A virtual scene construction method driven by real-scene 3D data

Abstract

The present invention relates to the technical field of virtual scene construction, and discloses a method for constructing a virtual scene driven by real-life three-dimensional data. The method obtains multi-source three-dimensional data of a target area, extracts geometric feature parameters of scene data units, calculates topological feature indicators and generates a construction input set, loads a pre-configured three-dimensional modeling algorithm to determine dynamic evolution parameters, annotates parameters to generate a three-dimensional scene map, and constructs a virtual scene after spatial consistency detection. If a parameter conflict is detected, the parameters are fused according to the priority of multi-source data. After the construction is completed, the spatial difference value between the scene data unit and the adjacent unit is monitored in real time, and the parameters are corrected and the scene is updated when the threshold is exceeded. The method makes full use of multi-source data, improves the accuracy and efficiency of virtual scene construction, effectively solves the problems existing in traditional methods, and has broad application prospects in urban planning, cultural heritage protection and other fields.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual scene construction, and in particular to a virtual scene construction method driven by real-scene three-dimensional data. Background Art

[0002] With the rapid development of computer technology, digital image processing technology, and 3D modeling technology, virtual scene construction has shown great application potential in many fields, such as urban planning, cultural heritage protection, game development, film and television production, virtual reality (VR) and augmented reality (AR). However, current virtual scene construction technology still faces many challenges.

[0003] Traditional methods for constructing virtual scenes mostly rely on manual modeling, which requires a significant investment of manpower, material resources, and time. Modelers need to create each type of object in the scene one by one, and everything from object shape design to texture rendering to scene layout planning requires meticulous manual work. When constructing large-scale, complex scenes, this manual modeling is extremely inefficient and difficult to guarantee the scene's authenticity and accuracy. For example, when constructing a virtual scene of a large city, it is necessary to precisely model every building, every street, every green space, etc. Manual operations are not only labor-intensive but also prone to omissions or errors, resulting in deviations between the virtual scene and the actual scene.

[0004] Although some virtual scene construction technologies use 3D scanning technology to obtain data, there are limitations in the data processing and scene construction process. On the one hand, the data obtained from a single data source is often not comprehensive. For example, relying solely on point cloud data obtained by LiDAR, although it can accurately measure the spatial position information of an object, it lacks the ability to obtain the object's texture and semantic information; and using only oblique photography image data, although it can obtain rich texture information, there are certain errors in the precise measurement of spatial coordinates. This makes it difficult to achieve the ideal effect in the fusion of geometric accuracy and texture details in the constructed virtual scene. On the other hand, when processing multi-source data, different data sources have different data formats, accuracy, resolution, etc. How to efficiently fuse this data and extract useful information has become a key issue that restricts the quality of virtual scene construction.

[0005] During the construction of virtual scenes, there are also deficiencies in simulating the dynamic evolution of objects within them. Existing technologies often struggle to accurately simulate the dynamic changes of objects under varying conditions, failing to meet the demands of certain applications that demand real-time and interactivity. For example, when simulating urban traffic, dynamic information such as vehicle trajectories, pedestrian paths, and the changes in buildings over time are difficult to accurately simulate, resulting in a lack of realism and immersion in the virtual scenes.

[0006] Furthermore, existing virtual scene construction technologies need to be strengthened in terms of scene consistency detection and optimization. Due to the accumulation of errors during data acquisition and processing, the constructed virtual scenes may exhibit problems such as object position conflicts and unnatural texture splicing, affecting the quality of the virtual scenes and the user experience. Furthermore, after scene construction is complete, the lack of effective dynamic monitoring and real-time correction mechanisms makes it impossible to promptly respond to changes in the scene, such as building demolition and construction, and environmental changes. Summary of the Invention

[0007] The purpose of the present invention is to provide a method for constructing a virtual scene driven by real-scene three-dimensional data to solve the problems raised in the above background technology.

[0008] To achieve the above-mentioned object, the present invention provides the following technical solution: a method for constructing a virtual scene driven by real-scene three-dimensional data, the method comprising:

[0009] Acquire multi-source 3D data of the target area, and extract geometric feature parameters of each scene data unit from a preset 3D modeling engine, wherein the geometric feature parameters include spatial coordinate attributes, texture attributes, and semantic attributes of the scene data unit;

[0010] For each scene data unit, a topological feature index of the scene data unit is calculated based on the geometric feature parameters, and a construction input set of the scene data unit is generated based on the multiple topological feature indexes; wherein the topological feature index is a comprehensive quantitative parameter that integrates the spatial coordinate attributes and texture attributes of each scene data unit; and the construction input set is a standardized data set composed of the selected multiple topological feature indexes;

[0011] For each scene data unit, a preconfigured three-dimensional modeling algorithm is loaded, and based on the spatial relationship between the plurality of feature indicators included in the construction input set and the processing rules of the three-dimensional modeling algorithm, a dynamic evolution parameter of the scene data unit during the construction process is determined;

[0012] In the three-dimensional modeling engine, each scene data unit is labeled with the dynamic evolution parameter to generate a three-dimensional scene atlas;

[0013] For each scene data unit, performing a spatial consistency test according to the dynamic evolution parameter to obtain a test result;

[0014] A virtual scene is constructed based on the detection result to generate a virtual scene construction result.

[0015] Preferably, the virtual scene is constructed based on the detection result to generate a virtual scene construction result, including: when the detection result indicates that there is a conflict between multiple dynamic evolution parameters contained in the scene data unit, the scene data unit is determined as an abnormal unit; the conflict parameters of the abnormal unit are obtained as a correction object, and the multi-source data priority corresponding to the conflict parameters is determined; and the scene data unit is parameter-fused based on the data priority to generate a virtual scene construction result.

[0016] Preferably, the multi-source three-dimensional data includes lidar point cloud data and oblique photography image data, the data priority of the lidar point cloud data is a first priority, and the data priority of the oblique photography image data is a second priority; performing parameter fusion on the scene data units based on the data priorities to generate a virtual scene construction result includes:

[0017] adjusting a construction input set corresponding to the oblique photography image data according to an order of the first priority and the second priority;

[0018] Extracting dynamic evolution parameters of the oblique photography image data in the adjusted construction input set based on the three-dimensional scene atlas;

[0019] The dynamic evolution parameters are combined with the three-dimensional scene atlas to perform parameter fusion to generate a virtual scene construction result.

[0020] Preferably, for each scene data unit, the topological feature index of the scene data unit is calculated based on the geometric feature parameters, and the construction input set is generated according to the multiple topological feature indexes, including:

[0021] For each scene data unit, determining a plurality of adjacent units that are spatially adjacent to the unit;

[0022] Calculating the correlation between the geometric feature parameters and the texture attributes of each adjacent unit, and determining a first feature unit with the highest correlation from the adjacent units based on a plurality of the correlation degrees;

[0023] Based on the first feature unit, determining a plurality of extension units adjacent to the feature unit, calculating a correlation between a geometric feature parameter and a texture attribute of each extension unit, and determining a next feature unit with the highest correlation from the extension units based on the plurality of correlation degrees;

[0024] Repeating the steps of determining a plurality of extension units adjacent to the feature unit, calculating the correlation degree of each extension unit, and determining the next feature unit with the highest correlation degree until all scene data units in the target area are traversed, so as to obtain a set of topological feature indicators of the scene data units;

[0025] The construction input set is generated based on the topological feature indicator set.

[0026] Preferably, the calculating of the correlation between the geometric feature parameters and the texture attributes of each adjacent unit includes:

[0027] respectively calculating a first correlation degree between the spatial coordinate attribute and the texture attribute of each of the adjacent units, and a second correlation degree between the spatial coordinate attribute and the semantic attribute of each of the adjacent units;

[0028] For each of the adjacent units, a weighted sum of the first degree of association and the second degree of association is used as the degree of association of the adjacent unit.

[0029] Preferably, the determining the first feature unit with the highest correlation degree from the adjacent units based on the plurality of correlation degrees includes:

[0030] storing a plurality of association degrees corresponding to a plurality of adjacent units into a feature candidate list, and sorting the association degrees in the feature candidate list in descending order to obtain a sorting result;

[0031] Based on the ranking result, the adjacent unit corresponding to the highest correlation degree is determined as the first feature unit.

[0032] Preferably, after constructing a virtual scene based on the detection result and generating a virtual scene construction result, the method further includes:

[0033] After the virtual scene construction result is generated, for each scene data unit, the spatial difference value between the scene data unit and adjacent units is monitored in real time;

[0034] When the spatial difference value exceeds a preset threshold, the adjacent unit is used as an abnormal reference unit, and the dynamic parameters of the scene data unit are modified based on the abnormal reference unit to generate an updated virtual scene construction result.

[0035] Preferably, the training process of the preconfigured 3D modeling algorithm includes:

[0036] Acquire a historical three-dimensional data set, and extract a mapping relationship between geometric feature parameters and dynamic evolution parameters in the data set;

[0037] Constructing an initial three-dimensional modeling algorithm according to the mapping relationship, and iteratively optimizing the algorithm using a meshing method;

[0038] When the error between the predicted parameters output by the algorithm and the measured parameters is lower than a preset threshold, it is determined that the algorithm configuration is complete.

[0039] Preferably, the three-dimensional modeling engine further includes a three-dimensional modeling layer of spatial grid division, texture mapping rules and illumination model. When generating the three-dimensional scene atlas, the dynamic evolution parameters are superimposed on the three-dimensional modeling layer for visual expression.

[0040] Preferably, the method for generating the three-dimensional modeling layer includes:

[0041] Collect terrain point cloud data, building surface data and ambient lighting data of the target area;

[0042] Performing gridding processing on the terrain point cloud data to generate a spatial grid division layer, performing texture mapping on the building surface data to generate a texture mapping rule layer, and simulating the ambient lighting data to generate a lighting model layer;

[0043] The spatial grid division layer, the texture mapping rule layer and the illumination model layer are subjected to spatial overlay analysis to form the three-dimensional modeling layer.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] In terms of data utilization, this method obtains multi-source three-dimensional data of the target area, such as lidar point cloud data and oblique photography image data, giving full play to the advantages of different data sources. Lidar point cloud data can accurately provide spatial coordinate information, while oblique photography image data is rich in texture information. By extracting the geometric feature parameters of each scene data unit from the preset three-dimensional modeling engine, covering spatial coordinate attributes, texture attributes and semantic attributes, in-depth mining and integration of multi-source data are achieved. Compared with the traditional single data source modeling method, this greatly enriches the basic data for scene construction, making the constructed virtual scene more realistic and accurate in geometric accuracy and texture details. For example, when constructing a virtual scene of a historical and cultural block, it can not only accurately restore the external dimensions and spatial position of the building, but also show the traces of time and decorative details on the surface of the building through texture attributes, providing strong support for the digital protection and display of cultural heritage.

[0046] The process of calculating topological feature indicators and generating a construction input set further optimizes the data processing method. For each scene data unit, the topological feature indicators are calculated by integrating its spatial coordinate attributes and texture attributes. This comprehensive quantitative parameter can better reflect the intrinsic connection between scene data units. By determining adjacent units, calculating the correlation degree, and screening feature units, the construction input set generated by the obtained topological feature indicator set is a standardized data set, providing a high-quality data foundation for subsequent modeling. In the construction of virtual scenes in urban areas, this method can accurately capture the spatial relationship between buildings and between roads and buildings, making the scene layout more reasonable and in line with the actual geographic space logic.

[0047] Preconfigured 3D modeling algorithms play a crucial role in the construction of virtual scenes. This algorithm acquires historical 3D datasets, extracts the mapping between geometric feature parameters and dynamic evolution parameters, and employs a gridding method for iterative optimization. Configuration is considered complete when the error between the predicted parameters output by the algorithm and the measured parameters falls below a preset threshold. This enables the algorithm to accurately determine the dynamic evolution parameters of scene data units during the construction process based on the spatial relationships between the feature indicators in the input set. For example, in a scenario simulating changes in urban traffic flow, the algorithm can accurately predict vehicle trajectories and speed changes (dynamic evolution parameters) based on traffic flow data (as part of the geometric feature parameters) for roads over different time periods. This allows for dynamic simulation of traffic scenarios, enhancing the realism and interactivity of virtual scenes.

[0048] Generating a three-dimensional scene atlas and performing spatial consistency detection effectively ensures the quality of the virtual scene. In the three-dimensional modeling engine, the dynamic evolution parameters of each scene data unit are annotated to generate a three-dimensional scene atlas, and the conflict of dynamic evolution parameters contained in the scene data unit is promptly discovered through spatial consistency detection. When a conflict is detected, the parameters of the scene data unit are fused according to the priority of multi-source data. For example, when processing conflicts between lidar point cloud data and oblique photography image data, the input set is adjusted and constructed according to the set priority order, and the dynamic evolution parameters are extracted and fused to ensure the consistency of information such as the position and texture of objects in the virtual scene, avoid problems such as model overlap and texture confusion, and enhance the visual effect and use value of the virtual scene.

[0049] Furthermore, after the virtual scene is constructed, the spatial difference between the scene data unit and its adjacent units is monitored in real time. Dynamic parameter corrections are made to the scene data unit when the difference exceeds a preset threshold. This mechanism enables the virtual scene to adapt to dynamic changes in the environment, such as the construction or demolition of buildings during urban development, allowing for timely updates of the virtual scene content to maintain accuracy and timeliness. This method has significant application value, both in urban planning, providing planners with real-time updated virtual scenes to assist decision-making, and in game development, creating more immersive and realistic gaming environments for players. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a working principle diagram of the virtual scene construction method driven by real-scene three-dimensional data according to the present invention;

[0051] Figure 2 A workflow diagram for constructing virtual scenes based on data prioritization for processing oblique photography imagery data;

[0052] Figure 3A workflow diagram for calculating topological feature indices and generating a construction input set;

[0053] Figure 4 The workflow diagram for calculating the correlation degree of adjacent units;

[0054] Figure 5 Workflow diagram generated for 3D modeling layers. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0056] See also Figure 1-Figure 5 The present invention provides a method for constructing a virtual scene driven by real-scene three-dimensional data, and the specific implementation steps are as follows:

[0057] Acquire multi-source 3D data of the target area. This data comes from a wide range of sources, such as point cloud data collected by LiDAR and image data acquired through oblique photography. Extract the geometric feature parameters of each scene data unit from a pre-set 3D modeling engine. These include spatial coordinate attributes for the scene data unit, which accurately determine its position in space; texture attributes, which reveal detailed surface features; and semantic attributes, which assign specific meaning to the data unit, such as distinguishing between buildings, roads, and vegetation.

[0058] For each scene data unit, a topological feature index is calculated based on the acquired geometric feature parameters. This index is a comprehensive quantitative parameter that integrates spatial coordinate attributes and texture attributes, and it can more comprehensively reflect the characteristics of the scene data unit. After calculating multiple topological feature indices, appropriate indices are selected to form a standardized data set, i.e., the input set, which provides the basic data for subsequent processing.

[0059] A preconfigured 3D modeling algorithm is loaded for each scene data unit. Based on the spatial relationships between multiple feature indicators contained in the construction input set and the processing rules of the 3D modeling algorithm, the dynamic evolution parameters of the scene data unit during the construction process are determined. These parameters determine how the scene data unit changes during the virtual scene construction process.

[0060] In the 3D modeling engine, each scene data unit is annotated with the above-determined dynamic evolution parameters to generate a 3D scene graph. This graph integrates the information of each scene data unit and provides intuitive visualization and data support for virtual scene construction.

[0061] For each scene data unit, a spatial consistency test is performed based on the annotated dynamic evolution parameters to obtain the test results. This test mainly determines whether the dynamic evolution parameters of the scene data unit are spatially coordinated and consistent, and whether there are any conflicts or unreasonable situations.

[0062] Based on the detection results, a virtual scene is constructed to generate a virtual scene construction result. If the detection results indicate that there is no conflict in the dynamic evolution parameters of the scene data unit, these parameters are directly used to construct the virtual scene; if there is a conflict, further processing is required to ensure the accuracy and rationality of the virtual scene.

[0063] The technical solution of the present invention is further described below through specific embodiments:

[0064] Example 1:

[0065] When constructing a virtual scene, if the detection results indicate that there are multiple conflicts in the dynamic evolution parameters contained in the scene data unit, the scene data unit is determined to be an abnormal unit. The conflict parameters of the abnormal unit are obtained as the correction object, and the multi-source data priority corresponding to the conflict parameters is determined. For example, if the multi-source three-dimensional data includes lidar point cloud data and oblique photography image data, and the data priority of the lidar point cloud data is the first priority, and the data priority of the oblique photography image data is the second priority, the parameters of the scene data units are fused according to the data priority to generate the virtual scene construction result. Specifically, first, according to the priority order, the construction input set corresponding to the oblique photography image data is adjusted to better match the lidar point cloud data; then, based on the three-dimensional scene atlas, the dynamic evolution parameters of the oblique photography image data in the adjusted construction input set are extracted; finally, these dynamic evolution parameters are combined with the three-dimensional scene atlas for parameter fusion to obtain a more accurate virtual scene construction result. In this way, the advantages of different data sources can be fully utilized to improve the quality of the virtual scene.

[0066] Suppose you want to build a virtual scene of a city center area. This area contains multiple scene elements such as multiple buildings, streets, and parks. You use lidar to collect point cloud data and oblique photography to obtain image data as multi-source 3D data for constructing the virtual scene.

[0067] During the construction process, it was detected that a certain scene data unit (assuming it corresponds to a building on a street corner) had multiple conflicting dynamically evolving parameters. The scene data unit corresponding to this building was identified as an abnormal unit, and its conflicting parameters were obtained as the target for correction. It is known that LiDAR point cloud data has the highest data priority, while oblique photography data has the second highest data priority.

[0068] Adjust the construction input set corresponding to the oblique photography data based on priority. For example, LiDAR point cloud data can more accurately reflect the outline and height information of the building, while oblique photography data has an advantage in texture detail. When making adjustments, first ensure that the spatial coordinate information of the oblique photography data is consistent with the LiDAR point cloud data. Based on the approximate location and shape of the building determined by the LiDAR point cloud data, fine-tune the spatial position and angle of the corresponding part of the oblique photography data to better integrate it with the LiDAR point cloud data.

[0069] Based on the adjusted build input set, the dynamic evolution parameters of the oblique photography data are extracted from the 3D scene atlas. The 3D scene atlas records various information about each scene data unit. Using a specific extraction algorithm, the dynamic evolution parameters of the oblique photography data, such as texture changes and lighting effects, are obtained under the current adjusted build input set.

[0070] These dynamically evolving parameters are then combined with the 3D scene atlas for parameter fusion. For example, the building geometry determined by LiDAR point cloud data is combined with the texture information from oblique photography data, giving the building both precise shape and realistic texture in the virtual scene. Simultaneously, taking lighting effects into account, the lighting-related parameters from both data types are comprehensively calculated to determine the final lighting parameters, resulting in a more natural lighting and shadow effect for the building in the virtual scene. These steps generate a virtual scene construction result, creating a more realistic and accurate street corner building model that, together with surrounding scene elements, forms a complete virtual scene of the city center.

[0071] Example 2:

[0072] When calculating topological feature indices for each scene data unit and generating a construction input set, the system first determines multiple spatially adjacent neighboring units. Next, it calculates the correlation between the geometric feature parameters and texture attributes of each neighboring unit. This correlation is calculated by calculating a first correlation between the spatial coordinate attributes and texture attributes, and a second correlation between the spatial coordinate attributes and semantic attributes of each neighboring unit. The weighted sum of these correlations is then used as the correlation for the neighboring unit. Based on these multiple correlations, the first feature unit with the highest correlation is determined from the neighboring units. Specifically, the correlations corresponding to the multiple neighboring units are stored in a feature candidate list, and the correlations in the feature candidate list are sorted in descending order. The neighboring unit with the highest correlation is determined as the first feature unit. Based on the first feature unit, multiple extended units adjacent to the feature unit are determined. The correlation between the geometric feature parameters and texture attributes of each extended unit is calculated. Based on the multiple correlations, the next feature unit with the highest correlation is determined from the extended units. These steps are repeated until all scene data units in the target area have been traversed, resulting in a set of topological feature indices for the scene data units. This set is then used to generate the construction input set. This method can more comprehensively and accurately reflect the relationship between scene data units and provide more reliable data for subsequent virtual scene construction.

[0073] Suppose you want to build a virtual scene of a large campus, which contains a variety of scene elements such as teaching buildings, libraries, playgrounds, green areas, etc. These elements are divided into scene data units.

[0074] When calculating the topological feature index of each scene data unit and generating the construction input set, a scene data unit of one of the teaching buildings (assuming it is scene data unit A corresponding to the entrance of the teaching building) is used as an example for explanation.

[0075] First, multiple spatially adjacent units of scene data unit A are identified. These may include those corresponding to the stairs, flower beds, and roads near the entrance to the teaching building. Next, the correlation between the geometric feature parameters and texture attributes of each adjacent unit is calculated. For example, for adjacent unit B corresponding to the stairs, the first correlation between its spatial coordinate attributes and texture attributes, as well as the second correlation between its spatial coordinate attributes and semantic attributes, are calculated. The correlation between spatial coordinate and texture attributes can be evaluated by considering how closely they describe the characteristics of the stairs. If the texture of the stairs exhibits a pattern related to the direction of the steps, and this pattern is highly consistent with the spatial coordinate changes of the stairs (such as height changes and positional direction), then the first correlation is high. The second correlation between spatial coordinate attributes and semantic attributes is calculated based on the connection between the spatial location of the stairs and the semantic concept of "stairs." If the location of the stairs aligns with the normal flow of people on campus and is highly associated with the semantic concept of "for students and teachers to go up and down the stairs," the second correlation will also be high. The weighted sum of these two correlations is used to obtain the correlation for adjacent unit B. The correlations of other adjacent units (such as those corresponding to flower beds and roads) are calculated using the same method.

[0076] The association degrees of all adjacent cells are stored in a feature candidate list, and then the association degrees in the list are sorted in descending order. Assume that after sorting, it is found that the association degree of adjacent cell B corresponding to the stairs is the highest, then adjacent cell B is determined as the first feature cell.

[0077] Based on the first feature unit B, multiple extension units adjacent to it are determined. These extension units may include scene data units corresponding to floors and corridors connected to stairs. The correlation between the geometric feature parameters and texture attributes of each extension unit is then calculated. For example, for extension unit C corresponding to a floor connected to the stairs, its correlation is calculated according to the previous method. Then, based on the multiple correlations, the next feature unit with the highest correlation is determined from the extension units. Assuming that the correlation of unit C is the highest among these extension units, unit C is determined as the next feature unit.

[0078] Repeat the steps of identifying multiple extended units adjacent to the feature unit, calculating the correlation, and determining the next feature unit with the highest correlation. With this repetition, all scene data units across the entire campus are gradually traversed, resulting in a set of topological feature indicators for scene data unit A. This set encompasses a series of indicators that are closely related to scene data unit A in terms of space and characteristics, comprehensively reflecting its connection with surrounding scene elements.

[0079] Finally, a construction input set is generated based on the obtained set of topological feature indicators. This construction input set serves as key data for the subsequent virtual scene construction, providing strong support for accurately constructing the virtual scene of the teaching building entrance and the entire campus, ensuring that the constructed virtual scene can more realistically reflect the actual layout and characteristics of the campus.

[0080] Example 3:

[0081] After the virtual scene is constructed, the spatial difference value between each scene data unit and its adjacent units is monitored in real time. When a spatial difference value exceeds a preset threshold, the adjacent unit is used as an abnormal reference unit, and the scene data unit is dynamically corrected based on the abnormal reference unit to generate an updated virtual scene construction result. This process can promptly detect possible problems in the virtual scene and continuously optimize the virtual scene through dynamic parameter correction to make it more in line with actual conditions. During the monitoring process, by continuously comparing the spatial position, geometric features and other information of the scene data unit with those of the adjacent units, once it is found that the difference exceeds the preset range, the correction program is immediately initiated to ensure the accuracy and stability of the virtual scene.

[0082] Suppose that a virtual scene of a large commercial plaza is being constructed. The scene contains multiple commercial buildings, parking lots, fountain landscapes and other elements. These elements are divided into numerous scene data units.

[0083] After the virtual scene is constructed, each scene data unit is monitored in real time. For example, scene data unit A corresponding to a landmark tower in a commercial plaza and scene data unit B at the adjacent parking lot entrance are selected.

[0084] Set a spatial difference value calculation method, assuming that the spatial difference value The calculation formula is: ,in, represents the spatial coordinates of scene data unit A, Represents the spatial coordinates of scene data unit B. This formula measures the degree of spatial difference between two scene data units by calculating the Euclidean distance between them in three-dimensional space.

[0085] Preset a threshold , assuming (Unit: meter, can be set according to the actual scene accuracy requirements). During the real-time monitoring process, the spatial difference value between scene data unit A and adjacent unit B is continuously calculated. .

[0086] At a certain moment, due to data updates or model adjustments in the virtual scene, the calculated spatial difference value meters, exceeding the preset threshold At this time, the adjacent unit B is used as the abnormal reference unit.

[0087] Dynamic parameter corrections are performed on scene data unit A based on the abnormal reference unit B. For example, an inspection revealed that the spatial coordinates of scene data unit A deviated during the data update process, resulting in a significant spatial difference from the adjacent unit B. Based on the accurate spatial location information of scene data unit B and the overall layout logic of the commercial plaza, the spatial coordinate parameters of scene data unit A are corrected. At the same time, considering that scene data unit A is a tower, its relationship with the surrounding environment, such as light, shadow, and occlusion, will also change due to changes in spatial position. Dynamic evolution parameters such as lighting and shadows are adjusted accordingly.

[0088] After the parameters of scene data unit A were corrected, the relationships and parameters between the various scene data units in the entire virtual scene were recalculated to generate an updated virtual scene construction result. This process improved the spatial relationship between the tower and the parking lot entrance in the virtual scene, making the virtual scene of the entire commercial plaza more realistic and enhancing its accuracy and realism.

[0089] Example 4:

[0090] When training a preconfigured 3D modeling algorithm, the algorithm first acquires a historical 3D dataset and extracts the mapping relationship between geometric feature parameters and dynamic evolution parameters from the dataset. Based on this mapping relationship, an initial 3D modeling algorithm is constructed, and the algorithm is iteratively optimized using a meshing method. During the iterative process, the algorithm's parameters and structure are continuously adjusted to gradually reduce the error between the predicted parameters output by the algorithm and the measured parameters. When the error between the predicted parameters output by the algorithm and the measured parameters falls below a preset threshold, the algorithm configuration is determined to be complete. The 3D modeling algorithm obtained through such training can more accurately process the construction input set of scene data units and determine more reasonable dynamic evolution parameters, thereby improving the quality and efficiency of virtual scene construction.

[0091] Suppose you want to build a virtual scene for a historical block with a typical architectural style. To achieve this goal, you need to train an accurate 3D modeling algorithm.

[0092] A large number of historical 3D datasets for this historic district were collected from a variety of sources, including previous street surveys and high-precision imagery. From these datasets, a mapping relationship between geometric feature parameters and dynamic evolution parameters was extracted. For example, for an ancient temple within the district, the geometric feature parameters include the temple's length, width, and height, as well as the dimensions and spatial positions of various architectural components (such as eaves and brackets). Dynamic evolution parameters include changes in the temple's appearance over time under changing light and shadow, and changes in the building's surface texture due to natural erosion. By analyzing data from a large number of similar temples and other buildings within the district, the corresponding relationships between these geometric feature parameters and dynamic evolution parameters were summarized. For example, it was found that the building's orientation and height affect its light and shadow effects at different times, which in turn affects the illumination-related parameters in the dynamic evolution parameters.

[0093] An initial 3D modeling algorithm is constructed based on the extracted mapping relationships. This algorithm preliminarily establishes rules for calculating and predicting dynamic evolution parameters based on geometric feature parameters. Next, a gridding method is used to iteratively optimize the algorithm. Taking the temple area as an example, the temple and the surrounding space within a certain range are divided into grids, with each grid serving as an independent analysis unit. During each iteration, the dynamic evolution parameters predicted by the algorithm within each grid are compared with the parameters obtained from actual measurements. For example, if there is a difference between the predicted light intensity of a building surface within a grid at a specific time and the actual measured light intensity, the parameters and rules related to light calculation in the algorithm are adjusted based on this difference. Through continuous iteration, the error between the predicted parameters output by the algorithm and the measured parameters is gradually reduced.

[0094] During the iterative optimization process, a preset threshold is set. For example, an error rate of less than 5% is considered to be an acceptable accuracy range. When the error between the predicted parameters output by the algorithm and the measured parameters is lower than this preset threshold, the algorithm configuration is determined to be complete. The three-dimensional modeling algorithm obtained through such training can more accurately determine reasonable dynamic evolution parameters based on the input geometric feature parameters when processing scene data units of historical blocks. For example, when constructing a virtual scene, for buildings in different positions and structures in the block, their appearance changes under different time and natural conditions can be accurately predicted based on their geometric features, thereby providing reliable technical support for the construction of highly restored and dynamically changing virtual scenes of historical blocks, allowing people to more realistically experience the style of historical blocks in virtual scenes.

[0095] Example 5:

[0096] The 3D modeling engine includes 3D modeling layers for spatial meshing, texture mapping rules, and a lighting model. When generating a 3D scene atlas, dynamically evolving parameters are overlaid onto the 3D modeling layers for visual representation. The 3D modeling layers are generated by first collecting terrain point cloud data, building surface data, and ambient lighting data for the target area. The terrain point cloud data is then gridded to generate a spatial meshing layer, transforming the terrain data into a grid format that is easy to process and analyze. Texture mapping is then performed on the building surface data to generate a texture mapping rule layer, enabling realistic texture effects on the building surfaces. Ambient lighting data is then simulated to generate a lighting model layer, simulating the scene's lighting conditions. Finally, the spatial meshing layer, texture mapping rule layer, and lighting model layer are spatially overlaid and analyzed to form the 3D modeling layer. This generated 3D modeling layer provides rich information for constructing virtual scenes, making them more realistic and vivid.

[0097] Taking the construction of a virtual scene of a large theme park as an example, when constructing the virtual scene of a theme park, it is necessary to use a 3D modeling engine to generate a 3D modeling layer containing a variety of information to achieve a more realistic and vivid virtual scene display.

[0098] First, data collection is performed, including the theme park's topographic point cloud data, building surface data, and ambient lighting data. For topographic point cloud data, professional LiDAR equipment is used to scan the entire theme park's terrain, obtaining precise three-dimensional coordinate information for every feature within the park. For example, the undulating terrain in the roller coaster area and the terrain changes around the artificial lake are all accurately recorded. Building surface data is acquired through on-site photography and measurement. For the various distinctive buildings within the theme park, such as fairytale castles and pirate ship-shaped entertainment facilities, the surface texture details of the buildings are captured in all directions, and the dimensions of each building component are measured. Ambient lighting data is collected using light sensors, which measure information such as light intensity and angle in various areas of the park at different time periods and under different weather conditions.

[0099] After data collection, the terrain point cloud data is gridded to generate a spatial grid layer. The massive amount of terrain point cloud data collected is divided into small grids according to specific rules. For example, a 1m x 1m grid is used as a unit. Each grid is assigned a corresponding terrain height value. This creates a spatial grid layer that clearly displays the undulating terrain of the park. This layer allows users to intuitively identify high-lying and low-lying areas within the theme park, providing a topographical framework for subsequent scene construction.

[0100] Texture mapping is performed on the building surface data to generate a texture mapping rule layer. For example, a previously captured high-definition photo of the castle's surface is processed and, based on the castle's actual structure and dimensions, the textures in the photo are accurately mapped onto the corresponding building model surface. Texture mapping rules are determined, such as the alignment of brick textures on the castle walls and the correspondence between door and window positions and the actual photo. This ensures that the castle in the virtual scene appears as realistic as in reality. This generated texture mapping rule layer imbues the building in the virtual scene with rich detail and a sense of realism.

[0101] Simulate ambient lighting data to generate a lighting model layer. Using professional lighting simulation software, based on collected lighting data for different time periods and weather conditions, simulate the corresponding lighting effects in the virtual scene. For example, during a sunny noon, simulate strong direct light to cast noticeable shadows on the building; during dusk, simulate soft, warm light to create a cozy atmosphere. By adjusting parameters such as light intensity, angle, and color, a lighting model layer that matches the actual situation is generated, making the lighting and shadow effects in the virtual scene more natural.

[0102] The spatial grid division layer, texture mapping rule layer, and lighting model layer are spatially overlaid and analyzed to form a 3D modeling layer. During this process, the information from each layer is integrated to interconnect the terrain, building, and lighting information. For example, the undulations of the terrain affect the angle of illumination and the distribution of shadows, the position and shape of buildings can block light, and textures present different effects under lighting. Through this spatial overlay analysis, the generated 3D modeling layer contains rich information, providing a comprehensive foundation for the subsequent generation of 3D scene maps and the construction of virtual scenes. When generating the 3D scene map, the dynamic evolution parameters of the scene data units are overlaid on this 3D modeling layer for visual expression, allowing virtual scene builders to more intuitively see the position, characteristics, and dynamic changes of each scene data unit in the entire theme park virtual scene, thereby constructing a highly restored, vivid and realistic theme park virtual scene.

[0103] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0104] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a virtual scene driven by real-scene three-dimensional data, characterized in that: The method comprises: Acquire multi-source 3D data of the target area, and extract geometric feature parameters of each scene data unit from a preset 3D modeling engine, wherein the geometric feature parameters include spatial coordinate attributes, texture attributes, and semantic attributes of the scene data unit; For each scene data unit, a topological feature index of the scene data unit is calculated based on the geometric feature parameters, and a construction input set of the scene data unit is generated based on the multiple topological feature indexes; wherein the topological feature index is a comprehensive quantitative parameter that integrates the spatial coordinate attributes and texture attributes of each scene data unit; and the construction input set is a standardized data set composed of the selected multiple topological feature indexes; For each scene data unit, a preconfigured three-dimensional modeling algorithm is loaded, and based on the spatial relationship between the plurality of feature indicators included in the construction input set and the processing rules of the three-dimensional modeling algorithm, a dynamic evolution parameter of the scene data unit during the construction process is determined; In the three-dimensional modeling engine, each scene data unit is labeled with the dynamic evolution parameter to generate a three-dimensional scene atlas; For each scene data unit, performing a spatial consistency test according to the dynamic evolution parameter to obtain a test result; A virtual scene is constructed based on the detection result to generate a virtual scene construction result, including: when the detection result indicates that there is a conflict between multiple dynamic evolution parameters contained in the scene data unit, the scene data unit is determined as an abnormal unit; the conflict parameters of the abnormal unit are obtained as a correction object, and the multi-source data priority corresponding to the conflict parameters is determined; and the parameters of the scene data unit are fused based on the data priority to generate a virtual scene construction result.

2. The method for constructing a virtual scene driven by real-scene three-dimensional data according to claim 1, characterized in that: The multi-source three-dimensional data includes laser radar point cloud data and oblique photography image data, the data priority of the laser radar point cloud data is a first priority, and the data priority of the oblique photography image data is a second priority; performing parameter fusion on the scene data units based on the data priorities to generate a virtual scene construction result includes: adjusting a construction input set corresponding to the oblique photography image data according to an order of the first priority and the second priority; Extracting dynamic evolution parameters of the oblique photography image data in the adjusted construction input set based on the three-dimensional scene atlas; The dynamic evolution parameters are combined with the three-dimensional scene atlas to perform parameter fusion to generate a virtual scene construction result.

3. The method for constructing a virtual scene driven by real-scene three-dimensional data according to claim 1, characterized in that: The step of calculating, for each scene data unit, a topological feature index of the scene data unit based on the geometric feature parameters, and generating the construction input set according to a plurality of the topological feature indexes includes: For each scene data unit, determining a plurality of adjacent units that are spatially adjacent to the unit; Calculating the correlation between the geometric feature parameters and the texture attributes of each adjacent unit, and determining a first feature unit with the highest correlation from the adjacent units based on a plurality of the correlation degrees; Based on the first feature unit, determining a plurality of extension units adjacent to the feature unit, calculating a correlation between a geometric feature parameter and a texture attribute of each extension unit, and determining a next feature unit with the highest correlation from the extension units based on the plurality of correlation degrees; Repeating the steps of determining a plurality of extension units adjacent to the feature unit, calculating the correlation degree of each extension unit, and determining the next feature unit with the highest correlation degree until all scene data units in the target area are traversed, so as to obtain a set of topological feature indicators of the scene data units; The construction input set is generated based on the topological feature indicator set.

4. The method for constructing a virtual scene driven by real-scene three-dimensional data according to claim 3, characterized in that: The calculation of the correlation between the geometric feature parameters and the texture attributes of each adjacent unit includes: respectively calculating a first correlation degree between the spatial coordinate attribute and the texture attribute of each of the adjacent units, and a second correlation degree between the spatial coordinate attribute and the semantic attribute of each of the adjacent units; For each of the adjacent units, a weighted sum of the first degree of association and the second degree of association is used as the degree of association of the adjacent unit.

5. The method for constructing a virtual scene driven by real-scene 3D data according to claim 3, characterized in that: The determining, based on the plurality of correlation degrees, a first feature unit having the highest correlation degree from the adjacent units comprises: storing a plurality of association degrees corresponding to a plurality of adjacent units into a feature candidate list, and sorting the association degrees in the feature candidate list in descending order to obtain a sorting result; Based on the ranking result, the adjacent unit corresponding to the highest correlation degree is determined as the first feature unit.

6. The method for constructing a virtual scene driven by real-scene 3D data according to claim 1, characterized in that: After constructing the virtual scene based on the detection result and generating the virtual scene construction result, the method further includes: After the virtual scene construction result is generated, for each scene data unit, the spatial difference value between the scene data unit and adjacent units is monitored in real time; When the spatial difference value exceeds a preset threshold, the adjacent unit is used as an abnormal reference unit, and the dynamic parameters of the scene data unit are modified based on the abnormal reference unit to generate an updated virtual scene construction result.

7. The method for constructing a virtual scene driven by real-scene 3D data according to claim 1, characterized in that: The training process of the preconfigured 3D modeling algorithm includes: Acquire a historical three-dimensional data set, and extract a mapping relationship between geometric feature parameters and dynamic evolution parameters in the data set; Constructing an initial three-dimensional modeling algorithm according to the mapping relationship, and iteratively optimizing the algorithm using a meshing method; When the error between the predicted parameters output by the algorithm and the measured parameters is lower than a preset threshold, it is determined that the algorithm configuration is complete.

8. The method for constructing a virtual scene driven by real-scene 3D data according to claim 1, characterized in that: The three-dimensional modeling engine also includes a three-dimensional modeling layer of spatial grid division, texture mapping rules and illumination model. When generating the three-dimensional scene atlas, the dynamic evolution parameters are superimposed on the three-dimensional modeling layer for visual expression.

9. The method for constructing a virtual scene driven by real-scene three-dimensional data according to claim 8, characterized in that: The method for generating the three-dimensional modeling layer includes: Collect terrain point cloud data, building surface data and ambient lighting data of the target area; Performing gridding processing on the terrain point cloud data to generate a spatial grid division layer, performing texture mapping on the building surface data to generate a texture mapping rule layer, and simulating the ambient lighting data to generate a lighting model layer; The spatial grid division layer, the texture mapping rule layer and the illumination model layer are subjected to spatial overlay analysis to form the three-dimensional modeling layer.

Images