A multi-source pipeline data registration and fusion labeling method for heterogeneous coordinate systems

Abstract

The application relates to the technical field of data processing, and discloses a multi-source pipeline data registration and fusion labeling method for heterogeneous coordinate systems, which comprises the following steps: obtaining multi-source pipeline data and source coordinate system description information to establish a coordinate transformation model. Pipeline nodes, pipeline sections, connectivity relationships and semantic attributes are extracted, source and target pipeline network topology semantic constraint graphs are constructed, a registration error model is established, parameters are jointly solved through iterative optimization, and a globally consistent registration result is obtained. Pipeline nodes and pipeline sections from different pipeline data sources and corresponding in spatial position, topology structure and semantic attributes are clustered into unified pipeline objects and assigned with unified identification. A quality-driven attribute fusion model is constructed, geometric attributes are corrected within engineering constraints, classified attributes are weighted, and finally, a fusion labeling result is obtained. The application obtains unified pipeline objects which are globally consistent, topologically correct, semantically unified and provided with quality labeling.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method for registration and fusion annotation of multi-source pipeline data for heterogeneous coordinate systems. Background Technology

[0002] In the informatization construction of urban underground infrastructure, pipeline data is evolving from single surveying results to a coexistence of multi-source heterogeneous data, including design drawings, as-built drawings, geophysical exploration results, BIM (Building Information Modeling) models, and operation and maintenance ledgers. These data sources often employ different coordinate systems, elevation benchmarks, and data formats. To achieve three-dimensional visualization management, digital twins, and intelligent operation and maintenance of urban pipeline networks, it is necessary to register and fuse multi-source pipeline data under a unified spatial benchmark. This requires ensuring the accuracy of the geometric positions of pipeline nodes and segments while maintaining the consistency of network connectivity and semantic attributes. Therefore, a demand has arisen for multi-source pipeline data registration and fusion annotation oriented towards heterogeneous coordinate systems.

[0003] However, existing technologies mostly focus on the geometric level of model coordinate unification or the 3D visualization of single-source pipeline models, which is insufficient to meet the requirements of fine registration and fusion annotation of multi-source pipeline data. For example, patent CN109685893B provides a spatial integrated modeling method, which achieves spatial integrated model output by unifying the coordinates of multiple above-ground and underground models to a preset coordinate system. However, it mainly focuses on model format conversion and overall coordinate conversion, lacking topological semantic constraint modeling for semantic attributes such as the connectivity of pipeline nodes and pipeline segments, pipeline type, and facility number. It also does not set up a coordinate transformation model that combines translation, rotation, and local weak deformation for local measurement distortion. At the same time, existing methods lack a registration error model and registration quality evaluation mechanism based on geometric residuals, topological consistency, and semantic consistency. This makes it difficult to perform attribute fusion and unified identification annotation of multi-source pipeline objects under a unified spatial coordinate system, resulting in problems such as location mismatch, incorrect connectivity, and attribute conflicts in multi-source pipeline data within the integrated management system.

[0004] Therefore, it is necessary to design a multi-source pipeline data registration and fusion annotation method for heterogeneous coordinate systems to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a multi-source pipeline data registration and fusion annotation method for heterogeneous coordinate systems, which aims to solve the problems of inconsistent geometric positions, conflicting pipeline network connectivity, inconsistent attribute information, and difficulty in forming a reliable unified pipeline database among multi-source pipeline data in the prior art.

[0006] This invention proposes a method for registration and fusion annotation of multi-source pipeline data for heterogeneous coordinate systems, including: Acquire multi-source pipeline data and source coordinate system description information, and establish a coordinate transformation model for each pipeline data source. The coordinate transformation model includes translation parameters, rotation parameters, and local weak deformation parameters. Based on the multi-source pipeline data, pipeline nodes, pipeline segments, connectivity relationships, and semantic attributes are extracted. Source pipeline network topology semantic constraint graphs and target pipeline network topology semantic constraint graphs are constructed. Semantically unique registration constraint features are selected for coarse registration. Based on the coarse registration results, a registration error model is established. The parameters of each coordinate transformation model are solved by iterative optimization. Local registration optimization is performed on local spatial regions where the registration error exceeds the limit to obtain globally consistent registration results. The registration error model includes geometric residuals, topological consistency, and semantic consistency. Based on the registration results, a registration quality evaluation value is calculated for each pipeline node and each pipeline segment. Pipeline nodes and pipeline segments from different pipeline data sources that correspond in spatial location, topology, and semantic attributes are clustered into a unified pipeline object, and a unified identifier is assigned to the unified pipeline object. For each unified pipeline object, a quality-driven attribute fusion model is constructed based on the registration quality evaluation value, data source accuracy, and business credibility. Within the engineering constraints, the geometric attributes are corrected, and the classification attributes are weighted to obtain the final fusion annotation result.

[0007] Furthermore, when acquiring multi-source pipeline data and source coordinate system description information, the following is included: The source coordinate system description information includes the coordinate system type, coordinate unit, coordinate axis direction, and elevation datum of each pipeline data source. Under a unified spatial coordinate system, the source coordinate system description information is associated with known control points in the engineering control network and pipeline nodes with unique semantic identifiers. The translation and rotation parameters of each pipeline data source are obtained based on the correspondence obtained from the association, and an initial coordinate transformation model is generated.

[0008] Furthermore, obtaining the local weak deformation parameters includes: Based on the spatial distribution of the pipeline nodes and pipeline segments and the division results of the engineering control network, each pipeline data source is divided into multiple local spatial regions or into multiple pipeline segments along the pipeline route, and local weak deformation parameters are preset for each local spatial region or each pipeline segment to characterize local weak deformation.

[0009] Furthermore, when constructing the source network topology semantic constraint graph and the target network topology semantic constraint graph, and selecting semantically unique registration constraint features for coarse registration, the following steps are taken: For each pipeline data source, the pipeline nodes representing well chambers, valves, inlets / outlets, and equipment foundations, as well as the pipeline segments connected to the pipeline nodes, are identified. The pipeline nodes are treated as graph nodes, and the pipeline segments and their connectivity are treated as graph edges. A source network topology semantic constraint graph and a target network topology semantic constraint graph are constructed. Semantic attributes are recorded on each graph node and each graph edge. The semantic attributes include pipeline type, medium type, pipe diameter, material, and facility number. From the source network topology semantic constraint diagram and the target network topology semantic constraint diagram, pipeline nodes with unique facility numbers and connectivity greater than a preset connectivity threshold in the topology structure are selected as semantically unique registration constraint features. The positional and connectivity relationships between these semantically unique registration constraint features are used as coarse registration constraints.

[0010] Furthermore, when establishing a registration error model based on the coarse registration results, it includes: For the matched pipeline nodes and pipeline segments in the source network topology semantic constraint diagram and the target network topology semantic constraint diagram, calculate the position deviation and burial depth deviation respectively, and use the position deviation and burial depth deviation as geometric residuals; After coarse registration, determine whether the connectivity between the corresponding pipeline nodes and the corresponding pipeline segments maintains the original connectivity, loop structure, and branch structure, and use the degree of connectivity disruption as an indicator of topology consistency. The semantic attributes of corresponding pipeline nodes and corresponding pipeline segments are compared, and the inconsistency of the semantic attributes is used as an indicator of semantic consistency. The geometric residual, topological consistency and semantic consistency are comprehensively evaluated according to a preset comprehensive weight to obtain the registration error model.

[0011] Furthermore, by iteratively optimizing the parameters of each coordinate transformation model and jointly solving them, local registration optimization is performed on the local spatial regions where the registration error exceeds the limit to obtain a globally consistent registration result, including: Based on the registration error model, the translation and rotation parameters of all pipeline data sources are globally iteratively optimized to reduce the overall geometric residual, topological consistency and semantic consistency error to below the first preset error threshold. Based on the comprehensive error of each pipeline node and each pipeline segment in the registration error model, a local spatial region where the registration error exceeds the limit is defined. Within the local spatial region, the local weak deformation parameters in the corresponding coordinate transformation model are adjusted. While maintaining the continuity of the connection between pipeline nodes and pipeline segments and ensuring that the burial depth and slope meet the engineering constraints, the adjustment range of the local weak deformation parameters is limited so that the comprehensive error within the local spatial region is reduced to below the second preset error threshold, thereby obtaining the globally consistent registration result.

[0012] Furthermore, when calculating the registration quality evaluation value for each pipeline node and each pipeline segment based on the registration results, the following is included: The geometric quality evaluation value is obtained based on the geometric residuals of the corresponding pipeline nodes and corresponding pipeline segments in the registration error model; the topological quality evaluation value is obtained based on the topological consistency of the corresponding pipeline nodes and corresponding pipeline segments in the registration error model; and the semantic quality evaluation value is obtained based on the semantic consistency of the corresponding pipeline nodes and corresponding pipeline segments in the registration error model. The geometric quality evaluation value, the topology quality evaluation value, and the semantic quality evaluation value are weighted according to the accuracy level of each pipeline data source to obtain the registration quality evaluation value of the corresponding pipeline node and the corresponding pipeline segment.

[0013] Furthermore, clustering pipeline nodes and pipeline segments from different pipeline data sources that correspond in spatial location, topology, and semantic attributes into a unified pipeline object, and assigning a unified identifier to the unified pipeline object, includes: For each pipeline node and pipeline segment in the pipeline data source, a set of candidate corresponding objects is determined based on the distance relationship between its spatial location and the spatial location of pipeline nodes and pipeline segments in other pipeline data sources, the upstream and downstream connectivity relationship and the similarity of the loop structure, as well as the similarity of the semantic attributes. Then, each candidate corresponding object in the set of candidate corresponding objects is matched according to the registration quality evaluation value of the candidate corresponding objects. In each set of candidate corresponding objects, the candidate corresponding object with the highest registration quality evaluation value and the matching evaluation result greater than the preset matching evaluation threshold is selected as the target corresponding object. Pipeline nodes and pipeline segments belonging to the same target corresponding object are clustered into the same unified pipeline object. A unique unified identifier is assigned to each unified pipeline object. For pipeline nodes and pipeline segments that do not meet the matching evaluation threshold, they are marked as objects to be confirmed and are not included in any unified pipeline object.

[0014] Furthermore, for each of the aforementioned unified pipeline objects, when constructing a quality-driven attribute fusion model based on the registration quality evaluation value, data source accuracy, and business reliability, the following is included: For the unified pipeline object, corresponding geometric attributes and classification attributes are collected from each pipeline data source. The registration quality evaluation value is used as a quality indicator reflecting registration reliability, the data source accuracy is used as a quality indicator reflecting measurement accuracy, and the business reliability is used as a quality indicator reflecting business reliability. Based on all the quality indicators, geometric attribute fusion weights and classification attribute fusion weights are determined for each pipeline data source. Pipeline data sources with registration quality evaluation values ​​lower than a preset quality threshold are used as alternative references and are not included in the determination of the target attribute values ​​of the unified pipeline object.

[0015] Furthermore, within the engineering constraints, geometric attributes are corrected, and classification attributes are weighted. The final fused annotation result includes: Based on the geometric attributes of each pipeline data source and the corresponding geometric attribute fusion weights, combined with engineering constraints, the spatial location, burial depth, and orientation of the unified pipeline object are corrected. The engineering constraints include design slope, design burial depth range, and minimum cover thickness. When the correction result exceeds the engineering constraints, the alternative geometric attributes that satisfy the engineering constraints are used as the unified geometric attributes. Based on the classification attributes of each pipeline data source and the corresponding classification attribute fusion weight, the classification attributes are weighted and voted on. The weighted voting result is used as the target classification attribute of the unified pipeline object. For classification attributes that still have multiple candidate values ​​after weighted voting, they are marked as attributes to be confirmed and the pipeline data source information participating in the voting is recorded.

[0016] Compared with existing technologies, the advantages of this invention are as follows: By constructing a coordinate transformation model that simultaneously includes translation parameters, rotation parameters, and local weak deformation parameters, and on this basis introducing a registration error model composed of source network topology semantic constraint diagrams and target network topology semantic constraint diagrams, combined with geometric residuals, topological consistency, and semantic consistency, as well as iterative optimization and local registration optimization mechanisms, it can eliminate the overall offset and local distortion caused by different coordinate systems, multiple measurements, and historical data in the global scope. Furthermore, it can achieve fine-grained multi-source pipeline data while maintaining the correctness of pipeline connectivity, loop structure, and facility semantic attributes. Precise registration: By calculating the registration quality evaluation value for each pipeline node and pipeline segment from the registration results, multi-source data are clustered into unified pipeline objects with unified identifiers. A quality-driven attribute fusion model is constructed using the registration quality evaluation value, data source accuracy, and business credibility. Within the engineering constraints, geometric attributes are corrected in a controlled manner, and classification attributes are weighted and fused. This ensures that the final fused annotation results have high consistency, high credibility, and traceability in spatial location, topology, and semantic attributes, thereby reducing the risks of location mismatch, connectivity errors, and attribute conflicts in traditional solutions. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a flowchart of a method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems provided in an embodiment of the present invention. Detailed Implementation

[0018] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] In the process of information-based construction of urban underground infrastructure, the heterogeneity of coordinate systems, elevation benchmarks, and data formats of multi-source pipeline data makes it difficult to accurately align the geometric positions of pipeline nodes and segments when registering under a unified spatial benchmark. This leads to inconsistencies in pipeline network connectivity and semantic attributes, resulting in technical problems such as position mismatches, broken connectivity, and attribute conflicts. Specifically, inaccurate geometric positions manifest as systematic deviations in the spatial coordinates of the same pipeline object from different data sources; connectivity errors result in non-physical breaks or redundant connections in the pipeline network topology after fusion; and semantic attribute conflicts arise from the inconsistency of key business information such as pipeline type and facility number among multi-source data, thus affecting the reliability of 3D visualization management of the pipeline network, digital twin construction, and intelligent operation and maintenance decisions.

[0020] For example, in a digital integrated utility tunnel construction project in a new urban area, the design drawings used a local independent coordinate system, the geophysical survey results used the WGS84 coordinate system, and the BIM model used a custom engineering coordinate system. Furthermore, the elevation benchmarks and data formats of the various data sources were incompatible. When attempting to integrate this data into a unified platform, the planar position of the same manhole deviated in different data sources, the burial depth information was inconsistent, and connectivity issues manifested as valves being incorrectly associated with non-adjacent pipeline segments. Simultaneously, pipe diameter attributes contradicted those in the design drawings and geophysical survey results. Further, local measurement distortions (such as pipeline bending caused by construction errors) were not effectively modeled, resulting in geometric residuals in some areas exceeding engineering tolerances after coarse registration. Topological consistency was reduced due to disrupted connectivity, and semantic consistency was compromised by facility numbering conflicts. Ultimately, this led to the 3D visualization system being unable to accurately recreate the actual pipeline network structure, and the operation and maintenance records becoming disconnected from the spatial model.

[0021] If the aforementioned problems are not resolved, pipeline data in the integrated management system will continue to suffer from defects such as inconsistent spatial benchmarks, broken topological logic, and contradictory semantic information. This will cause distortion in the mapping between the digital twin model and the physical pipeline network. Decision instructions generated by the intelligent operation and maintenance system based on erroneous data may deviate from actual operating conditions, thus posing a potential risk to the safe operation of urban infrastructure. Specifically, location mismatch will lead to construction positioning errors, incorrect connectivity will cause failures in pipeline hydraulic simulation, and attribute conflicts will make it impossible to trace the actual pipeline information in the operation and maintenance ledger, ultimately hindering the refined management and emergency response capabilities of the urban underground pipeline network.

[0022] For this purpose, please refer to Figure 1 As shown, this application proposes a method for registration and fusion annotation of multi-source pipeline data for heterogeneous coordinate systems, including: S100: Acquire multi-source pipeline data and source coordinate system description information, and establish a coordinate transformation model for each pipeline data source. The coordinate transformation model includes translation parameters, rotation parameters, and local weak deformation parameters.

[0023] S200: Based on multi-source pipeline data, pipeline nodes, pipeline segments, connectivity relationships, and semantic attributes are extracted to construct source pipeline network topology semantic constraint graphs and target pipeline network topology semantic constraint graphs. Semantically unique registration constraint features are selected for coarse registration. Based on the coarse registration results, a registration error model is established. The parameters of each coordinate transformation model are solved by iterative optimization. Local registration optimization is performed on local spatial regions where the registration error exceeds the limit to obtain globally consistent registration results. The registration error model includes geometric residuals, topological consistency, and semantic consistency.

[0024] S300: Based on the registration results, calculate the registration quality evaluation value for each pipeline node and each pipeline segment, and cluster pipeline nodes and pipeline segments from different pipeline data sources that correspond in spatial location, topology and semantic attributes into a unified pipeline object, and assign a unified identifier to the unified pipeline object.

[0025] S400: For each unified pipeline object, a quality-driven attribute fusion model is constructed based on the registration quality evaluation value, data source accuracy, and business credibility. Within the engineering constraints, geometric attributes are corrected, and classification attributes are weighted to finally obtain the fusion annotation result.

[0026] This embodiment relates to a method for registration and fusion annotation of multi-source pipeline data in heterogeneous coordinate systems. The coordinate transformation model refers to a spatial transformation mechanism established for different pipeline data sources. This can be achieved using a least-squares fitting method based on common control points or by initializing parameters through user-defined reference points. For example, it can calculate translation vectors using pipeline nodes with known coordinates as control points, or adjust the rotation matrix based on differences in coordinate axis directions. Its main purpose is to achieve geometric position transformation of pipeline data under a unified spatial reference. Furthermore, the local weak deformation parameter refers to an adjustable parameter used to compensate for local spatial distortion. This can be achieved using radial basis function interpolation or a polynomial deformation field model. For example, it can dynamically generate a local deformation mesh based on the pipeline node distribution density, or preset deformation coefficients based on terrain undulation data. Its main purpose is to achieve adaptive correction of local spatial deviations caused by construction errors or measurement anomalies. Specifically, a topological semantic constraint graph refers to a graph data structure that represents the logical structure and business attributes of pipelines. It can be implemented using automatic graph generation algorithms based on pipeline segment connection relationships or by constructing node associations based on geometric proximity. For example, pipeline intersections can be used as graph nodes with connecting pipelines as edges, or topological edges can be generated based on burial depth continuity. Its main purpose is to achieve a structured expression of pipeline network connectivity and semantic attributes. As a preferred implementation, a registration error model refers to a quantitative index system for comprehensively evaluating registration quality. It can be implemented using a weighted combination of geometric distance deviation and topological connectivity differences or statistical analysis of semantic attribute matching degrees. For example, the Euclidean distance between pipeline nodes can be calculated as the geometric residual, or the consistency of loop structures can be verified through graph isomorphism algorithms. Its main purpose is to achieve multi-dimensional error measurement of registration results in geometric, topological, and semantic dimensions. The registration quality evaluation value refers to a quantitative indicator reflecting the registration reliability of pipeline objects. It can be achieved using a normalized score output by an error model or a dynamic weighting method based on the historical accuracy of the data source. For example, a quality heatmap can be generated based on the geometric residual distribution, or the evaluation value weight can be adjusted by combining measurement equipment calibration records. Its main purpose is to achieve an objective assessment of the contribution of different pipeline data sources. Therefore, the quality-driven attribute fusion model refers to a decision-making mechanism for attribute integration based on the reliability of multi-source data. It can be implemented using median filtering correction for geometric attributes or majority voting strategies for categorical attributes. For example, a smoothing algorithm can be applied to pipeline routes within engineering constraints, or a weighted average can be applied to pipe diameter attributes. Its main purpose is to achieve optimized fusion of unified pipeline objects in terms of geometric accuracy and semantic consistency. In practical applications, the clustering process for unified pipeline objects can be based on a joint determination of spatial location similarity and topological matching degree. For example, the candidate object set can be determined by calculating the spatial overlap rate of pipeline segments, or the target corresponding object can be screened based on the similarity of upstream and downstream connectivity. Its main purpose is to achieve accurate alignment of multi-source data in geometric, topological, and semantic dimensions.This application combines coordinate transformation models with topological semantic constraints and introduces a multi-dimensional registration error evaluation mechanism, avoiding the topological breaks and semantic gaps caused by existing technologies that rely solely on geometric transformations. This achieves global consistency assurance of pipeline node locations, pipeline connectivity, and business attributes under a unified spatial coordinate system, ultimately solving the problems of location mismatch, connectivity errors, and attribute conflicts in multi-source pipeline data fusion.

[0027] In the process of information-based construction of urban underground infrastructure, multi-source pipeline data often suffers from spatial inconsistencies due to differences in coordinate system type, elevation datum, and data format. This leads to problems such as geometric position deviations of pipeline nodes, breaks in pipeline network connectivity, and semantic attribute conflicts. To address these issues, a method for registration and fusion annotation of multi-source pipeline data oriented towards heterogeneous coordinate systems has been implemented. This method first acquires multi-source pipeline data and their source coordinate system description information from design drawings, as-built drawings, geophysical exploration results, etc., including coordinate system type, coordinate unit, coordinate axis direction, and elevation datum. Based on this information, a coordinate transformation model is established for each pipeline data source. This model includes translation parameters, rotation parameters, and local weak deformation parameters. The local weak deformation parameters are used to characterize local spatial deformation caused by construction errors or terrain influences. As a preferred implementation, the local weak deformation parameters can specifically be elastic deformation coefficients for pipeline segments, with their numerical range dynamically set according to the engineering control network division results, thereby ensuring that the coordinate transformation process can adapt to local measurement distortions.

[0028] Furthermore, pipeline nodes, pipeline segments, connectivity relationships, and semantic attributes are extracted based on multi-source pipeline data. The semantic attributes include pipeline type, medium type, pipe diameter, material, and facility number. This leads to the construction of a source network topology semantic constraint graph and a target network topology semantic constraint graph. Pipeline nodes are used as graph nodes, pipeline segments and their connectivity relationships are used as graph edges, and semantic attributes are recorded on the graph nodes and edges. In this application, the multi-source pipeline data includes at least one data source serving as a reference network and at least one data source serving as a network to be registered. Specifically: the pipeline nodes and pipeline segments of the pipeline data source serving as the network to be registered are constructed as the source network topology semantic constraint graph in a unified spatial coordinate system; the pipeline nodes and pipeline segments of the pipeline data source serving as the reference network or the network data that has been registered and fused in the aforementioned steps are constructed as the target network topology semantic constraint graph in a unified spatial coordinate system. In specific implementations, a pipeline data source with high measurement accuracy and complete attribute information can be selected as the target network, and the remaining pipeline data sources are sequentially used as source networks for registration with the target network. As an optional implementation, the target pipeline topology semantic constraint map is constructed from an authoritative pipeline database already established in the urban pipeline information system, while the source pipeline topology semantic constraint map is constructed from newly acquired as-built drawings, geophysical survey results, and other pipeline data sources. Pipeline nodes with unique facility numbers and connectivity greater than a preset threshold are selected as semantically unique registration constraint features, and coarse registration is performed based on positional and connectivity relationships. A registration error model is established based on the coarse registration results. This model includes three types of indicators: geometric residual, topological consistency, and semantic consistency. Geometric residual is obtained by calculating the positional and burial depth deviations of matched pipeline nodes. Topological consistency is evaluated based on the degree of preservation of connectivity, loop structure, and branch structure. Semantic consistency is determined by comparing the semantic attribute differences between corresponding pipeline nodes and pipeline segments. As a specific embodiment, the geometric residual can be specifically calculated as a coordinate deviation of 0.15 meters and a burial depth deviation of 0.08 meters for a valve node in the east-west direction; this value is used as input to the error model for subsequent optimization.

[0029] By iteratively optimizing the parameters of the joint solution of each coordinate transformation model, the translation and rotation parameters of all pipeline data sources are first globally adjusted to reduce the combined error of geometric residuals, topological consistency, and semantic consistency to below a preset threshold. For local spatial regions where the registration error exceeds the limit, the local weak deformation parameters in the corresponding coordinate transformation model are adjusted. At the same time, the parameter adjustment range is limited under engineering constraints. For example, during pipeline slope correction, the change in burial depth is ensured to not exceed 0.5% of the allowable range of the design slope, thereby obtaining globally consistent registration results. Based on the registration results, a registration quality evaluation value is calculated for each pipeline node and pipeline segment. This evaluation value is generated by weighting the geometric quality evaluation value, topological quality evaluation value, and semantic quality evaluation value, where the weighting coefficient is determined according to the accuracy level of the data source. As one implementation method, during the clustering process, pipeline nodes from different data sources can be clustered into a unified pipeline object based on the conditions that the spatial distance is less than 0.3 meters, the upstream and downstream connectivity matching degree is higher than 80%, and the semantic attribute similarity is greater than 90%, and a unique unified identifier is assigned.

[0030] For each unified pipeline object, a quality-driven attribute fusion model is constructed based on the registration quality evaluation value, data source accuracy, and business reliability. The fusion weights for geometric and classification attributes are determined comprehensively by the aforementioned quality indicators. When correcting geometric attributes within engineering constraints, spatial positions are adjusted in conjunction with constraints such as design slope, design burial depth range, and minimum cover thickness. When the correction result exceeds the constraints, alternative geometric attributes are used. When weighting classification attributes, the target classification attribute is determined through weighted voting. For example, for the pipe diameter attribute, voting is conducted based on the fusion weights of each data source, and the candidate value with the highest weight is used as the unified result. Thus, this method eliminates geometric position mismatches caused by local measurement distortions through the design of local weak deformation parameters in the coordinate transformation model. The comprehensive application of the topological semantic constraint graph and the registration error model ensures the continuity of pipeline connectivity and the consistency of semantic attributes. The implementation of the quality-driven attribute fusion mechanism under engineering constraints avoids attribute conflicts and generates unified and reliable fusion annotation results, ultimately achieving fine registration and fusion annotation of multi-source pipeline data in heterogeneous coordinate systems.

[0031] Specifically, in some of the embodiments described above in this application, it is proposed to obtain multi-source pipeline data and source coordinate system description information to establish a coordinate transformation model. However, in its implementation process, the specific association mechanism between the source coordinate system description information and the known control points in the engineering control network and the pipeline nodes with unique semantic identifiers is not clearly defined. This leads to a lack of accuracy in the initial calculation of translation and rotation parameters, which in turn affects the reliability of the coordinate transformation model and makes it easy for geometric position deviations to occur in the initial registration of multi-source pipeline data.

[0032] In this regard, this application further proposes that when acquiring multi-source pipeline data and source coordinate system description information, the following methods should be included: The source coordinate system description information includes the coordinate system type, coordinate unit, coordinate axis direction, and elevation datum of each pipeline data source. Under a unified spatial coordinate system, the source coordinate system description information is associated with known control points in the engineering control network and pipeline nodes with unique semantic identifiers. The translation and rotation parameters of each pipeline data source are obtained based on the correspondence obtained from the association, and an initial coordinate transformation model is generated.

[0033] The source coordinate system description information refers to the metadata set used to fully define the coordinate system attributes of the pipeline data source. This includes elements such as coordinate system type, coordinate units, coordinate axis directions, and elevation datum. Clarifying these basic attributes avoids ambiguity during coordinate system transformation. Associating the source coordinate system description information with known control points in the engineering control network and pipeline nodes with unique semantic identifiers involves establishing a mapping relationship between the source coordinate system description information and external spatial datums and semantic entities. This can be achieved using least-squares fitting or iterative nearest-point algorithms to match corresponding point sets, aiming to leverage the geometric accuracy of the engineering control network and the topological reliability of semantically identified nodes for dual verification. Calculating translation and rotation parameters based on the associated correspondence involves deriving spatial displacement and rotation relationships based on reliable corresponding point sets. This can be done using rigid body transformation models or affine transformation models for mathematical calculations, ensuring that the parameters objectively reflect the true spatial transformation characteristics. Generating the initial coordinate transformation model involves integrating the calculated parameters into an executable coordinate transformation framework.

[0034] Specifically, the proposed solution establishes a dual verification mechanism based on geometric accuracy and topological semantics by associating the source coordinate system description information with known control points and pipeline nodes with unique semantic identifiers in the engineering control network under a unified spatial coordinate system. This mechanism utilizes the high-precision known control points of the engineering control network as external benchmarks while ensuring the reliability of the topological structure through semantically identified nodes (such as well chambers or valves). Subsequently, based on the reliable corresponding point set formed by this association, translation and rotation parameters are calculated mathematically to generate an initial coordinate transformation model. This eliminates the reliance on empirical estimations in the initial registration process and reduces the accumulation of errors caused by initial geometric position deviations.

[0035] As a preferred embodiment, the specific implementation of this application is as follows: In a city gas pipeline network data fusion project, the design drawing data source adopts a local independent coordinate system, while the geophysical exploration data source adopts the WGS84 coordinate system. First, the source coordinate system description information of both is obtained, including coordinate system type, coordinate unit, coordinate axis direction, and elevation datum. Then, under a unified spatial coordinate system (such as CGCS2000), the coordinates of the pipeline node marked "Valve A-001" in the design drawing are matched with known control points in the engineering control network, and corresponding nodes with the same facility number in the geophysical exploration results are associated. Based on these matched point pairs, the translation and rotation parameters of the design drawing data source relative to the unified coordinate system are obtained using the least squares method, generating an initial coordinate transformation model.

[0036] Through the above solution, this application solves the problem of inaccurate initial coordinate transformation model parameter acquisition, ensures the reliability of the registration starting point of multi-source pipeline data, avoids the accumulation of global registration error caused by initial geometric position deviation, and provides technical support for the accurate registration of multi-source pipeline data under a unified spatial reference.

[0037] In the information-based construction of urban underground infrastructure, multi-source pipeline data often suffers from insufficient registration accuracy due to coordinate system differences and measurement limitations. Traditional existing multi-source pipeline data registration methods typically only use translation and rotation parameters for overall correction when acquiring the coordinate transformation model. However, during the implementation process, the initial coordinate transformation model cannot correct local measurement distortions, leading to excessive registration errors in densely populated pipeline areas or unevenly divided engineering control networks, affecting the accurate fusion of pipeline network topology and semantic attributes.

[0038] In this regard, this application further proposes that when obtaining local weak deformation parameters, the following methods should be included: Based on the spatial distribution of pipeline nodes and pipeline segments and the division results of the engineering control network, each pipeline data source is divided into multiple local spatial regions or multiple pipeline segments along the pipeline route. Local weak deformation parameters are preset for each local spatial region or each pipeline segment to characterize local weak deformation.

[0039] In practical applications, the spatial distribution of pipeline nodes and segments refers to the geometric arrangement characteristics of nodes and segments in pipeline data. It can be used to identify key deformation areas in the pipeline layout using spatial clustering algorithms or density analysis methods. For example, dense node areas or curved sections can be located by calculating the distribution of node spacing or changes in segment curvature. The aim is to accurately capture areas prone to local distortion due to limitations of measurement equipment or environmental interference. The division result of the engineering control network can be understood as a spatial partitioning benchmark for a high-precision control point network. It can be achieved using grid partitioning based on control point density or adaptive region segmentation techniques. For example, the engineering control network can be divided into triangular or quadrilateral sub-regions based on the uniformity of control point coverage. The purpose is to provide a reliable reference framework for local region partitioning, avoiding the defect that the overall transformation parameters cannot adapt to local irregular deformations. Specifically, a local spatial region refers to a sub-region unit defined based on spatial location. It can be implemented using regular grids or irregular polygonal regions. For example, urban blocks can be divided into independent processing units according to geographical boundaries. The purpose is to optimize for wide-area nonlinear deformations, such as handling ground settlement caused by construction disturbances. Furthermore, pipeline segments can be understood as linear sub-units defined along the pipeline route. These segments can be implemented using fixed-length segmentation or dynamic segmentation techniques based on node locations. For example, the pipeline can be divided into continuous segments using valves or manholes as boundaries. The purpose is to optimize linear deformation, such as correcting errors caused by pipeline expansion or bending. Local weak deformation parameters refer to mathematical parameters used to quantify local geometric distortions. These can be implemented using affine transformation parameters or thin-plate spline interpolation coefficients, such as setting translational offsets or rotational fine-tuning amounts. The aim is to independently adjust the geometry of local regions while maintaining the continuity of topological connectivity.

[0040] Specifically, the proposed solution dynamically identifies key areas requiring local correction by combining the spatial distribution characteristics of pipeline nodes and segments with the division results of the engineering control network. First, based on spatial distribution analysis, dense pipeline areas or points of abrupt geometric changes are identified. Simultaneously, the high-precision zoning results of the engineering control network are referenced to ensure alignment between the divided areas and a reliable spatial framework. Second, a spatial region division strategy or a pipeline segment division strategy is selected based on the region's characteristics. The former is suitable for handling planar nonlinear deformation, while the latter is suitable for handling linear extension deformation. Finally, independent local weak deformation parameters are preset for each division unit, enabling the registration optimization process to selectively adjust local geometry. This maintains the continuity of the connectivity between pipeline nodes and segments while limiting the parameter adjustment range to meet engineering constraints, thereby correcting local measurement errors in the global iterative optimization process.

[0041] As a specific implementation method, the scheme of this application is implemented as follows: In the data processing of a city's water supply network, for commercial road sections with dense valves, dense areas with node spacing less than a threshold are first identified based on the spatial distribution of pipeline nodes. Combined with the sparse control point division results in the engineering control network, the road section is divided into multiple local spatial regions. Simultaneously, the main water transmission pipeline is divided into multiple pipeline segments along the pipeline route, with valves as boundaries. Two-dimensional affine transformation parameters are preset for each local spatial region as local weak deformation parameters, allowing for fine-tuning of translation and rotation during registration optimization. One-dimensional scaling parameters are preset for each pipeline segment to correct for length changes caused by thermal expansion and contraction of the pipeline. During the registration process, these parameters are dynamically adjusted according to the registration error model to ensure that the geometric residuals within the local areas are reduced to within the allowable range for engineering.

[0042] Through the above technical solution, this application corrects local measurement distortion caused by limitations of measurement equipment or environmental interference in densely populated pipeline areas or local spatial areas with uneven division of engineering control networks by dynamically dividing the area and preset local weak deformation parameters, thus avoiding the problem of registration error exceeding the limit. This ensures the accurate integration of pipeline network topology and semantic attributes, and provides a reliable geometric basis for unified spatial reference registration of multi-source pipeline data.

[0043] Specifically, in some of the embodiments described above in this application, a method is proposed to construct a source network topology semantic constraint graph and a target network topology semantic constraint graph and select semantically unique registration constraint features for coarse registration. However, in its implementation, since the semantic attributes of pipeline nodes and pipeline segments are not recorded and utilized in a structured manner, and the selection of registration constraint features lacks unique facility identification and topology stability guarantee, coarse registration is easily affected by semantic ambiguity and inconsistent connectivity, resulting in fuzzy feature matching, mismatched positions, and accumulated registration errors.

[0044] To address this, this application further proposes constructing source network topology semantic constraint graphs and target network topology semantic constraint graphs. When selecting semantically unique registration constraint features for coarse registration, the process includes: identifying pipeline nodes representing manholes, valves, inlets / outlets, and equipment foundations in each pipeline data source, and pipeline segments connected to these nodes; treating pipeline nodes as graph nodes and pipeline segments and their connectivity as graph edges; constructing source and target network topology semantic constraint graphs; and recording semantic attributes on each graph node and edge, including pipeline type, medium type, pipe diameter, material, and facility number. From the source and target network topology semantic constraint graphs, pipeline nodes with unique facility numbers and connectivity greater than a preset connectivity threshold in the topology are selected as semantically unique registration constraint features. The positional and connectivity relationships between these semantically unique registration constraint features are used as coarse registration constraints.

[0045] In practical applications, pipeline nodes representing well chambers, valves, inlets / outlets, and equipment foundations refer to key facility points in the pipeline network with clear business semantics. These nodes can be identified using predefined semantic tags or business identifiers based on pipeline data sources. This aims to avoid the risk of matching confusion caused by semantic overlap in ordinary pipeline nodes. Treating pipeline nodes as graph nodes and pipeline segments and their connectivity as graph edges involves constructing a structured data model, which can be implemented using adjacency lists or graph database technologies. Its purpose is to simultaneously capture the geometric location information and physical connection relationships of pipelines, thereby fully preserving the spatial layout and logical topological characteristics of the pipeline network. Semantic attributes refer to multi-dimensional information such as pipeline type, medium type, pipe diameter, material, and facility number. These can be stored in the attribute fields of graph nodes and graph edges. The purpose is to ensure semantic traceability through the uniqueness of facility numbers and reduce attribute conflicts. Pipeline nodes with unique facility numbers and connectivity greater than a preset connectivity threshold refer to nodes with globally unique identifiers and high connectivity in the pipeline network topology. This can be achieved by querying the facility number database and calculating node degree, aiming to select representative and highly interference-resistant feature points for registration. Using location and connectivity relationships as coarse registration constraints involves matching based on the relative spatial distribution and topological connection patterns between registration constraint features. This can be achieved using graph matching algorithms or spatial relationship reasoning, aiming to avoid connectivity breakage problems caused by relying solely on geometric location.

[0046] Specifically, the proposed solution first identifies key facility nodes and treats them as graph nodes, while simultaneously treating pipeline segments and their connectivity as graph edges, thus structurally storing the geometric and topological information of the pipeline network. Based on this, multi-dimensional semantic attributes are recorded on the graph nodes and edges to ensure the uniqueness of facility numbers as the basis for semantic traceability. Furthermore, based on the uniqueness of facility numbers and topological connectivity, highly reliable registration constraint features are selected to eliminate interference from duplicate identifiers and low-connectivity nodes. Finally, coarse registration is performed using the positional and connectivity relationships of these features, ensuring that the matching process not only relies on spatial coordinates but also incorporates the inherent topological logic of the pipeline network. This maintains the integrity of connectivity relationships despite differences in heterogeneous coordinate systems, providing a high-precision initial alignment basis for global registration.

[0047] As a preferred embodiment, the specific implementation of this application's solution is as follows: In gas pipeline network data registration, pipeline nodes marked as "valves" in the data source are identified. These nodes have unique facility numbers. Valve nodes are used as graph nodes, and the pipeline segments connecting them are used as graph edges to construct a source pipeline network topology semantic constraint graph. Attributes such as the medium type ("natural gas") and pipe diameter are recorded on the nodes. Valve nodes with unique facility numbers and connectivity greater than a preset connectivity threshold are selected as registration constraint features. Coarse registration is performed based on the relative positions and connectivity relationships of these features. For example, valve nodes with the same upstream and downstream connection relationships in two data sources are matched to ensure that the registration process conforms to the physical connectivity logic of the pipeline network.

[0048] By employing the above approach, this application avoids matching errors caused by semantic ambiguity and topological instability in the coarse registration stage, reduces problems such as feature matching ambiguity, position mismatch, and registration error accumulation, thereby improving the accuracy and reliability of multi-source pipeline data registration.

[0049] In some of the embodiments described above in this application, coarse registration based on semantically unique registration constraint features is proposed. However, in its implementation, there is a lack of a systematic error evaluation mechanism for the registration results. It is impossible to quantify multi-dimensional deviations such as geometric position deviation, topological structure destruction, and semantic attribute conflict. This makes it difficult to identify the registration error exceeding the limit in local spatial regions, thereby affecting the reliability of global optimization and fusion annotation. As a result, pipeline data still has problems such as position mismatch, broken connectivity, or attribute conflict under a unified spatial reference.

[0050] In response, this application further proposes that when establishing a registration error model based on the coarse registration results, the following should be included: Calculate the positional deviation and burial depth deviation for the matched pipeline nodes and pipeline segments in the source network topology semantic constraint diagram and the target network topology semantic constraint diagram, and use the positional deviation and burial depth deviation as geometric residuals.

[0051] After coarse registration, determine whether the connectivity between the corresponding pipeline nodes and the corresponding pipeline segments maintains the original connectivity, loop structure, and branch structure, and use the degree of connectivity disruption as an indicator of topology consistency.

[0052] The semantic attributes of corresponding pipeline nodes and corresponding pipeline segments are compared. Inconsistencies in semantic attributes are used as an indicator of semantic consistency. Geometric residuals, topological consistency, and semantic consistency are comprehensively evaluated according to preset comprehensive weights to obtain a registration error model.

[0053] Among them, geometric residual refers to the core indicator that reflects the geometric accuracy of pipeline data by quantifying spatial positional differences. It can be achieved by calculating horizontal positional deviation using Euclidean distance and vertical burial depth deviation using elevation difference. The purpose is to accurately capture the degree of misalignment between pipeline nodes and pipeline segments in three-dimensional space, avoiding the omission of local distortions caused by single-dimensional evaluation. Topological consistency indicators can be understood as quantitative parameters characterizing the integrity of the pipeline network's logical structure. It can be achieved by analyzing the maintenance status of node connectivity, loop closure, and branch connection relationships. The purpose is to identify the risk of pipeline network functional failure caused by registration. For example, the destruction of connectivity may interrupt the fluid transmission path, the destruction of loop structure may affect redundancy, and the destruction of branch structure may cause node connection errors. Semantic consistency indicators are technical parameters that measure the degree of matching of business attributes. They can be achieved by comparing semantic attribute fields item by item and calculating the degree of difference. The purpose is to detect conflicts in key business information such as pipeline type and medium type, and prevent business logic confusion caused by semantic errors. The registration error model refers to a comprehensive evaluation system that integrates multi-dimensional error indicators. It can be achieved by weighted fusion of geometric residuals, topological consistency and semantic consistency based on preset weights. The purpose is to provide a unified quantitative standard for registration quality and support the identification and optimization of error-prone areas.

[0054] Specifically, the proposed solution forms a multi-dimensional error assessment closed loop by simultaneously performing geometric residual calculation, topological consistency judgment, and semantic consistency comparison based on the coarse registration results. First, the spatial coordinates and burial depth data of the matched pipeline nodes and segments are extracted, and the positional and burial depth deviations are calculated as geometric residuals. This process ensures that the geometric error quantification covers both horizontal and vertical directions. Second, based on the graph structure characteristics of the source and target pipeline topology semantic constraint graphs, the connectivity, loop structure, and branch structure of the corresponding pipeline nodes and segments are verified to ensure consistency with the original topology, transforming structural damage into topological consistency indicators. Simultaneously, the matching status of semantic attribute fields is compared item by item, quantifying inconsistencies into semantic consistency indicators. Finally, the three types of indicators are integrated into a registration error model based on preset comprehensive weights. This model dynamically adjusts the weight ratios to adapt to different engineering scenarios; for example, it emphasizes geometric residual weights in high-precision areas and strengthens connectivity indicator weights in topology-sensitive scenarios, thereby providing a quantifiable comprehensive evaluation basis for global registration quality and supporting accurate identification of registration error exceeding limits during iterative optimization.

[0055] As a specific implementation method, the scheme of this application is implemented as follows: In the process of registering urban gas pipeline network data, for the valve nodes and connected pipeline segments matched after coarse registration, firstly, the planar position deviation and burial depth deviation in a unified coordinate system are calculated as geometric residuals. Secondly, it is checked whether the connectivity between the valve nodes and the upstream and downstream pipeline segments maintains the original topology, and whether the loop network is closed and the branch pipeline connection is complete. The discovered breakpoints or redundant connections are used as topology consistency indicators. At the same time, the semantic attributes such as gas pipeline material, pipe diameter, and pressure rating are compared, and the number of conflicts in the attribute fields is recorded as semantic consistency indicators. Finally, according to engineering requirements, the geometric residual weight is set to 0.5, the topology consistency weight to 0.3, and the semantic consistency weight to 0.2. The registration error model value is generated by weighted calculation. When the value exceeds the preset threshold, the corresponding local spatial area is automatically marked as the error excess area.

[0056] Through the above technical solutions, this application achieves a refined quantitative evaluation of coarse registration results, which can systematically identify multi-dimensional errors such as geometric position deviation, topological structure destruction and semantic attribute conflict, providing a reliable basis for global iterative optimization and local registration optimization, avoiding position mismatch, connectivity breakage or attribute conflict problems in pipeline data under a unified spatial reference, and ensuring that the fused annotation results meet the actual operation requirements of the project.

[0057] Specifically, in some of the embodiments described above in this application, a global iterative optimization based on the registration error model is proposed to reduce the overall error. However, in the process of its implementation, the registration error in the local spatial region may exceed the limit, resulting in geometric position deviations, disruption of connectivity, or inconsistent semantic attributes of pipeline nodes and pipeline segments, affecting the accuracy and reliability of the global registration results.

[0058] To address this, this application further proposes a method that iteratively optimizes the parameters of each coordinate transformation model to perform local registration optimization on local spatial regions where registration errors exceed limits, thereby obtaining globally consistent registration results. This includes: Based on the registration error model, the translation and rotation parameters of all pipeline data sources are globally iteratively optimized to reduce the overall geometric residual, topological consistency and semantic consistency error to below the first preset error threshold.

[0059] Based on the comprehensive error of each pipeline node and each pipeline segment in the registration error model, local spatial regions where the registration error exceeds the limit are divided. Within the local spatial regions, the local weak deformation parameters in the corresponding coordinate transformation model are adjusted. While maintaining the continuity of the connection between pipeline nodes and pipeline segments and ensuring that the burial depth and slope meet the engineering constraints, the adjustment range of the local weak deformation parameters is limited so that the comprehensive error within the local spatial regions is reduced to below the second preset error threshold, thereby obtaining a globally consistent registration result.

[0060] Global iterative optimization refers to the process of systematically adjusting the translation and rotation parameters of all pipeline data sources to minimize the overall error. This can be achieved using numerical optimization methods such as least squares, gradient descent, or conjugate gradient methods. The aim is to provide a globally consistent registration basis and avoid deviations caused by optimizing a single geometric index. Local spatial regions where registration errors exceed limits refer to areas requiring local optimization identified based on the comprehensive error distribution. These can be determined using spatial clustering algorithms based on error thresholds or region partitioning methods based on pipeline topology. The aim is to accurately locate problem areas for concentrated optimization resources. Local weak deformation parameters are parameters used to correct local measurement distortions. They can be characterized using polynomial deformation models, thin-plate spline interpolation, or radial basis functions. The aim is to address local nonlinear deformation problems in heterogeneous data sources. Engineering constraints refer to the physical and regulatory requirements that must be met in pipeline design, including design aspect, design burial depth range, and minimum cover thickness. The aim is to ensure that the optimization results conform to actual engineering conditions and prevent physical inconsistencies caused by over-adjustment.

[0061] Specifically, the proposed solution first adjusts translation and rotation parameters through global iterative optimization to reduce the combined errors in overall geometric residuals, topological consistency, and semantic consistency, providing a reliable foundation for local optimization. Subsequently, based on the combined error distribution of each pipeline node and segment in the registration error model, local spatial regions where registration errors exceed limits are identified and delineated, ensuring that optimization resources are concentrated in key areas. Within these local spatial regions, local weak deformation parameters in the corresponding coordinate transformation model are adjusted selectively, while strictly limiting the adjustment range to maintain the continuity of connectivity between pipeline nodes and segments, and to ensure that burial depth and slope meet engineering constraints, avoiding connectivity breaks or design code violations. Finally, through local optimization, the combined error within the region is reduced to below a second preset error threshold, achieving coordinated control of local and global errors, thereby obtaining globally consistent registration results. This phased optimization mechanism, through the orderly connection between global and local errors, solves the problem of global inconsistency caused by excessive local errors.

[0062] As a specific implementation method, the scheme of this application is implemented as follows: In the global iterative optimization stage, the Levenberg-Marquardt algorithm is used to iteratively solve the translation and rotation parameters of all pipeline data sources until the comprehensive error of geometric residuals, topological consistency, and semantic consistency is lower than a first preset error threshold. Then, based on the comprehensive error value of the pipeline nodes, the DBSCAN clustering algorithm is used to divide the nodes with errors exceeding the limit into multiple local spatial regions, each region covering pipeline nodes and pipeline segments with similar error characteristics. Within each local spatial region, a thin-plate spline deformation model is applied to adjust the local weak deformation parameters, and during the adjustment process, it is verified in real time whether the pipeline slope conforms to the design specifications and whether the burial depth is within a reasonable range, ensuring that the continuity of connectivity is not disrupted. In this way, the comprehensive error of the local region is controlled within a second preset error threshold, ultimately obtaining a globally consistent registration result.

[0063] Through the above-mentioned scheme, this application can solve the problem of excessive local spatial registration error, ensure the accurate geometric position, correct connectivity, and consistent semantic attributes of pipeline nodes and pipeline segments, thereby avoiding position mismatch and connectivity error, and improving the global consistency and reliability of multi-source pipeline data registration.

[0064] Specifically, in some of the embodiments described above in this application, a registration quality evaluation value is proposed to be calculated based on the registration results. However, in its implementation, the quality evaluation only relies on the geometric residual index, and does not fully integrate multi-dimensional constraints such as topological consistency and semantic consistency. Furthermore, it does not consider the impact of the accuracy differences of different pipeline data sources on the evaluation value, which results in the quality evaluation value failing to comprehensively and objectively reflect the registration reliability. Consequently, the clustering process of pipeline nodes and pipeline segments is easily interfered with by low-quality registration results, leading to errors in the unified pipeline object identifier allocation and attribute fusion deviations.

[0065] In response, this application further proposes steps for calculating registration quality evaluation values ​​for each pipeline node and each pipeline segment based on the registration results, including: The geometric quality evaluation value is obtained based on the geometric residuals of the corresponding pipeline nodes and corresponding pipeline segments in the registration error model.

[0066] The topology quality evaluation value is obtained based on the topology consistency between the corresponding pipeline nodes and the corresponding pipeline segments in the registration error model.

[0067] The semantic quality evaluation value is obtained based on the semantic consistency between the corresponding pipeline nodes and the corresponding pipeline segments in the registration error model.

[0068] Based on the accuracy level of each pipeline data source, the geometric quality evaluation value, topological quality evaluation value, and semantic quality evaluation value are weighted to obtain the registration quality evaluation value of the corresponding pipeline node and the corresponding pipeline segment.

[0069] Among them, the geometric quality evaluation value refers to an indicator that quantifies the accuracy of spatial location registration of pipeline nodes and segments based on geometric residuals. It can be achieved by weighted averaging of positional deviation and burial depth deviation or by standardized normalization, and its purpose is to directly reflect the accuracy of spatial location registration. The topological quality evaluation value refers to an indicator that assesses the integrity of the logical structure of the pipeline network based on topological consistency. It can be achieved by checking the degree of preservation of connectivity, loop structure, and branch structure, or by using connected component analysis in graph theory, and its purpose is to identify anomalies in the topological logic of the registered pipeline network. The semantic quality evaluation value refers to an indicator that captures pipeline attribute conflicts based on semantic consistency. It can be achieved by comparing the degree of difference in semantic attributes such as pipeline type, material, and facility number, or by using semantic similarity algorithms for quantitative evaluation, and its purpose is to accurately identify the risk of semantic attribute inconsistencies. Weighted evaluation values ​​refer to a mechanism for dynamically assigning weights to evaluation values ​​of different quality dimensions based on the accuracy level of the data source. It can be achieved by using the accuracy level as a weight coefficient for linear weighting or by nonlinear weighting based on fuzzy logic, and its purpose is to overcome evaluation bias caused by uneven data source accuracy and improve the objectivity and adaptability of the evaluation values.

[0070] Specifically, the proposed solution first extracts geometric residuals, topological consistency, and semantic consistency indices for corresponding pipeline nodes and segments from the registration error model. Then, these indices are converted into geometric quality evaluation values, topological quality evaluation values, and semantic quality evaluation values, respectively, to quantify registration quality across different dimensions. Finally, these quality evaluation values ​​are weighted and fused according to the accuracy level of each pipeline data source to obtain a comprehensive registration quality evaluation value. Through this multi-dimensional evaluation combined with accuracy weighting, the solution ensures that the quality evaluation value comprehensively and objectively reflects registration reliability, providing accurate quality basis for pipeline object clustering and attribute fusion.

[0071] As a specific implementation method, for a given pipeline node, the geometric quality evaluation value is quantified based on its positional deviation; the smaller the deviation, the higher the evaluation value. The topology quality evaluation value is determined based on whether the node maintains its original connectivity after registration; if the connectivity remains intact, the evaluation value is higher. The semantic quality evaluation value is assessed by comparing the consistency of semantic attributes such as pipeline type and material of the node across different data sources; consistent attributes result in a higher evaluation value. When weighting the evaluation values, high-precision data sources are assigned a larger weight, while low-precision data sources are assigned a smaller weight, thereby obtaining the comprehensive registration quality evaluation value for the pipeline node.

[0072] Through the above scheme, the registration quality evaluation value of this application can comprehensively integrate geometric, topological and semantic multi-dimensional constraints, and take into account the differences in data source accuracy, thereby objectively reflecting the registration reliability and reducing the identification allocation error and attribute fusion deviation in the clustering process of pipeline nodes and pipeline segments.

[0073] Specifically, in some of the embodiments described above in this application, a mechanism is proposed to cluster corresponding nodes and pipeline segments in multi-source pipeline data into unified objects and assign identifiers to achieve data fusion. However, in this process, simple matching based solely on spatial location, topology, and semantic attributes cannot filter high-quality correspondences, which may result in low-reliability matching in the clustering results, leading to problems such as incorrect unified identifiers or inconsistent fused data.

[0074] To address this, this application further proposes clustering pipeline nodes and segments from different pipeline data sources that correspond in spatial location, topology, and semantic attributes into a unified pipeline object. When assigning a unified identifier to the unified pipeline object, the process includes: for each pipeline node and segment in each pipeline data source, determining a set of candidate corresponding objects based on the distance relationship between their spatial location and the spatial location of pipeline nodes and segments in other pipeline data sources, the similarity of their upstream and downstream connectivity relationships and loop structures, and the similarity of their semantic attributes; and matching each candidate corresponding object in the candidate corresponding object set according to its registration quality evaluation value. In each candidate corresponding object set, the candidate corresponding object with the highest registration quality evaluation value and a matching evaluation result greater than a preset matching evaluation threshold is selected as the target corresponding object. Pipeline nodes and segments belonging to the same target corresponding object are clustered into the same unified pipeline object. A unique unified identifier is assigned to each unified pipeline object. Pipeline nodes and segments that do not meet the matching evaluation threshold are marked as objects to be confirmed and are not included in any unified pipeline object.

[0075] The candidate matching object set refers to the set of potential matching objects identified for a specific pipeline node or segment during the multi-source pipeline data fusion process. This set can be achieved using methods such as spatial distance threshold filtering, topological similarity calculation, or semantic attribute correlation analysis. The aim is to narrow the matching search range and ensure that candidate objects have reasonable correlations in three dimensions: geometric distribution, network topology, and business semantics. The matching evaluation result can be understood as a quantitative assessment indicator of the matching quality of candidate objects. It can be implemented using generalized methods such as geometric residual weighted scoring, topological consistency probability models, or semantic attribute similarity functions. The aim is to transform the registration quality evaluation value into an objective basis for matching reliability. The preset matching evaluation threshold specifically refers to the critical value used to filter low-reliability matches. It can be set as a dynamic threshold based on historical data statistics or a fixed threshold set according to engineering accuracy requirements. The aim is to exclude unreliable matches through a quality threshold mechanism. The registration quality evaluation value is a comprehensive indicator reflecting the registration reliability of pipeline nodes or segments. It can be implemented based on a weighted combination of geometric quality evaluation values, topological quality evaluation values, and semantic quality evaluation values. The aim is to provide quantitative compensation for differences in data source accuracy during the matching process. A unified pipeline object can be understood as a logical data structure representing the same physical pipeline entity after fusion. It can be implemented using a graph model or object-relational mapping (ORM) technology, aiming to establish a unified representation of pipeline entities across data sources. A unified identifier specifically refers to a unique identification code assigned to a unified pipeline object. It can be implemented using a globally unique identifier or a coding system generated based on business rules, aiming to support attribute fusion and data management. Objects awaiting confirmation refer to pipeline data elements that do not meet matching quality requirements. They can be marked as special states and stored in an independent data area, aiming to preserve the integrity of the original data while avoiding erroneous clustering.

[0076] Specifically, this application's solution first determines a set of candidate corresponding objects based on spatial distance relationships, upstream and downstream connectivity relationships, loop structure similarity, and semantic attribute similarity. This ensures that candidate objects have reasonable relevance in three key dimensions: geometric distribution, network topology, and business semantics, avoiding topological breaks or semantic conflicts caused by single-dimensional matching. Then, matching evaluation is performed based on the registration quality evaluation values ​​of the candidate corresponding objects, transforming geometric, topological, and semantic quality indicators into criteria for matching reliability, prioritizing the contributions of high-quality data sources. On this basis, a dual screening mechanism selects the candidate object with the highest registration quality evaluation value and a matching evaluation result greater than a preset threshold as the target corresponding object. The threshold setting filters unreliable matches, and the quality evaluation value ranking ensures the selection of the optimal solution. Finally, the matched objects are clustered into unified pipeline objects and assigned unique identifiers, achieving seamless integration of objects across data sources. Objects that do not meet the threshold are marked as pending confirmation, preserving data integrity while preventing erroneous clustering. This process, through the synergistic effect of multi-dimensional similarity analysis and quality-driven matching mechanism, ensures that the clustering process of pipeline nodes and pipeline segments satisfies both spatial location constraints and conforms to pipeline network topology and business semantic rules, thereby achieving highly reliable object identification under a unified spatial benchmark.

[0077] As a preferred embodiment, the specific implementation of this application is as follows: In the scenario of urban water supply network data fusion, for two pipeline data sources from design drawings and geophysical exploration results, firstly, for valve nodes in the design drawings, the spatial Euclidean distance, upstream and downstream connectivity similarity, and facility number matching degree between them and all valve nodes in the geophysical exploration results are calculated to form a set of candidate corresponding objects. Subsequently, according to the registration quality evaluation value calculated according to claim 7, each matching pair in the candidate set is weighted and scored. This score comprehensively considers the geometric residual weight of 0.4, the topological consistency weight of 0.3, and the semantic consistency weight of 0.3. When the score exceeds a preset threshold of 0.85, the matching pair with the highest score is selected as the target corresponding object. The DN200 valve nodes in the design drawings and the corresponding nodes in the geophysical exploration results are clustered into a unified pipeline object and assigned a unified identifier in UUID format. For fire hydrant nodes with scores below the threshold, they are marked as objects to be confirmed and temporarily stored in an independent data area, pending manual review to determine whether to include them in the fusion results. In this embodiment, the spatial location distance relationship is calculated using three-dimensional coordinate deviation, the upstream and downstream connectivity relationship is determined through path analysis of the pipeline topology map, the similarity of the loop structure is evaluated based on the loop closure degree, and the semantic attribute similarity is quantified by facility number editing distance.

[0078] Through the above solution, this application solves the problem of inaccurate object clustering in multi-source pipeline data fusion. The quality-driven matching mechanism improves the reliability of unified pipeline object identification, avoids identification errors and data inconsistencies caused by low-quality matching, ensures the engineering usability of fusion annotation results in terms of pipeline connectivity maintenance and attribute consistency, and at the same time, the mechanism of retaining objects to be confirmed ensures the integrity and traceability of the data processing process.

[0079] In practical applications, some embodiments of this application propose constructing a quality-driven attribute fusion model to achieve attribute fusion annotation of multi-source pipeline data. However, in its implementation, how to scientifically quantify the reliability of data sources to dynamically allocate fusion weights, and how to eliminate the interference of low-quality data sources on the fusion results, have not yet been fully resolved. Specifically, existing methods lack a comprehensive consideration mechanism for multiple dimensions of indicators such as registration quality evaluation value, data source accuracy, and business credibility when fusing geometric and classification attributes. This leads to overly subjective or singular weight allocation, which may allow low-quality data sources to directly participate in the determination of target attribute values, thereby causing problems such as attribute conflicts, distorted fusion results, or decreased business credibility. This makes it difficult to meet the stringent requirements for attribute consistency and reliability in intelligent operation and maintenance of urban pipeline networks.

[0080] In response, this application further proposes that, for each unified pipeline object, when constructing a quality-driven attribute fusion model based on registration quality evaluation values, data source accuracy, and business reliability, the following should be included: For a unified pipeline object, corresponding geometric and classification attributes are collected from each pipeline data source. The registration quality evaluation value is used as a quality indicator reflecting registration reliability, the data source accuracy is used as a quality indicator reflecting measurement accuracy, and the business reliability is used as a quality indicator reflecting business reliability. Based on all quality indicators, geometric attribute fusion weights and classification attribute fusion weights are determined for each pipeline data source. Pipeline data sources with registration quality evaluation values ​​lower than a preset quality threshold are used as alternative references and are not included in the determination of the target attribute values ​​of the unified pipeline object.

[0081] Among them, the registration quality evaluation value refers to a comprehensive index reflecting the reliability of pipeline data registration. It can be achieved by weighted averaging of geometric quality evaluation value, topological quality evaluation value, and semantic quality evaluation value, aiming to quantify the overall accuracy of the registration results in terms of spatial location, connectivity, and semantic attributes. Data source accuracy refers to a quality index reflecting the measurement accuracy of pipeline data. It can be achieved by pre-setting different accuracy levels according to the data source type (e.g., geophysical results measured by total station, design drawings calculated from drawings), aiming to distinguish the differences in the reliability of original measurements from different data sources. Business reliability refers to a quality index reflecting the business reliability of pipeline data. It can be achieved through expert scoring or statistical analysis of historical operation and maintenance records, aiming to integrate operation and maintenance experience of key business attributes such as valve facility numbers into the fusion process. The quality-driven attribute fusion model refers to an attribute fusion framework built based on multi-dimensional quality indicators. It can be implemented using the analytic hierarchy process (AHP) or fuzzy comprehensive evaluation method, aiming to establish an objective weight allocation mechanism. Geometric attribute fusion weights refer to the weighting coefficients used for spatial location correction. These weights are normalized based on the product of registration quality evaluation values, data source accuracy, and business reliability, aiming to ensure that highly reliable data dominates the geometric attribute fusion process. Categorical attribute fusion weights refer to the weighting coefficients used for semantic attribute fusion. They can be calculated using a similar method to geometric attribute fusion weights but emphasize business reliability. Their purpose is to improve the business applicability of categorical attributes such as material and pipeline type. Categorical attributes are those belonging to discrete categories in semantic attributes, including pipeline type, media type, and pressure rating. The preset quality threshold is a critical value used to filter low-quality data sources. It is set as the percentile of the overall error based on the historical registration error statistical distribution, aiming to eliminate interference from low-quality data sources in the fusion results.

[0082] Specifically, this application's solution establishes a comprehensive data foundation by collecting geometric and classification attributes of various pipeline data sources for a unified pipeline object. Registration quality evaluation values, data source accuracy, and business reliability are used as multi-dimensional quality indicators to comprehensively evaluate the performance of each pipeline data source in registration reliability, measurement accuracy, and business scenarios. Based on all quality indicators, geometric attribute fusion weights and classification attribute fusion weights are dynamically calculated, ensuring that weight allocation adapts to the inherent characteristics and registration status of the data sources. Simultaneously, a preset quality threshold is set, excluding pipeline data sources with registration quality evaluation values ​​below this threshold from the determination of target attribute values, retaining them only as alternatives. This mechanism, through the synergistic effect of multi-dimensional quality indicators, achieves objectivity and refinement in weight allocation, fully utilizing high-quality data sources to improve fusion accuracy while avoiding interference from low-quality data sources, thereby ensuring the robustness and business reliability of pipeline attribute fusion within a unified spatial coordinate system.

[0083] As a specific implementation method, the scheme of this application is implemented as follows: For a unified pipeline object of valves in a water supply pipeline, geometric attributes and classification attributes are collected from three data sources: design drawings, geophysical exploration results, and maintenance records. Among them, the registration quality evaluation value of the geophysical exploration results is high (due to its small geometric residuals and good topological consistency), the data source accuracy is set to a high-precision level (due to the use of total station measurements), and the business reliability is medium (due to the lack of the latest replacement records). The registration quality evaluation value of the maintenance records is medium (due to a small number of semantic inconsistencies), the data source accuracy is low (due to estimation based on drawings), but the business reliability is high (due to the recording of the latest maintenance information). According to the quality index calculation, the geophysical exploration results obtain a higher geometric attribute fusion weight, while the maintenance records obtain a higher classification attribute fusion weight. Since the registration quality evaluation value of the design drawings is lower than the preset quality threshold (due to its large geometric residuals), it is marked as a candidate reference and does not participate in the determination of the target attribute value. Finally, the spatial location is mainly corrected based on the geophysical exploration results, and the valve type is mainly determined based on the maintenance records.

[0084] Through the above-mentioned solution, this application solves the problems of reliability quantification and low-quality data interference in the fusion of multi-source pipeline data attributes, avoids attribute conflicts and fusion distortion caused by subjective weight allocation or low-quality data participation, improves the reliability of fusion results in terms of geometric location accuracy and semantic attribute consistency, and meets the strict requirements of urban pipeline network intelligent operation and maintenance for attribute consistency.

[0085] Specifically, in some embodiments described above in this application, geometric attributes are corrected and classification attributes are weighted within engineering constraints to fuse pipeline data. However, in this process, geometric attribute correction may lead to correction results exceeding engineering specifications such as design slope, burial depth, or minimum cover thickness due to neglecting actual engineering constraints, resulting in infeasible fusion results. Simultaneously, the weighted voting of classification attributes may fail to resolve attribute inconsistencies due to insufficient weight allocation or data source conflicts, leading to fusion errors or unresolved conflicts in key semantic attributes such as pipeline type and media type.

[0086] In response, this application further proposes correcting geometric attributes within engineering constraints and weighting classification attributes to obtain the final fused annotation result, including: Based on the geometric attributes of each pipeline data source and the corresponding geometric attribute fusion weights, combined with engineering constraints, the spatial location, burial depth, and orientation of the unified pipeline object are corrected. The engineering constraints include design slope, design burial depth range, and minimum cover thickness.

[0087] When the correction result exceeds the engineering constraints, the alternative geometric attributes that satisfy the engineering constraints are used as the unified geometric attributes. Based on the classification attributes of each pipeline data source and the corresponding classification attribute fusion weight, the classification attributes are weighted and voted on. The weighted voting result is used as the target classification attribute of the unified pipeline object. For classification attributes that still have multiple candidate values ​​after weighted voting, they are marked as attributes to be confirmed and the pipeline data source information participating in the voting is recorded.

[0088] In practical applications, geometric attribute fusion weights refer to weight coefficients dynamically generated based on registration quality evaluation values, data source accuracy, and business reliability. These weights can be implemented using normalized numerical ranges or weight models trained on historical data. The aim is to prioritize input from highly reliable data sources during the geometric attribute correction process, avoiding interference from low-quality data sources. Engineering constraints refer to mandatory specifications that must be followed in pipeline design and construction. These constraints can include a set of parameters such as design slope, design burial depth range, and minimum cover thickness. The purpose is to ensure that the fused geometric attributes conform to actual engineering feasibility and prevent safety hazards caused by correction results exceeding specifications. Specifically, geometric attribute correction refers to the process of optimizing and adjusting the spatial location, burial depth, and orientation of a unified pipeline object. This can be achieved using iterative optimization algorithms combining least squares methods with constraint conditions or boundary verification mechanisms based on engineering specifications. The goal is to achieve optimal geometric consistency in the correction results while satisfying engineering constraints. In practical applications, alternative geometric attributes refer to the replacement attribute values ​​used when the correction result exceeds engineering constraints. These can be implemented using boundary values ​​that satisfy engineering constraints or reference values ​​from historical engineering cases. The purpose is to maintain the continuity of pipeline connectivity and avoid the risk of pipeline breakage caused by forcibly applying excessive correction values. The classification attribute fusion weight refers to the weight coefficient reflecting the reliability of each pipeline data source in classification attribute fusion. It can be implemented using a similar calculation logic to the geometric attribute fusion weight, but with adaptive adjustments for the characteristics of classification attributes. The purpose is to improve the accuracy of fusion of classification attributes such as pipeline type and material. Specifically, weighted voting refers to the process of deciding on the classification attribute values ​​of each pipeline data source based on the classification attribute fusion weight. This can be implemented using a weighted majority voting algorithm or a voting mechanism based on confidence thresholds. The purpose is to suppress interference from low-quality data sources and ensure that the voting results reflect the semantic information of high-confidence data sources. In practical applications, attributes to be confirmed refer to classification attributes that still have multiple candidate values ​​after weighted voting. These can be marked as awaiting manual verification and the data source information recorded. The purpose is to provide a clear basis for manual intervention and avoid attribute errors caused by blind fusion.

[0089] Specifically, the proposed solution combines geometric attribute fusion weights with engineering constraints to correct the geometric attributes of a unified pipeline object. During the correction process, it monitors in real time whether the correction results exceed engineering specifications such as design slope, design burial depth, or minimum cover thickness. When a correction result exceeds the constraints, it automatically switches to alternative geometric attributes that meet the engineering constraints, thus ensuring that the geometric fusion result is within the engineering feasible range. Simultaneously, for classification attributes, the fusion weights of classification attribute values ​​from each pipeline data source are weighted and voted on to determine the target attribute value. Attributes with multiple candidate values ​​after voting are marked and recorded, enabling the attribute fusion process to automate most cases while providing an interface for manual intervention in complex issues. This mechanism, through the linkage of engineering constraints and weight correction, and the synergy between the alternative mechanism and voting conflict handling, solves the core problems of geometric correction exceeding limits and classification attribute conflicts.

[0090] As a specific embodiment, the solution of this application is implemented as follows: For the correction of the burial depth attribute of a water supply pipeline, a weighted average burial depth value is calculated based on the geometric attribute fusion weight of each pipeline data source, and correction is performed in conjunction with the designed burial depth range and minimum cover thickness. If the corrected burial depth value is lower than the minimum cover thickness, the candidate burial depth value that meets the minimum cover thickness is automatically adopted as the unified geometric attribute. For the fusion of pipeline material attributes, the candidate material values ​​such as "cast iron" and "PE" provided by each data source are weighted and voted on according to the classification attribute fusion weight. If the vote rate of "cast iron" is higher than that of other candidate values ​​in the voting results, the material is determined to be "cast iron". If the vote rates of "cast iron" and "PE" are close after voting and neither reaches the preset threshold, the material attribute is marked as an attribute to be confirmed and the data source information participating in the voting is recorded.

[0091] Through the above-described scheme, this application ensures the feasibility of geometric attribute correction within engineering constraints, avoiding pipeline safety risks caused by correction results exceeding limits. Simultaneously, by employing weighted voting and conflict marking mechanisms, the inconsistency issue in classification attribute fusion is resolved, improving the fusion accuracy of key semantic attributes such as pipeline type and media type. This achieves a dual improvement in both engineering feasibility and semantic consistency for multi-source pipeline data.

[0092] In the above embodiments, a coordinate transformation model that simultaneously includes translation parameters, rotation parameters, and local weak deformation parameters is constructed. Based on this, a registration error model consisting of source network topology semantic constraint graphs and target network topology semantic constraint graphs, combined with geometric residuals, topology consistency, and semantic consistency, as well as iterative optimization and local registration optimization mechanisms, can be introduced. This can eliminate the overall offset and local distortion caused by different coordinate systems, multiple measurements, and historical data in the global scope. It can also achieve fine-grained accurate registration of multi-source pipeline data while maintaining the correctness of pipeline connectivity, loop structure, and facility semantic attributes. By calculating the registration quality evaluation value for each pipeline node and pipeline segment from the registration results, multi-source data are clustered into unified pipeline objects with unified identifiers. A quality-driven attribute fusion model is constructed using the registration quality evaluation value, data source accuracy, and business credibility. Within the engineering constraints, geometric attributes are corrected in a controlled manner, and classification attributes are weighted and fused. This ensures that the final fused annotation results have high consistency, high credibility, and traceability in spatial location, topology, and semantic attributes, thereby reducing the risks of location mismatch, connectivity errors, and attribute conflicts in traditional solutions.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for registration and fusion annotation of multi-source pipeline data for heterogeneous coordinate systems, characterized in that, include: Acquire multi-source pipeline data and source coordinate system description information, and establish a coordinate transformation model for each pipeline data source. The coordinate transformation model includes translation parameters, rotation parameters, and local weak deformation parameters. Based on the multi-source pipeline data, pipeline nodes, pipeline segments, connectivity relationships, and semantic attributes are extracted. Source pipeline network topology semantic constraint graphs and target pipeline network topology semantic constraint graphs are constructed. Semantically unique registration constraint features are selected for coarse registration. Based on the coarse registration results, a registration error model is established. The parameters of each coordinate transformation model are solved by iterative optimization. Local registration optimization is performed on local spatial regions where the registration error exceeds the limit to obtain globally consistent registration results. The registration error model includes geometric residuals, topological consistency, and semantic consistency. Based on the registration results, a registration quality evaluation value is calculated for each pipeline node and each pipeline segment. Pipeline nodes and pipeline segments from different pipeline data sources that correspond in spatial location, topology, and semantic attributes are clustered into a unified pipeline object, and a unified identifier is assigned to the unified pipeline object. For each unified pipeline object, a quality-driven attribute fusion model is constructed based on the registration quality evaluation value, data source accuracy, and business credibility. Within the engineering constraints, the geometric attributes are corrected, the classification attributes are weighted, and finally the fusion annotation result is obtained. When establishing a registration error model based on the coarse registration results, the following should be included: For the matched pipeline nodes and pipeline segments in the source network topology semantic constraint diagram and the target network topology semantic constraint diagram, calculate the position deviation and burial depth deviation respectively, and use the position deviation and burial depth deviation as geometric residuals; After coarse registration, determine whether the connectivity between the corresponding pipeline nodes and the corresponding pipeline segments maintains the original connectivity, loop structure, and branch structure, and use the degree of connectivity disruption as an indicator of topology consistency. The semantic attributes of corresponding pipeline nodes and corresponding pipeline segments are compared, and the inconsistency of the semantic attributes is used as an indicator of semantic consistency. The geometric residual, topological consistency and semantic consistency are comprehensively evaluated according to a preset comprehensive weight to obtain the registration error model.

2. The method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems according to claim 1, characterized in that, When acquiring multi-source pipeline data and source coordinate system description information, the following is included: The source coordinate system description information includes the coordinate system type, coordinate unit, coordinate axis direction, and elevation datum of each pipeline data source. Under a unified spatial coordinate system, the source coordinate system description information is associated with known control points in the engineering control network and pipeline nodes with unique semantic identifiers. The translation and rotation parameters of each pipeline data source are obtained based on the correspondence obtained from the association, and an initial coordinate transformation model is generated.

3. The method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems according to claim 2, characterized in that, Obtaining the local weak deformation parameters includes: Based on the spatial distribution of the pipeline nodes and pipeline segments and the division results of the engineering control network, each pipeline data source is divided into multiple local spatial regions or into multiple pipeline segments along the pipeline route, and local weak deformation parameters are preset for each local spatial region or each pipeline segment to characterize local weak deformation.

4. The method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems according to claim 1, characterized in that, When constructing the source network topology semantic constraint graph and the target network topology semantic constraint graph, and selecting semantically unique registration constraint features for coarse registration, the following steps are included: For each pipeline data source, the pipeline nodes representing well chambers, valves, inlets / outlets, and equipment foundations, as well as the pipeline segments connected to the pipeline nodes, are identified. The pipeline nodes are treated as graph nodes, and the pipeline segments and their connectivity are treated as graph edges. A source network topology semantic constraint graph and a target network topology semantic constraint graph are constructed. Semantic attributes are recorded on each graph node and each graph edge. The semantic attributes include pipeline type, medium type, pipe diameter, material, and facility number. From the source network topology semantic constraint diagram and the target network topology semantic constraint diagram, pipeline nodes with unique facility numbers and connectivity greater than a preset connectivity threshold in the topology structure are selected as semantically unique registration constraint features. The positional and connectivity relationships between these semantically unique registration constraint features are used as coarse registration constraints.

5. The method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems according to claim 1, characterized in that, When obtaining a globally consistent registration result by iteratively optimizing and jointly solving the parameters of each coordinate transformation model, local registration optimization is performed on local spatial regions where registration errors exceed the limit, including: Based on the registration error model, the translation and rotation parameters of all pipeline data sources are globally iteratively optimized to reduce the overall geometric residual, topological consistency and semantic consistency error to below the first preset error threshold. Based on the comprehensive error of each pipeline node and each pipeline segment in the registration error model, a local spatial region where the registration error exceeds the limit is defined. Within the local spatial region, the local weak deformation parameters in the corresponding coordinate transformation model are adjusted. While maintaining the continuity of the connection between pipeline nodes and pipeline segments and ensuring that the burial depth and slope meet the engineering constraints, the adjustment range of the local weak deformation parameters is limited so that the comprehensive error within the local spatial region is reduced to below the second preset error threshold, thereby obtaining the globally consistent registration result.

6. The method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems according to claim 1, characterized in that, When calculating the registration quality evaluation value for each pipeline node and each pipeline segment based on the registration results, the following are included: The geometric quality evaluation value is obtained based on the geometric residuals of the corresponding pipeline nodes and corresponding pipeline segments in the registration error model; the topological quality evaluation value is obtained based on the topological consistency of the corresponding pipeline nodes and corresponding pipeline segments in the registration error model; and the semantic quality evaluation value is obtained based on the semantic consistency of the corresponding pipeline nodes and corresponding pipeline segments in the registration error model. The geometric quality evaluation value, the topology quality evaluation value, and the semantic quality evaluation value are weighted according to the accuracy level of each pipeline data source to obtain the registration quality evaluation value of the corresponding pipeline node and the corresponding pipeline segment.

7. The method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems according to claim 6, characterized in that, Clustering pipeline nodes and pipeline segments from different pipeline data sources that correspond in spatial location, topology, and semantic attributes into a unified pipeline object, and assigning a unified identifier to the unified pipeline object includes: For each pipeline node and pipeline segment in the pipeline data source, a set of candidate corresponding objects is determined based on the distance relationship between its spatial location and the spatial location of pipeline nodes and pipeline segments in other pipeline data sources, the upstream and downstream connectivity relationship and the similarity of the loop structure, as well as the similarity of the semantic attributes. Then, each candidate corresponding object in the set of candidate corresponding objects is matched according to the registration quality evaluation value of the candidate corresponding objects. In each set of candidate corresponding objects, the candidate corresponding object with the highest registration quality evaluation value and the matching evaluation result greater than the preset matching evaluation threshold is selected as the target corresponding object. Pipeline nodes and pipeline segments belonging to the same target corresponding object are clustered into the same unified pipeline object. A unique unified identifier is assigned to each unified pipeline object. For pipeline nodes and pipeline segments that do not meet the matching evaluation threshold, they are marked as objects to be confirmed and are not included in any unified pipeline object.

8. The method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems according to claim 1, characterized in that, For each of the unified pipeline objects, when constructing a quality-driven attribute fusion model based on the registration quality evaluation value, data source accuracy, and business reliability, the following are included: For the unified pipeline object, corresponding geometric attributes and classification attributes are collected from each pipeline data source. The registration quality evaluation value is used as a quality indicator reflecting registration reliability, the data source accuracy is used as a quality indicator reflecting measurement accuracy, and the business reliability is used as a quality indicator reflecting business reliability. Based on all the quality indicators, geometric attribute fusion weights and classification attribute fusion weights are determined for each pipeline data source. Pipeline data sources with registration quality evaluation values ​​lower than a preset quality threshold are used as alternative references and are not included in the determination of the target attribute values ​​of the unified pipeline object.

9. The method for multi-source pipeline data registration and fusion annotation for heterogeneous coordinate systems according to claim 8, characterized in that, Within engineering constraints, geometric attributes are corrected, and classification attributes are weighted to obtain the final fused annotation result, which includes: Based on the geometric attributes of each pipeline data source and the corresponding geometric attribute fusion weights, combined with engineering constraints, the spatial location, burial depth, and orientation of the unified pipeline object are corrected. The engineering constraints include design slope, design burial depth range, and minimum cover thickness. When the correction result exceeds the engineering constraints, the alternative geometric attributes that satisfy the engineering constraints are used as the unified geometric attributes. Based on the classification attributes of each pipeline data source and the corresponding classification attribute fusion weight, the classification attributes are weighted and voted on. The weighted voting result is used as the target classification attribute of the unified pipeline object. For classification attributes that still have multiple candidate values ​​after weighted voting, they are marked as attributes to be confirmed and the pipeline data source information participating in the voting is recorded.

Images