Real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scene

Abstract

The present application relates to the fields of surveying and mapping geographic information and three-dimensional terrain dynamic modeling, and specifically discloses a multi-source heterogeneous terrain data real-time fusion modeling method for surveying and mapping scenes, comprising the following steps: obtaining multi-source heterogeneous terrain data and device accuracy information and generating unified observation records, determining a space-time influence domain and a model unit according to a space-time overlap relationship and an error range, generating regionalized reliability according to the geometric difference between new observations and existing models, determining an update area and constraining local topological reconstruction with boundary nodes, and correcting the regionalized reliability or rejudging the model unit state according to the model residual. The present application limits the model area affected by new observations and implements differentiated updates according to the model unit state, reduces repeated processing of unchanged areas, maintains the connection relationship between the update area and adjacent grids through boundary constraints, and reduces the accumulation of false updates caused by measurement deviation through model residual feedback.

Description

Technical Field

[0001] This invention relates to the field of surveying and mapping geographic information and three-dimensional terrain dynamic modeling technology, and in particular to a method for real-time fusion modeling of multi-source heterogeneous terrain data for surveying and mapping scenarios. Background Technology

[0002] In the field of surveying and mapping geographic information and 3D terrain dynamic modeling technology, activities such as data acquisition, transmission, coordinate transformation, spatial registration, model organization, storage and retrieval, and display are usually carried out around the landform, land feature boundaries, control points, and 3D terrain models. Surveying instruments, UAV photography equipment, laser scanning equipment, mobile surveying terminals, and servers cooperate according to the relationship between field data acquisition, data transmission, coordinate calculation, and result retrieval.

[0003] Among them, real-time fusion modeling of multi-source heterogeneous terrain data for surveying and mapping scenarios refers to a method for unified three-dimensional representation of terrain information with different sources, formats and accuracies in surveying and mapping operations. Generally, GNSS receivers, total stations, UAV cameras or laser scanning equipment acquire measurement points, images and point cloud records, which are transmitted to computing devices via wireless links, data interfaces or storage media. The data is then converted into formats, unified in coordinates, registered and stitched, triangulated, textured, locally updated and stored according to coordinate reference, spatial overlap range and proximity of points.

[0004] Existing 3D terrain modeling performs coordinate unification, spatial registration, and grid construction on heterogeneous surveying data, and updates the model according to data batches or coverage areas. When continuous data is accessed, unchanged areas are still read and recalculated. The positioning accuracy and acquisition time of different devices lack regional correlation. When new observations differ from existing models in location, elevation, or slope, direct replacement or average fusion makes it difficult to distinguish between terrain changes and measurement deviations. Local reconstruction affects the connection relationship between the updated area boundary and adjacent grids. The update results lack subsequent observation correction, resulting in repeated calculations and local topological fluctuations. Summary of the Invention

[0005] The purpose of this invention is to provide a real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios. This method addresses the differences in coordinate reference, acquisition time, positioning accuracy, elevation accuracy, and spatial distribution between data from different continuously accessed surveying and mapping equipment. It limits the actual impact range of new observations on the 3D terrain model, distinguishes between data missing, model consistency, terrain changes, and pending confirmation states, and implements local model updates under boundary connection constraints. Simultaneously, it utilizes subsequent surveying and mapping data to provide feedback correction for the update results, thereby reducing redundant processing of unchanged model areas and local topological errors caused by measurement deviations.

[0006] To achieve the above objectives, this invention provides a real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios, comprising the following steps: Acquire multi-source heterogeneous terrain data, positioning accuracy, and elevation accuracy output by surveying equipment, and generate a unified observation record that includes spatial range, acquisition time, positioning accuracy, and elevation accuracy.

[0007] The spatiotemporal influence domain is determined based on the spatiotemporal overlap relationship and error range between the unified observation records and the three-dimensional terrain model, and the spatiotemporal influence domain is divided into model units according to the spatial index of the three-dimensional terrain model.

[0008] New observation records corresponding to each model unit are extracted from the unified observation records, and existing model records corresponding to each model unit are extracted from the three-dimensional terrain model. The position difference, elevation difference, and slope difference between the new observation records and the existing model records are calculated to obtain the geometric difference value. Regional confidence is generated based on the unified observation records and the geometric difference value.

[0009] When the existing model record is missing, the corresponding model unit is identified as a supplementary unit. When the geometric difference meets the consistency condition, the existing model record of the corresponding model unit is retained. When the geometric difference does not meet the consistency condition, and the corresponding difference meets the spatiotemporal continuity and multi-source support conditions, the corresponding model unit is identified as a reconstruction unit. When the spatiotemporal continuity and multi-source support conditions are not met, the fusion of the corresponding model unit is temporarily suspended.

[0010] The supplementary unit and the reconstruction unit are merged to form an updated region. The boundary nodes adjacent to the updated region and the non-updated model units are extracted. Using the boundary nodes as constraints, local topology reconstruction is performed on the vertices, edges, faces and terrain attributes inside the updated region.

[0011] The model residual of the updated region is calculated based on subsequent survey data. The regionalization confidence is corrected based on the model residual. When the model residual exceeds the residual condition, the state of the corresponding model unit is re-determined.

[0012] Furthermore, when generating the unified observation record, the surveying equipment identifier, original coordinate reference, acquisition time and terrain feature record corresponding to the multi-source heterogeneous terrain data are parsed respectively, and the terrain feature record is converted to the target coordinate reference used by the three-dimensional terrain model to form the target terrain feature record; The spatial range is determined based on the spatial coordinates recorded by the target terrain features. A source index pointing to the original data storage location is established, and the surveying equipment identifier, the acquisition time, the spatial range, the positioning accuracy, the elevation accuracy, the target coordinate reference, and the source index are written into the unified observation record.

[0013] Furthermore, according to the spatial index granularity of the three-dimensional terrain model, the target terrain feature records are spatially grouped and spatial index keys are generated. A time index key is generated according to the synchronous time window to which the acquisition time belongs. The spatial index key, the time index key, and the source index are written into the unified observation record. Multi-source association relationships are established for target terrain feature records with the same spatial index key and time index key but different source indices.

[0014] Furthermore, when determining the spatiotemporal influence domain, the spatial range is expanded in a plane according to the positioning accuracy, and the elevation error range is determined according to the elevation accuracy to form an error expansion range; A time correlation window is established based on the acquisition time, and model records that intersect with the error expansion range are retrieved from the spatial index of the three-dimensional terrain model; When the update time of the model record is within the time association window, the corresponding model region is determined as a candidate model region, and the union of the candidate model region and the error expansion range is determined as the spatiotemporal influence domain; when there is no model record that meets the time condition, the error expansion range is determined as the spatiotemporal influence domain.

[0015] Furthermore, when dividing the model units, spatial index nodes that intersect with the spatiotemporal influence domain are extracted, and spatial index nodes that are completely located within the spatiotemporal influence domain and meet the preset index granularity are determined as internal model units; The spatial index nodes that cross the boundaries of the spatiotemporal influence domain are subdivided step by step to the smallest spatial index granularity. The intersecting parts after subdivision are determined as boundary model units, and spatial index identifiers and adjacency relationships are written for the internal model units and the boundary model units.

[0016] Furthermore, when calculating the geometric difference, terrain feature records whose spatial index identifier is consistent with each of the model units and whose acquisition time is within the time correlation window are selected to form the new observation record; Read terrain data with consistent spatial index identifiers to form the existing model record; Establish observation pairs at the same location according to the spatial coordinate proximity relationship, calculate the position difference and the elevation difference based on each observation pair at the same location, and calculate the slope difference based on the local terrain surface fitted by the neighboring terrain points.

[0017] Furthermore, when generating the regional credibility, the accuracy credibility level is determined based on the positioning accuracy and the elevation accuracy, the timeliness credibility level is determined based on the interval between the acquisition time and the update time of the existing model record, and the position difference, the elevation difference and the slope difference are compared with the corresponding consistency limits to determine the geometric consistency level. An initial regionalized confidence level is formed based on the lowest level among the accuracy confidence level, the timeliness confidence level, and the geometric consistency level. The initial regionalized confidence level is adjusted when the determination of the direction of terrain change is consistent among new observation records from different sources, thus forming the regionalized confidence level.

[0018] Furthermore, when determining the state of a model unit, the supplementary unit is determined when the existing model record is missing; a stable unit is determined when the position difference, elevation difference, and slope difference do not exceed the corresponding consistency limit, and the regionalization confidence is not lower than the confidence limit. When at least one difference exceeds the corresponding consistency limit, the number of consecutive units with the same difference direction in adjacent model units and the number of newly added observation records with different sources but the same difference direction are counted. When both of these numbers reach the corresponding support limit, the reconstructed unit is determined; otherwise, it is determined as a unit to be confirmed.

[0019] Furthermore, during the local topology reconstruction, the spatially adjacent supplementary units and reconstruction units are connected and merged to form the updated region. Shared vertices and shared edges are extracted from the stable units adjacent to the updated region to form a set of boundary nodes. The spatial coordinates of each boundary node in the boundary node set and its connection relationship with the non-updated model unit are retained. Based on the newly added observation records corresponding to the supplementary unit and the filtered observation records corresponding to the reconstruction unit, the vertices, edges and faces inside the updated region are reconstructed and the corresponding terrain attributes are written.

[0020] Furthermore, according to the acquisition time sequence, the subsequent mapping data is correlated with the current vertex coordinates, elevation, and slope of the updated area to calculate the position residual, elevation residual, and slope residual. The model residual is then generated based on the direction of change and the number of times these three types of residuals exceed limits within the continuous observation window. When the model residual does not exceed the residual condition, the regionalization confidence is corrected and the current state is retained; When the model residual exceeds the residual condition, the corresponding model unit is reinstated into the state determination. The current round of feedback correction is terminated when none of the model residuals within the continuous observation window exceed the residual condition; the next round of feedback correction is initiated when any of the model residuals exceeds the residual condition again.

[0021] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, multi-source heterogeneous terrain data is correlated with positioning accuracy, elevation accuracy, and acquisition time through unified observation records. The spatiotemporal influence domain is limited to the fusion range based on spatial overlap, temporal correlation, and error range, so that the processing is focused on the model units affected by the new observations, reducing repeated readings and reconstructions of unchanged areas. Position difference, elevation difference, and slope difference, together with regionalization reliability, participate in the model unit state determination, enabling supplementary units and reconstructed units to enter the corresponding update path. Boundary nodes constrain the local topology reconstruction within the update area, maintaining the connection between the update area and the non-updated model units. The model residuals formed by subsequent surveying data are used to correct the regionalization reliability and trigger state re-determination, reducing the accumulation of erroneous updates caused by discrete measurement deviations. Attached Figure Description

[0022] Figure 1 This is a flowchart of the real-time fusion modeling method for multi-source heterogeneous terrain data of the present invention; Figure 2 This is a schematic diagram illustrating the unified observation record generation and multi-source correlation of the present invention; Figure 3 This is a schematic diagram illustrating the determination of the spatiotemporal influence domain and the division of model units in this invention; Figure 4 This is a schematic diagram illustrating the generation of geometric differences and regionalized credibility at the model unit level in this invention. Figure 5 This is a schematic diagram of the state determination and update region formation of the model unit in this invention; Figure 6 This is a schematic diagram of the boundary constraint local topology reconstruction and model residual feedback of the present invention. Detailed Implementation

[0023] This embodiment relates to a real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios. The multi-source heterogeneous terrain data includes photogrammetric points with three-dimensional spatial coordinates, laser point clouds, elevation points, contour line sampling points, and digital elevation model grid points, as well as image texture data that establishes a correspondence with the three-dimensional spatial coordinates.

[0024] A unified observation record is used to record spatial range, acquisition time, positioning accuracy, elevation accuracy, coordinate datum, data type, and source index. Model cells represent terrain regions in the current fusion process after the 3D terrain model is divided according to spatial indexes; Regionalized confidence level represents the confidence level of newly added observation records in participating in state determination and terrain updating within the corresponding model unit; Model residuals represent the differences between the updated 3D terrain model and subsequent survey data in terms of location, elevation, and slope.

[0025] For raw aerial photographs that contain only pixel values ​​and not three-dimensional spatial coordinates, the raw aerial photographs do not directly participate in the calculation of position difference, elevation difference and slope difference. They are only used as texture data to establish a relationship with the model surface after the corresponding terrain surface is formed. For photogrammetric points whose three-dimensional spatial coordinates and elevations have been obtained through photogrammetric processing, the photogrammetric points are used as terrain geometric data in subsequent fusion.

[0026] Contour data is sampled at intervals according to the spatial trajectory of contour lines, and each sampling point inherits the elevation attribute of the corresponding contour line.

[0027] The real-time fusion referred to in this embodiment means that as soon as each batch of surveying and mapping data arrives, the spatiotemporal influence domain retrieval, model unit state determination, and local topology update are triggered, without waiting for the completion of all surveying and mapping tasks before uniformly reconstructing the complete three-dimensional terrain model.

[0028] Real-time fusion is not limited to a fixed millisecond processing time. The processing of a single batch of data is performed in the order of arrival, and incremental processing continues after the next batch of data arrives.

[0029] In a set of feasible parameter configurations for hilly surveying areas, the surveying area ranges from 1.2 square kilometers to 2 square kilometers, with an elevation difference of 80 meters to 220 meters. The input data includes aerial photographic images with a ground resolution of 0.04 meters to 0.08 meters, laser point clouds with an average point density of 8 to 20 points per square meter, satellite positioning measurement data with a horizontal positioning accuracy of 0.02 meters to 0.05 meters and an elevation accuracy of 0.03 meters to 0.08 meters, and existing digital elevation models with a grid spacing of 0.5 meters to 1 meter.

[0030] The above values ​​constitute a set of execution parameters for this embodiment and do not represent test conclusions or performance results. When using other surveying accuracies and spatial resolutions, the spatial retrieval range, consistency limits, and residual limits should be adjusted according to the accuracy requirements of the corresponding surveying task, and the relationship between the positioning accuracy, elevation accuracy, and the corresponding judgment limits should be kept consistent.

[0031] Please see Figures 1 to 6 This embodiment provides a real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios, which is suitable for continuous access and local updates of photogrammetric points, laser point clouds, elevation points, contour line sampling points, digital elevation model grid points, and image texture data.

[0032] Using multi-source heterogeneous terrain data and equipment accuracy information output by surveying equipment as input, data from different sources are converted into unified observation records with unified spatial and temporal identifiers. The spatiotemporal influence domain of this round of processing is determined by the spatiotemporal relationship between the unified observation records and the current three-dimensional terrain model. Model units are formed within the spatiotemporal influence domain, and geometric difference calculation, regional credibility generation, model unit state determination, local topology reconstruction, and model residual feedback are completed.

[0033] Each step is executed in the order that the result of the previous processing is used as the input for the next processing. Supplementary units and reconstructed units enter the update region, stable units retain the existing model records, and differences in spatiotemporal continuity or multi-source support enter a suspended fusion state. Subsequent mapping data is used to verify the update region and trigger regional credibility correction or re-determination of model unit status.

[0034] During continuous mapping operations, a batch of multi-source heterogeneous terrain data can be generated with an access cycle of 30 to 120 seconds. The coverage area of ​​a single batch of data is 30,000 to 80,000 square meters. Based on 8 to 20 laser points per square meter, a single batch of laser point cloud contains approximately 240,000 to 1,600,000 laser points.

[0035] The number of aerial photographs is determined based on flight altitude, camera field of view, and overlap, and a fixed number of images is not a necessary condition for this embodiment.

[0036] The basic grid spacing of the 3D terrain model can be set from 0.5 meters to 1 meter, and a single batch of data only triggers the local model area that intersects with the newly added observation coverage area to participate in the processing.

[0037] S1 acquires multi-source heterogeneous terrain data, positioning accuracy, and elevation accuracy output by surveying equipment, and generates a unified observation record.

[0038] S101 performs source identification, field integrity check, and content parsing on the data output by the surveying equipment.

[0039] Multi-source heterogeneous terrain data respectively carry the three-dimensional coordinates and image texture index of photogrammetric points, the three-dimensional coordinates and reflection intensity of laser point clouds, the planar coordinates and elevation of elevation points, the spatial trajectory and elevation attributes of contour lines, or the grid position and elevation record in a digital elevation model. Each set of multi-source heterogeneous terrain data is also associated with the surveying equipment identification, original coordinate datum, acquisition time, positioning accuracy, and elevation accuracy.

[0040] Positioning accuracy and elevation accuracy are obtained from the accuracy field of the surveying equipment output file, equipment calibration record, or surveying task configuration record. When multiple accuracy sources exist for the same batch of data, the equipment output accuracy record directly associated with the corresponding acquisition time is used first; if the equipment output file does not contain an accuracy field, the accuracy value of the corresponding equipment and corresponding operation mode in the surveying task configuration record is used.

[0041] Data records that proceed to subsequent processing should include data type, source identifier, acquisition time, original coordinate reference, and coordinate fields that can determine spatial location.

[0042] Data records lacking any required field will not be included in the unified observation record for this round. For duplicate records with the same source identifier, collection time, and spatial coordinates, only one record will be retained. For records from the same source with different spatial coordinates at the same collection time, records will be retained separately according to different terrain points.

[0043] Aerial photographic images are identified by the number of flights taken and the image sequence number. Laser point clouds are identified by the scanning equipment and the scanning time period. Elevation points are identified by the station and the observation time.

[0044] Each parsed record retains the acquisition time accurate to the second. When the timestamp difference of continuously acquired data exceeds the corresponding acquisition period, the data is organized according to different data batches.

[0045] The obtained topographic feature records maintain the correspondence between the original data and the surveying equipment identification, original coordinate datum, acquisition time, positioning accuracy, and elevation accuracy, so that subsequent processing can trace back the source and accuracy conditions of each observation.

[0046] S102 transforms the terrain feature record from the original coordinate datum to the target coordinate datum used by the three-dimensional terrain model, forming the target terrain feature record.

[0047] Coordinate transformation parameters are obtained from the coordinate transformation configuration of the surveying project, control point results, or external orientation and positioning records corresponding to the acquisition equipment.

[0048] When using control point results to determine coordinate transformation parameters, the control points should simultaneously have coordinates under the original coordinate datum and coordinates under the target coordinate datum.

[0049] Coordinate transformation uses the spatial coordinates in the terrain feature records as the processing object. After transformation, the corresponding acquisition time, positioning accuracy, elevation accuracy, data type and surveying equipment identification are retained, without changing the source correspondence of the original data.

[0050] For photogrammetric points, the target terrain feature record maintains the correspondence between the three-dimensional spatial coordinates and the image texture index; for laser point clouds, elevation points and contour line sampling points, the target terrain feature record maintains the target spatial coordinates and elevation attributes of each terrain point. For digital elevation model data, the target terrain feature records maintain the correspondence between grid positions and elevation records.

[0051] After completing the coordinate transformation, use control points with the target coordinate reference coordinates to check the transformation results.

[0052] When the horizontal deviation of the control point does not exceed 0.05 meters and the vertical deviation does not exceed 0.08 meters, the target terrain features are recorded and marked as valid for coordinate transformation. When any deviation exceeds the corresponding boundary, the corresponding data batch will be marked as a coordinate transformation pending verification state, and the corresponding target terrain features will not be recorded as input for this round of geometric fusion.

[0053] Data in the coordinate transformation pending verification state retains the original source index, and the target terrain feature record is generated only after the coordinate transformation parameters are redefined.

[0054] The target coordinate datum and the spatial index of the 3D terrain model use the same spatial reference, enabling the target terrain feature records to directly participate in spatial extent determination and model unit retrieval.

[0055] S103, determine the spatial range based on the spatial coordinates recorded by the target terrain features, and establish a source index.

[0056] The spatial extent is composed of the minimum enclosing boundary of the target terrain features recorded in the plane direction and the recording range in the elevation direction, and is used to represent the terrain area that can be affected by the corresponding multi-source heterogeneous terrain data.

[0057] The source index points to the original data file and the location of the data blocks within the file corresponding to the target terrain feature record, and is associated with the surveying equipment identifier, so that the corresponding observation content can be retrieved according to the model unit during the local topology reconstruction stage.

[0058] For photogrammetric points, a single unified observation record can correspond to a set of photogrammetric points formed by an image or a group of photogrammetric points with a common time range; for laser point clouds, a single unified observation record can correspond to a point cloud data block with a coverage area of ​​2,500 square meters to 10,000 square meters; for elevation points and contour line sampling points, a single unified observation record can correspond to a set of records from the same station, the same acquisition time period, or the same spatial index range.

[0059] The source index can at least locate the original file, the start and end positions of the data blocks, and the mapping equipment identifier.

[0060] The mapping equipment identifier, data type, acquisition time, spatial range, positioning accuracy, elevation accuracy, target coordinate benchmark, and source index are written into the unified observation record. The unified observation record carries the spatial field, time field, accuracy field, data type field, and source field required for this round of fusion.

[0061] S104, organizes unified observation records according to the spatial index granularity of the three-dimensional terrain model.

[0062] The spatial index identifier is determined based on the spatial coordinates recorded by the target terrain features, and the spatial index identifier is written into the spatial index key. The time index key is generated according to the synchronization time window to which the acquisition time belongs.

[0063] The synchronization time window is used to organize observation records that can participate in the same round of time series comparison within the same time range without changing the original acquisition time of the observation records.

[0064] The first-level spatial index range of the 3D terrain model can be set to 50 meters by 50 meters, and the minimum spatial index granularity can be set to 5 meters by 5 meters. Each first-level spatial index node is divided into two sub-regions along two planar coordinate directions, thus forming four next-level spatial index nodes with equal areas. The division process is repeated until the side length of the sub-node reaches 5 meters.

[0065] The synchronization time window can be set to 5 minutes. Multi-source associations are established for target terrain feature records with the same spatial and temporal index keys but different source indices. Multi-source associations are established when at least two different types of data from photogrammetric points, laser point clouds, or ground elevation observations exist within the same minimum spatial index and in the same synchronization time window.

[0066] Multi-source correlation does not pre-average or replace data from different sources; it only records the correspondence between observations from different sources within the same spatial and temporal range, for subsequent multi-source support condition determination.

[0067] The unified observation record formed by S1 keeps the spatial location, acquisition time sequence, observation accuracy, data type and data source within the same record structure, providing input for spatiotemporal influence domain retrieval, geometric difference calculation and regional credibility generation.

[0068] S2. Based on the spatiotemporal overlap relationship and error range between the unified observation records and the three-dimensional terrain model, the spatiotemporal influence domain is determined, and the spatiotemporal influence domain is divided into model units according to the spatial index of the three-dimensional terrain model.

[0069] S201 reads the spatial range, acquisition time, positioning accuracy, and elevation accuracy from the unified observation record.

[0070] The planar error boundary corresponding to the positioning accuracy is extended outwards to form the planar extension boundary; the elevation error interval is determined according to the elevation error boundary corresponding to the elevation accuracy. The planar extension boundary and the elevation error interval together constitute the error extension range.

[0071] The horizontal expansion distance is no less than twice the positioning accuracy, and the vertical expansion distance is no less than twice the elevation accuracy. In a parameter configuration with a positioning accuracy of 0.03 meters and an elevation accuracy of 0.05 meters, the horizontal expansion distance is 0.10 meters outward, and the vertical expansion distance is 0.15 meters upward and downward, respectively. The horizontal expansion distance does not exceed the side length of the smallest model unit, and the vertical expansion distance does not exceed twice the elevation difference consistency limit used in this embodiment.

[0072] The error extension range is used to cover model areas affected by positioning and elevation errors that correspond to new observations, while excluding model areas that are not spatially related to new observations from this round of processing.

[0073] S202, establish a time correlation window based on the acquisition time of the newly added observation record, and retrieve model records that intersect with the error expansion range from the spatial index of the three-dimensional terrain model.

[0074] The time correlation window is set before the acquisition time of the newly added observation record, and the update time of the model record must not be later than the acquisition time of the newly added observation record. In this embodiment, the model record meets the time correlation condition when the difference between the update time of the model record and the acquisition time of the newly added observation record does not exceed 5 minutes; for construction surveying or slope surveying areas with high terrain change frequency, the time correlation window can be shortened to 2 minutes.

[0075] The regions corresponding to model records that meet the spatial intersection and temporal correlation conditions are identified as candidate model regions. If a model record intersects with the error expansion range but its update time is earlier than the start time of the temporal correlation window, it is not considered a candidate model region, but is still retained as an existing model record in the 3D terrain model for subsequent reading and comparison according to the spatial index.

[0076] S203, when a candidate model region exists, the union of the candidate model region and the error extension range is determined as the spatiotemporal influence domain; when no model record satisfies the time correlation condition, the error extension range is determined as the spatiotemporal influence domain.

[0077] The spatiotemporal influence domain is jointly defined by the coverage of newly added observations, measurement error boundaries, and the temporal correlation of existing model records, serving as the spatial boundary for this round of model unit division and fusion processing. Regions in the 3D terrain model located outside the spatiotemporal influence domain are not included in this round of geometric difference calculation and local topology reconstruction.

[0078] When the coverage area of ​​a single batch of data is between 30,000 and 80,000 square meters, the spatiotemporal influence domain corresponds to 12 to 32 primary spatial index ranges. Divided according to the smallest model unit of 5 meters by 5 meters, the spatiotemporal influence domain corresponds to approximately 1,200 to 3,200 smallest model units.

[0079] S204: Extract spatial index nodes that intersect with the spatiotemporal influence domain according to the spatial index hierarchy of the three-dimensional terrain model.

[0080] Spatial index nodes that are completely located within the spatiotemporal influence domain and have an index granularity of 5 meters by 5 meters are identified as internal model units. For spatial index nodes that cross the boundary of the spatiotemporal influence domain, they are subdivided step by step according to the four-equal division method described in S104 until the minimum spatial index granularity of 5 meters by 5 meters is reached. The subdivided nodes that intersect with the spatiotemporal influence domain are identified as boundary model units.

[0081] Spatial index identifiers are written for internal and boundary model units. When two model units share a complete edge, they are recorded as edge-adjacent; when two model units only touch at a corner, they are recorded as corner-adjacent. Subsequent spatial continuity and updated region connectivity determinations use edge adjacency; corner adjacency is only used for neighborhood queries and is not used as a basis for connectivity merging.

[0082] Spatial index identifiers are used to extract corresponding data from unified observation records and 3D terrain models, while adjacency relationships are used to determine the spatial continuity of differences and to extract boundary nodes of updated regions.

[0083] S2 limits the processing range by using error expansion range, unidirectional temporal correlation window and spatial index hierarchy, so that subsequent difference calculation and reconstruction operations are concentrated in the model cells affected by the new observations.

[0084] S3 extracts new observation records corresponding to each model unit from the unified observation records, extracts corresponding existing model records from the 3D terrain model, calculates geometric differences, and generates regionalized credibility.

[0085] S301. Based on the spatial index identifier of the model unit, select target terrain feature records with consistent spatial index keys and acquisition times within the synchronization time window from the unified observation records to form new observation records; read the current vertex coordinates, elevation, slope and terrain attribute records from the three-dimensional terrain model according to the same spatial index identifier to form existing model records.

[0086] New observation records retain their source index, data type, acquisition time, positioning accuracy, and elevation accuracy. Existing model records retain their model update time and spatial index identifier. For model cells for which no existing model records exist, a geometric comparison between new observation records and existing model records is not performed; instead, the record missing status is passed to S4.

[0087] For a 5m x 5m model unit, newly added laser point cloud observation records can include 200 to 500 topographic points, while photogrammetric point records can include 50 to 200 topographic points. New observation records are entered into geometric comparison when the number of newly added observation points within the model unit is no less than 20 and the effective coverage area reaches 30% of the model unit area. If the above conditions are not met, the corresponding newly added observation records are retained as records to be supplemented, without forming geometric differences, and the corresponding model unit is transferred to the pending confirmation process.

[0088] S302. For model units that have both new observation records and existing model records, establish observation pairs at the same location according to the spatial coordinate proximity relationship.

[0089] Using the target spatial coordinates in the newly added observation records as a reference, the model positions within the positioning accuracy range are retrieved from the existing model records. The planar retrieval radius for observation pairs at the same location can be set to two to three times the positioning accuracy, and limited to the range of 0.05 meters to 0.20 meters.

[0090] When a new observation point corresponds to multiple candidate model locations, they are first sorted in ascending order of planar distance, and then the model location with the smallest distance that has not yet been paired with other new observation points is selected. Once a model location has formed a pair of observations at the same location, it will not be repeatedly paired with other new observation points from the same source and at the same acquisition time.

[0091] When the number of observation pairs formed at the same location within a model unit reaches 60% of the number of newly added observation points, the corresponding model unit is marked as having complete geometric comparison; when it is less than 60%, the corresponding model unit is marked as having incomplete geometric comparison and enters the pending confirmation process.

[0092] S303 addresses the positional, elevation, and slope differences between observation pairs at the same location.

[0093] By comparing the planar position of the newly added observation with the position of the existing model, the planar distance of each pair of observations at the same position is obtained; by comparing the elevation of the newly added observation with the elevation of the existing model, the elevation difference with direction is obtained for each pair of observations at the same position.

[0094] Extract topographic points from the neighborhood of each observation pair at the same location. The neighborhood range is 1 to 2 meters around the observation point at the same location, and should contain at least three topographic points that are not on the same straight line. The plane that minimizes the overall vertical deviation of the neighborhood topographic points from the fitted plane is used as the local topographic surface, and the slope is determined based on the inclination direction and degree of the local topographic surface. If there are fewer than three neighborhood topographic points or the topographic points are on the same straight line, no slope difference is formed, and the corresponding model unit is marked as having incomplete geometric comparison and enters the pending confirmation process.

[0095] The median of the planar distances between all valid observation pairs at the same location within a model element is determined as the positional difference of the model element; the median of the absolute values ​​of all elevation differences with direction is determined as the elevation difference of the model element, and the direction of elevation change is determined by the sign of the median of the elevation differences with direction; the median of the absolute values ​​of all slope differences is determined as the slope difference of the model element, and the direction of slope change is determined by the median direction of the slope change. The positional difference, elevation difference, and slope difference together constitute the geometric difference at the model element level.

[0096] The consistency limit for positional difference is set at 0.15 meters, the consistency limit for elevation difference is set at 0.20 meters, and the consistency limit for slope difference is set at 3 degrees. For areas with gentle slope changes, such as roads and site areas, the consistency limit for slope difference can be set at 2 degrees; for hilly slopes or embankment areas, the consistency limit for slope difference can be set at 3 to 5 degrees.

[0097] S304 reads positioning accuracy, elevation accuracy, and acquisition time from the unified observation record and generates regional reliability.

[0098] The accuracy reliability level is divided into high reliability, medium reliability, and low reliability. A positioning accuracy of no more than 0.05 meters and an elevation accuracy of no more than 0.08 meters is classified as high reliability; a positioning accuracy greater than 0.05 meters but not exceeding 0.10 meters, or an elevation accuracy greater than 0.08 meters but not exceeding 0.15 meters, is classified as medium reliability; and anything exceeding the medium reliability level boundary is classified as low reliability.

[0099] When the interval between the data collection time and the update time of the existing model record is no more than 2 minutes, it is determined to be of high timeliness and reliability level; when the interval is greater than 2 minutes but no more than 5 minutes, it is determined to be of medium timeliness and reliability level; when it exceeds 5 minutes, the time association condition of this embodiment is not met.

[0100] The position difference, elevation difference, and slope difference are compared with their corresponding consistency limits. If none of the three differences exceed the corresponding consistency limit, the first geometric consistency level is determined; if one or two differences exceed the corresponding consistency limit, the second geometric consistency level is determined; if all three differences exceed the corresponding consistency limit, the third geometric consistency level is determined.

[0101] The first geometric consistency level corresponds to a high geometric confidence level, the second geometric consistency level corresponds to a medium geometric confidence level, and the third geometric consistency level corresponds to a low geometric confidence level. Following a unified order from high to low confidence level (high confidence level, medium confidence level, low confidence level), the accuracy confidence level, timeliness confidence level, and geometric confidence level are compared, and the lowest confidence level is used as the initial regionalized confidence level.

[0102] For new observation records with the same spatial and temporal index keys but different source indices, compare the directions of elevation and slope change from different sources. When the directions of elevation and slope change are consistent between at least two different sources, and there are no observations in the opposite direction with a high confidence level, increase the initial regionalization confidence by one level, but not exceeding the high confidence level. When there are observations in the opposite direction with a high confidence level, do not increase the initial regionalization confidence, and retain the corresponding model unit as a candidate for confirmation. When only there are observations in the opposite direction with a low confidence level, maintain the initial regionalization confidence.

[0103] Regionalized credibility is tied to model units and is not used as a fixed credibility level for the entire data source.

[0104] The geometric difference at the model unit level and the regional credibility generated by S3 are used together to determine the state of the model unit, avoiding the replacement of existing model records based solely on the acquisition time or the difference of a single terrain point.

[0105] S4 determines the state of model units based on geometric differences and regionalized credibility, forming supplementary units, stable units, reconstructed units, or units to be confirmed.

[0106] S401, Check if there is an existing model record in the model unit.

[0107] When existing model records are missing and new observation records that meet the S301 data integrity conditions exist, the corresponding model unit is designated as a supplementary unit. A supplementary unit indicates that the 3D terrain model lacks comparable and retainable terrain records at the corresponding spatial index location, and the new observation records serve as the source of local terrain data.

[0108] If existing model records are missing but newly added observation records do not meet the data integrity condition of S301, the corresponding model unit will not be identified as a supplementary unit. Instead, the newly added observation records will be retained and the corresponding model unit will be identified as a unit to be confirmed until subsequent observation records meet the data integrity condition.

[0109] S402, when existing model records exist, compare the position difference, elevation difference, and slope difference with the corresponding consistency limits.

[0110] If none of the three differences exceed the corresponding consistency limit, the existing model vertices, edges, and faces of the corresponding model unit are retained. When the regionalization confidence level is not lower than the medium confidence level, the corresponding model unit is identified as a stable unit, and the acquisition time, source index, and observation accuracy attributes in the existing model records are updated; when the regionalization confidence level is low confidence level, the existing model vertices, edges, and faces are still retained, but the corresponding model unit is identified as a unit to be confirmed, and the terrain attributes are not updated using newly added observation records with low confidence.

[0111] S403: When at least one of the location difference, elevation difference, or slope difference exceeds the corresponding consistency limit, perform a spatiotemporal continuity and multi-source support condition determination on the difference.

[0112] Spatial continuity is determined by counting the number of continuous model units with the same direction of elevation change and slope change based on the adjacent relationship of the edges. The spatial support limit is set to at least 3 adjacent model units. For a 5m by 5m model unit, the continuous change length corresponding to 3 continuous model units is not less than 15m.

[0113] Temporal continuity is determined based on the records of the same spatial index identifier in consecutive acquisition cycles. The temporal continuity condition is met when the same model unit maintains the same direction of elevation change in two or more consecutive acquisition cycles, and the interval between adjacent acquisition cycles does not exceed the upper limit of the corresponding data access cycle.

[0114] The multi-source support condition is determined based on the source index. The multi-source support condition is met when at least two different surveying sources have the same direction of elevation change within the corresponding model unit, and there are no high-confidence level observations with opposite directions of slope change.

[0115] S404. For model units that simultaneously meet the conditions of spatial continuity, temporal continuity, and multi-source support, and whose regional credibility is not lower than the medium credibility level, the corresponding model unit is identified as a reconstruction unit; for model units that do not meet any of the conditions, the corresponding model unit is identified as a unit to be confirmed and fusion is temporarily suspended.

[0116] The unit to be confirmed retains the correspondence between newly added observation records and existing model records, but is not written into the 3D terrain model in this round. After subsequent mapping data is written into the same spatial index key, S302 to S404 are executed again.

[0117] Supplementary units, stable units, reconstructed units, and units to be confirmed are mutually exclusive states, and the same model unit maintains only one state in the same round of processing.

[0118] S4 uses the absence of existing model records as a preliminary judgment, geometric consistency and regional credibility as the basis for judging the stable state, and spatiotemporal continuity and multi-source support as the basis for judging the reconstruction state, so that missing data supplementation, model preservation, terrain reconstruction and delayed fusion enter the corresponding processing paths respectively.

[0119] S5 merges the supplementary unit and the reconstructed unit into an update region, extracts the boundary nodes, and performs local topology reconstruction.

[0120] S501, based on the edge adjacency relationship of the model unit, connect and merge the spatially adjacent supplementary units and reconstructed units.

[0121] Supplementary and reconstructed units with consecutive edges form the same update region, while supplementary or reconstructed units without shared edges form different update regions. Stable units and units awaiting confirmation are not incorporated into the update region, where stable units are used to provide boundary connectivity, and units awaiting confirmation remain in a temporarily suspended fusion state.

[0122] A single update region can consist of 4 to 40 spatially adjacent model units, corresponding to an area of ​​100 to 1000 square meters. Two model units that only touch through a corner point are not connected and merged.

[0123] The updated region retains the spatial index identifier, state type, and corresponding newly added observation record of the included model units, enabling the reconstruction process to distinguish between the data insertion path of the supplementary units and the data filtering path of the reconstructed units.

[0124] S502, extract shared vertices and shared edges from the non-updating model units adjacent to the updated region to form a set of boundary nodes.

[0125] Prioritize extracting shared vertices and shared edges from stable cells; when updating the local boundary of a region where there are no stable cells but there are cells to be confirmed with complete existing topology, only the existing model shared vertices and shared edges of the cells to be confirmed are used as fixed boundaries, and newly added observation records in the cells to be confirmed are not used.

[0126] The boundary nodes in the boundary node set maintain their spatial coordinates, elevations, and connections with non-updated model units within the current 3D terrain model; shared edges maintain the existing connection order between boundary nodes. The boundary node set forms a closed boundary according to the connection order of shared edges.

[0127] When constructing a 3D terrain model based on a grid spacing of 0.5 to 1 meter, the spacing between adjacent boundary nodes remains the same as the original grid spacing, and the planar position and elevation of the boundary nodes are not moved during the local topology reconstruction process.

[0128] In this embodiment, the supplementary units are the missing regions within the 3D terrain model, and the updated regions are surrounded by non-updated model units with existing topology. For newly added regions located on the outermost edge of the 3D terrain model that cannot form a closed set of boundary nodes, local topology reconstruction is not performed in this round; it will be processed after obtaining surveying data that can determine the external boundaries.

[0129] S503, for supplementary units, the newly added observation records pointed to by the source index are used as internal terrain data; for reconstructed units, observation records with regional credibility reaching the medium credibility level and whose change direction is consistent with the determination direction of the reconstructed unit are selected from the newly added observation records that meet the conditions of spatiotemporal continuity and multi-source support to form the selected observation records.

[0130] The selected observation records retain their spatial coordinates, elevation, acquisition time, data type, and source index. New observation records that do not reach the medium confidence level or have inconsistent directions of change are not included in this round of reconstruction results but are retained in the unified observation records for subsequent feedback and correction.

[0131] If the effective coverage area of ​​the selected observation records within the reconstruction unit reaches more than 70% of the model unit area, and the number of effective topographic points per square meter is not less than 4, then the unit will be organized into internal topographic points. If the coverage area or point density conditions are not met, the corresponding reconstruction unit will be changed to a unit to be confirmed and removed from the update area in this round.

[0132] S504 uses the set of boundary nodes as external constraints to organize the newly added observation records corresponding to the supplementary unit and the filtered observation records corresponding to the reconstruction unit into terrain points within the updated region, and forms new vertices, edges and faces.

[0133] First, delete duplicate internal terrain points with identical spatial coordinates and elevations. When the horizontal distance between multiple internal terrain points is less than half the target point spacing, retain the internal terrain point with the higher regional credibility level. If the credibility levels are the same, retain the internal terrain point with the more recent acquisition time. Internal terrain points are organized with a target point spacing of 0.5 meters to 1 meter.

[0134] Shared edges between boundary nodes are set as non-crossable constraint edges. Triangular faces are created based on the planar positions of internal terrain points and boundary nodes. The three vertices of each triangular face are selected from internal terrain points or boundary nodes. The edges of the triangular face must not intersect with constraint edges, and the interior of the triangular face must not exceed the update region enclosed by the set of boundary nodes. Degenerate triangular faces with three vertices on the same straight line, triangular faces with duplicate vertices, and triangular faces that cross the boundary of the update region are deleted.

[0135] Internal vertices located at the boundary are connected to the boundary nodes according to the original shared edge order. The vertices, edges, and faces within the updated region are written to the spatial index positions corresponding to the 3D terrain model, and the corresponding terrain attributes in the elevation, acquisition time, source index, reflection intensity, or image texture index are written synchronously.

[0136] When a single 100-square-meter update area is organized with a target point spacing of 1 meter, it can form approximately 100 internal grid positions; when organized with a target point spacing of 0.5 meters, it can form approximately 400 internal grid positions. The actual number of vertices is determined based on the boundary shape, the effective distribution of shape points, and the results of duplicate point removal.

[0137] Before writing, the spatial index of all vertices in the update region is checked, the existence status of the two endpoints of all edges is checked, the integrity of the three vertices and three edges of all faces is checked, and the original connection relationship between boundary nodes and non-updated model units is checked. If all checks pass, the update result is written to the current 3D terrain model; if any check fails, the local topology before the update is retained, and the corresponding update region is marked as a state to be reconstructed.

[0138] S5 determines the reconstruction boundary by merging model units, using shared vertices and edges in non-updating model units as boundary constraints, and distinguishes the data sources of supplementary units and reconstruction units, so that local updates are limited to the update area, while maintaining the topological connection between the update area and the rest of the 3D terrain model.

[0139] S6. Calculate the model residual of the updated area based on subsequent survey data, correct the regionalization reliability based on the model residual, and re-determine the state of the corresponding model unit when the model residual exceeds the residual condition.

[0140] S601, read the subsequent surveying data obtained in the update area according to the acquisition time sequence, convert the subsequent surveying data into spatial coordinates, acquisition time, positioning accuracy, elevation accuracy, data type and source index records consistent with the unified observation record according to S101 to S104, and establish the corresponding relationship according to the spatial index identifier of the model unit in the update area.

[0141] Subsequent mapping data can be accessed in cycles ranging from 60 to 180 seconds, with each continuous observation window consisting of three consecutive acquisition cycles.

[0142] Within each cycle, if each model unit in the update region acquires no fewer than 20 valid subsequent observation points, and the area covered by the valid observation points reaches 50% of the model unit area, the model residual calculation for this cycle is performed; if the above conditions are not met, the acquired subsequent observation points are retained and awaited for supplementation in the next acquisition cycle, and this cycle is not included in the continuous observation window.

[0143] S602 uses the same one-to-one proximity matching rule as S302 to match the spatial location, elevation, and neighborhood slope in subsequent survey data with the current vertex coordinates, elevation, and slope in the updated model.

[0144] Comparing the planar positions of each corresponding observation pair forms the observation pair position residual; comparing the elevations forms the zone-direction elevation residual; and comparing the neighboring slopes forms the zone-direction slope residual.

[0145] The median of the planar position residuals of all valid observation pairs within the same model unit is determined as the model unit position residual; the median of the absolute value of the elevation residuals is determined as the model unit elevation residual, and the median direction of the elevation residuals is retained; the median of the absolute value of the slope residuals is determined as the model unit slope residual, and the median direction of the slope residuals is retained.

[0146] The position residual limit is set to 0.10 meters, the elevation residual limit is set to 0.15 meters, and the slope residual limit is set to 2.5 degrees. If the residual of the same type of model unit exceeds the corresponding residual limit in two or more of three consecutive data acquisition cycles, the model residual of the corresponding model unit is determined to have exceeded the residual condition. Similarly, if the position residual exceeds 0.20 meters, the elevation residual exceeds 0.30 meters, or the slope residual exceeds 5 degrees in any cycle, the model residual of the corresponding model unit is also determined to have exceeded the residual condition.

[0147] The model residuals of the updated region consist of the model residual states of each model unit within the updated region. The out-of-limit results of a few model units are not used to directly replace the state of the entire updated region. Only model units whose residuals exceed the residual conditions are re-included in the state determination; other model units retain their current state.

[0148] S603, when the model residuals of the model unit do not exceed the residual conditions, the regionalization confidence is adjusted according to the degree of closeness between the location residuals, elevation residuals and slope residuals and the corresponding residual limits.

[0149] Within three consecutive acquisition cycles, if the location residual does not exceed 0.05 meters, the elevation residual does not exceed 0.08 meters, and the slope residual does not exceed 1.5 degrees, the regionalization confidence level will be increased by one level, but not exceeding the high confidence level; if any residual reaches more than 80% of the corresponding residual limit but does not exceed the residual condition, the regionalization confidence level will be decreased by one level, but not lower than the low confidence level, and the current topology of the model unit will be retained.

[0150] When the model residual of a model unit exceeds the residual condition, the corresponding model unit is reintegrated into S4. Based on the corrected regionalization confidence, spatial continuity, temporal continuity, and multi-source support conditions, the supplementary state, stable state, reconstructed state, or pending confirmation state is redefined. If it is redefined as a reconstructed unit, it proceeds to S5, forming a new update region with adjacent supplementary or reconstructed units and performing local topology reconstruction.

[0151] S604: Within the continuous observation window, if the position residuals, elevation residuals, and slope residuals of all model units in the update area do not exceed the corresponding residual conditions, the current round of feedback correction for the update area is terminated, and the corrected regional credibility and model unit status are saved as the basis for comparison of subsequent surveying data.

[0152] If there are fewer than three valid acquisition cycles of continuous observation data, the residual records of the acquired cycles are retained, and the updated area is not marked as feedback correction complete. If the model residual of any model unit exceeds the residual condition again, the next round of feedback correction is initiated.

[0153] Feedback correction uses newly arrived subsequent survey data as cyclic input, takes the exceeding state and change direction of model unit-level position residual, elevation residual and slope residual as the basis for single update, and takes the condition that the residual condition has not been exceeded for three consecutive valid acquisition cycles as the termination condition for this round.

[0154] In a specific set of operational data, a 50m by 50m first-level spatial index area was selected as the processing object. This area was divided into 100 model units of 5m by 5m. The average point density of the laser point cloud was 12 points per square meter, and the first-level spatial index area contained approximately 30,000 laser points; the digital elevation model used a 1-meter grid spacing.

[0155] Both batches of laser point cloud data and photogrammetric point data showed an increase in elevation within 12 adjacent model cells. These 12 model cells were consecutively distributed in 4 columns and 3 rows, corresponding to an updated area of ​​300 square meters. The median positional difference among the 12 model cells did not exceed 0.08 meters, the median elevation difference ranged from 0.32 meters to 0.46 meters, and the median slope difference ranged from 3.2 degrees to 4.4 degrees. The elevation changes in both the laser point cloud and photogrammetric point data showed an increasing direction, and the direction of change was consistent across both batches of data.

[0156] The positioning accuracy is no more than 0.05 meters, the elevation accuracy is no more than 0.08 meters, and the interval between data acquisition and existing model updates is no more than 2 minutes. Both the accuracy and timeliness confidence levels are high. Because the elevation and slope differences exceed the corresponding consistency limits, the geometric confidence level is medium, and the initial regionalization confidence level is also medium. Since the two types of data from different sources change in the same direction and there are no observations in the opposite direction with high confidence levels, the regionalization confidence level is raised to high.

[0157] Twelve model units, each exceeding 15 meters in length and maintaining the same direction of elevation change across two consecutive acquisition cycles, were identified as reconstruction units due to their support from two different data sources. These twelve reconstruction units were merged into a single update region based on edge adjacency. Shared vertices and edges were extracted from the stable units around the update region, forming a continuous closed set of boundary nodes with a 1-meter grid spacing. The vertices, edges, and faces within the update region were then reconstructed without altering the spatial coordinates and elevations of the boundary nodes.

[0158] In the subsequent three consecutive valid data collection cycles, the median of the model unit position residuals did not exceed 0.05 meters, the median of the elevation residuals did not exceed 0.08 meters, and the median of the slope residuals did not exceed 1.5 degrees. Since the corresponding model residuals did not exceed the residual conditions, the current round of feedback correction was terminated, and the updated model unit status and regional credibility were saved.

[0159] Through sequential execution of S1 to S6, multi-source heterogeneous terrain data is converted into unified observation records that retain spatial, temporal, precision, data type, and source fields. Subsequently, the processing range is defined by the spatiotemporal influence domain and model units, and mutually exclusive model unit states are formed through model unit-level geometric differences, regional credibility, spatiotemporal continuity, and multi-source support conditions.

[0160] The supplementary and reconstructed units complete the local topology reconstruction within the updated area under the constraints of the boundary nodes. The model unit-level model residuals formed by subsequent surveying data continue to affect the regional credibility and model unit status, so that data access, status determination, local update and feedback correction form a closed processing chain.

[0161] The data types, field states, processing order, judgment conditions, index relationships, credibility levels, and feedback relationships in the above embodiments are used to illustrate the possible implementations of the present invention. Without departing from the technical concept of the present invention and with the support of the original application documents, those skilled in the art can make equivalent transformations to the data storage format, spatial index implementation format, and terrain surface data organization format. However, equivalent transformations should not delete or weaken the data relationships, judgment relationships, and temporal relationships necessary for forming the spatiotemporal influence domain, generating regional credibility, determining the model unit state, performing boundary constraint local topology reconstruction, and implementing model residual feedback. The scope of protection of the present invention is defined by the claims.

Claims

1. A real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios, characterized in that, Includes the following steps: Acquire multi-source heterogeneous terrain data, positioning accuracy, and elevation accuracy output by surveying equipment, and generate a unified observation record that includes spatial range, acquisition time, positioning accuracy, and elevation accuracy; The spatiotemporal influence domain is determined based on the spatiotemporal overlap relationship and error range between the unified observation records and the three-dimensional terrain model, and the spatiotemporal influence domain is divided into model units according to the spatial index of the three-dimensional terrain model; New observation records corresponding to each model unit are extracted from the unified observation records, and existing model records corresponding to each model unit are extracted from the three-dimensional terrain model. The position difference, elevation difference, and slope difference between the new observation records and the existing model records are calculated to obtain the geometric difference value. Regional confidence is generated based on the unified observation records and the geometric difference value. When the existing model record is missing, the corresponding model unit is identified as a supplementary unit. When the geometric difference meets the consistency condition, the existing model record of the corresponding model unit is retained. When the geometric difference does not meet the consistency condition and the corresponding difference meets the spatiotemporal continuity and multi-source support conditions, the corresponding model unit is identified as a reconstruction unit; otherwise, fusion is temporarily suspended. The supplementary unit and the reconstruction unit are merged into an update region. The boundary nodes adjacent to the update region and the non-updating model unit are extracted. Using the boundary nodes as constraints, local topology reconstruction is performed on the vertices, edges, faces and terrain attributes inside the update region. The model residual of the updated region is calculated based on subsequent survey data. The regionalization confidence is corrected based on the model residual. When the model residual exceeds the residual condition, the state of the corresponding model unit is re-determined.

2. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 1, characterized in that: The specific process for generating the unified observation record is as follows: The mapping equipment identifier, original coordinate reference, acquisition time, and terrain feature record corresponding to the multi-source heterogeneous terrain data are analyzed respectively. The terrain feature record is converted from the original coordinate datum to the target coordinate datum used by the three-dimensional terrain model to generate the target terrain feature record; The spatial range is determined based on the spatial coordinates recorded by the target terrain features, and a source index pointing to the corresponding multi-source heterogeneous terrain data storage location is established; The mapping equipment identifier, the acquisition time, the spatial range, the positioning accuracy, the elevation accuracy, the target coordinate datum, and the source index are written into the unified observation record.

3. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 2, characterized in that: The process of organizing the unified observation records is as follows: According to the spatial index granularity of the three-dimensional terrain model, the spatial coordinates of the target terrain features are grouped, and the corresponding spatial index identifiers are written into the spatial index keys. Generate a time index key according to the synchronization time window to which the acquisition time belongs; The spatial index key, the temporal index key, and the source index are written into the unified observation record, and a multi-source association relationship is established for target terrain feature records that have the same spatial index key and temporal index key but different source indices.

4. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 3, characterized in that: The process of determining the spatiotemporal domain of influence includes: Read the spatial range, acquisition time, positioning accuracy, and elevation accuracy from the unified observation record; The spatial range is expanded in plane according to the positioning accuracy, and the elevation error range is determined according to the elevation accuracy to generate the error expansion range; A time correlation window is established based on the acquisition time, and model records that intersect with the error expansion range are retrieved from the spatial index of the three-dimensional terrain model; When the update time of the model record is within the time association window, the model region corresponding to the model record is determined as a candidate model region, and the union of the candidate model region and the error expansion range is determined as the spatiotemporal influence domain. When there is no model record whose update time is within the time association window, the error expansion range is determined as the spatiotemporal influence domain.

5. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 4, characterized in that: The process of dividing the spatiotemporal influence domain into model units includes: Based on the spatial index hierarchy of the three-dimensional terrain model, extract the spatial index nodes that intersect with the spatiotemporal influence domain; Spatial index nodes that are completely located within the spatiotemporal influence domain and meet the preset index granularity are determined as internal model units; The spatial index nodes that cross the boundaries of the spatiotemporal influence domain are subdivided step by step until the minimum spatial index granularity of the three-dimensional terrain model is reached, and the part that intersects with the spatiotemporal influence domain after subdivision is determined as the boundary model unit. Write corresponding spatial index identifiers and adjacency relationships for the internal model units and the boundary model units to form each model unit.

6. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 5, characterized in that: The process of calculating the geometric difference corresponding to each of the model units includes: New observation records are generated by selecting terrain feature records from the unified observation records whose spatial index identifiers are consistent with the model unit and whose acquisition time is within the time correlation window. Read terrain data whose spatial index identifiers match the model unit from the 3D terrain model to form an existing model record; Based on the spatial coordinate proximity relationship, establish observation pairs at the same location between the newly added observation record and the existing model record, calculate the planar coordinate distance and elevation difference of each observation pair at the same location, and form the position difference and elevation difference respectively; Based on the observations at the same location, the local terrain surface is fitted to the terrain points in the neighborhood, the slope difference of the corresponding local terrain surface is calculated, and the position difference, the elevation difference and the slope difference are combined into the geometric difference value of the model unit.

7. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 6, characterized in that: The process of generating regionalized credibility includes: The positioning accuracy, elevation accuracy, and acquisition time are read from the unified observation record. The accuracy reliability level is determined based on the error range corresponding to the positioning accuracy and elevation accuracy. The timeliness reliability level is determined based on the time interval between the acquisition time and the update time of the existing model record. The position difference, elevation difference, and slope difference are compared with their corresponding consistency limits. If none of the three differences exceed the corresponding consistency limit, a first geometric consistency level is determined. If some differences exceed the corresponding consistency limit, a second geometric consistency level is determined. If all three differences exceed the corresponding consistency limit, a third geometric consistency level is determined. An initial regionalized confidence level is generated based on the lowest level among the accuracy confidence level, the timeliness confidence level, and the geometric consistency level. When newly added observation records from different sources show consistent determination of the direction of terrain change, the initial regionalized confidence level is increased to form a regionalized confidence level.

8. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 7, characterized in that: The process of determining the state of a model unit includes: When existing model records are missing, the corresponding model unit is identified as a supplementary unit; When the position difference, elevation difference, and slope difference do not exceed the corresponding consistency limit, and the regionalization confidence is not lower than the confidence limit, the corresponding model unit is determined as a stable unit and the existing model record is retained. When at least one of the location difference, elevation difference, or slope difference exceeds the corresponding consistency limit, the number of consecutive units with the same direction of difference in adjacent model units is counted, and the number of newly added observation records from different sources with the same direction of difference is counted. When the number of consecutive units and the number of newly added observation records reach the corresponding support limit, the corresponding model unit is determined as a reconstruction unit; otherwise, it is determined as a unit to be confirmed and fusion is temporarily suspended.

9. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 8, characterized in that: The process of performing local topology reconstruction includes: The spatially adjacent supplementary units and reconstructed units are connected and merged to form an updated region; Shared vertices and shared edges are extracted from the stable units adjacent to the updated region to form a set of boundary nodes; The spatial coordinates of each boundary node in the boundary node set and its connection relationship with the non-updated model unit are retained. Based on the newly added observation records corresponding to the supplementary unit and the filtered observation records corresponding to the reconstruction unit, the vertices, edges and faces inside the updated region are reconstructed. The reconstructed vertices, edges, faces, and terrain attributes are written into the updated region, while maintaining the topological connection between each boundary node and the non-updated model unit.

10. The real-time fusion modeling method for multi-source heterogeneous terrain data in surveying and mapping scenarios according to claim 1, characterized in that: The process of correcting regionalization confidence based on model residuals and re-evaluating model unit states includes: According to the collection time sequence, the subsequent mapping data are matched with the current vertex coordinates, elevation and slope of the updated area, and the position residual, elevation residual and slope residual are calculated respectively. The model residual is generated according to the change direction and number of times the three types of residuals exceed the limit within the continuous observation window. When the model residual does not exceed the residual condition, the regionalization confidence is adjusted according to the degree of closeness between the three types of residuals and the corresponding limit, and the current state of the model unit is retained; When the model residual exceeds the residual condition, the model unit corresponding to the excess residual is re-included in the state determination, and the state of the model unit is re-determined based on the corrected regional credibility, spatiotemporal continuity and multi-source support conditions. The current round of feedback correction is terminated when none of the model residuals within the continuous observation window exceed the residual condition; the next round of feedback correction is started when any of the model residuals exceeds the residual condition again.

Images