A low-altitude space-time three-dimensional environment twin method and system

Abstract

The application discloses a low-altitude space-time three-dimensional environment twin method and system, relates to the low-altitude twin technical field, and comprises the following steps: acquiring low-altitude multi-source environment observation data of a task area of an unmanned aerial vehicle, and constructing a space-time slicing voxel reference and an observation credible field; generating a low-altitude environment state tensor based on the space-time slicing voxel reference and the observation credible field, and performing time sequence consistent correction; and generating a passable state graph and an incremental update instruction according to the low-altitude environment state tensor and coupling navigation constraints. The method generates the passable state graph and the incremental update instruction by coupling the navigation constraints, realizes the conversion of the corrected low-altitude environment state tensor to the unmanned aerial vehicle flyable nodes, transfer edges and local update ranges, and enables the three-dimensional environment twin result to directly support low-altitude flight path calculation, flight safety evaluation and local refreshing of the task area.

Description

Technical Field

[0001] This invention relates to the field of low-altitude twin technology, specifically to a method and system for creating a three-dimensional environmental twin in low-altitude spatiotemporal space. Background Technology

[0002] 3D environmental twins in low-altitude applications essentially involve the acquisition, organization, storage, fusion, updating, and retrieval of multi-source environmental data within the mission area. This data typically includes remote sensing imagery, 3D point clouds, oblique photogrammetry models, visual reconstruction results, meteorological observation data, communication coverage data, airspace boundary data, and temporary obstacle data. These data differ significantly in data format, spatial resolution, temporal granularity, coordinate reference, sampling frequency, and update methods. Therefore, computer data processing is required to create a queryable, updatable, and reasonable 3D environmental data representation to support subsequent mission analysis, environmental retrieval, status assessment, and data service retrieval.

[0003] Existing 3D environmental twin methods primarily focus on geometric model construction or map data overlay, typically storing environmental data from different sources separately or performing simple registration using a unified coordinate system. However, in low-altitude mission areas, the same spatial location may simultaneously contain multiple observations from point clouds, imagery, meteorological, communication, and airspace data. These different observations may also suffer from time lag, spatial occlusion, resolution inconsistencies, and data gaps. If only static 3D models or single-layer overlays are used for data organization, it is difficult to accurately represent the occupancy status, environmental attribute status, and data reliability status of the same spatial unit at different time slices within the computer. This easily leads to inconsistent environmental twin data updates, difficulty in tracing historical states, and difficulty in timely identifying local data conflicts. Furthermore, existing low-altitude 3D environmental data processing methods typically manage geometric data, dynamic obstacle data, meteorological data, communication data, and airspace restriction data as independent data layers, lacking a unified data structure and state update mechanism for the low-altitude spatiotemporal environment. Due to the lack of voxel-level spatiotemporal indexing, observation reliability recording, and cross-time-slice consistency correction mechanisms, the system struggles to distinguish between real-world environmental changes, sensor observation errors, data latency, and state differences caused by local occlusion when receiving asynchronously updated data. Existing methods primarily generate 3D environmental data for visualization or static querying, failing to convert multi-source environmental states into unified, traversable state data and incrementally updated data. This results in inconsistent data structures, high update costs, and low real-time call efficiency in subsequent computer retrieval, path analysis, task simulation, and data distribution of low-altitude environment twins. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that existing low-altitude 3D environment twin methods lack a unified spatiotemporal reference for asynchronous multi-source observations, make it difficult to form a traceable and reliable fusion of occlusion observations and mutually exclusive observations, make it difficult for static 3D scenes to support UAV navigation constraint judgments, and address the issues of how to generate a passable state map and incremental update instructions for low-altitude flight processes.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a three-dimensional environment twin method for low-altitude spatiotemporal data, comprising acquiring low-altitude multi-source environmental observation data of the UAV mission area, constructing a spatiotemporal piecewise voxel benchmark and an observation reliability field; generating a low-altitude environmental state tensor based on the spatiotemporal piecewise voxel benchmark and the observation reliability field and performing temporal consistency correction; and generating a passable state diagram and incremental update instructions by coupling navigation constraints according to the low-altitude environmental state tensor.

[0007] As a preferred embodiment of the low-altitude spatiotemporal three-dimensional environment twin method described in this invention, the acquisition of low-altitude multi-source environmental observation data of the UAV mission area includes acquiring UAV payload observation data, ground perception data, airspace constraint data, meteorological observation data, and communication coverage data; using the boundary of the UAV mission area, mission start and end times, and mission altitude range as the dividing criteria, a voxel record containing voxel identifiers, spatial boundaries, time slice identifiers, data source identifiers, and coordinate transformation relationships is formed, and the voxel record is stored in a spatiotemporal index structure under a unified spatial coordinate system and a unified time reference.

[0008] As a preferred embodiment of the low-altitude spatiotemporal three-dimensional environment twin method described in this invention, the observation credibility field includes: reading the positioning residual, acquisition time, observation line-of-sight occlusion marker, sensor operating status marker, and field integrity marker respectively for low-altitude multi-source environmental observation data entering the same voxel record and the same time slice identifier; determining a time decay term based on the interval between the acquisition time and the time slice identifier, determining a spatial deviation term based on the positioning residual, determining a visibility deduction term based on the observation line-of-sight occlusion marker, and determining a data availability term based on the sensor operating status marker and the field integrity marker; combining the time decay term, spatial deviation term, visibility deduction term, and data availability term into a source credibility field value; when there are mutually exclusive observations from different data sources for the same voxel record, retaining each source credibility field value and the mutually exclusive field marker, and determining the master observation field value based on the sorting relationship between the source credibility field values; when there are no observations in the current time slice for the same voxel record, writing an unobserved field value instead of replacing it with a null value; and transferring the credibility field value to the low-altitude environmental state tensor along with the voxel record.

[0009] As a preferred embodiment of the low-altitude spatiotemporal three-dimensional environment twin method described in this invention, the generation of the low-altitude environment state tensor includes: using voxel identifiers as spatial dimension indexes, time slice identifiers as temporal dimension indexes, and various environmental fields as state dimension indexes; writing the location, velocity, shape boundary, wind direction, wind speed, visibility, communication coverage category, and airspace boundary relationship from low-altitude multi-source environmental observation data into the corresponding state dimension indexes; for environmental fields with the same name from different data sources, determining the writing order according to the reliable field values, and storing unwritten candidate fields in the candidate observation column. The table establishes adjacency relationships for spatially and temporally adjacent voxel records, including horizontal, vertical, and temporal adjacency. It also establishes occlusion propagation markers for occluded voxel records, which, together with the observation line-of-sight occlusion markers, determine whether the environment field is allowed to be directly overlaid. The generated low-altitude environment state tensor includes voxel occupancy, obstacle motion, wind field, visibility, communication coverage, spatial constraint, trusted field values, mutual exclusion field markers, and time-slice identifiers, retaining the source data source, write time, and trusted field value for each state dimension index.

[0010] As a preferred embodiment of the low-altitude spatiotemporal three-dimensional environment twin method described in this invention, the temporal consistency correction includes: reading the preceding state field from the low-altitude environment state tensor corresponding to the previous time slice identifier; generating the predicted state field of the current time slice identifier based on the obstacle motion field, wind field field, and temporal adjacency; performing a field-level comparison between the predicted state field and the observed state field of the current time slice identifier to form a state residual field; when the state residual field does not exceed a preset residual threshold, using the observed state field of the current time slice identifier as the corrected state field; when the state residual field exceeds the preset residual threshold, reading the... The system generates a confidence field value, a mutual exclusion field marker, an occlusion propagation marker, and a candidate observation list. Candidate state fields are selected in the following order: confidence field value priority, spatial adjacency continuity, temporal adjacency continuity, and no spatial boundary crossing. For candidate state fields that cannot satisfy spatial and temporal adjacency continuity, a status marker pending review is added. For candidate state fields that satisfy spatial and temporal adjacency continuity, the corresponding observation state field is replaced to form a corrected state field. The corrected state field covers the voxel occupancy field, obstacle motion field, wind field field, visibility field, communication coverage field, and spatial constraint field within the current time slice identifier.

[0011] As a preferred embodiment of the low-altitude spatiotemporal three-dimensional environment twinning method described in this invention, the navigation constraints include: UAV outline constraints, climb constraints, turning radius constraints, wind field tolerance constraints, communication maintenance constraints, airspace boundary constraints, and safety interval constraints; coupling navigation constraints includes reading the corrected state field in the low-altitude environment state tensor, and selecting voxel records that satisfy the voxel occupancy field, wind field field, communication coverage field, and airspace constraint field as candidate nodes; establishing transition edges between candidate nodes based on horizontal adjacency, vertical adjacency, temporal adjacency, and UAV outline constraints, and deleting transition edges that do not satisfy the climb constraints, turning radius constraints, wind field tolerance constraints, communication maintenance constraints, airspace boundary constraints, or safety interval constraints.

[0012] As a preferred embodiment of the low-altitude spatiotemporal three-dimensional environment twinning method described in this invention, the traversable state diagram and incremental update instructions include: traversable nodes, restricted nodes, prohibited nodes, traversable transition edges, restricted transition edges, updated voxel range, updated field name, update trigger reason, and version time; generating the traversable state diagram and incremental update instructions includes: writing traversable node markers to retained candidate nodes; writing restricted node markers to candidate nodes with pending verification state markers or mutually exclusive field markers; and writing prohibited node markers to candidate nodes with airspace boundary violations, voxel occupancy conflicts, communication coverage not meeting constraints, or wind field bearing constraints not meeting constraints. Stop node marking; write passable transfer edge markings for retained transfer edges, and write restricted transfer edge markings for transfer edges where any endpoint is a restricted node; collectively determine the voxel records whose state residual field exceeds the preset residual threshold, voxel records intersecting with the candidate track region, and voxel records adjacent to the current position of the UAV as the update voxel range; generate update field names based on the state dimension index that has changed within the update voxel range; generate update trigger reasons based on changes in source trusted field values, changes in state residual fields, changes in navigation constraints, and changes in candidate node markings; arrange the update voxel range, update field names, and update trigger reasons according to version time to form an incremental update instruction.

[0013] As a preferred embodiment of the low-altitude spatiotemporal three-dimensional environment twin system described in this invention, it includes: a patchwork field construction module, a tensor correction module, and a traffic update module; the patchwork field construction module is used to acquire low-altitude multi-source environmental observation data of the UAV mission area and construct a spatiotemporal patch voxel reference and an observation reliability field; the tensor correction module is used to generate a low-altitude environmental state tensor based on the spatiotemporal patch voxel reference and the observation reliability field and perform temporal consistency correction; the traffic update module is used to generate a trafficable state diagram and incremental update instructions based on the low-altitude environmental state tensor and coupled with navigation constraints.

[0014] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program being steps of a method for realizing a three-dimensional environmental twin in low-altitude spatiotemporal space.

[0015] A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of a low-altitude spatiotemporal three-dimensional environment twin method.

[0016] The beneficial effects of this invention are as follows: The low-altitude spatiotemporal 3D environment twinning method provided by this invention, by constructing a spatiotemporal piecewise voxel benchmark and an observation credibility field, realizes the organization of low-altitude multi-source environmental observation data under a unified spatial coordinate system and a unified time benchmark. This allows acquisition delay, positioning residuals, observation occlusion, and field missingness to be recorded with credible field values ​​and mutually exclusive fields, thus providing a traceable data foundation for subsequent low-altitude environment state tensor generation. By generating the low-altitude environment state tensor and performing temporal consistency correction, the same spatiotemporal expression is achieved among voxel occupancy, obstacle movement, wind field, visibility, communication coverage, and airspace constraints. This allows the state residual field to guide the selection of candidate state fields, thereby reducing the interference of asynchronous observation, occlusion propagation, and abrupt observations on the 3D environment twinning update. By coupling navigation constraints to generate a navigable state diagram and incremental update instructions, the conversion from the corrected low-altitude environment state tensor to UAV flyable nodes, transfer edges, and local update ranges is realized. This enables the 3D environment twinning results to directly support low-altitude trajectory calculation, flight safety evaluation, and local update of the mission area. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 The overall flowchart of a three-dimensional environmental twin method for low-altitude spatiotemporal space provided by the present invention is shown.

[0019] Figure 2 A schematic diagram of a computer device provided by the present invention. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0021] Reference Figure 1 As an embodiment of the present invention, a three-dimensional environment twinning method for low-altitude spatiotemporal space is provided, comprising:

[0022] S1: Acquire low-altitude multi-source environmental observation data of the UAV mission area, and construct a spatiotemporal segmented voxel benchmark and observation credibility field.

[0023] Furthermore, acquiring low-altitude multi-source environmental observation data of the UAV mission area includes acquiring UAV payload observation data, ground perception data, airspace constraint data, meteorological observation data, and communication coverage data; using the boundaries of the UAV mission area, mission start and end times, and mission altitude range as the dividing criteria, forming voxel records containing voxel identifiers, spatial boundaries, time slice identifiers, data source identifiers, and coordinate transformation relationships, and storing the voxel records in a spatiotemporal index structure under a unified spatial coordinate system and a unified time reference.

[0024] It should be noted that low-altitude multi-source environmental observation data specifically refers to environmental data collected around UAV low-altitude flight missions and transferable to the same mission area. UAV payload observation data originates from at least one of airborne vision, lidar, millimeter-wave radar, positioning devices, and inertial measurement units. Ground perception data originates from at least one of fixed camera equipment, ground radar, takeoff and landing field sensors, and roadside perception equipment. Airspace constraint data originates from at least one of mission planning documents, temporary control boundary documents, and no-fly zone boundary documents. Meteorological observation data originates from at least one of wind field, visibility, and precipitation observation documents. Communication coverage data originates from coverage measurement documents of ground communication equipment. The data format of low-altitude multi-source environmental observation data includes a timestamp field, spatial coordinate field, data source identifier field, observation object field, observation value field, and quality description field. The quality description field includes positioning residuals, sensor operating status markers, field integrity markers, and observation line-of-sight obstruction markers.

[0025] Furthermore, the spatiotemporal segmentation voxel reference specifically refers to the combination of three-dimensional voxels and time slices obtained by jointly dividing the UAV mission area in both spatial and temporal dimensions. For spatial segmentation, a unified spatial coordinate system is first established based on the mission area boundaries, then height layers are formed according to the mission altitude range, and subsequently, a planar grid is formed within each height layer. For temporal segmentation, time slices are formed based on the mission start and end times, environmental change frequency, and UAV control cycle. Each voxel record is bound to a voxel identifier, spatial boundary, and time slice identifier, enabling subsequent observation data to be fused using the same voxel record. In this embodiment, the voxel side length is set to 2m, and the time slice length is set to 5s. The voxel side length is derived from a comprehensive configuration of building boundaries, low-altitude obstacle shapes, and UAV positioning errors within the UAV mission area, while the time slice length is derived from a comprehensive configuration of the ground perception data refresh cycle, UAV control cycle, and low-altitude environmental change frequency. This setting avoids excessive indexing due to overly fine segmentation of the voxel side length, and avoids distortion of dynamic obstacle states due to excessively long aggregation times.

[0026] Furthermore, a unified spatial coordinate system is established based on the UAV mission area, with the origin, axis directions, and altitude reference determined by the mission planning document or the takeoff and landing site positioning reference. For coordinates from different data sources, the coordinate transformation relationship corresponding to the data source identifier field is first read, and then the spatial coordinate field is converted to coordinates under the unified spatial coordinate system. For data that cannot be converted, the corresponding data is written to the abnormal observation list and does not participate in the generation of the current time slice of the observation confidence field. The unified time reference is established based on the mission start time, and all acquisition times and effective times are converted to relative times. For data with acquisition delays, the corresponding delay compensation rule is read according to the data source identifier field, and the compensated time is matched to the corresponding time slice identifier. For data spanning two time slices, two candidate observation records are generated according to the overlap between the effective period and the time slice, so that the same observation value can remain traceable in the time dimension.

[0027] Furthermore, the observation confidence field includes: reading the positioning residual, acquisition time, observation line-of-sight occlusion marker, sensor operating status marker, and field integrity marker from low-altitude multi-source environmental observation data entering the same voxel record and the same time slice identifier; determining the time decay term based on the interval between the acquisition time and the time slice identifier, determining the spatial deviation term based on the positioning residual, determining the visibility deduction term based on the observation line-of-sight occlusion marker, and determining the data availability term based on the sensor operating status marker and the field integrity marker; combining the time decay term, spatial deviation term, visibility deduction term, and data availability term into the source confidence field value; when there are mutually exclusive observations from different data sources for the same voxel record, retaining each source confidence field value and the mutually exclusive field marker, and determining the master observation field value based on the sorting relationship between the source confidence field values; when there are no observations in the current time slice for the same voxel record, writing the unobserved field value instead of replacing it with a null value; and transferring the confidence field value to the low-altitude environmental state tensor along with the voxel record.

[0028] It should be noted that the source trust field value is calculated according to the following formula:

[0029]

[0030] in, Indicates data source In voxel recording and time slice identifier The source confidence field value is derived from the quality description field of the corresponding observation record, and the value ranges from 0 to 1. This indicates the availability of data, derived from sensor operating status and field integrity flags. A value of 1 indicates normal sensor operation and complete fields, while a value of 0 indicates abnormal sensor operation or missing critical fields. This represents the time decay term, which is derived from the time interval between the acquisition time and the time interval corresponding to the time slice identifier, and its value ranges from 0 to 1. The spatial deviation term is derived from the comparison between the positioning residual and the tolerance range of the voxel space boundary, and its value ranges from 0 to 1. This represents the visibility deduction, derived from the observation line-of-sight occlusion marker. It is set to 1 when there is no occlusion and to 0.6 when there is an occlusion propagation marker. , , The weights of the timeliness decay term, spatial deviation term, and visibility deduction term are respectively derived from the task configuration file. In this embodiment, they are set to 0.4, 0.35, and 0.25 respectively, and the sum of the three is 1.

[0031] The calculation process is as follows: The computer first reads the observation records entering the same voxel record according to the voxel identifier and time slice identifier. Then, it reads the acquisition time, positioning residual, observation line-of-sight occlusion marker, sensor working status marker, and field integrity marker from the observation records. Subsequently, it obtains the time decay item, spatial deviation item, visibility deduction item, and data availability item, and substitutes them into the source confidence field value calculation formula. When multiple data sources give different observation results for the same environmental field, they are sorted from high to low according to the source confidence field value. The observation result with the highest ranking and which has not triggered the abnormal observation list is written into the main observation field value, and the remaining observation results are written into the candidate observation list, while retaining the mutual exclusion field marker.

[0032] Furthermore, the timeliness decay term is determined by the interval between the acquisition time and the time interval corresponding to the time slice identifier. In this embodiment, the preset time tolerance threshold is set to 2 seconds, and the setting source is the task configuration value formed by the time slice length of 5 seconds, the second-level refresh cycle of ground perception data, and the UAV control cycle. When the acquisition time is no more than 2 seconds away from the center time of the time slice, the corresponding observation record is allowed as a candidate for the main observation field value. When the acquisition time is more than 2 seconds away from the center time of the time slice, the corresponding observation record is not directly written into the main observation field value, but written into the candidate observation list. This setting can avoid significantly lagging observation data from overwriting the new observation state within the current time slice, while retaining delayed data for subsequent time series consistency correction.

[0033] Furthermore, the spatial deviation term is determined by the relationship between the positioning residual and the voxel spatial boundary tolerance range. In this embodiment, the voxel spatial boundary tolerance range is set to 1.5m, which is derived from a task configuration combination of a voxel side length of 2m, a UAV positioning error upper limit of 0.5m, and a sensor ranging error upper limit of 1m. When the positioning residual is within 1.5m, the corresponding observation record is allowed as a candidate for the main observation field value. When the positioning residual exceeds 1.5m, the corresponding observation record retains the data source identifier field and the observation value field, but writes a spatial deviation flag. This setting can prevent observation results with large positioning errors from being incorrectly written to adjacent voxel records.

[0034] Furthermore, the visibility deduction is determined by the observation line-of-sight occlusion marker. For observation records generated by airborne vision or lidar, if the observation path intersects with building boundaries, terrain occlusion boundaries, or existing occupied voxels, an observation line-of-sight occlusion marker is generated, and the visibility deduction is set to 0.6. For observation records without occlusion, the visibility deduction is set to 1. This value is derived from the conservative deduction rule for occluded observations in the task configuration and is not considered a true experimental conclusion. With this setting, occluded observations will not be directly deleted, and their impact on the main observation field value is lower than that of unoccluded observations.

[0035] Furthermore, when mutually exclusive observations from different data sources exist for the same voxel record, the observation confidence field does not directly discard the mutually exclusive observations, but retains the source confidence field values ​​and mutual exclusion field markers. The mutual exclusion field markers are used to record conflicts between occupied and unoccupied, passable and inaccessible, and communication coverage satisfied and unsatisfied constraints for the same environment field within the same time slice. The ordering relationship between source confidence field values ​​is used to determine the master observation field value; unselected observation records are added to the candidate observation list.

[0036] It should be noted that this step does not simply overlay multi-source data onto a 3D mesh. Instead, it simultaneously sets spatial boundaries, time-slice identifiers, data source identifiers, and source reliability field values ​​in the voxel records, enabling traceability to be maintained for asynchronous acquisitions, occluded observations, and mutually exclusive observations. Through the source reliability field value calculation formula, timeliness, spatial bias, visibility, and data availability are converted into a unified field that can be processed by a computer, solving the problem of inconsistent time references, spatial references, and reliability expressions for low-altitude multi-source environmental observation data.

[0037] S2: Based on the spatiotemporal piecewise voxel benchmark and the observation confidence field, generate the low-altitude environment state tensor and perform temporal consistency correction.

[0038] Furthermore, generating the low-altitude environment state tensor involves using voxel identifiers as spatial dimension indexes, time slice identifiers as temporal dimension indexes, and various environmental fields as state dimension indexes. The location, velocity, shape boundary, wind direction, wind speed, visibility, communication coverage category, and airspace boundary relationships from low-altitude multi-source environmental observation data are written into the corresponding state dimension indexes. For environmental fields with the same name from different data sources, the writing order is determined according to the trusted field values, and unwritten candidate fields are stored in the candidate observation list. Adjacency relationships are established for spatially and temporally adjacent voxel records, including horizontal, vertical, and temporal adjacency. Occlusion propagation markers are established for occluded voxel records, and these markers, along with the observation line-of-sight occlusion markers, jointly determine whether environmental fields are allowed direct coverage. The generated low-altitude environment state tensor includes voxel occupancy fields, obstacle movement fields, wind field fields, visibility fields, communication coverage fields, airspace constraint fields, trusted field values, mutual exclusion field markers, and time slice identifiers, retaining the source data source, writing time, and trusted field value for each state dimension index.

[0039] It should be noted that the low-altitude environment state tensor specifically refers to a data structure formed by using voxel identifiers as spatial dimension indexes, time slice identifiers as temporal dimension indexes, and environment fields as state dimension indexes. Each state dimension index records the observation value, source data source, write time, and trusted field value. The voxel occupancy field indicates whether the voxel record is occupied by terrain, buildings, temporary obstacles, aircraft, or other occupant objects. The obstacle motion field indicates the direction and magnitude of the obstacle's position change within the current time slice. The wind field field indicates the wind direction, wind speed range, and wind shear marker within the voxel record. The visibility field indicates the attenuation state of visual or lidar observations within the voxel record. The communication coverage field indicates the link coverage identifier between the UAV and ground communication equipment. The airspace constraint field indicates whether the voxel record is located within a permitted flight boundary, temporary control boundary, or no-fly zone.

[0040] Furthermore, the low-altitude environment state tensor is generated according to the following formula:

[0041]

[0042] in, Voxel recording Time slice identifier Environment fields The corresponding low-altitude environment state tensor elements are derived from the spatiotemporal piecewise voxel reference and the observation confidence field. The voxel identifier is derived from the voxel record in the spatiotemporal segmentation voxel reference. This indicates a time slice identifier, derived from a spatiotemporal index structure under a unified time base. The environmental fields include at least the voxel occupancy field, obstacle motion field, wind field field, visibility field, communication coverage field, and airspace constraint field. Represents environment field The observed values ​​are derived from the value of the primary observation field with the highest writing order. This indicates the source data source, which originates from the data source identifier field of the corresponding observation record. Indicates the write time, which originates from the acquisition time or effective time of the observation record. This indicates the trusted field value, which originates from the source trusted field value obtained from S1. The field markers include at least one of the following: mutually exclusive field markers, occlusion propagation markers, unobserved field values, and pending verification status markers.

[0043] The computational process is as follows: First, the computer uses voxel identifiers and time-slice identifiers as two-dimensional search keys to read low-altitude multi-source environmental observation data within the same voxel record and time-slice identifier from the spatiotemporal index structure. Then, using the environmental field as the third-dimensional index, location, velocity, shape boundary, wind direction, wind speed, visibility, communication coverage category, and airspace boundary relationship are written into the corresponding state-dimensional index. For environmental fields with the same name, the writing order is determined according to the confidence field value from high to low, and the observation value with the highest ranking is written. Synchronously write its source data, write time, trusted field value, and field tag. .

[0044] Furthermore, for environmental fields with the same name from different data sources, observation records with the highest trusted field values ​​are written into the state dimension index, while observation records with lower trusted field values ​​are added to the candidate observation list. The candidate observation list retains the observation value, source data source, trusted field value, and mutually exclusive field markers. Adjacency relationships include horizontal adjacency, vertical adjacency, and temporal adjacency. Horizontal adjacency is determined by voxel records sharing a boundary within the same altitude layer; vertical adjacency is determined by voxel records whose projected regions intersect within different altitude layers; and temporal adjacency is determined by voxel records with the same voxel identifier and consecutive time slice identifiers. Adjacency relationships are written into the low-altitude environmental state tensor, enabling temporal consistency correction to simultaneously read spatial and temporal continuity.

[0045] Furthermore, the occlusion propagation marker is generated based on the occlusion marker of the observation line of sight and the adjacency relationship of voxels. If a voxel record is occluded by a building boundary, terrain boundary, or a voxel that has been confirmed to be occupied, then the voxel record located along the same occlusion path along the observation direction is written into the occlusion propagation marker. The occlusion propagation marker does not directly change the voxel occupancy field, but rather limits whether the corresponding environmental field is allowed to be covered by the current observation value, avoiding the unobserved state caused by occlusion being mistakenly written as an unoccupied state. In this embodiment, the maximum number of occlusion propagation layers is set to 3 voxels, and the setting is based on the voxel side length of 2m, the occlusion diffusion range configuration of airborne vision and lidar in low-altitude close-range observation, and the building boundary density of the mission area. This setting can limit the occlusion influence range and prevent a single occlusion marker from spreading to voxel records that are too far away.

[0046] Furthermore, the temporal consistency correction includes: reading the preceding state field from the low-altitude environmental state tensor corresponding to the previous time slice identifier; generating the predicted state field for the current time slice identifier based on the obstacle motion field, wind field field, and temporal adjacency; performing a field-level comparison between the predicted state field and the observed state field of the current time slice identifier to form a state residual field; when the state residual field does not exceed a preset residual threshold, using the observed state field of the current time slice identifier as the corrected state field; when the state residual field exceeds the preset residual threshold, reading the reliable field value, the mutually exclusive field marker, and the occlusion... Propagate the markers and candidate observation list, and filter candidate state fields in the following order: priority of reliable field values, spatial adjacency continuity, temporal adjacency continuity, and no cross-boundary of spatial domain. For candidate state fields that cannot satisfy spatial and temporal adjacency continuity, write them as pending verification status markers. For candidate state fields that satisfy spatial and temporal adjacency continuity, replace the corresponding observation state fields and form corrected state fields. The corrected state fields cover the voxel occupancy field, obstacle motion field, wind field field, visibility field, communication coverage field, and spatial constraint field within the current time slice identifier.

[0047] It should be noted that the timing consistency correction forms the state residual field according to the following formula:

[0048]

[0049] in, Voxel recording Time slice identifier Environment fields The corresponding state residual field is derived from the field-level comparison between the predicted state field and the observed state field. The predicted field value is derived from the preceding state field, obstacle motion field, wind field field, and time adjacency relationship corresponding to the previous time slice identifier. The observation status field value, representing the current time slice identifier, is derived from the master observation field value in the low-altitude environment state tensor. Represents environment field The range of the fields is derived from the task configuration file. The wind speed field specifies the allowed wind speed range configured for the task, while the position change field specifies the position change range determined by both the voxel side length and the time slice length. The environmental fields include at least the voxel occupancy field, obstacle motion field, wind field field, visibility field, communication coverage field, and airspace constraint field.

[0050] The computational process is as follows: The computer first reads the preceding state fields from the low-altitude environment state tensor corresponding to the previous time slice identifier, and then maps the preceding state fields to the current time slice identifier based on temporal adjacency. For the obstacle motion field, a predicted field value is formed based on the direction and magnitude of position change. For the wind field field, a predicted field value is formed based on the wind direction and speed of the previous time slice identifier and the adjacent voxel records of the current time slice identifier. Subsequently, the predicted field values ​​are compared with the current observation state fields at the field level, and normalized residual values ​​are calculated. If the normalized residual value does not exceed a preset residual threshold, the current observation state field is retained. If the normalized residual value exceeds the preset residual threshold, the reliable field values, mutually exclusive field markers, occlusion propagation markers, and candidate observation lists are read, and candidate state fields are selected in the order of reliable field value priority, spatial adjacency continuity, temporal adjacency continuity, and no spatial boundary crossing.

[0051] Furthermore, in this embodiment, the preset residual threshold for continuous numerical fields is set to 0.3, and the preset residual threshold for categorical and Boolean fields is set to 0.5. The preset residual threshold for continuous numerical fields is derived from the task configuration combination of field range, voxel side length of 2m, time slice length of 5s, and sensor error range. The preset residual thresholds for categorical and Boolean fields are derived from the discrete comparison rule of taking 0 for consistency and 1 for inconsistency. With this setting, continuous fields can tolerate small fluctuations in low-altitude observations, while categorical and Boolean fields can promptly identify abrupt changes in occupancy status, communication status, or airspace boundary relationships.

[0052] Furthermore, for the voxel occupancy field, the state residual field is set to 0 when the predicted field value matches the observed field value, and to 1 when the predicted field value does not match the observed field value. For the wind field field, the field range... The allowed wind speed range is set to 0 m / s to 12 m / s in this embodiment, derived from the wind field tolerance boundary in the UAV mission configuration. When the wind speed residual exceeds 0.30, the candidate observation list and adjacent voxel records are read to determine whether the wind field change has spatial continuity. This setting can distinguish between local sensor anomalies and sudden changes in the actual wind field.

[0053] Furthermore, when the state residual field exceeds a preset residual threshold, if the confidence field value of the candidate state field is higher than that of the current observed state field, and the candidate state field has continuous records in the horizontal, vertical, or temporal adjacency directions, and does not violate spatial boundary relationships, then the candidate state field replaces the current observed state field and forms the corrected state field. If the candidate state field cannot simultaneously satisfy both spatial and temporal adjacency continuity, it does not directly replace the current observed state field, but instead writes a status flag to be reviewed. The corrected state field continues to retain the source data source, writing time, confidence field value, and candidate observation list index, enabling the traceability of the environmental field source of each node and transition edge when generating a walkable state graph later.

[0054] It should be noted that this step is not a conventional smoothing process for continuous time slices, but rather a correction that incorporates voxel adjacency relationships, observation confidence fields, occlusion propagation markers, and spatial boundary non-crossing conditions. Through the low-altitude environment state tensor formula and the state residual field formula, voxel occupancy, obstacle movement, wind field, visibility, communication coverage, and spatial constraints are organized into a unified internal computer state structure. This allows for the differentiation and correction of unobserved states caused by real-world environmental changes, asynchronous data delays, and occlusion within the same spatiotemporal index structure.

[0055] S3: Generate a passable state diagram and incremental update instructions by coupling navigation constraints based on the low-altitude environment state tensor.

[0056] Furthermore, the navigation constraints include UAV outline constraints, climb constraints, turning radius constraints, wind field tolerance constraints, communication maintenance constraints, airspace boundary constraints, and safety interval constraints. Coupled navigation constraints include reading the corrected state field from the low-altitude environment state tensor, selecting voxel records that satisfy the voxel occupancy field, wind field field, communication coverage field, and airspace constraint field as candidate nodes, establishing transition edges between candidate nodes based on horizontal adjacency, vertical adjacency, temporal adjacency, and UAV outline constraints, and deleting transition edges that do not satisfy climb constraints, turning radius constraints, wind field tolerance constraints, communication maintenance constraints, airspace boundary constraints, or safety interval constraints.

[0057] It should be noted that navigation constraints specifically refer to the technical limitations corresponding to the actual flight capabilities and safe operating conditions of the UAV within the mission area. UAV profile constraints are determined by the UAV's airframe envelope and the mission payload envelope. Climb constraints are determined by the UAV's vertical motion capability boundaries. Turning radius constraints are determined by the UAV's horizontal maneuverability boundaries. Wind field tolerance constraints are determined by the wind field boundaries allowed by the UAV's mission configuration. Communication maintenance constraints are determined by link coverage identifiers and mission communication requirements. Airspace boundary constraints are determined by permitted flight boundaries, temporary control boundaries, and no-fly zones. Safety separation constraints are jointly determined by the UAV's external dimensions, positioning error, and voxel space boundary tolerance range.

[0058] Furthermore, candidate nodes are determined according to the following formula:

[0059]

[0060] in, Voxel recording and time slice identifier The corresponding candidate node labels are derived from the corrected low-altitude environment state tensor. This indicates a conditional indicator function. It returns 1 if all conditions within the parentheses are met, and 0 otherwise. This field represents the voxel occupancy, derived from the corrected status field. A value of 0 indicates no occupancy, and a value of 1 indicates occupancy. This indicates that the wind speed value in the wind field field is derived from the corrected status field. This represents the maximum permissible wind speed threshold corresponding to the constraints the wind field can withstand, and is derived from the UAV mission configuration file. The link coverage value corresponding to the communication coverage field is derived from the corrected status field. This represents the minimum communication coverage threshold corresponding to the communication maintenance constraint, derived from the communication requirements of the UAV mission. This field represents airspace boundary constraints and is derived from the corrected status field. A value of 1 indicates that the area is within the permitted flight boundary and has not entered the no-fly zone, while a value of 0 indicates that the airspace boundary constraints are not met. It represents the distance between the voxel record and the obstacle boundary or spatial restriction boundary, derived from the voxel spatial boundary and the corrected state field. The safety interval threshold is derived from the UAV's external dimensions, positioning error, and mission safety interval configuration.

[0061] The calculation process is as follows: The computer first reads the voxel occupancy field, wind field, communication coverage field, airspace constraint field, and safety interval related fields from the corrected low-altitude environmental state tensor. Then, it determines whether the voxel record meets the following conditions: unoccupied, wind speed not exceeding the maximum allowable wind speed threshold, communication coverage not lower than the minimum communication coverage threshold, airspace boundary not exceeding the limit, and safety interval not lower than the safety interval threshold. When all conditions are met, the corresponding voxel record is written to the candidate node set. When there is a status flag to be reviewed or a mutual exclusion field flag, it is not directly written to a passable node, but to a restricted node candidate set. When there is an occupancy conflict, airspace boundary exceeding the limit, communication coverage not meeting the constraint, or wind field bearing constraint not meeting the constraint, it is written to the prohibited node candidate set.

[0062] Furthermore, in this embodiment, the maximum permissible wind speed threshold Set to 10 m / s, with the setting source being the wind field tolerance boundary in the UAV mission configuration file. Minimum communication coverage threshold. Set to 0.65, with the source being the link-keeping requirements after normalization of the communication coverage field. Security interval threshold. The setting is 3m, and the source of this setting is a combination of the UAV's outer radius, the upper limit of positioning error, and the tolerance range of the voxel space boundary. With this setting, candidate nodes not only reflect whether the 3D environment is occupied, but also simultaneously reflect the limitations of wind field, communication, airspace, and safety interval on passability.

[0063] Furthermore, when generating transfer edges between candidate nodes, adjacent voxel records are read based on horizontal adjacency, vertical adjacency, and temporal adjacency, and the UAV's outline constraints are mapped to the node's occupied area. If the spatial boundary between two candidate nodes can accommodate the node's occupied area, and the transfer direction satisfies the climb constraint and turning radius constraint, then a candidate transfer edge is generated. If the intermediate voxel record traversed by the transfer edge has an occupancy conflict, airspace boundary overrun, communication coverage not meeting the constraint, or wind field tolerance not meeting the constraint, then the corresponding candidate transfer edge is deleted. In this embodiment, the maximum climb angle threshold is set to 25°, and the minimum turning radius threshold is set to 6m, which are derived from the vertical motion capability boundary and horizontal maneuverability boundary in the UAV mission configuration file. This setting allows the transfer edge to reflect whether adjacent voxel records satisfy the UAV's continuous motion capability, rather than simply connecting edges based on spatial adjacency.

[0064] The passability state diagram and incremental update instructions include passable nodes, restricted nodes, prohibited nodes, passable transfer edges, restricted transfer edges, update voxel range, update field name, update trigger reason, and version time. Generating the passability state diagram and incremental update instructions involves writing passable node markers to retained candidate nodes, writing restricted node markers to candidate nodes with pending verification state markers or mutually exclusive field markers, and writing prohibited node markers to candidate nodes with airspace boundary violations, voxel occupancy conflicts, communication coverage failures, or wind field tolerance failures. It also involves writing passable transfer edge markers to retained transfer edges and restricted transfer edge markers to transfer edges where any endpoint is a restricted node. The update voxel range is determined by voxel records whose state residual fields exceed a preset residual threshold, voxel records intersecting with candidate track regions, and voxel records adjacent to the UAV's current position. Update field names are generated based on the state dimension index that has changed within the update voxel range. Update trigger reasons are generated based on changes in source trusted field values, changes in state residual fields, changes in navigation constraints, and changes in candidate node markers. The update voxel range, update field name, and update trigger reason are arranged according to version time to form an incremental update instruction.

[0065] Furthermore, the passability state diagram and incremental update instructions include passable nodes, restricted nodes, prohibited nodes, passable transition edges, restricted transition edges, updated voxel range, updated field names, update trigger reasons, and version times. Generating the passability state diagram and incremental update instructions includes writing passable node markers to retained candidate nodes, writing restricted node markers to candidate nodes with pending verification status markers or mutually exclusive field markers, and writing prohibited node markers to candidate nodes with airspace boundary violations, voxel occupancy conflicts, communication coverage not meeting constraints, or wind field tolerance not meeting constraints. The instructions also include writing... Enter the passable transition edge marker, and write the restricted transition edge marker for any transition edge whose endpoint is a restricted node; determine the voxel records whose state residual field exceeds the preset residual threshold, voxel records that intersect with the candidate track region, and voxel records that are adjacent to the current position of the UAV as the update voxel range; generate the update field name based on the state dimension index that has changed within the update voxel range; generate the update trigger reason based on the changes in the source trusted field value, the changes in the state residual field, the changes in the navigation constraint, and the changes in the candidate node marker; arrange the update voxel range, the update field name, and the update trigger reason according to the version time to form the incremental update instruction.

[0066] It should be noted that when generating incremental update instructions, voxel records whose state residual fields exceed a preset residual threshold are read first, then voxel records intersecting with the track candidate region are read, and finally voxel records adjacent to the UAV's current position are read. These three types of voxel records are merged to form the updated voxel range. The track candidate region is determined by the mission start point, mission end point, mission altitude range, and mission time window, and does not rely on a manually selected single route. In this embodiment, voxel records adjacent to the UAV's current position include voxel records horizontally adjacent to one layer, vertically adjacent to one layer, and temporally adjacent to one time slice. The source of these voxel records is the voxel adjacency relationship and the reachable range of the UAV's adjacent time slices. This setting ensures that the incremental update range can cover the nearest safety impact area while avoiding updating the full 3D environment twin data each time.

[0067] Furthermore, the updated field names are determined by the state dimension index that has changed within the updated voxel range, including the voxel occupancy field, obstacle movement field, wind field field, visibility field, communication coverage field, airspace constraint field, trusted field value, mutual exclusion field marker, and pending verification state marker. The update trigger reason is determined by changes in the source trusted field value, changes in the state residual field, changes in navigation constraints, and changes in the candidate node marker. Each update trigger reason is bound to a voxel identifier and a time slice identifier.

[0068] Furthermore, the version time is jointly determined by the current time slice identifier and the update generation time. When generating an incremental update instruction, the updated voxel range, updated field name, update trigger reason, and version time are combined into an incremental record that can be written into the 3D environment twin result. For multiple incremental records generated within the same time slice for the same voxel record, they are arranged according to the version time and the index of the previous version is retained, enabling the task control flow to trace back the environmental state change process. The 3D environment twin result includes a spatiotemporal piecewise voxel baseline, an observation confidence field, a corrected low-altitude environmental state tensor, a navigable state map, and incremental update instructions.

[0069] It should be noted that this step does not involve attaching static no-fly boundaries to the 3D scene. Instead, it maps the UAV's outline, mobility, wind field, communication coverage, airspace boundaries, safety intervals, and temporal correction results together as nodes and transition edges. Using a candidate node selection formula, the corrected low-altitude environment state tensor is converted into a computer-searchable, walkable state graph. Through incremental update commands, voxel records with large state residuals, those close to the UAV's current position, or those intersecting with candidate track regions are prioritized for updating. This allows the low-altitude 3D environment twin results to be transformed from static display data into data processing results that can be updated with mission time slices.

[0070] One embodiment of the present invention provides a three-dimensional environment twin system for low-altitude spatiotemporal environments, including a piecewise field construction module, a tensor correction module, and a passage update module.

[0071] The segmented field construction module is used to acquire low-altitude multi-source environmental observation data of the UAV mission area and construct a spatiotemporal segmented voxel benchmark and observation confidence field; the tensor correction module is used to generate a low-altitude environmental state tensor based on the spatiotemporal segmented voxel benchmark and observation confidence field and perform time-series consistency correction; the access update module is used to generate a passable state diagram and incremental update instructions based on the low-altitude environmental state tensor and coupled with navigation constraints.

[0072] Reference Figure 2 This embodiment also provides a computer device applicable to the low-altitude spatiotemporal three-dimensional environment twinning method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the low-altitude spatiotemporal three-dimensional environment twinning method proposed in the above embodiment.

[0073] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0074] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for realizing a three-dimensional environment twin in low-altitude spatiotemporal space as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

Claims

1. A three-dimensional environment twinning method for low-altitude spatiotemporal space, characterized in that, include: Acquire low-altitude multi-source environmental observation data of the UAV mission area, and construct a spatiotemporal segmented voxel benchmark and observation reliability field; Based on the spatiotemporal piecewise voxel benchmark and the observation confidence field, a low-altitude environmental state tensor is generated and temporal consistency correction is performed. Based on the low-altitude environment state tensor, navigation constraints are coupled to generate a passable state diagram and incremental update instructions.

2. The three-dimensional environment twinning method for low-altitude spatiotemporal conditions as described in claim 1, characterized in that: The acquisition of low-altitude multi-source environmental observation data in the UAV mission area includes... Acquire UAV payload observation data, ground perception data, airspace constraint data, meteorological observation data, and communication coverage data; Based on the boundaries of the UAV mission area, the start and end times of the mission, and the mission altitude range, voxel records are formed that include voxel identifiers, spatial boundaries, time slice identifiers, data source identifiers, and coordinate transformation relationships. The voxel records are then stored in a spatiotemporal index structure under a unified spatial coordinate system and a unified time reference.

3. The three-dimensional environment twinning method for low-altitude spatiotemporal conditions as described in claim 2, characterized in that: The observation confidence field includes, For low-altitude multi-source environmental observation data that enter the same voxel record and the same time slice identifier, the positioning residual, acquisition time, observation line of sight occlusion marker, sensor working status marker and field integrity marker are read respectively. The time decay item is determined based on the interval between the acquisition time and the time slice identifier; the spatial deviation item is determined based on the positioning residual; the visibility deduction item is determined based on the observation line of sight occlusion mark; and the data availability item is determined based on the sensor working status mark and the field integrity mark. Combine the time-decrease item, spatial deviation item, visibility deduction item, and data availability item into the source trust field value; When there are mutually exclusive observations from different data sources for the same voxel record, retain the source confidence field values ​​and mutual exclusion field markers, and determine the master observation field value based on the sorting relationship between the source confidence field values; When there are no observations in the current time slice for the same voxel record, write the unobserved field value instead of replacing it with a null value; The trusted field value is passed to the low-altitude environment state tensor along with the voxel record.

4. The three-dimensional environment twinning method for low-altitude spatiotemporal conditions as described in claim 3, characterized in that: The generation of the low-altitude environment state tensor includes... Using voxel identifiers as spatial dimension indexes, time slice identifiers as time dimension indexes, and various environmental fields as state dimension indexes, the location, velocity, shape boundary, wind direction, wind speed, visibility, communication coverage category, and airspace boundary relationship in the low-altitude multi-source environmental observation data are written into the corresponding state dimension indexes. For environment fields with the same name from different data sources, the writing order is determined according to the trusted field value, and the candidate fields that have not been written are saved in the candidate observation list; Establish adjacency relationships for spatially and temporally adjacent voxel records. Adjacency relationships include horizontal adjacency, vertical adjacency, and temporal adjacency. An occlusion propagation marker is established for the occluded voxel record. The occlusion propagation marker and the observation line occlusion marker together determine whether the environmental field can be directly overwritten. The generated low-altitude environment state tensor includes voxel occupancy field, obstacle motion field, wind field field, visibility field, communication coverage field, airspace constraint field, trusted field value, mutual exclusion field marker, and time slice identifier. It retains the source data source, write time, and trusted field value of each state dimension index.

5. The three-dimensional environment twinning method for low-altitude spatiotemporal conditions as described in claim 4, characterized in that: The timing consistency correction includes, Read the preceding state field from the low-altitude environment state tensor corresponding to the previous time slice identifier, and generate the predicted state field of the current time slice identifier based on the obstacle motion field, wind field field, and time adjacency. The predicted state field is compared with the observed state field identified by the current time slice at the field level to form the state residual field; When the state residual field does not exceed the preset residual threshold, the observation state field identified by the current time slice is used as the corrected state field; When the state residual field exceeds the preset residual threshold, the reliable field value, mutual exclusion field marker, occlusion propagation marker and candidate observation list are read, and the candidate state fields are filtered in the order of reliable field value priority, spatial adjacency continuity, temporal adjacency continuity and spatial boundary non-crossing. For candidate state fields that cannot satisfy both spatial and temporal adjacency continuity, write a status flag to be reviewed; For candidate state fields that satisfy both spatial and temporal adjacency continuity, the corresponding observed state fields are replaced to form corrected state fields; The corrected status field covers the voxel occupancy field, obstacle motion field, wind field field, visibility field, communication coverage field, and airspace constraint field within the current time slice identifier.

6. The three-dimensional environment twinning method for low-altitude spatiotemporal conditions as described in claim 5, characterized in that: The navigation constraints include, Unmanned aerial vehicle (UAV) outline constraints, climb constraints, turning radius constraints, wind field tolerance constraints, communication maintenance constraints, airspace boundary constraints, and safety interval constraints; Coupled navigation constraints include reading the corrected state field from the low-altitude environment state tensor and using voxel records that satisfy the voxel occupancy field, wind field field, communication coverage field and airspace constraint field as candidate nodes. Transition edges between candidate nodes are established based on horizontal adjacency, vertical adjacency, temporal adjacency, and UAV outline constraints. Transition edges that do not meet the climb constraint, turning radius constraint, wind field tolerance constraint, communication maintenance constraint, airspace boundary constraint, or safety interval constraint are deleted.

7. The three-dimensional environment twinning method for low-altitude spatiotemporal conditions as described in claim 6, characterized in that: The walkable state diagram and incremental update instructions include, Passable nodes, restricted nodes, prohibited nodes, passable transition edges, restricted transition edges, update voxel range, update field name, update trigger reason and version time; The generation of the passable state diagram and incremental update instructions includes writing passable node markers to the retained candidate nodes, writing restricted node markers to the candidate nodes with pending verification state markers or mutually exclusive field markers, and writing prohibited node markers to the candidate nodes with airspace boundary crossing, voxel occupancy conflict, communication coverage not meeting constraints, or wind field bearing constraints not meeting constraints. Write a passable transition edge mark for the retained transition edges, and write a restricted transition edge mark for any transition edge whose endpoint is a restricted node; The voxel records whose state residual field exceeds the preset residual threshold, the voxel records that intersect with the track candidate region, and the voxel records that are adjacent to the current position of the UAV are collectively determined as the updated voxel range. Generate update field names based on the state dimension index that has changed within the updated voxel range; The update trigger reason is generated based on changes in the source trust field value, changes in the state residual field, changes in navigation constraints, and changes in candidate node labels; The update voxel range, update field name, and update trigger reason are arranged according to version time to form an incremental update instruction.

8. A three-dimensional environment twin system for low-altitude spatiotemporal environments, employing the three-dimensional environment twinning method for low-altitude spatiotemporal environments as described in any one of claims 1 to 7, characterized in that: Includes a segmented field construction module, a tensor correction module, and a passage update module; The segmented field construction module is used to acquire low-altitude multi-source environmental observation data of the UAV mission area and construct a spatiotemporal segmented voxel benchmark and a reliable observation field. The tensor correction module is used to generate a low-altitude environmental state tensor and perform temporal consistency correction based on the spatiotemporal piecewise voxel reference and the observation confidence field. The passage update module is used to generate a passable state diagram and incremental update instructions based on the low-altitude environment state tensor and coupled with navigation constraints.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the low-altitude spatiotemporal three-dimensional environment twin method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the three-dimensional environment twinning method for low-altitude spatiotemporal space as described in any one of claims 1 to 7.

Images