completed, the subsequent process is executed again. The semantic ambiguity area that originally appeared on the plot is reduced, and the soft conflict with the residential area is more accurately limited to the road connection and parking conversion area, rather than the entire plot being regarded as an incompatible space. The purpose of this step is to incorporate the experience gained from manual review into a set of rules for the sustainable use of the system, thereby enabling the database construction logic to be continuously optimized as the planning context and local standards evolve, and reducing repetitive semantic misjudgments.

[0035] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for intelligent database construction of land spatial planning based on multi-source heterogeneous data integration, characterized in that: The specific steps include: Step 1: Collect multi-source heterogeneous spatial data and text specification data; extract geometric boundaries, timestamps, coordinate system precision features, and feature attributes from the multi-source heterogeneous spatial data; and extract semantic ontology label features from the text specification data; divide the multi-source heterogeneous spatial data into several spatial grids according to a preset grid size. Step 2: Construct a planning ontology library, map the semantic ontology label features to the planning ontology library, identify differences through semantic comparison, generate semantic ambiguity region features, and bind the semantic ambiguity region features to their corresponding geometric boundaries; Step 3: Based on the timestamp and the coordinate system precision characteristics, and with a preset standard reference plane coordinate system as a reference, perform geometric correction and temporal smoothing on the multi-source heterogeneous spatial data to generate spatiotemporally consistent spatial features. Step 4: Spatial superposition of the semantic ambiguity region features bound to geometric boundaries, the spatial features of aligned standard classification nodes, and the spatiotemporal consistency spatial features; and conflict detection by combining the spatial overlap area and feature attributes between different elements using a preset topology operator to generate hard conflict features and soft conflict features. Step 5: Obtain the preset data source authority level. The data source authority level is determined by comprehensively considering standardization, approval effectiveness, data update procedures, and surveying accuracy. Based on the data source authority level, perform conflict resolution processing on the hard conflict features and the soft conflict features, record the number of conflict resolutions within each spatial grid, generate a logical closed-loop database, and output a data confidence heatmap based on the logical closed-loop database.

2. The intelligent database construction method for land spatial planning based on multi-source heterogeneous data integration as described in claim 1, characterized in that, Step one specifically includes: Collect multi-source heterogeneous spatial data, image data, and text specification data from different business departments through preset data interfaces; The format of the multi-source heterogeneous spatial data, image data and text standard data is parsed, and the corresponding data generation time is extracted as the timestamp. The spatial reference information of the multi-source heterogeneous spatial data is analyzed, the error parameters of the spatial reference information are extracted as the coordinate system accuracy features, the business fields are extracted as the feature attributes, and the vector outer contour is extracted as the geometric boundary. Natural language processing is performed on the text specification data to extract classification standard vocabulary as semantic ontology label features; The multi-source heterogeneous spatial data is divided into several spatial grids according to the preset grid size.

3. The intelligent database construction method for land spatial planning based on multi-source heterogeneous data integration as described in claim 2, characterized in that, Step two specifically includes: Construct a planning ontology library, which includes preset standard classification nodes and the relationships between the nodes; Calculate the semantic similarity between the semantic ontology label features and the standard classification nodes; When the semantic similarity is higher than a preset similarity threshold, the semantic ontology label features are aligned to the corresponding standard classification node; When the semantic similarity is lower than or equal to the preset similarity threshold, the semantic ontology label feature is marked as the semantic fuzzy region feature, and an association mapping between the semantic fuzzy region feature and the corresponding geometric boundary is established.

4. The intelligent database construction method for land spatial planning based on multi-source heterogeneous data integration as described in claim 3, characterized in that, Step three specifically includes: Based on the accuracy characteristics of the coordinate system, the position offset of the multi-source heterogeneous spatial data relative to the preset standard reference plane coordinate system is calculated; Based on the position offset, the node coordinates of the multi-source heterogeneous spatial data are translated and affine transformed to complete the geometric correction process. The multi-source heterogeneous spatial data after the geometric correction process is completed is sorted by time sequence according to the timestamp. Calculate the boundary change rate of spatial elements at adjacent time nodes, perform interpolation fitting on boundary nodes that exceed the preset change rate threshold, and retain the original coordinates of boundary nodes that do not exceed the preset change rate threshold to complete the temporal smoothing process and output the spatiotemporal consistent spatial features.

5. The intelligent database construction method for land spatial planning based on multi-source heterogeneous data integration according to claim 4, characterized in that, Step four specifically includes: The semantically ambiguous region features bound to geometric boundaries, the spatial features of aligned standard classification nodes, and the spatiotemporally consistent spatial features are geometrically intersected in the same coordinate space to generate an overlay layer. In the overlay layer, the preset topology operator is invoked to calculate the spatial overlap area between different elements; When the spatial overlap area is greater than 0 and the feature attributes include mutually exclusive red line attributes, the hard conflict feature is determined and generated. When the spatial overlap area is greater than 0 and the feature attribute is a non-mutually exclusive but functionally incompatible attribute, the soft conflict feature is determined and generated. When the spatial overlap area is greater than 0 and the element attributes are mutually compatible, it is determined to be a regular overlap and there is no conflict; when the spatial overlap area is equal to 0, there is no conflict; wherein, the mutually exclusive red line attribute corresponds to the normative control requirements that cannot coexist; the non-mutually exclusive but functionally incompatible attribute corresponds to the situation that is not absolutely prohibited but is difficult to coordinate functionally; the mutually compatible attribute corresponds to the situation that can coexist functionally.

6. The intelligent database construction method for land spatial planning based on multi-source heterogeneous data integration according to claim 5, characterized in that, Step five involves performing conflict resolution processing on the hard-conflict features and the soft-conflict features based on the authority level of the data source, specifically including: Read the authority level of the data source corresponding to each of the two conflicting elements that generate the hard conflict feature; Compare the authority levels of the two data sources, retain the geometric boundaries and attributes of the conflicting elements with the higher authority level, trim or delete the overlapping parts of the conflicting elements with the lower authority level, and accumulate the number of conflict resolutions. For the soft conflict feature, a preset attribute compatibility mapping table is queried to obtain the transition attribute corresponding to the high-level feature attribute, the attribute of the low-level feature is modified to the transition attribute, and the number of conflict resolutions is accumulated. Elements that have been cropped, deleted, or had their attributes modified are stored in the database, forming the logical closed-loop database.

7. The intelligent database construction method for land spatial planning based on multi-source heterogeneous data integration according to claim 6, characterized in that, The confidence heatmap of the data output based on the aforementioned logical closed-loop database specifically includes: Extract the number of conflict resolutions occurring within each spatial grid in the logical closed-loop database and the coordinate system precision features of the original data; Obtain preset conflict resolution frequency weights and accuracy weights. Then, perform a weighted summation of the conflict resolution frequency and the coordinate system accuracy feature based on the conflict resolution frequency weights and accuracy weights to calculate the comprehensive confidence score for each spatial grid. Specifically, the conflict resolution frequency is negatively correlated with the comprehensive confidence score, and the coordinate system accuracy feature is positively correlated with the comprehensive confidence score. The weighted calculation includes mapping the conflict resolution frequency to a basic conflict resolution score. The mapping formula is Map the coordinate system precision features to a basic precision score. The mapping formula is ; The overall confidence score is mapped to a preset color gradient band; The spatial grid is rendered based on the color gradient band, and the data confidence heatmap is generated and output.

8. The intelligent database construction method for land spatial planning based on multi-source heterogeneous data integration according to claim 7, characterized in that, Also includes: Receive externally input error correction instructions for the logical closed-loop database, and extract error tag data from the error correction instructions; The erroneous label data is used as training samples, and the association weights in the planning ontology are updated using the backpropagation algorithm. Specifically, the planning ontology is mapped into a semantic network graph structure consisting of classification nodes and associations. The weight parameters of the network connections corresponding to the associations are used to calculate the loss error between the predicted classification probability and the target classification label based on the error label data. The loss error is then propagated along the association path of the semantic network graph structure using the chain rule of differentiation and the backpropagation algorithm, thereby updating the association weights. Steps two through five are re-executed using the updated planning ontology library to achieve adaptive iteration of the library building logic.

Images