Real-time three-dimensional reconstruction method for underwater topography of hydraulic filling area based on telemetry data

Abstract

The present application relates to the technical field of three-dimensional reconstruction, and specifically discloses a real-time three-dimensional reconstruction method for underwater topography in a hydraulic filling area based on telemetry data, which comprises: acquiring multi-source time-series telemetry point cloud data and performing adaptive filtering, respectively outputting an effective skeleton point set with a first density and an effective fine point set with a second density, and the first density being less than the second density. After receiving the telemetry point cloud data, the technical scheme of the present application completes double-dimensional fine partitioning of the topography grid by calculating the space-time joint topological gradient of the grid unit and combining the spatial structure features and the time-series activity. In actual construction operations, the system can distinguish the topography area with characteristic changes and the topography area maintaining gentle according to the above, and dispatches the matched data processing logic according to different areas.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional reconstruction technology, and in particular to a method for real-time three-dimensional reconstruction of underwater topography in reclamation areas based on telemetry data. Background Technology

[0002] In the 3D reconstruction of underwater topography in reclamation areas, underwater point cloud data is typically acquired using telemetry equipment such as multibeam echo sounders. Then, graphical algorithms such as triangulation are used to convert the point cloud into a surface mesh model for construction monitoring. Most existing reconstruction methods employ a globally unified mesh processing mechanism, meaning that upon receiving time-series point cloud data, the entire topology mesh is synchronously reconstructed or processed using a unified downsampling algorithm.

[0003] In actual reclamation construction scenarios, underwater topography exhibits spatial heterogeneity and temporal evolution differences. Areas near the silt removal site show abrupt topographic changes and high-frequency evolution characteristics due to continuous sediment accumulation or erosion; while areas far from the site have relatively flat terrain and tend towards solidification and stability. However, existing global unified grid processing mechanisms fail to differentiate calculations based on the evolutionary state of different regions. This leads to unnecessary grid topology update calculations being performed on stable regions when processing time-series input data, resulting in computational resource consumption.

[0004] Meanwhile, due to the limited global computing power distribution in abrupt regions, it is difficult to generate high-density local meshes specifically for these areas. This approach struggles to balance reconstruction efficiency with local mesh morphology characteristics, easily leading to technical issues such as data processing delays and unclear representation of terrain boundaries in specific construction areas. Summary of the Invention

[0005] This invention aims to at least partially address one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a real-time three-dimensional reconstruction method for underwater topography in reclamation areas based on telemetry data, in order to improve the computational scheduling efficiency during the underwater topography model generation process.

[0006] To achieve the above objectives, a first aspect of the present invention proposes a method for real-time three-dimensional reconstruction of underwater topography in reclamation areas based on telemetry data, comprising:

[0007] Acquire multi-source time-series telemetry point cloud data and perform adaptive filtering to output an effective skeleton point set with a first density and an effective detail point set with a second density, wherein the first density is less than the second density.

[0008] Based on the first frame of valid point set in the multi-source time-series telemetry point cloud data, a basic macroscopic three-dimensional mesh surface covering the entire domain is generated, and initial spatiotemporal attributes are assigned to each mesh unit;

[0009] The spatiotemporal joint topological gradient of each grid cell is calculated. Based on the spatial structural characteristics, the terrain is divided into multiple structural types, including smooth and uniform regions, abrupt change regions, and stable regions. Based on the temporal activity, it is divided into life cycle states, including basic state, evolutionary state, and solidified state, thus completing the two-dimensional fine partitioning.

[0010] For the grid cells in the smooth and uniform region of the base state, perform adaptive temporal elevation updates while maintaining the grid topology connectivity.

[0011] For the mesh cells in the mutation region of the evolutionary state, perform local mesh topology update to generate high-density local patches;

[0012] For the mesh cells in the stable region that are in the solidified state, a half-side folding algorithm is used to perform progressive mesh simplification, reduce the mesh density and fall back to the basic mesh resolution;

[0013] All high-density local patches are topologically stitched together with the simplified mesh and the base mesh to obtain a topologically complete global 3D surface mesh, which is then rendered in real time.

[0014] To achieve the above objectives, a second aspect of the present invention proposes a real-time three-dimensional reconstruction system for underwater topography in reclamation areas based on telemetry data, comprising:

[0015] The data acquisition and filtering module is used to acquire multi-source time-series telemetry point cloud data and perform adaptive filtering, and output an effective skeleton point set with a first density and an effective detail point set with a second density, wherein the first density is less than the second density.

[0016] The basic grid initialization module is used to generate a basic macroscopic three-dimensional grid surface covering the entire domain based on the first frame of valid points in the multi-source time-series telemetry point cloud data, and to assign initial spatiotemporal attributes to each grid cell;

[0017] The dual-dimensional fine partitioning module is used to calculate the spatiotemporal joint topological gradient of each grid cell. Based on spatial structural features, the terrain is divided into multiple structural types, including smooth and uniform regions, abrupt change regions, and stable regions. Based on temporal activity, it is divided into life cycle states, including basic state, evolutionary state, and solidified state, thus completing the dual-dimensional fine partitioning.

[0018] The base state update module is used to perform adaptive temporal elevation updates on the grid cells in the smooth and uniform region of the base state, while maintaining the grid topology connection relationship unchanged.

[0019] The evolutionary state update module is used to perform local mesh topology updates on mesh cells within the mutation region of the evolutionary state to generate high-density local patches;

[0020] The solidified state simplification module is used to perform progressive mesh simplification on the mesh cells in the stable region of the solidified state using a half-side folding algorithm, thereby reducing the mesh density and falling back to the basic mesh resolution;

[0021] The Topology Stitching and Rendering Output module is used to topologically stitch all high-density local patches with the simplified mesh and the base mesh to obtain a topologically complete global 3D surface mesh and perform real-time rendering output.

[0022] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described method for real-time three-dimensional reconstruction of underwater topography in reclamation areas based on telemetry data.

[0023] The method in this embodiment of the invention, after receiving telemetry point cloud data, calculates the spatiotemporal joint topological gradient of grid cells and combines spatial structural features and temporal activity to complete a two-dimensional fine partitioning of the terrain grid. In actual construction operations, the system can distinguish between terrain areas with characteristic changes and those that remain flat, and schedule matching data processing logic for different areas. For flat and uniform areas, only the elevation is updated while maintaining the original grid topology; for areas with abrupt changes in evolution, local topology updates are performed to preserve high-density construction details; for stable areas in a solidified state, progressive grid simplification is performed to reduce data density; and finally, global topology stitching is performed.

[0024] This mechanism allocates differentiated computing resources based on the physical stage of the terrain, reduces repetitive topological subdivision calculations for regions without state changes, alleviates the data load of overall rendering of 3D surface meshes while maintaining the geometric accuracy of local mutation features, and improves the computational scheduling efficiency in the underwater terrain model generation process. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the real-time three-dimensional reconstruction method for underwater topography in reclamation areas based on telemetry data provided by the present invention.

[0026] Figure 2 This is a three-dimensional scatter plot of the adaptive filtering and dual-density separation effect of the original multi-source time-series telemetry point cloud data in the real-time three-dimensional reconstruction method of underwater topography in reclamation areas based on telemetry data provided by the present invention.

[0027] Figure 3 This invention provides a method for real-time three-dimensional reconstruction of underwater topography in reclamation areas based on telemetry data, which includes a dual-dimensional fine-grained pseudo-color heat map of underwater topography based on spatial structural features and temporal activity.

[0028] Figure 4This is a schematic diagram of the incremental vertex insertion and constrained Delaunay triangulation topological evolution in the real-time three-dimensional reconstruction method of underwater topography in reclamation areas based on telemetry data provided by the present invention.

[0029] Figure 5 This is a theoretical data curve showing the evolution of the expected value of one-dimensional soil consolidation settlement with physical settlement time in the real-time three-dimensional reconstruction method of underwater topography in reclamation areas based on telemetry data provided by this invention.

[0030] Figure 6 This is a schematic diagram of the elevation adaptive compensation process of the solidified stable zone based on distance inverse weighting in the real-time three-dimensional reconstruction method of underwater topography in reclamation area based on telemetry data provided by the present invention;

[0031] Figure 7 This is a three-dimensional view of the surface distribution of hydrodynamic shear stress in the evolutionary blind zone of the real-time three-dimensional reconstruction method of underwater topography in reclamation areas based on telemetry data provided by the present invention.

[0032] Figure 8 This is a schematic diagram illustrating the implementation of the real-time three-dimensional reconstruction system for underwater topography in reclamation areas based on telemetry data provided by the present invention.

[0033] Figure 9 This is a schematic diagram of the electronic device provided by the present invention. Detailed Implementation

[0034] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0035] The following description, with reference to the accompanying drawings, outlines an embodiment of the present invention of a real-time three-dimensional reconstruction system, method, and electronic device for underwater topography in reclamation areas based on telemetry data.

[0036] Example 1:

[0037] like Figure 1 As shown in the figure, this embodiment provides a real-time three-dimensional reconstruction method for underwater topography in reclamation areas based on telemetry data. This method is mainly applied to complex underwater construction scenarios such as water conservancy projects, waterway dredging, and reclamation.

[0038] In actual reclamation operations, sludge discharge pipelines continuously release a mixture of sediment and mud into designated waters, causing drastic changes in the underwater topography within a short period. To accurately and digitally control the construction progress, it is typically necessary to use telemetry equipment such as multibeam echo sounders mounted on survey vessels to periodically scan and measure the underwater topography.

[0039] The method provided in this embodiment aims to solve the problems of computing power bottleneck and local feature loss when converting massive temporal point cloud data into three-dimensional terrain mesh by deeply integrating the underlying algorithms of computer graphics with dynamic data processing logic.

[0040] Specifically, the real-time three-dimensional reconstruction method for underwater topography in reclamation areas based on telemetry data provided in this embodiment includes the following steps:

[0041] Step S1: Acquisition and adaptive filtering separation of multi-source time-series telemetry point cloud data.

[0042] In the initial stage of system operation and during continuous monitoring, multi-source time-series telemetry point cloud data is first acquired and adaptively filtered to output an effective skeleton point set with a first density and an effective detail point set with a second density, wherein the first density is less than the second density.

[0043] Specifically, the survey vessel navigates along the planned survey lines within the survey area. Its onboard multibeam echo sounder, real-time dynamic differential positioning system, and attitude sensors work together to output raw point clouds with spatial three-dimensional coordinates and echo intensity. Due to the presence of a large amount of suspended sediment and air bubbles in the underwater reclamation area, these non-seabed sediment disturbances generate a significant amount of acoustic noise.

[0044] Therefore, the system extracts the normalized echo signal-to-noise ratio and neighborhood elevation stability coefficient of each measurement point in the multi-source time-series telemetry point cloud data, and calculates the confidence weight of each measurement point by weighting it according to the preset signal-to-noise ratio weight ratio and elevation stability weight ratio. The calculation logic here is limited by formula (1):

[0045] (1)

[0046] in, Indicates the first The credibility weight of each measurement point; This indicates the preset signal-to-noise ratio weighting ratio; Indicates the first Normalized echo signal-to-noise ratio at each measurement point; This indicates the preset elevation stability weight ratio; Indicates the first The neighborhood elevation stability coefficient of each measuring point.

[0047] After calculating the confidence weight, the system removes outlier noise points based on a preset three-condition judgment rule. The three conditions are: the absolute value of the deviation of the measurement point's elevation from the mean elevation of its nearest neighbor is greater than three times the standard deviation of the neighbor elevation; the spatial density of the measurement point's neighborhood is lower than a preset density ratio of the average density of the entire point cloud; and the confidence weight is less than a preset weight threshold.

[0048] These three conditions jointly verify the authenticity of point clouds from three dimensions: elevation abruptness, spatial isolation, and acoustic-physical properties. If any measuring point meets any of the three conditions, the system will classify it as suspended sediment noise or multipath interference noise and discard it.

[0049] Optionally, after removing the measurement points that satisfy any of the three conditions, the system separates and outputs the effective skeleton point set and the effective detail point set according to a preset spatial sampling frequency.

[0050] In practical engineering, the effective skeleton point set is usually generated by a large spatially thinned bounding box, which preserves the macroscopic low-frequency contour of the terrain and has a small data volume (i.e., the first density); while the effective detail point set retains more measuring points and is used to characterize local small sand slopes or scour pits, and has a larger data volume (i.e., the second density).

[0051] like Figure 2 The simulation results demonstrate the system's adaptive filtering and dual-density separation processing after acquiring raw time-series telemetry point cloud data. In the figures, the spatial horizontal and vertical coordinates are set to range from 0 to 100, and the underwater elevation in the vertical direction is distributed between -15 and -5. All coordinate axes are in meters. By zooming in on the scatter plots, the spatial distribution of the three different types of data point clouds can be clearly observed.

[0052] The gray dots in the figure represent suspended sediment and multipath interference noise that are free above and below the main terrain surface and are identified and removed by the system based on the three-condition judgment rule.

[0053] The dark blue scattered points constitute an effective skeleton point set, which is widely distributed in areas with gentle terrain gradient changes with low spatial density, effectively preserving the macroscopic terrain outline and significantly reducing data transmission redundancy in the base state region.

[0054] The orange-red scattered dots constitute an effective set of detail points, which are concentrated in the evolutionary region with significant sand wave characteristics and elevation changes with extremely high spatial density, accurately depicting the microscopic morphology of underwater sediment deposition.

[0055] The simulation results objectively confirm that the method effectively ensures the geometric detail accuracy of the critical operation area while reducing the overall data throughput. However, in the transition zone where the dark blue point set and the orange-red point set meet, there are still a few spatially uneven discrete phenomena and minor data breaks. This reflects that the system's adaptive spatial morphology filtering mechanism has a few digital discretization errors when dealing with the smoothing determination of the boundary of extremely complex underwater terrain. This indicates that there is still room for further improvement in the transition smoothing algorithm of multi-scale density grids.

[0056] Step S2: Initialization and spatiotemporal attribute allocation of the basic macroscopic three-dimensional mesh surface.

[0057] After obtaining the filtered effective point set, the system generates a basic macroscopic three-dimensional grid surface covering the entire domain based on the first frame of effective point set in the multi-source time-series telemetry point cloud data, and assigns initial spatiotemporal attributes to each grid cell.

[0058] Specifically, to construct the basic mathematical framework of the underwater terrain, the system employs the Delaunay triangulation algorithm to globally subdivide the effective point set of the first frame, generating the basic macroscopic 3D mesh surface. Delaunay triangulation ensures that the generated triangular mesh conforms to the characteristics of an empty circle, avoiding the generation of excessively narrow and distorted triangles, thus providing a good geometric quality foundation for subsequent graphics rendering and volume calculation.

[0059] Furthermore, the system extends the spatiotemporal properties to each grid cell within the basic macroscopic three-dimensional grid surface.

[0060] Unlike traditional static 3D modeling, which only includes geometric coordinates and normal vectors, the spatiotemporal attributes in this embodiment include terrain complexity, spatiotemporal activity, lifecycle state markers, and stable continuous frame count. During the system initialization phase, since there is no historical frame data for comparison, the system initializes the spatiotemporal activity and stable continuous frame count of all mesh cells to a preset starting value (usually 0), and initializes the lifecycle state markers to the base state.

[0061] This step allocates storage space for a state machine in the memory data structure for each grid instance, enabling the grid to track its lifecycle evolution over time.

[0062] Step S3: Calculation of spatiotemporal joint topological gradient and two-dimensional fine partitioning.

[0063] The system calculates the spatiotemporal joint topological gradient of each grid cell, divides the terrain into multiple structural types including gentle and uniform regions, abrupt change regions and stable regions based on spatial structural characteristics, and divides it into life cycle states including basic state, evolutionary state and solidified state based on temporal activity, thus completing a two-dimensional fine partitioning.

[0064] Specifically, for each newly received frame of data, the system projects and matches the effective skeleton point set to the corresponding grid cell, constructs a two-dimensional surface structure tensor matrix using the grid cell as the unit, and performs eigenvalue decomposition. The structure tensor matrix can effectively reflect the divergence of local terrain normal vectors. The system extracts a first eigenvalue and a second eigenvalue through decomposition, wherein the first eigenvalue is not less than the second eigenvalue. Subsequently, the ratio of the first eigenvalue to the average side length of the corresponding grid cell is used as the spatial topological gradient.

[0065] To further quantify the dramatic evolution of terrain over time, the system calculates the temporal topological gradient based on the average elevation of grid cells from a preset number of consecutive historical frames, and uses the weighted sum of the spatial and temporal topological gradients as the spatiotemporal joint topological gradient. Its calculation is constrained by formula (2):

[0066] (2)

[0067] in, Indicates the first The spatiotemporal joint topological gradient of each grid cell; Fixed-configuration weights representing the spatial gradient; Indicates the first Spatial topological gradient of each grid cell; Fixed-configuration weights representing the time gradient; Indicates the first The temporal topological gradient of each grid cell.

[0068] Based on this, the system calculates the temporal activity of the current frame's grid cells using a weighted temporal smoothing mechanism, based on the spatiotemporal activity of the previous frame's grid cells and the normalized value of the spatiotemporal joint topological gradient of the current frame. This smoothing mechanism is defined by formula (3):

[0069] (3)

[0070] in, Indicates the first The grid cell in the first... Temporal activity of frames; This represents the preset exponential decay memory factor; Indicates the first The grid cell in the first... Temporal activity of frames; Indicates the first The normalized value of the spatiotemporal joint topological gradient of each grid cell in the current frame.

[0071] It is important to note that after obtaining the aforementioned feature values, the system compares the first and second feature values ​​obtained from the decomposition of the two-dimensional surface structure tensor with the first and second gradient thresholds, respectively, and the second gradient threshold is greater than the first gradient threshold. This constitutes the basis for dividing the spatial dimensions:

[0072] If all of the first feature values ​​are lower than the first gradient threshold, it indicates that the terrain of the area is extremely flat (such as a deep-water still flow area far away from the dredging and dumping outlet), and then it is classified as the flat and uniform area.

[0073] If the first feature value is higher than the second gradient threshold, it is further divided into a linear steep slope area (showing a sharp change in one direction, such as the edge of an underwater trench) or a planar gradual change area (showing a slope-like change in multiple directions, such as the slope of a newly filled sand dune) based on the ratio between the first feature value and the second feature value.

[0074] If both the first feature value and the second feature value are higher than the second gradient threshold and the local elevation of the corresponding grid is an extreme value, then it is divided into a point-like abrupt change zone (such as a newly thrown isolated rock or mud lump), wherein the linear steep slope zone, the planar gradual change zone and the point-like abrupt change zone together constitute the abrupt change zone.

[0075] Simultaneously, the system performs a state machine transition partitioning along the time dimension: classifying the lifecycle states of grid cells with temporal activity levels higher than the high activity threshold as the evolutionary state; classifying the lifecycle states of grid cells with temporal activity levels lower than the low activity threshold and stable continuous frame counts reaching a preset stable frame count threshold as the solidified state; and statistically classifying the lifecycle states of the remaining grid cells that do not fall into the determination conditions of the evolutionary state and the solidified state as the basic state.

[0076] Through two-dimensional fine partitioning, the system successfully decomposes the complex three-dimensional mesh into a set of independent states with different physical evolution characteristics, laying the foundation for subsequent classification calculations.

[0077] like Figure 3 This paper presents a pseudo-color heatmap of underwater topography with fine-grained two-dimensional partitioning. The horizontal and vertical spatial coordinates in the map range from 0 to 100, and the coordinate axis unit is meters. The system uses two-dimensional planar color block mapping to present the current life cycle status of each grid cell.

[0078] The large green areas in the diagram represent the basic, flat zones, which are distributed in water bodies with minimal topographic undulations. The red areas represent the evolutionary, abrupt change zones, concentrated in the core construction sections between horizontal spatial coordinates 60 and 80, exhibiting high activity characteristics influenced by mud discharge.

[0079] The blue area represents the solidified stable zone, located between 15 and 40 in the horizontal spatial coordinates. This indicates that the area has completed geomorphic physical evolution and entered a long-term consolidation and deposition stage.

[0080] The contour curves at the boundaries of different colored blocks show that the system effectively identifies and partitions the continuous terrain grid based on the extracted spatial structure features and temporal activity indicators.

[0081] However, at the boundaries between different state regions, a small number of scattered, isolated, heterochromatic patches still exist, and the curved edges of the partition boundaries exhibit a certain degree of sawtooth-like undulating deformation. This indicates that in the current spatiotemporal joint judgment logic, the classification algorithm will still make local isolated misjudgments when faced with small physical disturbance signals at the boundary of the underlying flow field. Subsequent technical solutions need to be specifically upgraded and optimized in terms of boundary space smoothing and neighborhood state clustering algorithms to improve partitioning accuracy.

[0082] Step S4: Advance prediction and transition processing of the mesh in the evolutionary mutation region.

[0083] In actual reclamation operations, due to the rapid diffusion of slurry injection, the boundary of the abrupt change zone physically spreads within the time interval between two adjacent frames of telemetry data. Therefore, this embodiment adds an additional predictive buffering mechanism for grid cells within the abrupt change zone in the evolved state before performing local grid topology updates.

[0084] Specifically, the system extracts the center coordinates of the bounding boxes of abrupt change regions in multiple consecutive frames to estimate the terrain expansion rate, and uses the terrain expansion rate to predict the position and expansion direction of the abrupt change region in the next frame.

[0085] In the predicted expansion direction, the grid cells in the corresponding region are pre-subdivided according to rules to generate a transition grid with a density that is a preset multiple of the base grid. This step allows the 3D model to prepare sufficient computational grid density in the potential expansion region before the mudflow actually arrives, preventing large-scale triangular breakage and reconstruction caused by subsequent abrupt changes in point cloud data.

[0086] In addition, since the grid surface generated directly from the original telemetry data often appears rough and jagged, which does not conform to the depositional morphology of natural underwater sediment, the system extracts the feature vector of the topographic orientation and uses the orientation-weighted Laplace smoothing algorithm to perform anisotropic smoothing on local patches.

[0087] In this algorithm, a first smoothing weight is assigned along the direction of the terrain's orientation, and a second smoothing weight is assigned along the direction of a steep terrain change perpendicular to the terrain's orientation, with the first smoothing weight being greater than the second smoothing weight. This anisotropic design ensures that while smoothing the mesh to eliminate noise, the geometric edge features of underwater sand slopes and steep banks are strictly preserved, avoiding excessive blurring of the features.

[0088] Step S5: Categorize and update the execution logic, that is, differentiate the processing for the basic state, the evolved state and the fixed state.

[0089] After partitioning and preprocessing, the system implements drastically different graphical operation update algorithms for grid sets in different states.

[0090] Logical branching 1: Elevation update for the base state flat and uniform region.

[0091] For the grid cells within the smooth, uniform region in the base state, adaptive temporal elevation updates are performed while maintaining the grid topology connections unchanged. Since only slow sediment deposition or erosion occurs in this region, there is no need to expend computational resources to destroy and rebuild the triangular network structure.

[0092] Specifically, the system obtains the confidence weight of the corresponding measuring point, the planar Euclidean distance factor between the measuring point and the corresponding grid vertex, and the terrain complexity factor of the neighborhood of the grid vertex.

[0093] Subsequently, the confidence weight, the planar Euclidean distance factor, and the terrain complexity factor are multiplied to construct the interpolation weight of the target measurement point to the target grid vertex, which is limited by formula (4):

[0094] (4)

[0095] in, Indicates the first The target measurement point for the first Interpolation weights for each target grid vertex; Indicates the first The credibility weight of each measurement point; Represents an exponential function with the natural constant as its base; Indicates the first The measuring point and the first The planar Euclidean distance factor of each grid vertex; This indicates the set reference Gaussian decay radius parameter; Indicates the first The terrain complexity factor of the neighborhood of each grid vertex.

[0096] After calculating the interpolated elevation of the corresponding grid vertex using this weighted average method, the calculated interpolated elevation and the historical elevation of the target grid vertex are subjected to time-series smoothing processing to update the final elevation of the grid cell vertices in the smooth and uniform region. This achieves smooth tracking of underwater deformation without changing the graphics topology.

[0097] Logical branching 2: Local topology update for the mutation region of the evolutionary state.

[0098] For the mesh cells within the abrupt change zone in the aforementioned evolutionary state, a local mesh topology update is performed to generate high-density local patches. This region is the core area for dredged reclamation and must reflect extremely high-frequency topographic changes with the highest mesh density.

[0099] Specifically, for the mutation regions determined to be continuously evolving, the effective set of details within the local bounding box of the current frame is extracted as a new effective set of details, and new vertices are inserted one by one into this new set of details. In computer graphics, vertex insertion often leads to triangular mesh intersections or overlaps; therefore, slope constraint checks are performed before inserting new vertices. Based on soil mechanics principles, if it is determined that the inclination angles of all triangular faces in the neighborhood of the inserted vertex do not exceed the preset natural angle of repose of the fill soil (for example, the natural angle of repose of fine sand is usually around 30 degrees; exceeding this angle will physically cause landslides, which is considered a data anomaly), then topology updates are achieved through incremental vertex insertion and edge flipping.

[0100] For newly added mutation zones determined to be non-continuously evolving (i.e., new work surfaces that show drastic changes for the first time), constrained Delaunay triangulation is performed to generate the high-density local patches. In order to maximize the restoration of terrain semantics, the linear steep slope areas are meshed along the terrain direction, the face-shaped gradual change areas are meshed isotropically uniformly, and the point-shaped mutation areas are meshed radially with extreme points as the center. During the triangulation process, the natural angle of repose of the dredged fill is also embedded as a hard constraint to eliminate triangular patches with excessive dip angles.

[0101] like Figure 4 This demonstrates the topological evolution effect after incremental vertex insertion and constrained triangulation in the mutation region of the evolutionary state. The spatial horizontal and vertical coordinates in the figure range from 10 to 60, and the underwater elevation is distributed in the interval of -12 to -5, all in meters.

[0102] The light blue framed area on the outside of the image represents a flat surface grid in its basic state. Its grid density is low and its undulations are gentle, reflecting the underwater sedimentary landforms that have not been directly affected by the silt removal operation.

[0103] The raised red frame in the center of the image represents a high-density local patch generated by performing local constraint triangulation on the newly added abrupt change region. The system significantly improved the mesh density in this region through incremental vertex insertion, reconstructing the steep morphology formed by the rapid deposition of dredged fill sediment.

[0104] In the sloping area where the red high-density grid meets the light blue basic grid, there are blank areas marked by black scattered dots at the edges. This is the physical feedback mechanism for the system to perform slope constraint verification.

[0105] When the inclination angle of a local triangular facet exceeds the preset natural repose angle of the dredged fill, the system fails to meet the empty circle characteristic verification, and it is judged as a deformed triangular facet that is prone to collapse and is removed, thus forming the void in the super-boundary slope constraint removal area shown in the figure.

[0106] From an overall topological perspective, this processing mechanism effectively constrains the spatial reconstruction process of graphics by adhering to the geotechnical physical and mechanical boundaries. However, several narrow, sharp triangles still remain at the outer edges of high-density local patches, indicating an excessively large area gradient span in these areas. This suggests that the current edge-flipping topology update logic still exhibits slight aspect ratio degradation when dealing with stitched boundaries where mesh density changes drastically. Further refinement of the isotropic mesh smoothing mechanism is needed to improve the triangulation quality of the boundary transition regions.

[0107] Logical Diversion 3: Progressive simplification for the solidified stable region.

[0108] For mesh cells within the stable region in the solidified state, a progressive mesh simplification algorithm is used to reduce mesh density and fall back to the base mesh resolution. Retaining a high-density mesh in a solidified region that has been completed and remains unchanged for a long time would be a huge waste of video memory and the rendering bus.

[0109] Specifically, the system extracts the mesh geometric error calculated based on the quadratic error measure and the structure preservation error calculated based on the matching degree between the mesh edge direction and the terrain orientation. The mesh geometric error and the structure preservation error are weighted and calculated according to their respective preset weight coefficients, and used as the composite folding error of the target mesh edge. The calculation is performed using formula (5):

[0110] (5)

[0111] in, Indicates the first The composite folding error of the target mesh edges; Preset weighting coefficients representing mesh geometry errors; Indicates the first Mesh geometry error of the target mesh edge; Preset weighting coefficients representing structural retention error; Indicates the first The structure of the target mesh edges maintains the error.

[0112] Sort the composite folding errors from smallest to largest, and perform half-side folding operations step by step from the geometric center of the stable region in the solidified state to the boundary of the stable region, that is, merge the two vertices of an edge into one vertex, thereby eliminating the two adjacent triangles.

[0113] To prevent model collapse due to oversimplification, a preset proportional constraint is set to ensure that the number of folded edges in a single frame does not exceed the total number of edges in the corresponding stable region, and the boundary vertex positions of the stable region are kept fixed. When the average edge length of the mesh in the corresponding stable region expands to a preset multiple threshold of the edge length of the basic mesh, the folding simplification operation is stopped and the state of the mesh cells in the stable region is reset to the basic state.

[0114] Step S6: Seamless global topology stitching and real-time rendering output.

[0115] After completing the basic state elevation update, the evolved state topology subdivision, and the solidified state topology simplification, cracks or overhangs inevitably exist at the boundaries of various local sub-mesh densities. Therefore, it is necessary to topologically stitch all high-density local patches with the simplified mesh and the basic mesh to obtain a topologically complete global 3D surface mesh and perform real-time rendering output.

[0116] Specifically, the system extracts the boundary vertices at the intersection of the high-density local patch and the basic grid. The trapezoidal prism method is used to calculate the earthwork volume of the corresponding local area before and after the update, respectively, to obtain the relative volume deviation rate. The calculation logic is shown in formula (6):

[0117] (6)

[0118] in, Indicates the first The relative volume deviation rate of the local junction area; Indicates the first Calculated earthwork volume of a local junction region after topology update and stitching; Indicates the first The original earthwork volume of a local junction area before topological update stitching; This represents the absolute value operator.

[0119] In engineering applications, the core measurement basis for land reclamation projects lies in the volume of earthwork. Therefore, when the relative volume deviation rate exceeds a preset volume deviation threshold (e.g., 0.5%), a forced linear elevation compensation adjustment is performed on the corresponding boundary vertices until the relative volume deviation rate is less than or equal to the volume deviation threshold. This step ensures that the topology stitching process in computer graphics never sacrifices the rigor of the engineering volume measurement, ultimately achieving a seamless topology stitching output of a three-dimensional surface mesh.

[0120] Finally, the output 3D surface mesh is combined with a view distance-related Level of Detail (LOD) dynamic switching mechanism. By removing or merging triangular faces with view distance values ​​greater than a preset view distance threshold, real-time rendering is performed, thus presenting an extremely smooth and realistic real-time dynamic 3D underwater terrain scene on the display screen of the command center.

[0121] In summary, the technical solution disclosed in this embodiment represents a breakthrough in underlying logic compared to traditional 3D reconstruction methods involving global traversal, full partitioning, or full updating in the prior art. This solution closely integrates the physical and objective laws governing underwater topographic changes in reclamation areas, introducing a spatiotemporal joint topological gradient and lifecycle state model. By dynamically sensing massive input point clouds, the system can intelligently identify high-frequency evolution zones with drastic topographic changes and determine which areas belong to stable zones where cost reduction and efficiency improvement can be implemented.

[0122] This differentiated and refined asynchronous processing mechanism not only greatly releases the parallel computing resources of computer processors and shortens the grid refresh latency between adjacent data frames, but more importantly, it minimizes the redundant video memory usage in non-operation areas while ensuring that the high-resolution terrain details in the core work area are not lost. Ultimately, it achieves high real-time performance and high physical fidelity 3D terrain reconstruction in ultra-large-scale underwater engineering scenarios.

[0123] Example 2:

[0124] In actual large-scale underwater reclamation or channel dredging operations, due to limitations such as the speed of survey vessels, the effective scanning and width of multibeam echo sounders, and maritime traffic control, telemetry equipment can often only focus on tracking active construction areas where mud pumps are currently discharging mud. This results in areas where reclamation has been completed, where topographic features no longer undergo drastic geometric changes, and which have been classified as solidified by the system, remaining in a state of lack of multi-source time-series telemetry point cloud data updates for extended periods, creating a data acquisition time blind spot.

[0125] However, in actual geotechnical engineering, the silt mixture formed by hydraulic reclamation is a typical two-phase fluid. Under the continuous action of its own weight and the hydrostatic pressure of the overlying water, the pore water inside this type of soil will be continuously discharged over time, the soil skeleton will be compressed, and this will macroscopically manifest as the continuous subsidence of the underwater topographic surface. Existing computer graphics 3D modeling methods, in the absence of external sensor data input, usually default to maintaining the spatial coordinates of the mesh vertices unchanged.

[0126] This purely digital static processing method completely deviates from the real laws of physical and mechanical evolution. Over time, the three-dimensional mesh surface rendered by the system will be much higher than the actual physical surface. Ultimately, when performing seamless stitching of the global topology and earthwork volume calculation, it will produce extremely serious positive volume deviation, which will greatly affect the objectivity and accuracy of construction measurement and settlement.

[0127] To address this technical problem with significant engineering practical implications, this embodiment explicitly states in its key technical solution that, for grid cells in the solidified state that lack updates from multi-source time-series telemetry point cloud data, before performing progressive grid simplification using the half-side folding algorithm, a grid vertex elevation deformation compensation step based on a physical consolidation prior model is also included.

[0128] This step breaks with the conventional logic of computer graphics—that updates only occur when there is no data—and introduces geotechnical mechanics mechanisms as the self-driving force for 3D mesh deformation. The method in this embodiment will be explained step by step:

[0129] Specifically, this solution first needs to address the technical issue of aligning digital time with physical time. The steps include: extracting the number of stable, continuous frames from the spatiotemporal attributes of the grid cell, and converting the number of stable, continuous frames into physical settling time parameters using the system's preset frame rate conversion parameters.

[0130] In the system's underlying data structure, each mesh cell maintains an independent state machine object. When the system determines that a mesh cell has exited the evolving state and entered the solidified state through two-dimensional fine partitioning, the stable continuous frame count contained in the spatiotemporal attributes of that mesh cell will be incremented after each system rendering cycle without receiving any new valid detail points for that region. This stable continuous frame count is essentially an integer counter reflecting the passage of digital time.

[0131] However, the time parameters required by the geotechnical mechanics model must have standard physical dimensions. Therefore, the system is pre-configured with preset frame rate conversion parameters corresponding to the telemetry data acquisition frequency and the graphics engine refresh rate. By multiplying the system's accumulated stable continuous frame count with this preset frame rate conversion parameter, the system can accurately calculate the actual physical duration experienced by the solidified grid cell since the last real data update. This physical settlement time parameter is the core independent variable driving the subsequent evolution of the consolidation settlement model over time, enabling the system to map not only the terrain in the spatial dimension but also the progression of its physical state in the temporal dimension.

[0132] Optionally, after obtaining accurate physical time boundary conditions, the system then enters the geometric deformation calculation process based on the laws of geotechnical mechanics. This step specifically involves: obtaining the initial void ratio and compression coefficient parameters of the fill soil in the construction area where the grid unit is located; inputting the initial void ratio parameters, the compression coefficient parameters, and the physical settlement time parameters as calculation variables into a preset one-dimensional soil consolidation settlement model; and calculating the expected physical settlement of the grid unit.

[0133] In practice, a thorough geological survey and geotechnical tests are conducted on the hydraulic fill construction area before construction begins. The system uses the spatial index of the plane coordinates of the current grid cell to accurately retrieve the actual physical parameters of the hydraulic fill soil at that specific location by calling the geographic information system database.

[0134] Among them, the initial void ratio parameter characterizes the basic ratio of pore volume to solid particle volume in the soil, and determines the potential compression space limit of the soil; the compression coefficient parameter characterizes the ease with which the volume of this specific type of dredged fill soil shrinks under stress.

[0135] After obtaining the above geological parameters, the system constructs a one-dimensional soil consolidation settlement model to solve for the expected settlement value. The underlying logic of this calculation process is rigorously defined by the following formula (7):

[0136] (7)

[0137] In the formula, Indicates the first The expected physical settlement of each grid cell; Indicates the first The compression coefficient parameter of the fill soil in the area corresponding to each grid cell; Indicates the first The initial void ratio parameters of the dredged fill soil in the corresponding area for each grid cell; Indicates the first The parameters of the effective overlying stress load borne by each grid cell; Indicates the first The exponential function of the thickness of the dredged fill layer in the region where each grid cell is located; The consolidation time factor parameter represents the rate of soil drainage consolidation; This indicates the corresponding number obtained from the aforementioned calculation. Physical settlement time parameters for each grid cell.

[0138] Through the calculation of formula (7), the system transforms the elevation change that was originally obtained by scanning with external sensors into an analytical solution that is predicted in advance through internal physical mechanisms. The calculated expected physical settlement is a scalar value that accurately reflects the thickness of the soil volumetric compression caused by the discharge of pore water in the vertical direction.

[0139] like Figure 5 This figure presents theoretical data curves showing the evolution of the expected value of one-dimensional soil consolidation settlement over physical settlement time. The horizontal axis represents physical settlement time, ranging from 0 to 180 days, and the vertical axis represents the expected physical settlement, ranging from 0 to 900 millimeters.

[0140] The blue solid curve in the figure represents the trajectory of the theoretical subsidence displacement of the vertex, calculated by the system using a one-dimensional soil consolidation settlement model when the grid cell in the solidified state lacks external multi-source time-series telemetry point cloud data updates.

[0141] Based on the logical deduction in the embodiment, it can be seen from the waveform transformation characteristics of the blue solid curve that in the initial 0 to 50 days of the physical deduction, a large amount of pore water is discharged, and the curve shows a steep upward trend with a very large slope. The settlement amount accumulates rapidly to about 690. This band reflects that the three-dimensional grid vertices in this technical solution can sensitively respond to the early rapid self-weight compaction mechanism of the dredged fill.

[0142] Between 50 and 180 days, the slope of the tangent line of the curve gradually decreases and becomes flat, exhibiting typical volume shrinkage convergence characteristics. This indicates that the scheme transforms the objective evolution law of engineering geology into the driving force for grid adaptive sinking, maintaining the fidelity of volume evolution of digital terrain in long-term data blind areas.

[0143] In the later stages where the blue solid curve tends to stabilize, small, step-like oscillations still remain, and the overall curve exhibits a slow, linear upward offset that has not fully converged horizontally. This waveform defect, reflecting the actual R&D state, exposes discretized digital truncation errors generated by the system when performing large-span time conversions using preset frame rate conversion parameters. It also indicates that the basic one-dimensional model has not accurately decoupled the complex secondary creep displacement components of deep soil, suggesting that the subsequent three-dimensional reconstruction system still has room for improvement in terms of time step smoothing compensation algorithms and the fusion depth of multi-dimensional physical constitutive models.

[0144] For example, in order to integrate the physical scalars calculated above into the processing pipeline of 3D computer graphics, the system then performs the following operation: converting the expected physical settlement into the deformation displacement vector of the corresponding mesh vertex along the negative direction of gravity, and performing adaptive sinking update on the vertex elevation value of the mesh cell in the solidified state based on the deformation displacement vector.

[0145] In a Cartesian coordinate system, the direction of soil consolidation settlement under its own weight is strictly parallel to the direction of gravity. Therefore, the system uses vectorization to assign a directional attribute to this scalar value, generating a deformation displacement vector pointing in the negative direction of gravity in the spatial coordinate system.

[0146] Subsequently, the system traverses all vertex data structures within the solidified mesh cell, performing vector addition on the extracted original vertex coordinates and the deformation displacement vector. This operation directly modifies the depth coordinates of the vertices in video memory, resulting in a sunken visual and geometric effect on the 3D mesh surface.

[0147] This adaptive sinking update mechanism successfully eliminates the absolute dependence of traditional 3D reconstruction systems on external sensor data, enabling the system to present dynamic terrain evolution characteristics that conform to natural laws even in data blind spots, and maintaining the consistency of the digital twin with the real physical world in terms of spatial volume.

[0148] It should be noted that after completing the physical deformation compensation of the solidified mesh vertices, the computational power optimization requirements of the computer graphics field must also be taken into account, that is, the progressive mesh simplification operation mentioned in Example 1 must be performed.

[0149] However, traditional topology simplification algorithms may destroy the physical deformation features accumulated in the early stages. Therefore, this technical solution specifically stipulates that when the half-side folding algorithm is executed to reduce the mesh density, the elevation values ​​of the new vertices generated by folding and collapsing the original mesh vertices at both ends are set as the weighted average of the elevation values ​​of the original mesh vertices at both ends after adaptive sinking and updating by the deformation displacement vector, calculated based on a preset distance ratio.

[0150] In computer 3D graphics, the half-edge folding algorithm is a mainstream method to reduce mesh resolution. Its core action is to select an edge in the mesh, move the original mesh vertices at both ends of the edge towards the target collapse position, and finally merge them into a new vertex, thereby eliminating the related triangular facets.

[0151] In this process, the elevation value of the newly generated vertex is crucial. If the value is chosen arbitrarily, the previously carefully calculated physical settlement characteristics will be directly erased. Therefore, the system constructs weights by calculating the spatial distance between the original vertex and the target position of the new vertex. This weight calculation and elevation assignment mechanism is clearly defined by the following formula (8):

[0152] (8)

[0153] In the formula, Indicates the first The elevation value of the new vertex generated when the half-folding operation is performed again; This represents the preset spatial distance between the first original mesh vertex and the target collapse position when the p-th half-folding operation is performed; This represents the preset spatial distance between the second original mesh vertex and the target collapse position when the p-th half-side folding operation is performed; This represents the elevation value of the first original mesh vertex after adaptive sinking and updating when the p-th half-side folding operation is performed; This represents the elevation value of the second original mesh vertex after adaptive sinking and re-updating when the p-th half-side folding operation is performed.

[0154] As can be seen from the architecture of formula (8), the system adopts a distance-inverse weighting logic mechanism. When the target collapse position is spatially closer to the first original grid vertex, the preset spatial distance value of the first original grid vertex will be smaller, resulting in an increased elevation weight ratio. This mechanism ensures that even if the underlying graphic topology undergoes drastic folding and simplification, the local terrain subsidence characteristics and elevation distribution centroid caused by soil physical consolidation are still accurately inherited and preserved by the new vertex, avoiding terrain deformation distortion caused by reduced grid density.

[0155] like Figure 6 This diagram illustrates a cross-sectional view of the adaptive compensation process for half-side folding elevation in the solidified stable region based on distance inverse weighting. The horizontal axis represents the horizontal projected distance in meters, and the vertical axis represents the absolute elevation in meters.

[0156] The blue dashed line in the figure represents the initial grid profile without settlement, and the green dotted line represents the subsided grid profile after undergoing one-dimensional physical consolidation deformation. From the waveform transformation, it can be seen that the elevation of the vertex at the horizontal coordinate of 10 meters decreases from -3.5 meters to -4.0 meters, and the elevation of the vertex at the horizontal coordinate of 20 meters decreases to -5.5 meters.

[0157] The solid red line in the figure shows the final folded compensation profile after the system performs progressive mesh simplification using a half-side folding algorithm on a stable region in a solidified state. During the folding process, the two original subsidence vertices located at x-coordinates of 10 meters and 20 meters are collapsed and merged into a new vertex at x-coordinate of 16 meters. Since this new location is 6 meters from the left vertex and 4 meters from the right vertex, the system assigns a weight of 40% to the left elevation and 60% to the right elevation based on the inverse distance weighting logic, ultimately calculating the new vertex's elevation to be -4.9 meters. This graphical transformation process demonstrates that this technical solution effectively preserves the elevation centroid fidelity of the terrain's physical subsidence while reducing mesh node density.

[0158] However, comparing the ideal subsidence shape of the red solid line in the figure with the gray dashed line representing objective deformation in nature, it can be found that within the large span of 0 to 16 meters on the horizontal coordinate, there is a significant geometric gap between the simplified grid profile formed by directly connecting linear lines and the non-linearly curved gray dashed line. This separation phenomenon on the waveform reveals that the existing spatial linear folding connection mechanism still loses a small amount of envelope volume accuracy due to linear truncation error when dealing with seabed topography with complex curvature. This indicates that the system still needs to further introduce a non-linear curvature preservation factor or a higher-order spline fitting mechanism in the topological stitching logic of the half-side folding algorithm to achieve technical upgrades and improvements.

[0159] In summary, considering the limitations of existing technologies in underwater 3D reconstruction, it can be found that most existing conventional 3D reconstruction systems only play the role of data visualization and rendering tools, are severely limited by the sampling frequency and coverage of front-end hardware data, lack an effective automatic update mechanism for data acquisition blind spots, and are prone to outputting static models that lag behind actual physical deformations.

[0160] The technical solution in this embodiment combines the one-dimensional soil consolidation mechanism of geotechnical engineering with the half-fold topology simplification algorithm of computer graphics. When faced with solidified areas lacking telemetry data, the system uses spatiotemporal attributes to track physical time, retrieves geological parameters to drive the consolidation settlement model, and spontaneously performs deformation compensation on the vertices of the three-dimensional mesh by calculating the expected physical subsidence. In the subsequent topology collapse to save computing power, distance-weighted logic is used to effectively preserve this deformation compensation feature.

[0161] This method, which deeply embeds the laws of the objective physical world into the digital twin foundation, endows the system with the ability to automatically correct and evolve in data-free environments. Ultimately, when the regional grid is stitched together with the global grid and used for earthwork measurement, the calculated earthwork volume closely approximates the actual state of the dredged fill area after physical compaction. This effectively eliminates the volume surplus calculation error caused by neglecting soil consolidation in traditional algorithms, providing a highly accurate, reliable, and mechanically and physically realistic three-dimensional digital decision-making basis for ultra-large-scale water conservancy and waterway projects.

[0162] Example 3:

[0163] In the actual construction operations of underwater reclamation projects and large-scale waterway dredging projects, after the system performs local mesh topology updates according to the method in the above embodiments and generates high-density local patches, although it can restore the terrain details of the core construction area with high spatial resolution, it still faces the inherent bottleneck of physical sampling frequency in the time dimension.

[0164] Specifically, due to the limitations of the physical working mechanism of acoustic telemetry equipment (such as multibeam echo sounders), the propagation of sound waves in water and the reception of echoes have inherent time overhead, and the navigation maneuver of the survey vessel requires periodic time scheduling. This inevitably results in a data acquisition interval of several seconds or even tens of seconds between two adjacent frames of multi-source time-series telemetry point cloud data.

[0165] Within this time blind zone, the sludge discharge pipes of the reclamation equipment are continuously spraying a high-concentration, high-pressure mixture of mud and sand underwater. When this high-energy sludge discharge flow comes into contact with the newly deposited, dense localized areas, it generates a very strong hydrodynamic scouring effect, causing significant geometric displacement of the underwater topography within seconds.

[0166] If a computer-based 3D reconstruction system merely passively waits for the next frame of measurement data, the surface mesh output by the rendering engine will remain stagnant during this period. This phenomenon, where the 3D model representation lags behind the actual physical deformation, can not only cause construction personnel to misjudge the current state of underwater topography and siltation, but also easily lead to serious equipment damage accidents caused by subsequent construction equipment (such as cutter heads or deployed deep-sea exploration instruments) failing to avoid actually bulging or collapsing silt deposits in time.

[0167] In order to effectively solve the technical problems of model lag and collision risk caused by the blind spot of telemetry data acquisition time, this embodiment introduces a mechanism that couples fluid mechanics and computer graphics.

[0168] In this embodiment, after performing local mesh topology updates and generating high-density local patches, the method further includes an inter-frame mesh morphology prediction and compensation step based on hydrodynamic scouring effects. This step breaks through the limitations of traditional data-driven modeling and achieves feedforward prediction of physical mechanisms in the absence of data. Specifically:

[0169] The system first needs to build a data bridge connecting the physical equipment used in construction operations with the computer's virtual grid. This step involves: acquiring the real-time sludge discharge velocity vector and the spatial coordinates of the sludge discharge outlet from the construction equipment; and, combined with the current grid topology of the high-density local area, constructing a real-time hydrodynamic shear stress field acting on each triangular facet of the high-density local area.

[0170] In the digital deployment of actual construction environments, the ends of the dredged and discharged sludge pipelines are equipped with high-precision inertial navigation units and Doppler velocity profilers. These industrial sensors can synchronously send the current absolute spatial coordinates of the discharge port and the velocity vector data of the sludge injection to the central processing system at extremely high frequencies (such as tens of hertz).

[0171] After obtaining the aforementioned physical input parameters, the system maps them into a virtual three-dimensional graphics space. Since the system has previously generated a high-resolution, high-density local patch, which consists of a large number of interconnected triangular patches, and each triangular patch has a definite spatial normal vector and surface area.

[0172] The system calculates the drag shear force exerted by the water flow on the surface of each discrete triangular facet by calculating the projection of the sludge discharge velocity vector onto the normal vector direction of each triangular facet, and combining this with the boundary layer momentum transfer principle in fluid mechanics. To ensure that the calculation process has rigorous mathematical logic support, this embodiment uses the following formula (9) to quantify the force on a single facet in the stress field:

[0173] (9)

[0174] In the formula, Indicates the action on the first Real-time hydrodynamic shear stress of a triangular facet; This indicates the density parameter of the mud fluid in the construction operation area; Indicates the first Surface drag coefficient of each triangular facet; This represents the obtained real-time sludge discharge velocity vector; Indicates the first The unit normal vector of a triangular facet; This represents the operator for taking the absolute value of the dot product of vectors.

[0175] This formula accurately characterizes the differences in force distribution of grid patches with different spatial orientations in the same flow field by performing dot product operation on the flow field parameters and grid geometric parameters, thereby constructing a real-time hydrodynamic shear stress field covering the entire abrupt change region.

[0176] For example, after generating a physically meaningful stress field, the system needs to introduce a critical stability mechanism from geotechnical mechanics to determine grid deformation. This step involves extracting the critical initiation shear stress parameters of the corresponding area of ​​dredged fill. The real-time hydrodynamic shear stress borne by each triangular facet in the hydrodynamic shear stress field. With the critical starting shear stress parameter Numerical comparison analysis was performed.

[0177] In sediment dynamics theory, underwater deposited soil particles do not move immediately under the shear force of water flow. Only when the externally applied hydrodynamic shear force overcomes the particle's own weight, the cohesion between particles, and underwater frictional resistance will the soil begin to displace from the grid base. This critical threshold is called the critical initiation shear stress.

[0178] The system retrieves the critical initiation shear stress parameters corresponding to the current evolutionary state of the region from the regional geological survey database entered before the project commences. Subsequently, the logic control unit in the system background will perform traversal calculations on the generated stress field, comparing the real-time stress state of each discrete unit with the physical threshold one by one.

[0179] Furthermore, based on the results of the above numerical comparison analysis, the system enters the stage of precise extraction and classification of graphical elements. Its core step is: screening out the real-time hydrodynamic shear stress... Greater than the critical starting shear stress parameter The target triangular facet is identified, and the eroded mesh vertices contained in the target triangular facet are extracted.

[0180] In this step, the system uses a Boolean decision filter to split the entire high-density local patch into two parts: patches in a stable state and patches about to undergo physical slippage and erosion. Triangular patches that are subjected to less stress, are in a masked position, or are in a velocity attenuation zone are considered safe and their spatial coordinates are maintained and do not participate in the inter-frame deformation calculation.

[0181] The selected target triangular facets indicate that, under the current flow field conditions, the physical sediment attached to them will inevitably be lost and displaced. This is because, in the computer 3D graphics pipeline, all geometric transformations (Translation, Rotation, Scaling) ultimately need to be applied to the most basic unit of the mesh, namely the vertex.

[0182] Therefore, the system further deconstructs the marked target triangular facets into a set of vertices through the topology lookup interface of the data structure, extracts the eroded mesh vertices that are directly subjected to dynamic erosion and are about to undergo coordinate changes, and prepares the computational objects for the next matrix transformation.

[0183] Specifically, the core deformation prediction and vertex transformation mechanism of this embodiment lies in the following: during the time interval between the acquisition of multi-source time-series telemetry point cloud data from two adjacent frames, the direction of the sludge discharge velocity vector is used as the displacement guiding direction, and the maximum real-time hydrodynamic shear stress local extremum is transmitted to the vertex. With the critical starting shear stress parameter The product of the difference and the preset displacement ratio coefficient is used as the offset displacement step size to perform predictive offset of spatial coordinates on the eroded mesh vertices.

[0184] This step fully demonstrates the engineering wisdom of using mechanical feedback to drive computer graphics. In the blind zone where acoustic depth sounding data is lacking, the system, based on the logic of real physical deduction, assumes that the direction of movement of sediment after being washed away must follow the direction of the dominant flow velocity field.

[0185] Therefore, the system directly extracts the direction vector of the sludge discharge velocity vector as the unit direction vector guiding the movement of the scoured grid vertices in the three-dimensional Cartesian coordinate system. Simultaneously, the distance by which the grid vertices shift (i.e., the offset step size) should not be an arbitrary fixed value, but rather should be positively correlated with the excess hydrodynamic force experienced by the region. The greater the shear stress exceeding the critical threshold, the more severe the scouring and the more serious the collapse or slippage of the local terrain.

[0186] To achieve precise control over this three-dimensional vertex spatial offset, the system calls the following formula (10) for calculation:

[0187] (10)

[0188] In the above formula, Indicates the first Predictive offset vectors of eroded grid vertices; This represents the preset displacement scaling factor input during system initialization; Indicates the relationship with the first The local extremum of the maximum real-time hydrodynamic shear stress transmitted to a vertex among all target triangular facets directly adjacent to a vertex of a scourted mesh. This represents the extracted critical starting shear stress parameter; This represents the magnitude of the real-time sludge discharge velocity vector.

[0189] When performing predictive offsets of spatial coordinates, the system projects the calculated predictive offset vector onto the local terrain tangent plane where the vertex is located. Then, it adds the original 3D coordinates of the eroded mesh vertex to the projected tangential offset vector, thus completing the real-time update calculation of the vertex's spatial position in the memory data. This tangential constraint operation ensures that vertex displacement in computer graphics strictly follows the physical logic of sediment erosion, preventing terrain mesh clipping or overhang failure.

[0190] This offset processing method enables geometric nodes in computer graphics to acquire force response characteristics similar to those of real-world sediment particles.

[0191] like Figure 7 A three-dimensional view of the surface distribution of hydrodynamic shear stress within the evolutionary blind zone is shown. The values ​​of the horizontal and vertical coordinates in the figure range from 0 to 50, and the underwater elevation is distributed in the range of -12 to -8. The unit of each coordinate axis is meters.

[0192] The 3D surface with color mapping in the figure represents a high-density local patch generated after performing a local mesh topology update on the abrupt change region. The transition of its surface color from cool to warm tones intuitively presents the numerical distribution of the real-time hydrodynamic shear stress field. A dedicated color bar is added above the chart to precisely calibrate the specific shear stress values ​​represented by different colors.

[0193] By combining the undulations of the three-dimensional surface with the mapping relationship of the color bars, it can be seen that in the water-facing area of ​​the terrain near the horizontal coordinate 15 and the vertical coordinate 20, the shear stress is highly concentrated and reaches a peak value of 80 to 90 (presenting a corresponding warm-toned peak value).

[0194] Based on this, the system extracts grid vertices where the local shear stress exceeds a set critical starting shear stress value to provide a physical criterion for predictive spatial coordinate offset. The figure clearly reflects the spatial distribution characteristics of the calculated water flow shear stress during the data acquisition interval. However, the presence of small, high-frequency color spots indicates slight numerical oscillations in the current discretized stress field calculation results, suggesting room for further optimization and upgrades in the spatial smoothing processing of the fluid mesh mapping algorithm.

[0195] It is also important to note that after the underlying data structure completes the update of the vertex matrix parameters, it must be converted into feedback at the visualization level. Therefore, the system executes the following: reconnecting the offset vertices of the scourted mesh to form a predicted deformation patch, and during the acquisition time interval, outputting the predicted deformation patch as the surface mesh of the current evolution state to the rendering engine for temporary real-time rendering display, until the next frame of multi-source time-series telemetry point cloud data is received and replacement correction is performed using the actual elevation.

[0196] In the graphics rendering pipeline, since only the coordinates of some vertices are extrapolated and the system still retains the original mesh's index buffer mechanism, the system can reconnect the displaced eroded mesh vertices with the surrounding fixed vertices that have not been displaced, based on the original triangulation topology, and quickly construct a predicted deformed patch that presents the erosion pit or flow accumulation morphology.

[0197] During the scanning window of the telemetry equipment, the rendering engine will push this physical model, which is autonomously generated by the system, as a high-priority temporary display layer to the construction command and control screen. At this time, the underwater terrain displayed on the terminal is no longer a static image in a stagnant state, but a dynamic evolution process of terrain being scoured, eroded, and stripped away in real time, accompanied by the fluctuation of the sludge discharge flow field.

[0198] This temporary real-time rendering display will continue until the multibeam echo sounder completes its current scheduling cycle and transmits the accurate physical elevation point cloud data obtained from the new frame scan to the system backend. At that time, the system will terminate the feedforward prediction mechanism, use the acquired real elevation data to forcibly replace and correct the prediction grid, eliminate accumulated inference errors, and begin the next data collection and blind zone prediction cycle.

[0199] In summary, considering the shortcomings of existing underwater 3D reconstruction technologies, such as low data update frequency, model lag behind reality, and lack of dynamic flow field coupling perception capabilities, the technical solution constructed in this embodiment, which includes hydrodynamic scour effect assessment and mesh morphology prediction, provides an effective solution.

[0200] Specifically, by intermittently inserting a grid offset calculation layer driven by real fluid dynamics and geotechnical mechanics principles into the traditional acoustic measurement data stream, the computer-based 3D reconstruction system gains the ability to autonomously maintain the predictive capability of terrain evolution during data input fault periods. This significantly refines the temporal granularity of the final digital twin 3D model presented to construction decision-makers, effectively filling the cognitive blind spots caused by hardware scanning intervals. This forward-looking model extrapolation method can help the construction scheduling system predict potential underwater sandbank landslides and dramatic terrain changes several seconds or tens of seconds in advance, greatly improving the safety early warning capability for construction equipment and significantly reducing the risk of underwater equipment bottoming out or colliding due to model lag, demonstrating high engineering practical value.

[0201] Example 4:

[0202] like Figure 8 As shown, this embodiment provides a real-time three-dimensional reconstruction system for underwater topography in reclamation areas based on telemetry data. This embodiment is a system embodiment corresponding to the above-described method embodiment.

[0203] In existing underwater topography reconstruction operations in reclamation areas, the back-end data processing system typically employs a globally unified grid processing mechanism. Specifically, after the survey area receives time-series point cloud data transmitted back from the front-end telemetry equipment, the existing system often performs synchronous topological subdivision calculations on the global topographic grid or uses a unified downsampling algorithm for compression.

[0204] However, real-world reclamation areas exhibit significant spatial heterogeneity and temporal evolution differences. Areas near work sites, such as discharge pipes, experience extremely high-frequency geometric evolution due to continuous sediment accumulation; while deeper water areas or completed areas far from work sites remain in a relatively stable consolidation state. Existing systems fail to allocate computational resources based on terrain evolution, resulting in a large number of redundant grid topology updates being performed on stable regions with no significant changes during data processing, leading to a severe waste of computational resources. Simultaneously, because computational power is globally distributed, areas requiring high-precision characterization of abrupt changes cannot obtain sufficient local grid density.

[0205] This processing mechanism caused severe data processing delays and resulted in blurred underwater topographic boundaries in specific construction areas, making it difficult to meet the real-time requirements for collision avoidance and accurate measurement during construction operations.

[0206] To address the aforementioned technical issues, this embodiment provides a real-time three-dimensional reconstruction system for underwater topography in reclamation areas based on telemetry data. This system is deployed on an industrial-grade server or edge computing node in the construction command center, and includes:

[0207] The data acquisition and filtering module is used to acquire multi-source time-series telemetry point cloud data and perform adaptive filtering, outputting an effective skeleton point set with a first density and an effective detail point set with a second density, wherein the first density is less than the second density. The basic grid initialization module is used to generate a basic macroscopic three-dimensional grid surface covering the entire domain based on the first frame of effective point set in the multi-source time-series telemetry point cloud data, and assign initial spatiotemporal attributes to each grid cell. The two-dimensional fine partitioning module is used to calculate the spatiotemporal joint topological gradient of each grid cell, divide the terrain into multiple structural types including smooth and uniform regions, abrupt change regions, and stable regions based on spatial structural features, and divide it into lifecycle states including basic state, evolving state, and solidified state based on temporal activity. The system completes two-dimensional fine partitioning; the base state update module performs adaptive temporal elevation updates on mesh cells in the smooth and uniform region of the base state while maintaining the mesh topology connectivity; the evolution state update module performs local mesh topology updates on mesh cells in the abrupt change region of the evolution state to generate high-density local patches; the solidified state simplification module performs progressive mesh simplification on mesh cells in the stable region of the solidified state using a half-edge folding algorithm to reduce mesh density and fall back to the base mesh resolution; and the topology stitching and rendering output module stitches all high-density local patches with the simplified mesh and the base mesh to obtain a topologically complete global 3D surface mesh and performs real-time rendering output.

[0208] In practical underwater engineering applications, the above functional modules are deeply integrated with specific physical devices and data flow links. The specific operating mechanism is as follows:

[0209] The data acquisition and filtering module directly interfaces with external sensors mounted on the surface survey vessel via a hardware link. Specifically, it receives underwater depth point clouds generated by a multibeam echo sounder (MBES), centimeter-level absolute spatial coordinates provided by a real-time dynamic carrier phase differential (RTK-GNSS) device, and ship roll, pitch, and yaw data provided by an attitude reference system (AHRS) via a wireless bridge or satellite communication link. Upon receiving the raw composite data, the module uses a built-in signal processing algorithm to filter out acoustic anomalies and reflection noise caused by underwater suspended sediment and sludge bubbles. Based on different spatial sampling intervals, it separates the point cloud stream into a first density effective skeleton point set for constructing macroscopic topography and a second density effective detail point set for characterizing local micro-topography, providing a clean data source for subsequent modeling.

[0210] The basic mesh initialization module runs in the system's host memory space. When the effective point set of the survey area is acquired for the first time, this module calls on the computing resources of the central processing unit (CPU) to construct a global basic macroscopic three-dimensional mesh surface through a spatial partitioning algorithm.

[0211] Meanwhile, this module instantiates an attribute object for each generated triangular mesh cell in the system's relational database or in-memory data structure, assigning initial spatiotemporal attribute parameters such as terrain complexity and spatiotemporal activity, thus establishing the skeletal foundation of the digital twin model.

[0212] The aforementioned two-dimensional fine-grained partitioning module is essentially the core logic scheduling engine of the system. This module continuously reads the latest input data and analyzes the topological gradient of the local mesh through matrix operations. In practical applications, this module can accurately identify which coordinate regions are within the mud spraying coverage area of ​​the dredger (i.e., the abrupt change zone of the evolutionary state), which regions are in a deep-water calm flow environment (i.e., the smooth and uniform zone of the basic state), and which regions are historical construction areas that have been completed and verified (i.e., the stable zone of the solidified state). By establishing this state machine model, the system avoids the indiscriminate global calculations found in traditional methods from the bottom up.

[0213] For the base state update module, for those gently sloping and uniform regions determined to be in the base state, the module directly accesses the Vertex Buffer Objects (VBOs) in the graphics processing unit (GPU) memory. Since the underwater topography in this region only undergoes slow sedimentation, the module only updates the Z-axis elevation coordinates of the vertices numerically, strictly locking the topological connectivity relationships in the Index Buffer Objects (IBOs), thus achieving zero consumption of topological computation power in this region while ensuring elevation accuracy.

[0214] The evolutionary state update module focuses on the active water area where construction operations are being carried out. When a dense set of valid detail points in the area is received, the module schedules concurrent computation threads to perform local constrained triangulation. Since dredging operations lead to steep underwater sand slopes, the module incorporates the natural angle of repose of the dredged fill as a physical geometric constraint during computation, eliminating unreasonable distorted meshes and generating high-density local patches to reflect the diffusion boundary and deposition morphology of underwater mudflows with the highest data fidelity.

[0215] The solidified state simplification module's practical function lies in long-term memory capacity management. As the construction cycle lengthens, the area of ​​the solidified stable region continuously expands. This module periodically scans mesh regions that have not evolved for a long time using a background thread, and merges redundant mesh vertices using a quadratic error measure and a half-edge folding algorithm. This prevents the system from experiencing memory overflow or system crashes due to years of data accumulation when facing large-scale land reclamation projects that often cover several square kilometers.

[0216] The topology stitching and rendering output module is responsible for seamlessly stitching together local meshes that have undergone different processing logics at spatial boundaries while maintaining volume conservation. After stitching, the data is pushed into the GPU's rendering pipeline. At the hardware output end, the real-time 3D scene generated by this module is projected onto the ultra-high-definition visualization screen in the construction command center for management personnel to make macro-level decisions. Simultaneously, it is distributed via data interface to the navigation control terminals of various construction vessels and the Electronic Chart System (ECDIS). Combined with distance-dependent dynamic level of detail (LOD) display technology, the system can provide smooth underwater 3D visual feedback on various terminals.

[0217] In summary, the real-time 3D reconstruction system disclosed in this embodiment achieves a technical effect of great engineering value through the collaborative work of the aforementioned modules. Based on the calculated spatiotemporal joint topological gradient, the system completes the two-dimensional fine partitioning of the terrain mesh, accurately matching computer computing resources with the physical evolution state of the underwater terrain.

[0218] In actual operation, the system updates only the elevation and maintains the mesh topology in smooth and uniform areas, avoiding redundant reconstruction overhead. For evolutionary abrupt change areas at the core of construction, it performs local topology updates, preserving critical geometric collision avoidance boundaries. For solidified stable areas, it performs progressive mesh simplification, proactively releasing redundant memory usage. This mechanism effectively solves the problems of wasted computing power and rendering latency caused by global unified processing in existing technologies. While maintaining the geometric accuracy of local abrupt change features, it significantly reduces the data load of overall 3D surface mesh rendering, effectively improving the scheduling efficiency of 3D terrain data processing and the operational stability of the system in large-scale underwater reclamation projects.

[0219] Example 5:

[0220] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0221] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0222] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0223] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0224] The memory 103 stores a computer program corresponding to the real-time three-dimensional reconstruction method for underwater topography in reclamation areas based on telemetry data according to the above embodiments of the present invention. This computer program is executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0225] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0226] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for real-time three-dimensional reconstruction of underwater topography in reclamation areas based on telemetry data, characterized in that, include: Acquire multi-source time-series telemetry point cloud data and perform adaptive filtering to output an effective skeleton point set with a first density and an effective detail point set with a second density, wherein the first density is less than the second density. Based on the first frame of valid point set in the multi-source time-series telemetry point cloud data, a basic macroscopic three-dimensional mesh surface covering the entire domain is generated, and initial spatiotemporal attributes are assigned to each mesh unit; The spatiotemporal joint topological gradient of each grid cell is calculated. Based on the spatial structural characteristics, the terrain is divided into multiple structural types, including smooth and uniform regions, abrupt change regions, and stable regions. Based on the temporal activity, it is divided into life cycle states, including basic state, evolutionary state, and solidified state, thus completing the two-dimensional fine partitioning. For the grid cells in the smooth and uniform region of the base state, perform adaptive temporal elevation updates while maintaining the grid topology connectivity. For the mesh cells in the mutation region of the evolutionary state, perform local mesh topology update to generate high-density local patches; For the mesh cells in the stable region that are in the solidified state, a half-side folding algorithm is used to perform progressive mesh simplification, reduce the mesh density and fall back to the basic mesh resolution; All high-density local patches are topologically stitched together with the simplified mesh and the base mesh to obtain a topologically complete global 3D surface mesh, which is then rendered in real time.

2. The method according to claim 1, characterized in that, The process of acquiring multi-source time-series telemetry point cloud data and performing adaptive filtering to output an effective skeleton point set with a first density and an effective detail point set with a second density includes: Extract the normalized echo signal-to-noise ratio and neighborhood elevation stability coefficient of each measurement point in the multi-source time-series telemetry point cloud data, and calculate the confidence weight of each measurement point by weighting it according to the preset signal-to-noise ratio weight ratio and elevation stability weight ratio. Outlier noise points are eliminated based on a preset three-condition judgment rule. The three conditions are: the absolute value of the deviation of the elevation of the measuring point from the average elevation of its nearest neighbor is greater than three times the standard deviation of the neighborhood elevation; the spatial density of the measuring point's neighborhood is lower than a preset density ratio of the average density of the point cloud in the entire region; and the confidence weight is less than a preset weight threshold. After removing the measurement points that meet any of the three conditions, the effective skeleton point set and the effective detail point set are separated and output according to the preset spatial sampling frequency.

3. The method according to claim 1, characterized in that, The process involves generating a basic macroscopic 3D mesh surface covering the entire region based on the first frame of valid points in the multi-source time-series telemetry point cloud data, and assigning initial spatiotemporal attributes to each mesh cell, including: The Delaunay triangulation algorithm is used to globally partition the effective point set of the first frame to generate the basic macroscopic three-dimensional mesh surface. The spatiotemporal attributes are extended to each grid cell within the basic macroscopic three-dimensional grid surface. The spatiotemporal attributes include terrain complexity, spatiotemporal activity, lifecycle state marker, and stable continuous frame count. During the system initialization phase, the spatiotemporal activity and stable continuous frame count of all grid cells are initialized to a preset starting value, and the lifecycle state flag is initialized to the base state.

4. The method according to claim 3, characterized in that, The calculation of the spatiotemporal joint topological gradient of each of the grid cells includes: The effective skeleton point set is projected and matched to the corresponding grid cell. A two-dimensional surface structure tensor matrix is ​​constructed using the grid cell as the unit and eigenvalue decomposition is performed to extract the first eigenvalue and the second eigenvalue, wherein the first eigenvalue is not less than the second eigenvalue. The ratio of the first eigenvalue to the average side length of the corresponding grid cell is used as the spatial topological gradient; The temporal topological gradient is calculated based on the average elevation of grid cells in a preset number of consecutive historical frames, and the weighted sum of the spatial topological gradient and the temporal topological gradient is used as the spatiotemporal joint topological gradient. Based on the spatiotemporal activity of the grid cells in the previous frame and the normalized value of the spatiotemporal joint topological gradient in the current frame, the temporal activity of the grid cells in the current frame is calculated by weighting using a temporal smoothing mechanism.

5. The method according to claim 4, characterized in that, The terrain is divided into multiple structural types based on spatial structural features, including gentle and uniform regions, abrupt change regions, and stable regions. Furthermore, based on temporal activity, it is divided into lifecycle states, including basic state, evolving state, and solidified state, thus completing a two-dimensional fine-grained partitioning process, including: The first eigenvalue and the second eigenvalue obtained by the two-dimensional surface structure tensor decomposition are compared with the first gradient threshold and the second gradient threshold, respectively, and the second gradient threshold is greater than the first gradient threshold. If all of the first feature values ​​are lower than the first gradient threshold, then the region is divided into the smooth and uniform region. If the first feature value is higher than the second gradient threshold, it is further divided into a linear steep slope region or a planar gradual change region according to the ratio between the first feature value and the second feature value. If both the first feature value and the second feature value are higher than the second gradient threshold and the local elevation of the corresponding grid is an extreme value, then it is divided into a point-like abrupt change region, wherein the linear steep slope region, the planar gradual change region and the point-like abrupt change region together constitute the abrupt change region; The lifecycle states of grid cells whose temporal activity is higher than the high activity threshold are classified as the evolutionary state; the lifecycle states of grid cells whose temporal activity is lower than the low activity threshold and whose stable continuous frame count reaches a preset stable frame count threshold are classified as the solidified state; the lifecycle states of the remaining grid cells that do not fall into the determination conditions of the evolutionary state and the solidified state are statistically classified as the basic state.

6. The method according to claim 2, characterized in that, The step of performing adaptive temporal elevation updates on grid cells within the smooth, uniform region of the base state includes: Obtain the confidence weight of the corresponding measurement point, the planar Euclidean distance factor between the measurement point and the corresponding grid vertex, and the terrain complexity factor of the neighborhood of the grid vertex; The confidence weight, the planar Euclidean distance factor, and the terrain complexity factor are multiplied to construct the interpolation weight of the target measurement point to the target grid vertex, and the interpolation elevation of the corresponding grid vertex is calculated by weighted average method. The calculated interpolated elevation and the historical elevation of the target grid vertex are subjected to time-series smoothing processing to update the final elevation of the grid cell vertices in the smooth and uniform region.

7. The method according to claim 1, characterized in that, The step of performing local mesh topology updates on mesh cells within the mutation region of the evolved state to generate high-density local patches includes: For the mutation region determined to be continuously evolving, the effective minutiae set within the local bounding box of the current frame is extracted as the new effective minutiae set, and new vertices are inserted one by one into the new effective minutiae set; before inserting the new vertices, slope constraint verification is performed. If it is determined that the tilt angle of all triangular facets in the neighborhood of the inserted vertex does not exceed the preset natural repose angle of the dredged fill, then the topology is updated by incremental vertex insertion and edge flipping. For newly added mutation zones determined to be non-continuously evolving, constrained Delaunay triangulation is performed to generate the high-density local patches. Specifically, linear steep slope zones are meshed along the terrain direction, isotropic uniform meshing is used for face-shaped gradual change zones, and radial meshing is performed with extreme points as the center for point-shaped mutation zones. During the meshing process, the natural angle of repose of the dredged fill is embedded as a hard constraint to eliminate triangular patches with excessive dip angles.

8. The method according to claim 1, characterized in that, The progressive mesh simplification using the half-side folding algorithm includes: Extract the mesh geometric error calculated based on the quadratic error measure and the structure preservation error calculated based on the matching degree between the mesh edge direction and the terrain orientation. Then, calculate the weighted sum of the mesh geometric error and the structure preservation error according to their respective preset weight coefficients, and use it as the composite folding error of the target mesh edge. Sort the composite folding errors from smallest to largest, and perform half-folding operations step by step from the geometric center of the stable region in the solidified state to the boundary of the stable region; Set a preset proportional constraint that the number of folded edges in a single frame does not exceed the total number of edges in the corresponding stable region, and keep the boundary vertex positions of the stable region fixed; when the average edge length of the mesh in the corresponding stable region expands to a preset multiple threshold of the edge length of the basic mesh, stop the folding simplification operation and reset the state of the mesh cells in the stable region to the basic state.

9. The method according to claim 1, characterized in that, Before performing a local mesh topology update on the mesh cells within the mutation region of the evolved state, the method further includes: The center coordinates of the bounding boxes of abrupt change regions in multiple consecutive frames are extracted to estimate the terrain expansion rate, and the terrain expansion rate is used to predict the position and expansion direction of the abrupt change region in the next frame. In the predicted expansion direction, the grid cells in the corresponding region are subdivided according to rules in advance to generate a transition grid with a density that is a preset multiple of the base grid; Feature vectors of terrain orientation are extracted, and anisotropic smoothing is performed on local patches using a direction-weighted Laplacian smoothing algorithm. A first smoothing weight is assigned along the terrain orientation, and a second smoothing weight is assigned along the terrain abrupt change direction perpendicular to the terrain orientation. The value of the first smoothing weight is greater than the value of the second smoothing weight.

10. The method according to claim 1, characterized in that, The process involves topologically stitching all high-density local patches with the simplified mesh and the base mesh to obtain a topologically complete global 3D surface mesh and performing real-time rendering output, including: Extract the boundary vertices at the intersection of the high-density local patch and the base mesh; The trapezoidal prism method was used to calculate the earthwork volume of the corresponding local area at the junction before and after the update, and the relative volume deviation rate was obtained. When the relative volume deviation rate exceeds the preset volume deviation threshold, the elevation linear compensation adjustment is performed on the corresponding boundary vertex until the relative volume deviation rate is less than or equal to the volume deviation threshold, and the topology seamless stitching is completed to output a three-dimensional surface mesh. The output 3D surface mesh is combined with a view distance-dependent hierarchical detail dynamic switching mechanism, and real-time rendering output is performed by culling or merging triangular faces with view distance values ​​greater than a preset view distance threshold.

Images