A method and system for local updating of edge optimization of a three-dimensional model

Abstract

The application provides a three-dimensional model local updating joint edge optimization method and system, and belongs to the technical field of city real scene three-dimensional model updating. The method obtains a benchmark period three-dimensional model, a new period local three-dimensional model and a to-be-updated area contour line, sets a joint edge sampling profile line along the to-be-updated area contour line, establishes corresponding surface points and calculates a joint edge residual error, and forms a joint edge residual error sequence; convergence analysis is performed based on the joint edge residual error sequence to determine a joint edge boundary, residual error constraint weights are determined according to a residual error attenuation process, a new period local three-dimensional model joint edge area is adjusted, and local model replacement is completed; after updating, the joint edge residual error is recalculated, and local joint edge adjustment is performed on a joint edge abnormal section. The application can realize adaptive determination of the joint edge boundary, and improve the joint edge continuity, geometric consistency and model fusion quality of the three-dimensional model after local updating.

Description

Technical Field

[0001] This invention belongs to the field of urban 3D modeling technology, specifically relating to a method and system for optimizing the edge connection of a 3D model for local updates. Background Technology

[0002] Urban reality 3D models are crucial foundational data for smart cities, natural resource management, digital twins, urban planning, and infrastructure management. They are typically generated using processing techniques such as multi-view imagery dense matching, point cloud fusion, surface reconstruction, and texture mapping. As urban development continues, the construction, demolition, and renovation of buildings, as well as the constant changes in spatial elements such as roads, bridges, and terrain, gradually reduce the timeliness of existing urban reality 3D models. To maintain consistency between the model data and the actual urban environment, continuous updates to existing 3D models are necessary.

[0003] For large-scale urban models, reprocessing all aerial imagery and reconstructing the entire survey area every time local changes occur not only consumes significant computational resources but also incurs high costs for manual verification, data replacement, and data dissemination. Therefore, current engineering practices typically employ a local incremental update approach. This involves first identifying the spatial area where changes have occurred, then reconstructing the corresponding local 3D model using newly acquired data, and finally replacing the existing urban 3D model with the newly created local model. This reduces overall update costs and improves update efficiency.

[0004] In local incremental updates, model edge alignment is a crucial factor affecting update quality. Local updates not only require obtaining a new local 3D model but also accurately determining the actual replacement boundary between the old and new models, ensuring a continuous, smooth, and consistent geometric connection between the local model and the original model at the replacement boundary. Currently, engineering projects typically expand the update area outward by a fixed distance based on changing feature areas, building boundaries, or manually defined areas, or directly use regular tile boundaries as the model replacement range. Then, model trimming, vertex interpolation, surface smoothing, and texture gradation are performed at the edges of the update area to reduce the seams between the old and new models. However, these methods primarily rely on preset rules to determine the replacement boundary and do not directly establish a correspondence between the boundary position and the actual geometric differences between the old and new models. Therefore, in complex urban environments, this can easily affect the quality of local updates.

[0005] Analysis reveals that existing technologies for local updating of 3D urban reality models have at least the following shortcomings: (1) Fixed-distance buffer zones cannot reflect the spatial distribution characteristics of edge joint errors. For the same area of ​​change, different directions may be adjacent to different features such as roads, building roofs, building facades, vegetation cover areas and terrain slope areas. Using a uniform outward expansion distance to form the update range can easily cause some areas to be over-updated, while other areas are still located in positions where the geometry of the new and old models are significantly different, resulting in obvious seams.

[0006] (2) Using regular tiles or data block boundaries as model replacement boundaries can easily cut through continuous structures such as buildings, bridges and retaining walls. Even if the old and new models use the same coordinate reference and resolution, due to differences in mesh vertex distribution, triangulation method and local reconstruction results, problems such as cracks, overlapping surfaces, steps and abrupt changes in normal direction can still easily occur.

[0007] (3) The overall registration accuracy cannot truly reflect the local edge quality. After the coordinate registration of the new and old models is completed, the overall root mean square error is usually used to evaluate the registration effect. However, there may still be systematic deviations, local distortions and abnormal errors caused by occlusion in local areas. It is difficult to accurately determine the appropriate local replacement boundary based solely on the overall registration result.

[0008] (4) Judging the edge quality based on the single-point distance threshold is easily affected by outliers and noise. Factors such as vehicles, tree canopies, hole edges, shadows and sparse points may cause local residual abnormalities. It is impossible to guarantee that the neighborhood area as a whole has stable edge conditions based solely on the distance threshold of individual sampling points.

[0009] (5) A model transition zone with a uniform width is difficult to adapt to the geometric complexity of different boundary areas. Flat areas usually only require a short distance to complete the model transition, while building roofs, facades and complex structures require a larger adjustment range. A transition method with a fixed width and fixed weight can easily lead to some areas being overly smooth or having insufficient joints.

[0010] (6) The existing edge quality is usually checked after the model replacement is completed. When it is found that the local edge quality does not meet the requirements, it is often necessary to expand the entire update area or regenerate the complete tile, which leads to repeated calculation of areas that have met the requirements and reduces update efficiency.

[0011] (7) Although some schemes introduce multiple processing modules such as model registration, geometric fusion, texture restoration, quality inspection and result management, each module adopts different constraints. There is a lack of a unified constraint mechanism between boundary determination, model fusion and quality evaluation, making it difficult to form a consistent technical route that runs through the entire process of local model replacement.

[0012] Therefore, existing technologies require a local update method for urban real-scene 3D models that establishes a unified constraint mechanism around the edge-joining error of the local model. This method would enable the same edge-joining residual to be used simultaneously for boundary replacement determination, model transition adjustment, and edge-joining quality control. Furthermore, when the edge-joining quality exceeds the limit, only the corresponding boundary segment would be locally corrected, thereby improving the edge-joining quality and update efficiency of local incremental updates of urban real-scene 3D models. Summary of the Invention

[0013] This invention provides a method and system for optimizing the edge connection during local updates of 3D models, in order to solve the problems of poor edge connection quality and insufficient continuity of model fusion in the local update process of 3D models in the prior art.

[0014] To address the aforementioned technical problems, the present invention discloses the following technical solutions: One aspect of the present invention provides a method for local update edge optimization of a three-dimensional model, comprising: Obtain the baseline 3D model, the new local 3D model, and the outline of the area to be updated; Multiple outwardly extending edge sampling profiles are set along the outline of the area to be updated; Based on the sampling profiles of each edge, the corresponding surface points of the baseline period 3D model and the new period local 3D model are determined. The edge residuals are determined according to the deviation between the corresponding surface points, and the edge residual sequence corresponding to each sampling profile is formed. The edge residual sequence is subjected to convergence analysis to determine the edge position on each edge sampling profile, and the edge positions are connected to form a closed edge boundary surrounding the contour line to be updated. The residual constraint weights are determined based on the residual decay process between the contour line to be updated and the corresponding edge position on each edge sampling profile. The model surface of the new local 3D model at the junction boundary is adjusted according to the residual constraint weight, and the adjusted new local 3D model replaces the reference period 3D model within the junction boundary. The edge joint residuals are redefined in the updated 3D model, and local edge joint adjustments are made to the abnormal edge joint sections based on the edge joint residuals.

[0015] Optionally, the step of setting multiple outwardly normal-extending edge sampling profiles along the contour line of the area to be updated includes: Candidate edge bands are constructed outward from the outline of the area to be updated; Discretize the contour line of the area to be updated to obtain multiple contour points; Starting from each of the aforementioned contour points, multiple edge sampling profiles are formed by extending along the outward normal direction at the corresponding contour point within the candidate edge band.

[0016] Optionally, the step of setting multiple sampling positions on each edge sampling profile and determining the corresponding surface points of the baseline period 3D model and the new period local 3D model based on each sampling position includes: Multiple sampling positions are set along the sampling profile of each of the aforementioned joint edges at preset sampling intervals; Based on each sampling location, a reference surface point is determined on the reference period 3D model. The surface normal direction at the reference surface point is obtained, and the corresponding surface point is searched along the surface normal direction on the surface of the new period local 3D model to form a corresponding surface point pair.

[0017] Optionally, the step of determining the joint residual based on the deviation between corresponding surface points to form a joint residual sequence corresponding to each joint sampling profile includes: The original edge deviation corresponding to each sampling position is determined based on the positional deviation between each corresponding surface point pair. The system deviation component corresponding to the sampling profile is determined based on the original edge deviation of multiple sampling positions on the same edge sampling profile. The edge residuals of each sampling position are determined based on the original edge deviation and the corresponding system deviation component, and the edge residual sequence corresponding to each edge sampling profile is formed according to the arrangement order of each sampling position on the corresponding edge sampling profile.

[0018] Optionally, the step of performing convergence analysis on the edge residual sequence to determine the edge position on each edge sampling profile includes: Convergence analysis is performed along the residual sequences corresponding to the sampling profiles of each edge to determine the residual convergence interval that meets the preset convergence conditions. For each edge sampling profile, the edge position is determined within the first residual convergence interval outward from the outline of the area to be updated.

[0019] Optionally, the convergence analysis performed along the residual sequences corresponding to each of the sampling profiles to determine the residual convergence intervals that satisfy the preset convergence conditions includes: Sliding window statistical analysis was performed on the edge residual sequence of each edge sampling profile; Based on the statistical results of the edge residuals of the corresponding sequence intervals of each sliding window, it is determined whether each sequence interval meets the preset convergence condition, and the sequence intervals that meet the preset convergence condition consecutively for no less than a preset number are determined as the residual convergence intervals.

[0020] Optionally, determining the residual constraint weights based on the residual decay process between the contour line to be updated and the corresponding edge position on each edge sampling profile includes: For each edge sampling profile, the first position that meets the preset attenuation condition is determined according to the corresponding edge residual sequence as the transition start point, and the corresponding edge position is taken as the transition end point. Connecting the transition starting points on each edge sampling profile line forms an inner transition boundary, and the inner transition boundary and the edge boundary enclose an edge transition zone. For each sampling position within the transition zone, calculate the ratio of the cumulative residual from the sampling position to the corresponding edge position to the cumulative residual from the corresponding transition start point to the edge position, and determine the residual constraint weight corresponding to the sampling position based on the ratio.

[0021] Optionally, adjusting the model surface of the new-period local 3D model at the junction boundary according to the residual constraint weights, and replacing the baseline 3D model within the junction boundary with the adjusted new-period local 3D model, includes: Based on the residual constraint weights corresponding to each sampling position within the transition zone, determine the adjustment amount of the corresponding points on the surface of the new local 3D model; Based on the adjustment amount of the corresponding points on each model surface, the surface of the new local 3D model in the transition zone of the dock edge is adjusted; The updated 3D model is obtained by replacing the reference period 3D model within the boundary with the adjusted new local 3D model, while retaining the reference period 3D model outside the boundary.

[0022] Optionally, the step of redetermining the edge joint residuals of the updated 3D model and making local edge joint adjustments to abnormal sections based on the edge joint residuals includes: Within the transition zone, the corresponding surface points of the baseline 3D model and the updated 3D model are redefined; Based on the positional deviation between the corresponding surface points that have been redefined, the residual sequence of each edge sampling profile in the edge transition zone is determined, and the edge residual sequence is subjected to convergence analysis. Continuous segments in the edge-joining residual sequence that do not meet the preset convergence conditions are regarded as edge-joining abnormal segments. For each abnormal edge segment, the first residual convergence interval is found outward and the edge position is redefined. The edge position is then used to re-adjust the local edge of the 3D model located in the abnormal segment.

[0023] Another aspect of this invention discloses a three-dimensional model local update edge optimization system, comprising: The data acquisition module is configured to acquire the baseline 3D model, the new local 3D model, and the outline of the area to be updated; The edge sampling profile setting module is configured to set multiple edge sampling profiles extending outward in the normal direction along the outline of the area to be updated; The edge residual determination module is configured to determine the corresponding surface points of the baseline period 3D model and the new period local 3D model based on each edge sampling profile, determine the edge residual according to the deviation between the corresponding surface points, and form the edge residual sequence corresponding to each edge sampling profile. The edge boundary forming module is configured to perform convergence analysis on the edge residual sequence, determine the edge position on each edge sampling profile, and connect each edge position to form a closed edge boundary surrounding the contour line to be updated. The residual constraint module is configured to determine the residual constraint weights based on the residual decay process between the contour line to be updated and the corresponding edge position on each edge sampling profile. The model update module is configured to adjust the model surface of the new local 3D model at the junction boundary according to the residual constraint weight, and replace the reference 3D model within the junction boundary with the adjusted new local 3D model. The local edge adjustment module is configured to redetermine the edge residuals of the updated 3D model and perform local edge adjustment on the abnormal edge sections based on the edge residuals.

[0024] This invention provides a method and system for optimizing the edge joint of a local update of a three-dimensional model. By setting edge sampling profiles along the contour line of the area to be updated, corresponding surface points between the baseline three-dimensional model and the new local three-dimensional model are established, and an edge joint residual sequence is constructed. Based on the convergence analysis of the edge joint residual sequence, the edge joint boundary is adaptively determined. This avoids the problem of unreasonable edge joint positions caused by using fixed edge joint boundaries, and enables the edge joint boundary to better adapt to the actual geometric changes of the local model, thereby improving the accuracy and rationality of the edge joint boundary determination.

[0025] This invention further determines the residual constraint weights based on the edge residual decay process, adaptively adjusts the new local 3D model of the edge region, and re-analyzes the edge residuals after the model update, implementing local edge adjustment in abnormal edge sections, thereby forming a closed-loop optimization mechanism for edge quality. This effectively improves the geometric continuity and model fusion quality of the 3D model after local update, reduces geometric abrupt changes and splicing traces in the edge region, and improves the overall accuracy and reliability of the local update of the 3D model.

[0026] The summary section is provided to present the chosen concepts in a simplified form, which will be further described in the detailed description below. The summary section is not intended to identify essential or necessary features of this disclosure, nor is it intended to limit the scope of this disclosure. Attached Figure Description

[0027] The above and other objects, features and advantages of this disclosure will become more apparent from the accompanying drawings, in which like reference numerals generally denote like parts.

[0028] Figure 1 This is a flowchart illustrating a method for optimizing the local update of edges in a three-dimensional model, as provided in an embodiment of the present invention. Figure 2 An implementation provided by an embodiment of the present invention Figure 1 A flowchart illustrating step S200; Figure 3 This is a schematic diagram of a process for determining corresponding surface points of a baseline three-dimensional model and a new local three-dimensional model, provided by an embodiment of the present invention. Figure 4This is a schematic diagram of a process for obtaining the residual sequence of each edge sampling profile according to an embodiment of the present invention; Figure 5 A schematic flowchart illustrating the process of determining the position of each joint on a sampling profile line according to an embodiment of the present invention; Figure 6 An implementation provided by an embodiment of the present invention Figure 5 A flowchart illustrating step S401; Figure 7 An implementation provided by an embodiment of the present invention Figure 1 A flowchart of step S500; Figure 8 An implementation provided by an embodiment of the present invention Figure 1 A flowchart illustrating step S600; Figure 9 An implementation provided by an embodiment of the present invention Figure 1 A flowchart illustrating step S700; Figure 10 This is a schematic diagram of a three-dimensional model local update edge optimization system provided in an embodiment of the present invention. Detailed Implementation

[0029] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While 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 so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0030] The term "comprising" and its variations as used herein signify open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "at least partially based on". The terms "one example embodiment" and "one embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0031] Figure 1 This is a flowchart illustrating a method for optimizing the local update of a three-dimensional model according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes the following steps: Step S100: Obtain the baseline 3D model, the new local 3D model, and the outline of the area to be updated.

[0032] The baseline 3D model is a city 3D model that has been established and put into use before the local update, used to represent the spatial morphology of the target area at the baseline time; the new local 3D model is a local 3D model reconstructed based on the new data, used to represent the spatial morphology of the target area at the update time; the outline of the area to be updated is used to define the area range where actual changes have been confirmed and which needs to be updated using the new local 3D model.

[0033] In one embodiment of this invention, the baseline 3D model and the new local 3D model can be a triangular mesh model, a surface point cloud model, a digital surface model, or other 3D surface models capable of performing surface point queries. When the two use different data formats, they can be converted into a unified 3D surface representation before performing subsequent steps to establish the correspondence between model surfaces.

[0034] In addition to covering the area enclosed by the outline of the area to be updated, the new local 3D model should extend beyond the outline of the area to be updated by a predetermined range, so that the new local 3D model and the baseline 3D model form a spatial overlap area around the area to be updated. The spatial overlap area is used to subsequently establish corresponding surface points of the old and new models, calculate edge residuals, and determine the actual edge positions.

[0035] After acquiring the baseline 3D model and the new local 3D model, their spatial reference information is checked to ensure they use the same plane coordinate datum and elevation datum. If they are not yet under a unified coordinate system, coordinate alignment can be performed through control point registration, feature matching, or other conventional registration methods to ensure the overall registration error is within a preset range. It should be noted that coordinate alignment is used to ensure the reliable establishment of subsequent surface correspondences and does not directly replace the subsequently determined edge residuals with the overall registration error.

[0036] The outline of the area to be updated can be manually delineated based on the actual range of change, or generated based on the change detection results of multiple images, point clouds, or 3D models. It can also be provided by urban planning, construction management, or 3D model update systems. This invention does not limit the specific method of obtaining the outline of the area to be updated.

[0037] The outline of the region to be updated is a closed outline used to completely enclose the region to be updated. For an update task consisting of multiple separate changing regions, multiple closed outlines can be determined separately, and subsequent edge optimization processes can be performed for each closed outline.

[0038] In one embodiment of the present invention, the contour line of the area to be updated can also be preprocessed, including removing duplicate contour points, repairing contour line breaks, eliminating self-intersecting segments, and smoothing local abnormal polylines, so as to obtain a topologically continuous and directionally ordered contour line of the area to be updated, providing a basis for subsequently laying out edge sampling profiles along it.

[0039] Step S200: Set multiple outward-extending edge sampling profiles along the outline of the area to be updated.

[0040] This step is used to establish a unified spatial reference for the subsequent edge optimization process. By constructing candidate edge regions around the area to be updated and laying edge sampling profiles along the outline of the area to be updated, the subsequent establishment of surface correspondence, edge residual calculation, residual convergence analysis, edge boundary determination, model edge adjustment, and post-update edge quality verification are all based on the same spatial index, thereby ensuring that the entire edge optimization process uses a unified data foundation.

[0041] In one embodiment of the present invention, such as Figure 2 As shown, step S200 can be implemented in the following manner: Step S201: Construct candidate edge bands outward based on the outline of the area to be updated.

[0042] After obtaining the outline of the area to be updated, the outline is used as the inner boundary of the candidate edge band, and candidate edge bands are constructed along the outer side of the outline. The candidate edge band is located outside the area to be updated. Its inner boundary does not enter the area to be updated that has been confirmed to have changed, and its outer boundary is located in the area that is effectively covered by both the baseline 3D model and the new local 3D model. This is used to limit the search range for subsequent edge positions.

[0043] The candidate edge band has a preset search width to ensure that the edge search can avoid areas where the edge of the region to be updated is still affected by changes, while also covering areas where there is a stable correspondence between the old and new models. When there are model missing areas, insufficiently covered areas, or insufficient local overlap within the candidate edge band, the candidate edge band can be locally expanded according to the model coverage of the corresponding area to ensure that subsequent edge sampling profiles can obtain effective model surface information.

[0044] The candidate edge band is not the final model replacement boundary, but is used to limit the search range of edge positions. It provides a spatial basis for establishing the correspondence of the model surface, calculating the edge residual, identifying the residual convergence interval, and determining the adaptive edge boundary. Therefore, the final edge boundary is determined by the results of the subsequent edge residual analysis, rather than directly using the boundary of the candidate edge band.

[0045] Step S202: Discretize the contour line of the area to be updated to obtain multiple contour points.

[0046] The contour line of the area to be updated is discretized according to a preset sampling interval, generating multiple contour points arranged sequentially along the contour line. Each contour point is established with a corresponding relationship according to the connection order of the contour line and maintains a unified number, providing the starting position for the subsequent generation of edge sampling profiles.

[0047] The sampling spacing between contour points is adaptively adjusted based on the local geometric features of the contour line. When the contour line is located in relatively straight areas such as road edges or straight building boundaries, the spacing between contour points can be appropriately increased to reduce the amount of subsequent calculations. When the contour line is located at building corners, areas with concave and convex changes, or areas with complex model structures, the spacing between contour points should be appropriately decreased to improve the accuracy of local boundary representation, so that subsequent edge connections can more accurately reflect the changes in the real spatial structure.

[0048] Contour points are not only used to generate edge sampling profiles, but also serve as a unified boundary index throughout the edge optimization process. This ensures that the subsequently determined edge positions can be connected in sequence according to the contour points to form closed edge boundaries, and guarantees that the updated edge quality verification can still be performed along the same spatial position.

[0049] Step S203: Starting from each contour point, extend along the outward normal direction at the corresponding contour point within the candidate edge band to form multiple edge sampling profiles.

[0050] For each contour point, the corresponding outward normal direction is calculated based on the local tangential direction of the contour line of the area to be updated at that contour point. Then, taking the corresponding contour point as the starting point, the outward normal direction is extended to the outside of the candidate edge band until the outer boundary of the candidate edge band, forming the corresponding edge sampling profile.

[0051] Each edge sampling profile passes through the candidate edge zone outward from the region to be updated, so that each edge sampling profile covers the spatial range from the changing region to the stable region, thus fully reflecting the surface difference change process between the old and new models along the outward normal direction.

[0052] Each edge sampling profile is uniquely numbered and established in a one-to-one correspondence with its corresponding contour point. When establishing the model surface correspondence, sampling positions are set along each edge sampling profile. When calculating edge residuals, each edge sampling profile serves as the organizational unit for the residual sequence. During residual convergence analysis, edge positions are determined on each edge sampling profile. When generating edge boundaries, the corresponding edge positions are connected according to the arrangement order of the edge sampling profiles. After model updates are completed, the model surface correspondence is re-established along the original edge sampling profiles, and the updated edge residuals are calculated to verify the local edge quality. Therefore, the same edge sampling profile runs throughout the entire edge optimization process without establishing a new spatial reference system, ensuring that edge boundary determination, model adjustment, and edge quality verification all use a unified spatial index, improving the consistency and traceability of the entire local update process.

[0053] Step S300: Based on each edge sampling profile, determine the corresponding surface points of the baseline period 3D model and the new period local 3D model, determine the edge residuals according to the deviation between the corresponding surface points, and form the edge residual sequence corresponding to each edge sampling profile.

[0054] This step establishes the correspondence between the old and new models and converts the surface differences between them into a sequence of edge residuals distributed along the edge sampling profiles. This invention does not directly use the Euclidean distance or overall registration error between models as the basis for edge matching. Instead, it establishes local surface correspondences based on each edge sampling profile and, after eliminating local system biases, forms edge residuals that truly reflect the consistency of local surfaces.

[0055] In one embodiment of the present invention, such as Figure 3 As shown, the corresponding surface points of the baseline 3D model and the new local 3D model can be determined using the following methods: Step S301: Set multiple sampling positions along each edge sampling profile according to the preset sampling interval.

[0056] Along each edge sampling profile, multiple sampling positions are sequentially set from the area to be updated outwards to the outside of the candidate edge band according to a preset sampling interval. Each sampling position is assigned a corresponding number according to the order of arrangement along the profile direction. The sampling interval can be set according to the model resolution, model accuracy, or model density, so that adjacent sampling positions can fully reflect the spatial changes of the old and new models along the edge sampling profile direction.

[0057] All edge sampling profiles use the same sampling interval. When the local model structure is more complex or the model surface varies greatly, the sampling interval of the corresponding region can be appropriately reduced to improve the sampling accuracy of local surface differences; for regions with relatively gentle structures, the sampling interval can be appropriately increased to reduce the amount of subsequent calculations.

[0058] Step S302: Determine the reference surface points on the reference period three-dimensional model based on each sampling location, obtain the surface normal direction at the reference surface points, and search for corresponding surface points on the surface of the new period local three-dimensional model along the surface normal direction to form corresponding surface point pairs.

[0059] For each sampling location, a corresponding reference surface point is determined on the surface of the reference period 3D model, and the unit surface normal direction at that reference surface point is obtained. Subsequently, using the surface normal direction as the search direction, a new period surface point corresponding to the reference surface point is searched on the new period local 3D model surface, thereby establishing a corresponding surface point pair.

[0060] In this context, both the reference surface point and the corresponding surface point can be located inside the model's triangular face, without being required to correspond to model vertices, thereby reducing the impact of different model triangulation methods on the establishment of correspondence relationships.

[0061] Using the surface normal direction at the reference surface point as the search direction, the nearest model intersection point or the nearest surface point is searched along both the positive and negative normal directions. When multiple candidate surface points exist, the surface point closest to the reference surface point and whose surface normal direction meets the preset angle requirement can be selected as the corresponding surface point. If the distance of a candidate surface point exceeds the preset search distance, the corresponding position is located at the edge of a hole in the model, the difference in normal direction exceeds the preset range, or a reliable correspondence cannot be established, the corresponding sampling position is marked as an invalid sampling point and will not participate in the subsequent edge residual statistics.

[0062] After establishing corresponding surface point pairs, the validity of the correspondence, the local normal change, and the sampling coverage status are recorded synchronously for each sampling location. Among them, the validity of the correspondence, the local normal change, and the sampling coverage status are used to determine whether the correspondence is reliable, and only serve as validity constraints for subsequent edge residual statistics, and are not used as the sole basis for determining the edge boundary.

[0063] In one embodiment of the present invention, such as Figure 4 As shown, the edge residual sequence corresponding to each edge sampling profile can be obtained in the following way.

[0064] Step S303: Determine the original edge deviation corresponding to each sampling position based on the positional deviation between each corresponding surface point pair.

[0065] For each valid corresponding surface point pair, the original edge deviation of the corresponding sampling position is determined based on the positional relationship between the corresponding surface point of the new local three-dimensional model and the reference surface point of the reference three-dimensional model.

[0066] The difference vector between corresponding surface points is projected onto the surface normal direction at the reference surface point to obtain a directional normal deviation as the original edge-joining deviation. Using the normal deviation can preserve the relative offset direction between the old and new models, reduce the impact of differences in tangential sampling density and model triangulation on the deviation calculation results, and make the original edge-joining deviation more realistically reflect the local surface differences between the two models.

[0067] Step S304: Determine the system deviation component corresponding to the sampling profile of the same edge based on the original edge deviation of multiple sampling positions on the same edge sampling profile.

[0068] Even if the new local 3D model and the baseline 3D model are registered with unified coordinates, there may still be systematic deviations such as overall uplift, overall sinking, slow tilting or low-frequency drift in local areas. Therefore, this invention does not directly use the original edge deviation as the basis for edge judgment. Instead, it performs an overall analysis of the original edge deviations corresponding to multiple sampling positions on the same edge sampling profile to determine the systematic deviation components corresponding to the edge sampling profile.

[0069] Based on the original edge deviations corresponding to multiple sampling positions along the same edge sampling profile, the systematic bias component is determined through methods such as low-frequency trend estimation, robust fitting, or local trend fitting. This ensures that local outliers, tree occlusion, model holes, or local reconstruction errors do not dominate the systematic bias estimation result. The systematic bias component describes the overall slowly changing low-frequency offset trend along the same edge sampling profile, while local abrupt changes are retained in subsequent edge residuals.

[0070] Step S305: Determine the edge residual of each sampling position based on the original edge deviation and the corresponding system deviation component, and form the edge residual sequence corresponding to each edge sampling profile according to the arrangement order of each sampling position on the corresponding edge sampling profile.

[0071] For each sampling location, the corresponding original edge deviation is subtracted from the corresponding system deviation component to obtain the edge residual for that sampling location. The edge residual represents the degree of local surface inconsistency that still exists between the old and new models after removing local low-frequency system deviations. It mainly reflects local reconstruction errors, differences in the actual structure, occlusion effects, and areas near the boundary that are still affected by changes.

[0072] Subsequently, the edge residuals corresponding to each sampling position on the same edge sampling profile are organized sequentially from the area to be updated to the outside of the candidate edge band, forming the edge residual sequence corresponding to that edge sampling profile.

[0073] Furthermore, sliding window statistics can be performed on the residual sequence along the sampling profile at the junction to calculate statistical information such as the absolute mean of the residuals, the degree of dispersion of the residuals, the rate of change of the residuals between adjacent windows, and the proportion of effective samples for each window. The absolute mean of the residuals reflects the average surface deviation between the old and new models; the degree of dispersion of the residuals reflects local stability; the rate of change of the residuals describes the decay trend of the residuals along the sampling profile at the junction; and the proportion of effective samples ensures the reliability of the residual statistical results. All the above statistical results are derived from the same residual sequence at the junction and serve as the basis for subsequently identifying the residual convergence interval and determining the junction boundary, without establishing a new independent evaluation index system.

[0074] Step S400: Perform convergence analysis on the edge residual sequence to determine the edge position on each edge sampling profile, and connect each edge position to form a closed edge boundary surrounding the contour line to be updated.

[0075] This step is used to identify the location where the local surface differences between the new local 3D model and the baseline 3D model tend to stabilize, based on the spatial variation of the joint residuals along each joint sampling profile from the inside out, and to use this location as the actual joint position. Compared to using a fixed buffer distance, regular tile boundaries, or candidate joint zone outer boundaries as model replacement boundaries, this invention determines the joint position based on the convergence of the joint residuals on different joint sampling profiles, so that the formed joint boundary can adapt to the actual error distribution of the new and old models in different directions.

[0076] In one embodiment of the present invention, such as Figure 5 As shown, the joint positions on each joint sampling profile can be determined using the following method: Step S401: Perform convergence analysis along the residual sequences corresponding to the sampling profiles of each edge to determine the residual convergence interval that meets the preset convergence conditions.

[0077] For each edge sampling profile, the edge residual sequence corresponding to that profile is analyzed sequentially from one side of the outline of the area to be updated towards the outside of the candidate edge band. Since the area near the area to be updated is usually still affected by actual changes, local reconstruction differences, or model surface inconsistencies, the edge residuals in this region typically exhibit characteristics such as large amplitude, significant fluctuations, or unstable correspondences. As the sampling position gradually moves away from the area to be updated, if the new local 3D model and the baseline 3D model gradually enter a region where stable edge joining is possible, the amplitude, dispersion, and spatial variation of the edge residuals gradually decrease and remain stable within a certain range.

[0078] This invention determines the residual convergence interval by identifying the interval where the residual transitions from an unstable state to a continuous stable state. The residual convergence interval indicates that, along the corresponding edge sampling profile direction, the local surface difference between the old and new models has been reduced to an acceptable range, and this stable state has spatial continuity, thus making it suitable as a candidate range for the edge location.

[0079] In one embodiment of the present invention, such as Figure 6 As shown, step S401 can be implemented in the following way: Step S4011: Perform sliding window statistical analysis on the edge residual sequence of each edge sampling profile.

[0080] Multiple sliding windows are set along the residual sequence corresponding to each edge sampling profile. Each sliding window covers multiple sampling positions arranged consecutively on the edge sampling profile. The sliding windows are moved sequentially with a preset step size in the direction from the contour line of the area to be updated to the outside of the candidate edge band, thereby obtaining multiple sliding windows corresponding to different sequence intervals.

[0081] The window length and movement step size of the sliding window can be set according to the spatial resolution of the 3D model, the sampling interval, the nominal accuracy of the model, and the complexity of the local structure. The window length should be able to cover a sufficient number of valid sampling positions to avoid the random deviation of a single sampling position having an excessive impact on the convergence judgment. The movement step size can be smaller than the window length, so that adjacent windows partially overlap, thereby continuously reflecting the change process of the edge residual along the edge sampling profile direction.

[0082] For each sliding window, the edge residual statistics are calculated based on the edge residuals at each valid sampling position within the window. The edge residual statistics may include at least the residual amplitude statistics, residual dispersion statistics, residual variation statistics, and valid sample statistics.

[0083] The residual amplitude statistic is used to characterize the overall surface deviation between the new-period local 3D model and the baseline 3D model within the window. The mean of the absolute values ​​of the edge residuals within the window can be used as the residual amplitude statistic, i.e.:

[0084] Where k represents the edge sampling profile number, and j represents the sliding window number. This represents the j-th sliding window on the k-th edge sampling profile. This represents the edge residual corresponding to the i-th sampling position on the k-th edge sampling profile. The smaller the absolute mean of the residual, the lower the average inconsistency between the old and new model surfaces within the corresponding sequence interval.

[0085] Residual dispersion statistics are used to characterize the degree of fluctuation in the marginal residuals within a window. The standard deviation, variance, median absolute deviation, or other dispersion statistics of the marginal residuals within the window can be used. When using the standard deviation, it can be expressed as:

[0086] The smaller the residual dispersion, the more concentrated the edge residuals are at each sampling position within the window, and the more stable the local surface correspondence between the old and new models. If the absolute mean of the residuals is low but the dispersion is large, there may be individual large deviations, surface fluctuations, or unstable correspondences. It is not advisable to determine the edge position solely based on the lower average value.

[0087] The residual change statistic is used to characterize whether the statistical results of the residuals at the junction of adjacent sliding windows are still changing significantly along the sampling profile direction. The residual change rate can be determined based on the difference between the absolute mean of the residuals of the current sliding window and the previous sliding window, for example:

[0088] in, This represents the spatial distance between corresponding positions of adjacent sliding windows. The smaller the residual change rate, the more gradual the change in the edge residual along the profile direction has become. If the residual amplitude has decreased but there is still a significant change between adjacent windows, it indicates that the edge residual may still be in the decay process and has not yet entered the stable edge region.

[0089] The effective sample statistic is used to characterize the proportion of sampling locations within a sliding window that can reliably establish a relationship between corresponding surface points. The effective sample proportion ρ(k,j) can be the ratio of the number of effective sampling locations to the total number of sampling locations within the window. When the model surface has holes, vegetation occlusion, sparse data, inconsistent surface normals, or candidate corresponding points exceed the maximum search distance, the corresponding sampling locations are marked as invalid and do not participate in the calculation of residual amplitude and residual dispersion. By setting an effective sample proportion constraint, it is possible to avoid misjudging the edge converging residuals when there are only a few accidental effective corresponding points.

[0090] In other implementations, statistical measures such as the median residual, maximum residual residual, residual quantile, degree of normal variation, or continuous effective sampling length can be used to supplement the description of the same edge residual sequence, depending on the actual 3D model type and accuracy requirements. It should be noted that the above statistical measures all originate from the same edge residual sequence formed in step S300 and are used to characterize the stable state of the edge residual sequence from different perspectives, rather than constituting a separate edge boundary judgment mechanism independent of the edge residuals.

[0091] Step S4012: Based on the statistical results of the edge residuals of the corresponding sequence intervals of each sliding window, determine whether each sequence interval meets the preset convergence condition, and determine the sequence intervals that meet the preset convergence condition continuously by no less than a preset number as the residual convergence intervals.

[0092] For each sliding window, the statistical results of the edge residuals of the corresponding sequence interval are compared with the preset convergence threshold to determine whether the sequence interval meets the preset convergence condition.

[0093] In one embodiment of this invention, the preset convergence conditions include: the residual amplitude statistic within the sliding window is not greater than a preset residual amplitude threshold, the residual dispersion statistic is not greater than a preset dispersion threshold, the residual change statistic is not greater than a preset change threshold, and the effective sample proportion is not less than a preset effective proportion threshold. That is, when a sliding window simultaneously satisfies the following conditions, the sequence interval corresponding to that sliding window is determined as a stable sequence interval satisfying the preset convergence conditions: μ(k,j)≤τμ; σ(k,j)≤τσ; g(k,j)≤τg; ρ(k,j)≥τρ; Where τμ represents the residual amplitude threshold, τσ represents the residual dispersion threshold, τg represents the residual variation threshold, and τρ represents the effective sample proportion threshold.

[0094] The residual amplitude threshold is used to limit the allowable average surface deviation between the old and new models; the residual dispersion threshold is used to limit the residual fluctuation within the sequence interval; the residual change threshold is used to determine whether the edge residual has changed from a state of continuous change to a relatively stable state; and the effective sample proportion threshold is used to ensure that the corresponding statistical results have sufficient data support.

[0095] Each convergence threshold can be set according to the nominal accuracy of the 3D model, data resolution, model reconstruction quality, and specific application scenarios. In one implementation, the edge residual samples of a known stable region outside the candidate edge band can be pre-selected, and the corresponding convergence threshold can be determined based on the mean, standard deviation, median, or median absolute deviation of the residuals within that stable region. This allows the preset convergence conditions to adapt to the actual noise level and accuracy of the current model, rather than being limited to fixed values.

[0096] To avoid a single sliding window being incorrectly identified as a stable confluence region due to occasional low residuals or a small number of outliers being excluded, this invention does not directly use a single sequence interval that meets the preset convergence conditions as the residual convergence interval. Instead, it further determines whether there are at least a preset number of continuous stable sequence intervals along the confluence sampling profile direction.

[0097] Specifically, when at least K sliding windows arranged consecutively along the same sampling profile all satisfy the preset convergence condition, the sequence range jointly covered by these consecutively arranged sliding windows is determined as the residual convergence interval, where K is the preset number of consecutive stable windows. By using continuous window constraints, the impact of localized accidental low residuals, single-point noise, or short-distance data gaps on convergence judgment can be reduced, ensuring that the determined residual convergence interval has sufficient spatial continuity and stability.

[0098] When a sliding window fails to meet any preset convergence condition, it is marked as an unstable window, and the cumulative counting of the current consecutive stable windows is interrupted. The process then restarts from subsequent sliding windows until a residual convergence interval satisfying the consecutive number requirement is identified within the search range of the candidate edge band, or the analysis of all sliding windows of the edge sampling profile is completed.

[0099] A single edge sampling profile may contain one or more residual convergence intervals. For example, due to local occlusion, model holes, or complex structures, the edge residual may temporarily decrease in a certain region, then increase again, and finally stabilize in a more outer region. In this case, the residual convergence intervals can be recorded in order from the inside to the outside along the edge sampling profile. This provides a basis for selecting the first residual convergence interval and for searching for the next residual convergence interval when there are local edge anomalies after the update.

[0100] When a sampling profile for a junction does not have a residual convergence interval that meets preset conditions within the maximum search range of the candidate junction zone, the sampling profile for a junction is marked as a non-converged profile. Non-convergence may be caused by factors such as consistently large differences between the old and new models within the search range, insufficient model overlap, low effectiveness of corresponding surface points, or continuous changes in local structure. For non-converged profiles, the outer boundary of the candidate junction zone is not directly determined as a valid junction location to avoid mistaking the artificial boundary of the search range for the actual stable junction location.

[0101] For a small number of scattered non-converged profiles, the corresponding edge locations to be checked can be determined by combining the residual convergence intervals of adjacent edge sampling profiles, edge locations, and boundary continuity requirements. For cases where multiple consecutive edge sampling profiles have not formed residual convergence intervals, the candidate edge zones of the corresponding local area can be expanded, nearby areas with stable surface features can be selected, or it can be indicated that the coverage of the new local 3D model in this area is insufficient and new model data needs to be added.

[0102] Step S402: For each edge sampling profile, determine the edge position within the first residual convergence interval outward from the contour line of the area to be updated.

[0103] For each edge sampling profile that has identified at least one residual convergence interval, the residual convergence intervals are sorted according to the direction from the outline of the area to be updated to the outside of the candidate edge band, and the first residual convergence interval encountered along this direction is selected as the target residual convergence interval.

[0104] Choosing the first residual convergence interval ensures that the surface differences between the old and new models meet the requirements for stable edge connection, while making the edge location as close as possible to the actual change area. This avoids including a large portion of the already stable and non-renewable baseline 3D model in the replacement area, reducing the local update range and the corresponding data processing volume.

[0105] Within the first residual convergence interval, the edge position on the corresponding edge sampling profile can be determined according to preset selection rules. The edge position can be selected as the starting position, the center position, or a position offset outward from the starting position by a preset safety distance within the first residual convergence interval.

[0106] In the embodiments disclosed in this invention, a position within the first residual convergence interval that is close to the contour line of the area to be updated and maintains a preset safety margin from the starting position of the interval can be selected as the edge position. This avoids the edge position being at a critical position where the convergence condition is just met but the stability is insufficient, and also prevents the edge position from expanding excessively outward. The edge positions determined on each edge sampling profile can also be called edge anchor points.

[0107] For non-converged profiles whose edge locations to be verified are determined based on adjacent edge sampling profiles, the edge locations to be verified should be located within areas effectively covered by both the new local 3D model and the baseline 3D model, and constrained by adjacent effective edge locations and boundary continuity. The edge locations to be verified can be further adjusted when forming closed edge boundaries and performing edge quality verification after updates.

[0108] After determining the edge positions on each edge sampling profile, the edge positions are sequentially connected according to the arrangement order of the corresponding contour points on the contour line of the area to be updated, thus obtaining the initial edge boundary. Since each edge position is determined by the residual convergence state on different edge sampling profiles, the distance from each edge position to the contour line of the area to be updated can be different, resulting in an irregular shape of the initial edge boundary that adapts to the actual spatial distribution of edge error.

[0109] The initial boundary is closed and corrected to form a closed boundary that surrounds the outline of the area to be updated. During the closure correction process, it should be ensured that the closed boundary completely surrounds the outline of the area to be updated, does not self-intersect, does not cross the uncovered areas of the reference period 3D model or the new period local 3D model, and that the boundary segments between adjacent boundary positions remain continuous.

[0110] When smoothing or correcting the curvature of the initial boundary, each boundary position is only allowed to move within the target residual convergence interval of the corresponding boundary sampling profile. This prevents the boundary position from being moved back to a high residual or unstable region in order to improve the geometric smoothness of the boundary. When the boundary continuity requirement conflicts with the innermost boundary position on a boundary sampling profile, a more outer position can be selected within the residual convergence interval corresponding to that boundary sampling profile, without moving beyond that residual convergence interval towards the area to be updated.

[0111] The resulting closed boundary is not a fixed-width outer boundary of the area to be updated, but is determined by the convergence state of the boundary residuals on each boundary sampling profile. In directions where the boundary residuals converge quickly, the closed boundary is relatively close to the outline of the area to be updated; in directions where the boundary residuals decay slowly, the surface structure is complex, or there are few effective corresponding samples, the closed boundary extends outward accordingly, thus matching the actual replacement range of the model with the boundary error distribution between the old and new models.

[0112] Step S500: Determine the residual constraint weights based on the residual decay process between the contour line to be updated and the corresponding edge position on each edge sampling profile.

[0113] This step converts the spatial decay process of the edge residuals along each edge sampling profile from the inside out into constraint weights used when adjusting the model surface. This ensures that both the edge transition range and the model adjustment intensity are determined by the actual surface differences between the new local 3D model and the baseline 3D model. Compared to setting linear transition weights at a fixed distance, this invention determines the transition starting point and residual constraint weights based on the decay rate of the edge residuals along different edge sampling profiles, thereby forming a non-uniform width edge transition zone adapted to the differences in the local model.

[0114] The edge residual decay process refers to the gradual decrease in the edge residual as it moves along the edge sampling profile from the outline of the area to be updated to the corresponding edge position. The residual initially exhibits large amplitude, significant fluctuations, or instability, eventually entering the residual convergence interval. This process can be characterized by the edge residual at each sampling position, the statistical value of the residual amplitude corresponding to the sliding window, the rate of change of residuals between adjacent windows, or the cumulative change in the edge residual.

[0115] In one embodiment of the present invention, such as Figure 7 As shown, step S500 can be implemented in the following manner: Step S501: For each edge sampling profile, determine the first position that meets the preset attenuation condition based on the corresponding edge residual sequence as the transition start point, and take the corresponding edge position as the transition end point.

[0116] For each edge sampling profile, the edge residual sequence is analyzed between the outline of the area to be updated and the edge position on the edge sampling profile to identify the position where the edge residual begins to change continuously from an unstable state to a residual convergence state.

[0117] Specifically, the edge residual status of each sampling position or sliding window can be checked sequentially from the outline of the area to be updated to the edge position along the edge sampling profile. When the edge residual after a certain position shows an overall downward trend, and remains downward or does not increase in the opposite direction exceeding a preset range within a continuous number of sampling positions or sliding windows, the position is determined as the position that meets the preset attenuation condition.

[0118] Preset attenuation conditions may include: the magnitude of the edge residual decreases relative to the previous sampling position or the previous sliding window; the overall edge residual decreases across multiple consecutive sampling positions or sliding windows; the direction of change of the edge residual is towards the corresponding residual convergence interval; or the decrease in the edge residual reaches a preset proportion. Preset attenuation conditions are used to identify the position where the edge residual transitions from a state of significant fluctuation to a state of continuous attenuation, rather than requiring that the edge residual at each adjacent sampling position decreases strictly monotonically.

[0119] In one implementation, a preset attenuation condition can be met if the absolute mean of the residuals in the current sliding window is less than the absolute mean of the residuals in the previous sliding window, and the absolute mean of the residuals in multiple consecutive subsequent sliding windows generally shows a decreasing trend. To reduce the impact of local noise or individual outliers on the identification of transition starting points, the edge residual sequence can be smoothed, or the residuals can be judged to have entered a continuous decreasing phase based on the fitting slope of consecutive windows.

[0120] If multiple locations satisfying the preset attenuation condition exist on the same edge sampling profile, the first location satisfying the preset attenuation condition encountered along the outline of the area to be updated towards the edge location is selected as the transition starting point. Selecting the first location ensures that the edge transition zone covers the entire range where the edge residual begins to attenuate significantly, avoiding excessively concentrated geometric changes caused by adjusting the model only in a short segment near the edge location.

[0121] In other implementations, the transition starting point can be determined by tracing back from the corresponding edge position towards the outline of the area to be updated, and identifying the position between the last obviously unstable window and its outer continuously decaying window. Regardless of whether an outward search or an inward tracing from the edge position is used, the determined transition starting point is used to characterize the spatial position where the edge residual begins to continuously tend towards stability.

[0122] The junction position determined in step S402 on the junction sampling profile is taken as the transition endpoint. The transition endpoint is located within the first residual convergence interval and is used to define the position where the model surface is gradually constrained by the baseline 3D model and finally continuously connects with the junction boundary.

[0123] Therefore, each edge sampling profile corresponds to a transition start point and a transition end point. Due to differences in model structure, data quality, and edge residual decay rates in different directions, the distance between the transition start point and the transition end point on each edge sampling profile can vary. In regions where the edge residual decays rapidly, the distance between the transition start point and the transition end point is shorter; in regions where the edge residual decays slowly or where the local structure is complex, the distance between the two increases accordingly.

[0124] When a sampling profile for a junction does not show a clear, continuous decay phase between the outline of the area to be updated and the junction position, the transition starting point can be determined based on the position of the last clearly unstable window in the junction residual sequence. Alternatively, the transition starting point to be reviewed can be determined by combining the transition starting points of adjacent sampling profiles and boundary continuity requirements. The determined transition starting point should be located between the outline of the area to be updated and the corresponding junction position, and should not cross the junction position.

[0125] Step S502: Connect the transition starting points on each edge sampling profile to form the transition inner boundary, and the transition inner boundary and the edge boundary enclose the edge transition area.

[0126] According to the arrangement order of the contour points corresponding to each edge sampling profile on the contour line of the area to be updated, the transition starting points of each edge sampling profile are connected sequentially to form the transition inner boundary. The transition inner boundary is located inside the closed edge boundary and is used to define the position where the new local 3D model begins to accept edge adjustment.

[0127] When connecting the transition points, the initial connection results can be smoothed, curvature corrected, or locally positioned to avoid excessive changes in the distance between adjacent transition points, which could result in jagged boundaries. During these adjustments, each transition point should be kept on its corresponding edge sampling profile, and the adjusted transition points should not cross the corresponding edge position or enter the main area to be updated, where the original structure of the new model needs to be preserved.

[0128] When the distance between transition starting points on adjacent edge sampling profiles is large, or when there are edge sampling profiles where transition starting points cannot be reliably identified in certain areas, interpolation can be performed based on adjacent valid transition starting points. This interpolation, combined with the local orientation of the contour line of the area to be updated, the edge of the model structure, or the corresponding edge position, determines the intermediate transition inner boundary. This interpolation ensures the continuity of the transition inner boundary without altering the transition range determined by the edge residual attenuation process on each valid edge sampling profile.

[0129] The transition zone is formed by the inner boundary of the transition and the closed boundary formed in step S400. The inner boundary of the transition is the inner boundary of the transition zone, and the closed boundary is the outer boundary of the transition zone.

[0130] The transition zone defines the spatial range within which the surface of the new local 3D model undergoes residual constraint adjustments. The model region within the inner boundary of the transition zone primarily retains the original geometry of the new local 3D model to preserve information about the areas where changes have actually occurred. The model surface within the transition zone gradually transitions to the corresponding surface of the baseline 3D model based on the residual constraint weights. The model surface at the closed transition boundary is subject to stronger constraints from the baseline model to ensure a continuous connection between the updated model and the baseline 3D model retained outside the closed transition boundary.

[0131] Since the distance between the transition starting point and the junction position on each edge sampling profile is determined by the respective edge residual decay process, the resulting edge transition zone has different widths along the closed edge boundary direction. In areas with gentle surface structures and rapid convergence of edge residuals, the edge transition zone is narrower; in areas with building corners, retaining walls, complex terrain, or where the difference between old and new models decays slowly, the edge transition zone is wider, thus forming non-uniform width edge transition zones.

[0132] By forming a non-uniform width transition zone, the adjustment range of the model surface can be matched with the spatial influence range of the actual edge residual, avoiding unnecessary model deformation introduced by the fixed width transition zone in the stable region, or obvious steps, cracks or normal abrupt changes in the complex region due to insufficient transition distance.

[0133] Step S503: For each sampling position in the transition zone, calculate the ratio of the cumulative residual from the sampling position to the corresponding edge position to the cumulative residual from the corresponding transition start point to the edge position, and determine the residual constraint weight corresponding to the sampling position based on the ratio.

[0134] For each edge sampling profile, the sampling positions located between the corresponding transition start point and the edge position are determined as sampling positions within the edge transition zone, and the residual constraint weights are calculated based on the cumulative edge residuals between each sampling position and the edge position.

[0135] For any sampling position on the k-th edge sampling profile, the absolute values ​​of the edge residuals of all valid sampling positions between that sampling position and the corresponding edge position can be accumulated to obtain the cumulative edge residual from that sampling position to the edge position. The absolute values ​​of the edge residuals of all valid sampling positions between the corresponding transition start point and the edge position are then accumulated to obtain the cumulative edge residual for the entire transition section.

[0136] The cumulative edge residual can be discretized and summed based on the absolute values ​​of the edge residuals at each sampling location, or it can be approximated by integrating the absolute values ​​of the edge residuals based on the sampling interval. For example, when the sampling intervals are the same, the absolute values ​​of the edge residuals within the corresponding intervals can be directly summed. When the sampling intervals are not exactly the same, the edge residuals can be weighted and accumulated based on the actual distance between adjacent sampling locations.

[0137] At the transition starting point, the cumulative edge residual from the sampling position to the edge position is equal to the cumulative edge residual of the entire transition section, so the corresponding ratio is close to or equal to 1. As the sampling position moves along the edge sampling profile towards the edge position, the cumulative edge residual in the remaining section gradually decreases, and the corresponding ratio gradually decreases. At the edge position, the remaining cumulative edge residual is close to or equal to 0, so the corresponding ratio is close to or equal to 0.

[0138] In this embodiment of the invention, the ratio can be directly determined as the retention weight of the new local 3D model. Thus, the residual constraint weight near the transition starting point is close to 1, and the model surface mainly maintains the geometric position of the new local 3D model; the residual constraint weight near the junction gradually approaches 0, and the model surface gradually converges to the corresponding surface of the baseline 3D model.

[0139] In other implementations, the residual constraint weights can be determined by normalizing, smoothing, or nonlinearly mapping the comparison values. For example, a monotonic function can be used to map the comparison values, ensuring that the weights remain continuous and monotonically non-increasing between the transition start point and the junction position. Nonlinear mapping can be set according to the smoothness of the model surface and the requirements for preserving local structure, but the mapped residual constraint weights should still be directly derived from the ratio of the cumulative amount of the junction residuals, rather than reusing a fixed distance function unrelated to the junction residuals.

[0140] When the edge residual decreases rapidly in the transition section, the cumulative amount of edge residual is mainly concentrated on the side near the transition starting point. The residual constraint weight will decrease rapidly over a short distance, allowing the model to complete the edge transition within a narrow range. When the edge residual decreases slowly, the cumulative amount of edge residual is distributed along a longer section. The residual constraint weight decreases accordingly and gradually allows the model to complete the surface adjustment within a wider range.

[0141] For model surface points or mesh vertices within the transition zone that are not directly located on the edge sampling profile, the residual constraint weights corresponding to adjacent sampling positions can be interpolated based on their spatial relationship with adjacent edge sampling profiles to obtain the residual constraint weights of the model surface point or mesh vertex. During interpolation, the interpolation ratio can be determined based on the distance from the model surface point to the adjacent edge sampling profile, its position along the closed edge boundary direction, or its triangular mesh relationship, so that the residual constraint weights continuously change between adjacent edge sampling profiles.

[0142] When certain sampling locations are marked as invalid sampling locations, interpolation can be performed based on the edge residuals of adjacent valid sampling locations on the same edge sampling profile, or the cumulative edge residual of the corresponding interval can be calculated using the residual statistics corresponding to the sliding window, thus avoiding weight interruption due to a small number of invalid sampling locations. If the continuous invalid segments exceed a preset length, the corresponding region can be marked as a region to be reviewed, and the model adjustment intensity of the region can be determined by combining the residual constraint weights of adjacent edge sampling profiles.

[0143] In the above manner, the spatial width of the edge transition zone is determined by the position where the edge residual begins to decay and the position where the edge residual converges. The model adjustment intensity at each position inside the edge transition zone is determined by the cumulative decay process of the edge residual. Thus, the determination of the edge boundary, the generation of the transition range, and the adjustment of the model surface are all based on the same edge residual sequence, ensuring that there is a continuous and unified constraint relationship between each technical step.

[0144] Step S600: Adjust the model surface of the new local 3D model at the junction boundary according to the residual constraint weight, and replace the reference period 3D model within the junction boundary with the adjusted new local 3D model.

[0145] This step is used to progressively adjust the surface of the new local 3D model within the transition zone of the junction according to the residual constraint weights determined in step S500, so that the new local 3D model gradually connects with the base period 3D model from the inner boundary of the transition to the junction boundary. After the surface adjustment is completed, the base period 3D model within the junction boundary is replaced with the adjusted new local 3D model, thereby obtaining a 3D model with completed local incremental update.

[0146] In this embodiment of the invention, the new-period local 3D model located within the transition inner boundary mainly retains its original geometry and is used to represent the new-period spatial state within the area to be updated. The new-period local 3D model located within the transition zone undergoes surface adjustments to varying degrees based on the residual constraint weights. The new-period local 3D model located at the boundary gradually matches the baseline 3D model retained outside the boundary. Thus, while maintaining the true change information of the new-period model, it is possible to reduce abrupt elevation changes, surface misalignments, cracks, and normal discontinuities at the junction of the old and new models.

[0147] In one embodiment of the present invention, such as Figure 8 As shown, step S600 can be implemented in the following manner: Step S601: Determine the adjustment amount of the corresponding points on the surface of the new local three-dimensional model based on the residual constraint weights corresponding to each sampling position in the transition zone.

[0148] For each sampling position within the transition zone, obtain the reference surface point on the reference period three-dimensional model established in step S302 and the corresponding surface point on the new period local three-dimensional model, and determine the candidate adjustment vector of the corresponding surface point based on the positional relationship between the reference surface point and the corresponding surface point.

[0149] In one implementation, the displacement in the opposite direction from the reference surface point to the corresponding surface point of the new local three-dimensional model can be determined as a candidate adjustment vector. That is, based on the difference vector between the new surface point and the reference surface point, the adjustment direction and the maximum adjustment amount that make the new surface point move towards the reference surface point can be determined.

[0150] For example, the surface points corresponding to the new local 3D model can be adjusted along the surface normal direction at the reference surface point to reduce the impact of tangential movement on the model's planar position, structural edges, and triangular mesh connections. For the i-th sampling position on the k-th edge sampling profile, its candidate normal adjustment amount can be determined based on the original edge deviation or edge residual corresponding to that sampling position.

[0151] When the residual constraint weights determined in step S503 represent the retention weights of the new local 3D model, the degree to which the corresponding surface points move towards the surface of the baseline 3D model can be determined based on the complementary amounts of the residual constraint weights. For example, the adjustment amount of the corresponding points on the surface of the new local 3D model can be expressed as:

[0152] in, This represents the adjustment amount of the new surface point corresponding to the i-th sampling position on the k-th edge sampling profile. This indicates the residual constraint weight corresponding to the sampling location. This represents the reference surface point on the 3D model during the reference period. This represents the corresponding surface point on the new local 3D model.

[0153] When using the normal adjustment method, the difference vector between the reference surface point and the new surface point can be projected onto the surface normal direction of the reference surface point, and then the normal adjustment amount can be determined according to the residual constraint weight.

[0154] Therefore, at the transition starting point, the residual constraint weight is close to 1, corresponding to a small or zero adjustment amount, ensuring that the new local 3D model at this location basically maintains its original geometry. As the sampling position gradually approaches the junction, the residual constraint weight gradually decreases, corresponding to a gradually increasing adjustment amount. At the junction, the surface of the new local 3D model is strongly constrained by the surface of the baseline 3D model, thus achieving a continuous connection with the baseline 3D model outside the junction boundary.

[0155] It should be noted that the adjustment amount does not require all new period surface points in the transition zone to be completely moved to the surface of the baseline period 3D model. Instead, the candidate adjustment vector is scaled according to the residual constraint weights so that the new period local 3D model gradually transitions to the baseline period 3D model in the transition zone.

[0156] To avoid excessive deformation of the model surface caused by local anomalies, a maximum adjustment threshold can be set for the adjustment amount of each corresponding surface point. When the adjustment amount calculated based on the corresponding surface point pair exceeds the maximum adjustment threshold, the adjustment amount can be limited to the maximum allowable range, or the corresponding sampling position can be marked as a position to be verified. The maximum adjustment threshold can be determined based on the model's nominal accuracy, local structural scale, and allowable edge joint error range.

[0157] For the regions marked as invalid sampling locations in step S302, interpolation can be performed based on the adjustment amounts corresponding to adjacent valid sampling locations on the same edge sampling profile, or estimation can be performed based on valid adjustment amounts that are spatially close on adjacent edge sampling profiles. If consecutive invalid regions exceed a preset range, the adjustment intensity of the region can be reduced, or the region can be reserved for subsequent edge quality verification steps.

[0158] Step S602: Adjust the surface of the new local three-dimensional model in the transition zone of the interface according to the adjustment amount of the corresponding points of each model surface.

[0159] Based on the adjustment amount of each corresponding surface point determined in step S601, the surface of the new local three-dimensional model in the transition zone of the joint edge is deformed and adjusted so that the model surface transitions continuously from the inner boundary of the transition to the joint edge boundary.

[0160] When the new local 3D model is represented by a triangular mesh model, the edge sampling positions and corresponding surface points may be located inside the triangular faces, and not necessarily coincide with the vertices of the triangular mesh. Therefore, the adjustment amount of the corresponding surface points can be transferred to the relevant triangular mesh vertices based on the centroid coordinates of the triangular faces where the corresponding surface points are located, the distance between adjacent vertices, or the local mesh topology.

[0161] In one implementation, when a corresponding surface point is located inside a triangular face, the adjustment amount of the corresponding surface point can be allocated to the three vertices based on the centroid coordinates of the corresponding surface point relative to the three vertices of the triangular face. For a model vertex that is simultaneously affected by multiple sampling positions, the adjustment amounts transmitted from multiple sampling positions can be weighted and fused to obtain the final adjustment amount of the model vertex.

[0162] Weighted fusion can be performed based on the distance from the model vertex to each sampling position, the residual constraint weights corresponding to each sampling position, the reliability of the relationship between corresponding surface points, and the consistency of local normals. Sampling positions that are closer to the model vertex and have a higher reliability of the correspondence can have a larger influence weight.

[0163] For model vertices in the transition zone that are not directly located near the edge sampling profile, the adjustment amount of the sampling position on the adjacent profile or the corresponding point on the model surface can be interpolated based on their spatial position between adjacent edge sampling profiles to obtain the adjustment amount of the corresponding model vertex. By performing continuous interpolation between profiles, the model adjustment amount can be kept continuous along the edge boundary direction, avoiding strip-like abrupt changes between adjacent edge sampling profiles.

[0164] When adjusting the surface of a new local 3D model, local continuity constraints can be applied simultaneously to limit the variation in adjustment between adjacent model vertices. Local continuity constraints can be set based on the difference in adjustment between adjacent vertices, the change in the normal of the triangular face, or the change in local curvature, thereby preventing individual model vertices from forming spikes, pits, or flipped triangular faces due to abnormal adjustment.

[0165] For example, based on satisfying the residual constraint weights at each sampling location, the adjustment amount of the model vertices is locally smoothed. The smoothing process is only used to eliminate discontinuous changes between adjacent vertices and does not change the overall adjustment trend determined by the edge residual decay process.

[0166] When performing local smoothing, the inner regions with larger residual constraint weights can be set as stronger new-period model preservation constraints, while the outer regions near the boundary can be set as stronger baseline model matching constraints. This avoids the smoothing process spreading the effects of adjustments outside the boundary to the area to be updated, thus preserving the true change structure expressed by the new-period local 3D model.

[0167] For structural areas with significant geometric abrupt changes, such as building edges, road edges, retaining walls, steps, or other areas with obvious geometric changes, the adjustment amount can be limited to propagate across the structural edge by combining model structural lines or local normal differences, preventing incorrect smoothing of real structural abrupt changes. For the same continuous surface area, the adjustment amount can be allowed to propagate continuously between adjacent model vertices to improve the smoothness of edge transitions.

[0168] After adjusting the model surface, the vertex positions, triangle normals, triangle side lengths, and local topological state within the transition zone can be recalculated. When there is triangle degradation, triangle flipping, model self-intersection, or local curvature exceeding a preset range, the adjustment amount of the corresponding model vertices can be limited or redistributed to ensure that the adjusted new local 3D model has an effective geometric topological structure.

[0169] When a new local 3D model is represented using point clouds, digital surface models, or other surface data structures, adjustments can be applied based on the corresponding data structure. For example, for point cloud models, the spatial positions of points within the transition zone can be directly adjusted, and neighborhood interpolation can be performed on points without direct adjustments. For regular grid models, the grid elevation can be adjusted based on the residual constraint weights corresponding to each grid position, and continuity constraints can be applied to adjacent grids.

[0170] Through the above surface adjustments, the new local 3D model within the transition inner boundary maintains its original geometric state, and the surface of the new model in the transition zone gradually transitions to the surface of the reference 3D model. The surface of the new model at the boundary connects continuously with the outer reference 3D model.

[0171] Step S603: Replace the reference period 3D model within the boundary with the adjusted new period local 3D model, and retain the reference period 3D model outside the boundary to obtain the updated 3D model.

[0172] After completing the surface adjustment of the new local 3D model in the transition zone, the closed boundary determined in step S400 is used as the model replacement range. The model data located within the closed boundary in the reference 3D model is removed or marked as data to be replaced, and the model data located within the closed boundary in the adjusted new local 3D model is written to the corresponding position.

[0173] At the same time, the baseline period 3D model outside the closed boundary is retained, so that the updated 3D model is composed of the adjusted new period local 3D model inside the boundary and the baseline period 3D model outside the boundary.

[0174] Before replacing the model, the baseline 3D model and the adjusted new local 3D model can be trimmed based on the closed boundary. For the triangular faces intersecting the closed boundary, the intersecting triangular faces can be divided according to the closed boundary, so that the trimmed new local 3D model and the retained baseline 3D model have the same or a mapping relationship between their boundary vertices.

[0175] When the new local 3D model and the baseline 3D model have different mesh densities or triangulation structures at their boundary, local retriangulation can be performed on the model edges near the boundary to connect the boundary vertices of the new local 3D model with the retained boundary vertices of the baseline 3D model. Local retriangulation should be limited to the vicinity of the boundary, without altering the already confirmed main structure of the new model within the boundary, nor the main structure of the retained baseline model outside the boundary.

[0176] In one implementation, the boundary point sequences at the junction of the adjusted new local 3D model and the baseline 3D model can be extracted separately. A correspondence between the boundary points is established according to the junction sampling profile number or the arc length order along the closed junction boundary, and connecting triangular faces are generated based on this correspondence. For cases with different numbers of boundary points, the connection can be achieved by adding boundary interpolation points, merging boundary points that are too close together, or using locally constrained triangulation.

[0177] During model replacement, it should be ensured that there are no duplicate surfaces, overlapping triangles, holes, or unconnected edges at the boundary. For small local holes formed due to model trimming and retriangulation, surface patching can be performed based on the surrounding model vertices and surface normals.

[0178] When the 3D model also includes textures, colors, semantic tags, component attributes, or other auxiliary information, the corresponding attributes can be updated synchronously. Within the boundary, the texture or attribute information carried by the new local 3D model is used, while outside the boundary, the texture or attribute information of the baseline 3D model is retained. For the transition zone, the old and new textures, colors, or attributes can be progressively blended according to the residual constraint weights or the correspondence of the model surface.

[0179] For example, for triangular mesh models with texture coordinates, after completing geometric adjustments and local retriangulation, the texture coordinates of the affected triangular faces can be recalculated, and the texture source can be determined based on the new and baseline images at the corresponding locations. In cases where there are significant brightness or color differences near the edge boundaries, brightness correction or color gradation can be performed in the edge transition zone to reduce the impact of texture seams on the update results.

[0180] After obtaining the updated 3D model, the contour lines of the area to be updated, candidate edge bands, edge sampling profiles, transition inner boundaries, closed edge boundaries, residual constraint weights, and model replacement range used in this local update can be recorded for subsequent edge quality verification, local anomaly tracking, and incremental updates.

[0181] Therefore, this invention utilizes the same edge residual sequence to sequentially complete the edge boundary determination, edge transition zone construction, residual constraint weight generation, and model surface adjustment, enabling the new local 3D model to maintain real change information while smoothly and continuously merging with the baseline 3D model at the edge boundary, and avoiding edge steps, cracks, and local surface distortions caused by using fixed transition distances or simple boundary splicing.

[0182] Step S700: Redetermine the edge joint residuals for the updated 3D model, and make local edge joint adjustments for abnormal edge joint sections based on the edge joint residuals.

[0183] This step is used to verify the edge quality of the 3D model after the local update. Since processes such as model surface adjustment, local mesh reconstruction, model trimming, and boundary stitching may cause changes in the edge state of some areas, this embodiment of the invention does not directly output the final update result after completing the model update. Instead, it re-establishes the model correspondence within the edge transition zone and re-evaluates the updated edge quality. When a large edge residual is found in a local area, or when the edge residual has not reached a stable convergence state, only the edge position and local model of the abnormal area are readjusted, without re-executing the entire local update process, thereby further improving the model fusion quality and local update efficiency.

[0184] In one embodiment of the present invention, such as Figure 9 As shown, step S700 can be implemented in the following manner: Step S701: Re-determine the corresponding surface points of the baseline 3D model and the updated 3D model within the transition zone.

[0185] For the updated 3D model obtained in step 603, the surface correspondence between the baseline 3D model and the updated 3D model is re-established in the transition zone.

[0186] In this embodiment of the invention, the edge sampling profile established in step 302 is still used as a unified spatial reference. The corresponding relationship is re-established on each edge sampling profile according to the original sampling position to ensure that the edge residuals before and after the update have consistent spatial position and statistical basis.

[0187] Specifically, for each sampling location, a corresponding reference surface point is re-determined on the surface of the baseline 3D model, and the surface normal direction at the corresponding reference surface point is obtained. Then, along this surface normal direction, the corresponding surface point is searched again on the updated 3D model surface to form a new pair of corresponding surface points.

[0188] Since the updated 3D model has completed local model replacement and edge transition area surface adjustment, the newly established correspondence can truly reflect the actual edge state after the model update, rather than the difference between the old and new models before the update.

[0189] When the local mesh structure of the 3D model changes after the update, it is not required that the corresponding surface points maintain the same triangular mesh topology as before the update. It is only required that the corresponding surface points are located on the model surface corresponding to the same sampling position, so as to ensure that a unified edge quality evaluation can still be performed under different mesh structures.

[0190] If a corresponding sampling location cannot form a valid corresponding surface point due to local holes, occlusion, missing models, or the inability to establish a reliable correspondence, the corresponding sampling location will be marked as an invalid sampling location and treated as an invalid sample in the subsequent edge residual statistics process.

[0191] Step S702: Based on the positional deviation between the corresponding surface points that have been redefined, determine the edge residual sequence corresponding to each edge sampling profile in the edge transition zone, and perform convergence analysis on the edge residual sequence.

[0192] Based on the newly established corresponding surface point pairs, calculate the positional deviation between the corresponding surface points of the updated 3D model and the corresponding reference surface points of the baseline 3D model, and determine the edge residuals of the corresponding sampling positions.

[0193] In this embodiment of the invention, the method of steps S303 to S305 is still followed to separate the systematic deviation of the newly obtained position deviation and form the edge residual sequence corresponding to each edge sampling profile, so that the updated edge residual and the edge residual formed in step S300 are calculated in the same way.

[0194] Subsequently, the newly formed edge residual sequence is subjected to sliding window statistical analysis again. Following the convergence analysis method used in steps S401 and S402, the residual amplitude statistics, residual discrete statistics, residual change statistics, and effective sample ratio corresponding to each sliding window are calculated to determine whether each sequence interval meets the preset convergence conditions.

[0195] Since the convergence analysis method is consistent with that in step S400, the changes in the edge residuals on each edge sampling profile before and after the model update can be directly compared. When the overall edge residuals after the update further decrease and the position of the residual convergence interval remains stable, it indicates that the edge adjustment has achieved the expected effect; when a local area still cannot form a stable convergence interval or the convergence interval moves significantly outward, it indicates that there are still edge anomalies in the corresponding area, and further local correction is required.

[0196] Step S703: The continuous segments in the edge residual sequence that do not meet the preset convergence conditions are regarded as edge abnormal segments.

[0197] For each edge sampling profile, the edge status corresponding to each sliding window or sampling position is judged based on the convergence analysis results obtained again in step 702.

[0198] When the statistical results of the edge residuals in a certain continuous region fail to meet the preset convergence conditions, that is, when the residual amplitude is continuously greater than the preset residual amplitude threshold, the residual dispersion is continuously greater than the allowable range, the residual change is still in a state of significant fluctuation, or the proportion of effective samples is continuously insufficient, the corresponding continuous segment is identified as an edge abnormal segment.

[0199] In this embodiment of the invention, an abnormal segment is determined only when no less than a preset number of sampling positions or sliding windows fail to meet the preset convergence condition, so as to reduce the possibility of misjudgment caused by a single abnormal sampling point, local noise or random error.

[0200] Abnormal junction sections can include areas with abnormal junction quality caused by insufficient model surface adjustment, local model reconstruction errors, boundary retriangulation, local geometric changes caused by texture fusion, model clipping errors, and local mesh deformation.

[0201] For multiple adjacent or overlapping abnormal border sections, they can be merged according to spatial continuity to form a unified abnormal processing area, thereby reducing the number of repeated adjustments.

[0202] When all edge sampling profiles form a stable residual convergence interval and there are no abnormal edge segments, the updated 3D model is considered to meet the edge quality requirements, and the local model update is completed.

[0203] Step S704: For each abnormal edge segment, find the first residual convergence interval outward and redetermine the edge position, and use the edge position to readjust the local edge of the 3D model located in the abnormal segment.

[0204] For each abnormal junction segment, based on the junction sampling profile where the abnormal segment is located, continue searching for the updated junction residual sequence in the direction outward from the outline of the area to be updated.

[0205] When a residual convergence interval that meets the preset convergence conditions reappears outside the abnormal section, this residual convergence interval is determined as the new target convergence interval, and a new edge position is redefined within this target convergence interval.

[0206] The method for redetermining the edge location is consistent with step S402, that is, the first residual convergence interval encountered from the abnormal section outward is selected as the new edge location, and a preset safety distance is appropriately reserved to ensure that the new edge location is located inside the stable region.

[0207] When multiple adjacent edge sampling profiles have abnormal edge sections, the corresponding edge positions can be redefined simultaneously, and the local closed edge boundary and the corresponding transition inner boundary can be regenerated. Only the abnormal area needs to be re-established for the local edge transition area, without having to regenerate the edge boundary of the entire area to be updated.

[0208] Subsequently, based on the redefined junction position, the residual constraint weights corresponding to the abnormal section are recalculated, and following the methods in steps S601 and S602, only the local surface of the new local 3D model corresponding to the abnormal section is readjusted, so that the abnormal area gradually transitions to the new stable junction position.

[0209] After completing the local edge adjustment, steps S701 to S703 can be executed again to re-verify the edge quality of the abnormal area. If new abnormal edge segments still exist, the above local adjustment process can be repeated until the corresponding area forms a residual convergence interval that meets the preset convergence conditions, or the preset maximum number of adjustments is reached.

[0210] Since the edge positions are redefined and local adjustments are made only for the abnormal edge sections, while keeping the model of the rest of the area unchanged, the redundant calculations caused by the overall re-update can be avoided, the efficiency of local updates can be improved, and the geometric continuity, edge stability and overall fusion quality of the updated 3D model can be further improved.

[0211] Figure 10 This is a schematic diagram of a three-dimensional model local update edge optimization system provided in an embodiment of the present invention, as shown below. Figure 10 As shown, the system includes the following modules: Data acquisition module 1 is used to acquire the baseline period 3D model, the new period local 3D model, and the outline of the area to be updated.

[0212] Edge sampling profile setting module 2 is used to set multiple edge sampling profiles extending outward in the normal direction along the outline of the area to be updated.

[0213] The edge residual determination module 3 is used to determine the corresponding surface points of the baseline period 3D model and the new period local 3D model based on each edge sampling profile, determine the edge residual according to the deviation between the corresponding surface points, and form the edge residual sequence corresponding to each edge sampling profile.

[0214] The edge boundary formation module 4 is used to perform convergence analysis on the edge residual sequence, determine the edge position on each edge sampling profile, and connect each edge position to form a closed edge boundary surrounding the contour line to be updated.

[0215] The residual constraint module 5 is used to determine the residual constraint weight based on the residual decay process between the contour line to be updated and the corresponding edge position on each edge sampling profile.

[0216] Model update module 6 is used to adjust the model surface of the new local 3D model at the junction boundary according to the residual constraint weight, and replace the reference 3D model within the junction boundary with the adjusted new local 3D model.

[0217] The local edge adjustment module 7 is used to redetermine the edge residuals of the updated 3D model and perform local edge adjustment on the abnormal edge sections based on the edge residuals.

[0218] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for local update and edge optimization of a three-dimensional model, characterized in that, include: Obtain the baseline 3D model, the new local 3D model, and the outline of the area to be updated; Multiple outwardly extending edge sampling profiles are set along the outline of the area to be updated; Based on the sampling profiles of each edge, the corresponding surface points of the baseline period 3D model and the new period local 3D model are determined. The edge residuals are determined according to the deviation between the corresponding surface points, and the edge residual sequence corresponding to each sampling profile is formed. The edge residual sequence is subjected to convergence analysis to determine the edge position on each edge sampling profile, and the edge positions are connected to form a closed edge boundary surrounding the contour line to be updated. The residual constraint weights are determined based on the residual decay process between the contour line to be updated and the corresponding edge position on each edge sampling profile, including: For each edge sampling profile, the first position that meets the preset attenuation condition is determined according to the corresponding edge residual sequence as the transition start point, and the corresponding edge position is taken as the transition end point. Connecting the transition starting points on each edge sampling profile line forms an inner transition boundary, and the inner transition boundary and the edge boundary enclose an edge transition zone. For each sampling position within the transition zone, calculate the ratio of the cumulative residual from the sampling position to the corresponding edge position to the cumulative residual from the corresponding transition start point to the edge position, and determine the residual constraint weight corresponding to the sampling position based on the ratio. The model surface of the new local 3D model at the junction boundary is adjusted according to the residual constraint weight, and the adjusted new local 3D model replaces the reference period 3D model within the junction boundary. The edge joint residuals are re-determined for the updated 3D model, and local edge joint adjustments are made to abnormal sections based on the edge joint residuals, including: Within the transition zone, the corresponding surface points of the baseline 3D model and the updated 3D model are redefined; Based on the positional deviation between the corresponding surface points that have been redefined, the residual sequence of each edge sampling profile in the edge transition zone is determined, and the edge residual sequence is subjected to convergence analysis. Continuous segments in the edge-joining residual sequence that do not meet the preset convergence conditions are regarded as edge-joining abnormal segments. For each abnormal edge segment, the first residual convergence interval is found outward and the edge position is redefined. The edge position is then used to re-adjust the local edge of the 3D model located in the abnormal segment.

2. The method for local update and edge optimization of a three-dimensional model according to claim 1, characterized in that, The provision of multiple outwardly extending edge sampling profiles along the contour line of the area to be updated includes: Candidate edge bands are constructed outward from the outline of the area to be updated; Discretize the contour line of the area to be updated to obtain multiple contour points; Starting from each of the aforementioned contour points, multiple edge sampling profiles are formed by extending along the outward normal direction at the corresponding contour point within the candidate edge band.

3. The method for local update and edge optimization of a three-dimensional model according to claim 2, characterized in that, The determination of corresponding surface points between the baseline 3D model and the new local 3D model based on each edge sampling profile includes: Multiple sampling positions are set along the sampling profile of each of the aforementioned joint edges at preset sampling intervals; Based on each sampling location, a reference surface point is determined on the reference period 3D model. The surface normal direction at the reference surface point is obtained, and the corresponding surface point is searched along the surface normal direction on the surface of the new period local 3D model to form a corresponding surface point pair.

4. The method for local update and edge optimization of a three-dimensional model according to claim 3, characterized in that, The step of determining the joint residual based on the deviation between corresponding surface points and forming a joint residual sequence corresponding to each joint sampling profile includes: The original edge deviation corresponding to each sampling position is determined based on the positional deviation between each corresponding surface point pair. The system deviation component corresponding to the sampling profile is determined based on the original edge deviation of multiple sampling positions on the same edge sampling profile. The edge residuals of each sampling position are determined based on the original edge deviation and the corresponding system deviation component, and the edge residual sequence corresponding to each edge sampling profile is formed according to the arrangement order of each sampling position on the corresponding edge sampling profile.

5. The method for local update and edge optimization of a three-dimensional model according to claim 1, characterized in that, The convergence analysis of the edge residual sequence to determine the edge position on each edge sampling profile includes: Convergence analysis is performed along the residual sequences corresponding to the sampling profiles of each edge to determine the residual convergence interval that meets the preset convergence conditions. For each edge sampling profile, the edge position is determined within the first residual convergence interval outward from the outline of the area to be updated.

6. The method for local update and edge optimization of a three-dimensional model according to claim 5, characterized in that, The convergence analysis is performed on the residual sequences corresponding to the sampling profiles along each of the aforementioned edges to determine the residual convergence intervals that satisfy the preset convergence conditions, including: Sliding window statistical analysis was performed on the edge residual sequence of each edge sampling profile; Based on the statistical results of the edge residuals of the corresponding sequence intervals of each sliding window, it is determined whether each sequence interval meets the preset convergence condition, and the sequence intervals that meet the preset convergence condition consecutively for no less than a preset number are determined as the residual convergence intervals.

7. The method for local update and edge optimization of a three-dimensional model according to claim 1, characterized in that, The step of adjusting the model surface of the new local 3D model at the junction boundary according to the residual constraint weight, and replacing the reference period 3D model within the junction boundary with the adjusted new local 3D model, includes: Based on the residual constraint weights corresponding to each sampling position within the transition zone, determine the adjustment amount of the corresponding points on the surface of the new local 3D model; Based on the adjustment amount of the corresponding points on each model surface, the surface of the new local 3D model in the transition zone of the dock edge is adjusted; The updated 3D model is obtained by replacing the reference period 3D model within the boundary with the adjusted new local 3D model, while retaining the reference period 3D model outside the boundary.

8. A local update edge optimization system for a three-dimensional model, characterized in that, include: The data acquisition module is configured to acquire the baseline 3D model, the new local 3D model, and the outline of the area to be updated; The edge sampling profile setting module is configured to set multiple edge sampling profiles extending outward in the normal direction along the outline of the area to be updated; The edge residual determination module is configured to determine the corresponding surface points of the baseline period 3D model and the new period local 3D model based on each edge sampling profile, determine the edge residual according to the deviation between the corresponding surface points, and form the edge residual sequence corresponding to each edge sampling profile. The edge boundary forming module is configured to perform convergence analysis on the edge residual sequence, determine the edge position on each edge sampling profile, and connect each edge position to form a closed edge boundary surrounding the contour line to be updated. The residual constraint module is configured to determine the residual constraint weights based on the residual decay process between the contour line to be updated and the corresponding edge position on each edge sampling profile, including: For each edge sampling profile, the first position that meets the preset attenuation condition is determined according to the corresponding edge residual sequence as the transition start point, and the corresponding edge position is taken as the transition end point. Connecting the transition starting points on each edge sampling profile line forms an inner transition boundary, and the inner transition boundary and the edge boundary enclose an edge transition zone. For each sampling position within the transition zone, calculate the ratio of the cumulative residual from the sampling position to the corresponding edge position to the cumulative residual from the corresponding transition start point to the edge position, and determine the residual constraint weight corresponding to the sampling position based on the ratio. The model update module is configured to adjust the model surface of the new local 3D model at the junction boundary according to the residual constraint weight, and replace the reference 3D model within the junction boundary with the adjusted new local 3D model. The local edge adjustment module is configured to redetermine the edge residuals of the updated 3D model and perform local edge adjustment on abnormal edge sections based on the edge residuals, including: Within the transition zone, the corresponding surface points of the baseline 3D model and the updated 3D model are redefined; Based on the positional deviation between the corresponding surface points that have been redefined, the residual sequence of each edge sampling profile in the edge transition zone is determined, and the edge residual sequence is subjected to convergence analysis. Continuous segments in the edge-joining residual sequence that do not meet the preset convergence conditions are regarded as edge-joining abnormal segments. For each abnormal edge segment, the first residual convergence interval is found outward and the edge position is redefined. The edge position is then used to re-adjust the local edge of the 3D model located in the abnormal segment.

Images