A method, system, product, and medium for seamless and rapid fusion of multi-dimensional 3D data

Abstract

The present application provides a method, system, product and medium for seamless and rapid fusion of multivariate three-dimensional data, and relates to the field of electrical digital data processing technology. By acquiring multi-source heterogeneous data and unifying them into the coordinate system of the geographic base layer, a benchmark reference for data fusion is established. The geographic base layer serves as a benchmark and provides a stable spatial reference for subsequent data fusion. The surface smoothing algorithm is used to preprocess the real-life three-dimensional model and the monomer model data, reducing the impact of data noise on the fusion accuracy. Accurate alignment of different types of three-dimensional data is achieved through the coordinated use of feature point extraction and iterative nearest point algorithm. This feature-based iterative optimization method avoids the error accumulation problem caused by direct overall fusion, improves the fusion accuracy, and ensures the geometric consistency of the fused scene.

Description

Technical Field

[0001] The present application relates to the field of electronic digital data processing technology, and in particular to a method for seamless and rapid fusion of multivariate three-dimensional data. Background Art

[0002] With the rapid development of smart city construction, urban planning and management need to integrate many different types of 3D data. These data include real-life 3D models collected by surveying and mapping, individual models of planning and design, digital elevation models and digital orthophotos from topographic surveys, and BIM models and CAD drawings in engineering construction. Each type of data has different format standards and coordinate systems.

[0003] In related technologies, the overall fusion method is usually adopted to import all types of 3D data into the processing system at the same time, and perform data fusion based on the idea of ​​global matching. First, all data are batch-coordinate converted, then the corresponding relationship between different types of data is established through spatial topological relationship analysis, and finally the mesh deformation algorithm is used to adjust the model as a whole.

[0004] However, this overall fusion processing method will lead to excessive data coupling. When processing a large amount of heterogeneous 3D data, due to the lack of benchmark reference and processing order, it is easy to cause error accumulation. Especially when fusing terrain data, real scene models and engineering data, due to the different accuracy and importance of different types of data, the unified processing method often causes the position of important nodes to shift, affecting the overall fusion effect.

[0005] Therefore, in the related art, when there is too much three-dimensional data, it is easy to cause error accumulation, resulting in poor fusion effect. Summary of the invention

[0006] The present application provides a method, system, product and medium for seamless and rapid fusion of multivariate three-dimensional data, which are used to avoid the error accumulation problem caused by direct overall fusion and improve the fusion accuracy.

[0007] In the first aspect, the present application provides a method for seamless and rapid fusion of multi-dimensional three-dimensional data, including: obtaining real-life three-dimensional model data, individualized model data, digital elevation model data and digital orthophoto data, wherein the real-life three-dimensional model data is in osgb format and the individualized model data is in obj format; performing registration processing on the digital elevation model data and the digital orthophoto data to generate a geographic base layer of the three-dimensional scene; converting the real-life three-dimensional model data and the individualized model data to the coordinate system of the geographic base layer, so that the real-life three-dimensional model data and the individualized model data are aligned with the spatial position of the geographic base layer; using a surface smoothing algorithm to process the real-life three-dimensional model data and the individualized model data in the three-dimensional scene; receiving BIM data and CAD data; converting the BIM data and CAD data to the coordinate system of the geographic base layer; extracting feature points in the real-life three-dimensional model data, the individualized model data, the BIM data and the CAD data as point cloud data, using an iterative closest point algorithm to calculate the optimal transformation matrix between the point cloud data, and performing spatial transformation on the three-dimensional data according to the optimal transformation matrix until the distance error between the feature points is less than a preset distance threshold, thereby generating fused three-dimensional scene data.

[0008] By adopting the above technical solutions, a benchmark reference for data fusion is established by acquiring multi-source heterogeneous data and unifying them into the coordinate system of the geographic base layer. The geographic base layer serves as a benchmark and provides a stable spatial reference for subsequent data fusion. The surface smoothing algorithm is used to preprocess the real-life 3D model and the monomer model data, reducing the impact of data noise on fusion accuracy. The precise alignment of different types of 3D data is achieved through the combined use of feature point extraction and iterative nearest point algorithm. This feature-based iterative optimization method avoids the error accumulation problem caused by direct overall fusion, improves fusion accuracy, and ensures the geometric consistency of the fused scene.

[0009] In combination with some embodiments of the first aspect, in some embodiments, after the step of generating fused three-dimensional scene data, the method also includes: identifying missing areas of the real-life three-dimensional model data and generating characteristic entropy data of the missing areas; extracting geometric features of the complete area of ​​the real-life three-dimensional model data and generating characteristic flow field data, wherein the characteristic flow field data includes characteristic propagation direction data and characteristic propagation intensity data; propagating the geometric features of the complete area to the missing area according to preset propagation rules; the preset propagation rules include: the characteristic propagation intensity data decays nonlinearly with increasing propagation distance; the characteristic propagation intensity data increases with increasing characteristic entropy data; and the propagation process of the geometric features is constrained by the characteristic propagation direction data.

[0010] By adopting the above technical solution, the geometric complexity of the missing area is quantified through the characteristic entropy data, providing an accurate reference for subsequent feature propagation. Combined with the propagation direction and intensity information in the characteristic flow field data, an adaptive feature propagation mechanism is established. The characteristic propagation intensity will decay nonlinearly with the increase of propagation distance. This attenuation characteristic is consistent with the continuity characteristics of the actual scene. At the same time, the propagation intensity will increase with the increase of characteristic entropy, ensuring that more propagation supplementary information can be obtained in areas with complex geometric features, and finally achieving the natural restoration of the missing area and maintaining the visual continuity of the scene.

[0011] In combination with some embodiments of the first aspect, in some embodiments, before the step of identifying missing areas of the real-life three-dimensional model data and generating characteristic entropy data of the missing areas, the method also includes: setting an initial granularity parameter, wherein the initial granularity parameter is a preset maximum grid edge length; based on the initial granularity parameter, performing the step of identifying missing areas of the real-life three-dimensional model data and generating characteristic entropy data of the missing areas to obtain the real-life three-dimensional model data after propagation; reducing the initial granularity parameter by a preset step size; treating the real-life three-dimensional model data after propagation as the real-life three-dimensional model data, performing the step of identifying missing areas of the real-life three-dimensional model data and generating characteristic entropy data of the missing areas to obtain the real-life three-dimensional model data after propagation, until the initial granularity parameter reaches the preset minimum grid edge length.

[0012] By adopting the above technical solutions, a multi-scale feature propagation mechanism from coarse to fine is established. By setting the initial maximum grid edge length, feature propagation is completed at a larger scale to ensure the correctness of the overall morphology. As the grid edge length gradually decreases, feature propagation gradually focuses on more detailed geometric features. This progressive processing method not only improves computational efficiency, but also maintains the continuity of geometric features at different scales, ultimately achieving high-quality missing area repair effects.

[0013] In combination with some embodiments of the first aspect, in some embodiments, the step of using a surface smoothing algorithm to process real-life three-dimensional model data and individualized model data in a three-dimensional scene specifically includes: establishing a neighbor index for each grid point in the surface grid of the real-life three-dimensional model data and the individualized model data, and generating neighborhood index data for the grid point; determining the adjacent points of the grid point based on the neighborhood index data; calculating the local curvature of the grid point based on the grid point and the corresponding adjacent points; determining the grid point whose local curvature is greater than a preset curvature threshold as an elevation mutation point, and determining the grid point whose local curvature is not greater than the preset curvature threshold as a non-elevation mutation point; calculating the regional average elevation data of the grid point and the adjacent points to obtain initial smoothed elevation data; determining the weight coefficient data of the elevation mutation point based on the local curvature, wherein the weight coefficient data approaches 1 as the local curvature increases; for non-elevation mutation points, using the initial smoothed elevation data for smoothing; for elevation mutation points, using the weight coefficient data for smoothing.

[0014] By adopting the above technical solution, the neighborhood index data of the grid points is established, which provides a basis for local feature analysis. The calculation of local curvature enables the system to accurately identify elevation mutation points, which often represent important terrain features. For non-elevation mutation points, the regional average elevation is directly used for smoothing, while for elevation mutation points, the weight coefficient based on local curvature is used for adaptive smoothing. This differentiated processing method not only retains the main features of the terrain, but also improves the overall surface smoothness, effectively balancing the relationship between detail preservation and smoothing.

[0015] In combination with some embodiments of the first aspect, in some embodiments, for elevation mutation points, after the step of smoothing using weight coefficient data, the method also includes: extracting the elevation reference value of the position corresponding to the smoothed grid point from the geographic base layer to obtain reference elevation data; comparing the elevation data of the smoothed grid point with the reference elevation data to generate elevation deviation data; setting elevation constraint weights according to the elevation deviation data to generate constraint weight data, wherein the constraint weight data decreases as the elevation deviation data increases; and correcting the elevation data of the grid point based on the constraint weight data to generate corrected grid data.

[0016] By adopting the above technical solutions, a reliable reference system is established by introducing the geographic base layer as the elevation benchmark. By calculating the elevation deviation data, the system can quantitatively evaluate the accuracy of the smoothing process. The introduction of constraint weight data realizes an adaptive correction mechanism. When the elevation deviation is large, the correction range is limited by a smaller constraint weight, and vice versa, a greater degree of freedom for correction is allowed. This correction mechanism based on reference and constraints ensures the accuracy of the elevation data while avoiding distortion caused by over-correction.

[0017] In combination with some embodiments of the first aspect, in some embodiments, the step of using an iterative nearest point algorithm to calculate the optimal transformation matrix between point cloud data specifically includes: extracting corner point features and contour feature points from the point cloud data to generate feature point set data; using a scale-invariant feature transformation algorithm to describe the feature point set data to generate feature descriptor data; performing feature point matching on the feature descriptor data based on a nearest neighbor search algorithm to generate initial matching point pair data; using a random sampling consistency algorithm to optimize the initial matching point pair data to generate optimized matching point pair data; calculating coordinate transformation matrix data based on the optimized matching point pair data; applying the coordinate transformation matrix data to the point cloud data to calculate the distance error data between the transformed feature points; determining whether the distance error data is less than a preset distance threshold; if it is less than the preset distance threshold, determining the coordinate transformation matrix data as the optimal transformation matrix; if it is not less than the preset distance threshold, returning to the step of calculating the coordinate transformation matrix data based on the optimized matching point pair data.

[0018] By adopting the above technical solution, the key features of the point cloud data are obtained by extracting corner points and contour feature points. The scale-invariant feature transformation algorithm is used for feature description to ensure the stability of feature matching. The use of the random sampling consistency algorithm effectively removes the wrong matching point pairs and improves the accuracy of subsequent transformation matrix calculations. Through the iterative optimization process, continuous adjustments are made until the preset threshold requirements are met, and finally a high-precision spatial transformation matrix is ​​obtained, achieving accurate alignment of point cloud data.

[0019] In combination with some embodiments of the first aspect, in some embodiments, the step of converting real-life three-dimensional model data and monomerized model data to the coordinate system of a geographic base layer specifically includes: obtaining digital elevation model data and digital orthophoto data; resampling the digital elevation model data to obtain resampled elevation data, so that the spatial resolution data of the resampled elevation data is consistent with that of the digital orthophoto data; using a control point matching algorithm to perform geographic registration processing on the resampled elevation data and the digital orthophoto data to generate registered geographic data; constructing triangular mesh model data based on the resampled elevation data in the registered geographic data; and mapping the digital orthophoto data in the registered geographic data as texture data to the triangular mesh model data.

[0020] By adopting the above technical solutions, the spatial resolution of elevation data and image data is unified through resampling processing, laying the foundation for subsequent registration. The use of control point matching algorithm ensures the accuracy of geographic registration. A triangular mesh model is constructed based on the registered elevation data, and the orthophoto data is used as texture mapping. This processing method not only ensures geometric accuracy, but also provides a realistic visual effect. Finally, the high-precision construction of the geographic base layer is achieved, providing a reliable spatial reference for subsequent data fusion.

[0021] In the second aspect, the present application provides a system for seamless and rapid fusion of multivariate three-dimensional data, the system comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code comprises computer instructions, and one or more processors call the computer instructions to enable the system for seamless and rapid fusion of multivariate three-dimensional data to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0022] In the third aspect, the present application provides a computer program product comprising instructions, which, when run on a system for seamless and rapid fusion of multivariate three-dimensional data, enables the system for seamless and rapid fusion of multivariate three-dimensional data to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0023] In a fourth aspect, the present application provides a computer-readable storage medium comprising instructions, which, when executed on a system for seamless and rapid fusion of multivariate three-dimensional data, enables the system for seamless and rapid fusion of multivariate three-dimensional data to execute the method described in the first aspect and any possible implementation of the first aspect.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0025] 1. By acquiring multi-source heterogeneous data and unifying them into the coordinate system of the geographic base layer, a benchmark reference for data fusion is established. The geographic base layer serves as a benchmark and provides a stable spatial reference for subsequent data fusion. The surface smoothing algorithm is used to preprocess the real-life 3D model and the monomer model data, reducing the impact of data noise on fusion accuracy. The precise alignment of different types of 3D data is achieved through the combined use of feature point extraction and iterative nearest point algorithm. This feature-based iterative optimization method avoids the error accumulation problem caused by direct overall fusion, improves fusion accuracy, and ensures the geometric consistency of the fused scene.

[0026] 2. The geometric complexity of the missing area is quantified through the feature entropy data, providing an accurate reference for subsequent feature propagation. An adaptive feature propagation mechanism is established by combining the propagation direction and intensity information in the feature flow field data. The feature propagation intensity will decay nonlinearly with the increase of propagation distance. This attenuation characteristic is consistent with the continuity characteristics of the actual scene. At the same time, the propagation intensity will increase with the increase of feature entropy, ensuring that more propagation supplementary information can be obtained in areas with complex geometric features, and finally achieving the natural restoration of the missing area and maintaining the visual continuity of the scene.

[0027] 3. Establishing neighborhood index data of grid points provides a basis for local feature analysis. The calculation of local curvature enables the system to accurately identify elevation mutation points, which often represent important terrain features. For non-elevation mutation points, the regional average elevation is directly used for smoothing, while for elevation mutation points, the weight coefficient based on local curvature is used for adaptive smoothing. This differentiated processing method not only retains the main features of the terrain, but also improves the overall surface smoothness, effectively balancing the relationship between detail preservation and smoothing. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of a method for seamless and rapid fusion of multivariate three-dimensional data in an embodiment of the present application;

[0029] Figure 2 This is a specific flow chart of step S103 in the embodiment of the present application;

[0030] Figure 3 This is a specific flow chart of step S104 in the embodiment of the present application;

[0031] Figure 4 This is a specific flow chart of step S107 in the embodiment of the present application;

[0032] Figure 5 This is another flow chart of the method for seamless and rapid fusion of multivariate three-dimensional data in an embodiment of the present application;

[0033] Figure 6 It is an exemplary hardware structure diagram of the multivariate three-dimensional data seamless and rapid fusion system in the embodiment of the present application. DETAILED DESCRIPTION

[0034] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to and includes any or all possible combinations of one or more listed items.

[0035] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0036] See also Figure 1 , Figure 1 This is a flow chart of a method for seamless and rapid fusion of multivariate three-dimensional data in an embodiment of the present application;

[0037] S101, obtaining real-scene 3D model data, monomerized model data, digital elevation model data and digital orthophoto data, wherein the real-scene 3D model data is in osgb format and the monomerized model data is in obj format;

[0038] Among them, real-life 3D model data refers to 3D model data with real textures obtained through real-life 3D modeling technology, which is often used to represent the geometry and texture information of actual scenes such as buildings and terrain. Monomerized model data refers to model data after complex 3D models are divided into independent objects, which is used to represent independent buildings or structures. Digital elevation model data refers to raster data representing ground elevation, which is used to describe the ups and downs of terrain. Digital orthophoto data refers to geometrically corrected aerial or satellite remote sensing images, which are used to provide real image information of the earth's surface. The osgb format is a data format optimized for large-scale 3D scenes, and the obj format is a general 3D model file format.

[0039] S102, performing registration processing on the digital elevation model data and the digital orthophoto data to generate a geographic base layer of the three-dimensional scene;

[0040] Among them, the registration process refers to the process of unifying the spatial position information of different data sources into the same coordinate system.

[0041] In some embodiments, the digital elevation model data is projected and resampled so that it has the same spatial resolution and coordinate system as the digital orthophoto. It is precisely aligned by matching control points to ensure that the two types of data correspond precisely in spatial position. The aligned data is organized into a geographic base layer of the three-dimensional scene as a baseline reference for subsequent data fusion.

[0042] In some specific embodiments, it is necessary to first select a plurality of easily identifiable control points based on geographic features, such as obvious landmarks such as building corners, road intersections, and characteristic buildings. For each control point, it is necessary to accurately locate and record its coordinate values ​​in the digital elevation model and orthophoto respectively. It is recommended to select at least 4-6 evenly distributed control points to ensure the accuracy of the transformation. Subsequently, a quadratic polynomial coordinate transformation equation is established based on these control point pairs. The equation needs to consider the transformation relationships such as translation, rotation, and scaling. The optimal transformation parameters, including the six parameters of the affine transformation matrix, are calculated by the least squares method. Finally, the calculated transformation parameters are used to perform spatial correction on the entire data set, and the correction accuracy is evaluated by residual analysis, which is not limited here.

[0043] In some specific embodiments, firstly, feature points with significant features are extracted from two data using algorithms such as SIFT or SURF. These feature points usually include edge points, corner points and other locations with obvious local features. Then, the correspondence between the feature points is established through feature descriptor matching algorithms, such as KNN (K nearest neighbor) or RANSAC, and the wrong matching points are removed. Based on the matched feature point pairs, the rigid body transformation matrix that minimizes the sum of square distances between corresponding points is calculated using methods such as SVD decomposition. The calculated transformation matrix is ​​applied to the entire data set for spatial transformation. Then, the Euclidean distance between the transformed feature point pairs is calculated to evaluate the registration accuracy. If the accuracy does not meet the preset threshold (usually set to 1 / 2 or 1 / 3 of the data resolution), the feature point matching is updated based on the current result, and the above process is repeated until the required accuracy is reached or the maximum number of iterations is reached, which is not limited here.

[0044] S103, converting the real-scene 3D model data and the individualized model data into the coordinate system of the geographic base layer, so that the real-scene 3D model data and the individualized model data are aligned with the spatial position of the geographic base layer;

[0045] In some embodiments, the original coordinate systems of the real-scene 3D model data and the individualized model data are determined, and the differences between them and the coordinate system of the geographic base layer are analyzed. Coordinate transformation parameters are established, including transformation relationships such as translation, rotation, and scaling. Through a unified transformation algorithm, the two types of data are converted to the coordinate system of the geographic base layer, and spatial correction is performed to ensure accurate position alignment.

[0046] See also Figure 2 , Figure 2 This is a specific flow chart of step S103 in the embodiment of the present application;

[0047] In some embodiments, step S103 specifically includes:

[0048] S1031. Acquire digital elevation model data and digital orthophoto data;

[0049] S1032, resampling the digital elevation model data to obtain resampled elevation data, so that the spatial resolution data of the resampled elevation data is consistent with the spatial resolution data of the digital orthophoto data;

[0050] The resampling process refers to the process of converting the spatial resolution of raster data. In some embodiments, the original spatial resolution of the digital elevation model data and the digital orthophoto data is first analyzed to determine the resolution conversion ratio to be performed. Then, a suitable resampling algorithm is selected, taking into account the data characteristics and accuracy requirements. Finally, the resampling operation is performed to generate elevation data with the same spatial resolution as the orthophoto.

[0051] S1033, using a control point matching algorithm to perform geo-registration processing on the resampled elevation data and the digital orthophoto data to generate geo-registration data;

[0052] The control point matching algorithm refers to a calculation method for spatial position correspondence using control points.

[0053] In some embodiments, firstly, obvious feature points of objects are selected as control points in the two data sets to ensure that the control points are evenly distributed and representative. Then, the corresponding relationship between the control points is established and the optimal spatial transformation parameters are calculated. Finally, the entire data set is transformed to achieve accurate alignment of spatial positions.

[0054] In some specific embodiments, 8-12 evenly distributed characteristic control points are selected from the digital elevation model and the orthophoto respectively, such as building corners, road intersections, ridgeline inflection points and other easily identifiable land features; professional measurement tools or software are used to accurately record the spatial coordinate values ​​corresponding to each pair of control points in the two data, and a coordinate correspondence table is established; based on the coordinate values ​​of the control point pairs, a quadratic polynomial equation group including translation, rotation and scaling transformations is established; the optimal parameter values ​​of the transformation equations are calculated using the least squares method to minimize the residual sum of squares between the control point pairs; the calculated transformation parameters are applied to the entire digital elevation model data to achieve spatial position correction; by calculating the residual values ​​and root mean square errors of the control points after the transformation, it is evaluated whether the registration accuracy meets the requirements, and if necessary, control points are added or the control point positions are reselected, which is not limited here.

[0055] In some specific embodiments, a feature detection algorithm such as SIFT, SURF or ORB is used to extract a large number of key point sets with significant features from digital elevation models and orthophotos, and these feature points usually include locations where local gradient changes are obvious; the feature description vector of each feature point is calculated, and a preliminary correspondence between feature points is established using a nearest neighbor matching algorithm such as FLANN, and the RANSAC algorithm is used to eliminate false matches; based on the correspondence between the screened feature points, the optimal spatial transformation matrix is ​​calculated through an iterative optimization method such as the LM algorithm to minimize the Euclidean distance between corresponding points; the optimized transformation matrix is ​​applied to the entire data set to complete the spatial transformation; the registration errors between all feature point pairs after the transformation are calculated, including the average error, maximum error and standard deviation; if the registration accuracy does not reach a preset threshold (such as 1 / 2 of the pixel size), the feature point matching relationship is updated, and the optimization process is repeated until the accuracy requirement is met or the maximum number of iterations is reached.

[0056] S1034, constructing triangular mesh model data based on the resampled elevation data in the registered geographic data;

[0057] The triangular mesh model data represents a terrain surface model composed of triangular patches.

[0058] In some embodiments, the regular grid points of the resampled elevation data are first converted into scattered point data. Then, a triangulation algorithm is used to connect the scattered points to form an irregular triangulated network. Finally, the triangulated network is optimized to ensure that the mesh quality meets the modeling requirements.

[0059] S1035 , mapping the digital orthophoto data in the registered geographic data as texture data to the triangular mesh model data.

[0060] In some embodiments, the correspondence between the triangular mesh vertices and the orthophoto pixels is first established. Then the texture coordinates of each triangular facet are calculated to ensure seamless texture connection. Finally, the orthophoto data is mapped to the 3D model surface according to the calculated texture coordinates.

[0061] In some specific embodiments, the corresponding texture UV coordinates are first calculated based on the geographic coordinates of the triangular mesh vertices to ensure the continuity and accuracy of the coordinate mapping; then a mapping relationship from the vertex coordinates to the texture space is established for each triangular face to generate a texture mapping matrix; then the orthophoto data is allocated to each triangular face according to the mapping relationship, which is not limited here.

[0062] In some specific embodiments, the area to be processed is divided into multiple overlapping rectangular texture blocks according to a certain size (such as 512×512 pixels), ensuring that there is a 10-20% overlap area between adjacent blocks; then, the mapping parameters from the two-dimensional image space to the three-dimensional model surface are calculated for each texture block separately, and local texture coordinates are optimized; the calculated texture blocks are spliced ​​according to the spatial position relationship to ensure seamless connection between the texture blocks.

[0063] It can be seen that the spatial resolution of elevation data and image data is unified through resampling processing, laying the foundation for subsequent registration. The use of control point matching algorithm ensures the accuracy of geographic registration. A triangular mesh model is constructed based on the registered elevation data, and the orthophoto data is used as texture mapping. This processing method not only ensures geometric accuracy but also provides a realistic visual effect. Finally, the high-precision construction of the geographic base layer is achieved, providing a reliable spatial reference for subsequent data fusion.

[0064] S104, using a surface smoothing algorithm to process the real-scene 3D model data and the monomerized model data in the 3D scene;

[0065] The surface smoothing algorithm refers to a mathematical processing method for improving the surface quality of a three-dimensional model. In some embodiments, the model surface is first meshed to establish a neighborhood relationship of mesh points. Then, the local curvature characteristics of each mesh point are calculated to identify areas that need to be focused on. Finally, an adaptive smoothing algorithm is used to process the area based on the characteristics of different areas to ensure that key features are retained while improving the overall surface quality.

[0066] See also Figure 3 , Figure 3 This is a specific flow chart of step S104 in the embodiment of the present application;

[0067] In some specific embodiments, S104 specifically includes:

[0068] S1041, establishing a neighbor index for each grid point in the surface grid of the real-scene 3D model data and the monomerized model data, and generating neighborhood index data of the grid point;

[0069] Among them, the surface mesh represents the model surface structure composed of polygonal patches, the grid points refer to the vertices that constitute the mesh, the neighbor index represents the spatial adjacency relationship between the grid points, and the neighborhood index data is used to represent the organizational structure of the neighboring points around each grid point.

[0070] It should be noted that both the real-scene 3D model data and the individualized model data can be processed using steps S1041 to S1047, which will not be described in detail here.

[0071] In some embodiments, a spatial index structure is established for each grid point, and the neighborhood range of each point is determined by using a spatial partitioning or distance calculation method, and the neighborhood information is organized into a data structure that is easy to query. This index structure can quickly obtain the neighboring point information of any grid point.

[0072] S1042, determining the neighboring points of the grid point according to the neighborhood index data;

[0073] In some embodiments, based on the established neighborhood index data, a neighborhood query is performed on each grid point, all grid points directly connected to the current grid point are obtained according to the query results, and information of these points is stored as a neighboring point set.

[0074] S1043, calculating the local curvature of the grid point according to the grid point and the corresponding adjacent points;

[0075] The local curvature indicates the degree of curvature of the mesh surface near the point.

[0076] In some embodiments, the geometric features of the surface around the grid point are calculated based on the spatial position relationship between the grid point and its adjacent points. The curvature value of the surface at the point is estimated by analyzing the normal vector change, height difference and other features of the point, which reflects the degree of undulation of the local surface.

[0077] In some specific embodiments, the elevation value of the target grid point and the elevation values ​​of the adjacent points are obtained; the absolute value of the height difference between the target point and each adjacent point is calculated; the average value of these height differences is calculated as a local height change index; the index is divided by the horizontal horizontal distance from the adjacent point to the target point to obtain the height change rate; and the change rate is used as an approximate representation of the local curvature, which is not limited here.

[0078] In some specific embodiments, a triangular face formed by a target grid point and its adjacent points is obtained; the normal vector of each face is calculated; the angle between these normal vectors is calculated; the average value of the normal vector angle is calculated as a measure of local curvature; the larger the value, the greater the local curvature and the more drastic the terrain change, which is not limited here.

[0079] S1044, determining the grid points whose local curvature is greater than a preset curvature threshold as elevation mutation points, and determining the grid points whose local curvature is not greater than the preset curvature threshold as non-elevation mutation points;

[0080] S1045, calculating the regional average elevation data of the grid point and the adjacent points to obtain initial smoothed elevation data;

[0081] In some embodiments, for each grid point, the elevation characteristics of the local area are calculated based on its neighboring point set. The initial smoothed elevation value of the point is obtained by weighted averaging the elevation values ​​of all points in the neighborhood, which reflects the overall elevation trend of the local area.

[0082] S1046, determining weight coefficient data of the elevation mutation point according to the local curvature, wherein the weight coefficient data approaches 1 as the local curvature increases;

[0083] The weight coefficient data indicates the degree of preservation of elevation mutation points during smoothing. The weight coefficient approaches 1 as the local curvature increases, indicating that the points with greater curvature retain their original features more during smoothing.

[0084] In some embodiments, a mapping relationship from curvature to weight is established based on the local curvature value of the elevation mutation point. By designing a suitable functional relationship, the point with a larger curvature obtains a weight value closer to 1, thereby retaining more of its original features in the subsequent smoothing process.

[0085] S1047. For non-elevation mutation points, the initial smoothed elevation data is used for smoothing; for elevation mutation points, the weight coefficient data is used for smoothing.

[0086] In some embodiments, for non-elevation mutation points, the initial smoothed elevation data is directly used to replace the original elevation value to achieve a complete smoothing effect; for elevation mutation points, the original elevation value and the initial smoothed elevation value are weighted combined according to the weight coefficient, which not only maintains the mutation characteristics but also achieves moderate smoothing.

[0087] It can be seen that establishing neighborhood index data of grid points provides a basis for local feature analysis. The calculation of local curvature enables the system to accurately identify elevation mutation points, which often represent important terrain features. For non-elevation mutation points, the regional average elevation is directly used for smoothing, while for elevation mutation points, the weight coefficient based on local curvature is used for adaptive smoothing. This differentiated processing method not only retains the main features of the terrain, but also improves the overall surface smoothness, effectively balancing the relationship between detail preservation and smoothing.

[0088] In some embodiments, after step S1047, the method further includes:

[0089] S1048, extracting the elevation reference value of the corresponding position of the smoothed grid point from the geographic base layer to obtain reference elevation data;

[0090] In some embodiments, the smoothed grid point coordinates are converted into a geographic coordinate system, and the corresponding position of each grid point is located in the geographic base layer. The standard elevation value of the position is obtained from the geographic base layer through a spatial interpolation method as a reference for subsequent elevation correction.

[0091] S1049, comparing the elevation data of the smoothed grid points with the reference elevation data to generate elevation deviation data;

[0092] S10410, setting an elevation constraint weight according to the elevation deviation data to generate constraint weight data, wherein the constraint weight data decreases as the elevation deviation data increases;

[0093] The elevation constraint weight represents the control parameter of the elevation correction strength. The constraint weight data refers to a set of values ​​used to adjust the correction degree.

[0094] In some embodiments, a decreasing function is designed based on the elevation deviation data so that points with larger deviations receive smaller constraint weights. This setting ensures that areas with larger deviations are more strongly corrected, while areas with smaller deviations remain relatively stable.

[0095] S10411. Correct the elevation data of the grid points based on the constraint weight data to generate corrected grid data.

[0096] In some embodiments, the elevation value of each grid point is appropriately adjusted toward the reference elevation value according to the constraint weight data. The magnitude of the adjustment is controlled by the constraint weight (e.g., the constraint weight data is multiplied by the elevation data), ensuring that the corrected model not only maintains the original morphological features, but also better conforms to the actual terrain.

[0097] It can be seen that the introduction of the geographic base layer as the elevation benchmark has established a reliable reference system. By calculating the elevation deviation data, the system can quantitatively evaluate the accuracy of the smoothing process. The introduction of constraint weight data realizes an adaptive correction mechanism. When the elevation deviation is large, the correction amplitude is limited by a smaller constraint weight, and vice versa, a greater degree of freedom for correction is allowed. This correction mechanism based on reference and constraints ensures the accuracy of the elevation data while avoiding distortion caused by over-correction.

[0098] S105, receiving BIM data and CAD data;

[0099] Among them, BIM data refers to building information model data, which is used to represent the geometry, function and attribute information of buildings. CAD data refers to computer-aided design data, which is used to represent the geometry and size information of engineering designs.

[0100] S106, converting the BIM data and the CAD data into a coordinate system of a geographic base layer;

[0101] In some embodiments, the original coordinate system characteristics of BIM and CAD data are first analyzed to determine the conversion relationship with the coordinate system of the geographic base layer. Then a suitable conversion model is established, taking into account factors such as projection transformation and scale transformation. Finally, the coordinate conversion is performed, and the accuracy is evaluated and necessary adjustments are made.

[0102] It should be noted that the principle and process of this step are similar to those of step S103. The relevant principles and processes can be referred to step S103 and will not be repeated here.

[0103] S107, extracting feature points from real-scene 3D model data, monomerized model data, BIM data and CAD data as point cloud data, using an iterative closest point algorithm to calculate the optimal transformation matrix between the point cloud data, performing spatial transformation on the 3D data according to the optimal transformation matrix until the distance error between the feature points is less than a preset distance threshold, and generating fused 3D scene data.

[0104] Among them, feature points represent key points with significant features in three-dimensional data. Point cloud data refers to a data set consisting of a large number of spatial coordinate points. The iterative closest point algorithm is an iterative optimization algorithm for point cloud registration. The optimal transformation matrix represents the spatial transformation parameters that can minimize the distance error between point clouds. The preset distance threshold refers to the maximum allowed registration error.

[0105] In some embodiments, feature points are first extracted from various types of 3D data and converted into a unified point cloud format. Then, the transformation parameters are continuously optimized and calculated by iterative closest point algorithm until the preset accuracy requirements are met. Finally, the optimal transformation matrix is ​​applied to complete the accurate fusion of all data and generate the final 3D scene.

[0106] See also Figure 4 , Figure 4 This is a specific flow chart of step S107 in the embodiment of the present application;

[0107] In some embodiments, the step of using an iterative closest point algorithm to calculate the optimal transformation matrix between point cloud data specifically includes:

[0108] S1071, extracting corner point features and contour feature points from the point cloud data to generate feature point set data;

[0109] Among them, the corner feature represents the significant turning point or corner position in the point cloud data. The contour feature point refers to the key point set that describes the outline of the object. The feature point set data represents the data set containing all the extracted feature points.

[0110] In some embodiments, the input point cloud data is subjected to corner detection and contour feature extraction, and points with significant features are identified by calculating local geometric characteristics, such as curvature, normal vector change, etc. These feature points are organized into sets to provide basic data for subsequent feature matching.

[0111] S1072, using a scale-invariant feature transformation algorithm to perform feature description on the feature point set data to generate feature description sub-data;

[0112] The scale-invariant feature transform algorithm represents a stable feature description method, such as the SIFT algorithm. Feature descriptor data refers to a data set that uses numerical vectors to represent local characteristics of feature points.

[0113] In some embodiments, the local neighborhood of each feature point is analyzed, and its gradient information at different scales and directions is calculated to generate feature descriptors with rotation and scale invariance. These descriptors can effectively characterize the local geometric characteristics of the feature points and provide a reliable basis for subsequent feature matching.

[0114] In some specific embodiments, in some embodiments, the SIFT algorithm first determines a 16×16 pixel feature description window centered on the feature point, and divides the window into 4×4 sub-regions. For each sub-region, the gradient amplitude and direction of the pixel point are calculated, and the gradient histograms of 8 directions are counted. Finally, the gradient histograms of all sub-regions are connected in series to form a 128-dimensional feature descriptor, and normalized to obtain the final feature descriptor data. This description method has rotation invariance and scale invariance, and can effectively describe the local features of the feature point.

[0115] For example: calculate the gradient amplitude and direction of pixels in the neighborhood of the feature point; count the gradient histograms in 36 directions to determine the main direction; rotate and align the description window with the main direction as a reference; divide the window into 4×4 sub-regions; calculate the 8-directional gradient histogram for each sub-region; concatenate 16 histograms to form a 128-dimensional vector; and normalize the vector.

[0116] S1073, performing feature point matching on the feature descriptor data based on a nearest neighbor search algorithm to generate initial matching point pair data;

[0117] Among them, the nearest neighbor search algorithm refers to the method of finding the most similar points in the feature space, such as: ANN algorithm, KD tree algorithm. Feature point matching refers to establishing the association relationship between corresponding feature points in different point clouds. The initial matching point pair data represents the initially established feature point correspondence relationship set.

[0118] In some embodiments, in the feature descriptor space, the most similar target feature point is searched for each feature point to establish an initial feature point correspondence relationship. The matching process considers the distance metric between feature descriptors to ensure that the matching point pairs have similar local features.

[0119] In some specific embodiments, FLANN is selected as the index structure, index parameters including the number of trees and branching factors are set, feature descriptor data are added to the index structure and the construction of the index tree is completed; then search parameters and accuracy requirements are set, a k-nearest neighbor search is performed for each query feature (usually k=2), the distance between feature vectors is calculated and the nearest neighbor result is returned.

[0120] S1074, optimizing the initial matching point pair data using a random sampling consistency algorithm to generate optimized matching point pair data;

[0121] Among them, the random sampling consensus algorithm represents a robust algorithm for eliminating false matches.

[0122] In some embodiments, by randomly selecting a subset of matching point pairs, estimating the transformation model, and evaluating the consistency of other point pairs with the model, the optimal transformation model and the corresponding set of matching point pairs are screened out through multiple iterations, effectively removing false matches.

[0123] In some specific embodiments, the optimization process of the RANSAC algorithm generally includes: first, randomly extracting the minimum number of sample point pairs from the initial matching point pair set (such as 8 pairs of points required for basic matrix estimation and 4 pairs of points required for homography matrix estimation) for model parameter estimation; then using the extracted sample points to calculate the geometric transformation model (such as basic matrix, homography matrix, etc.), and using the model to calculate the projection error of all matching point pairs, and determining the number of inliers supporting the current model as the consistency score according to the set error threshold; then repeating the above random sampling and model estimation process, continuously updating the optimal model and the maximum inlier set through multiple iterations until the preset number of iterations is reached or the termination condition is met; finally, outputting the transformation model with the highest consistency score and its corresponding inlier set as the optimized matching result. This is not limited here.

[0124] S1075, calculating coordinate transformation matrix data based on the optimized matching point pair data;

[0125] In some embodiments, the optimized matching point pairs are used to establish a linear equation system, and the rigid body transformation matrix is ​​solved by the least square method or other optimization methods. The matrix contains rotation and translation information and can accurately describe the spatial correspondence between the two point clouds.

[0126] In some specific embodiments, the covariance matrix is ​​constructed using the coordinate information of the matching point pairs, and the matrix is ​​subjected to singular value decomposition (SVD) to obtain three matrices U, S, and V. The optimal rotation matrix is ​​then calculated based on the decomposition results, and the rotation matrix is ​​used to solve the translation vector. Finally, the rotation matrix and the translation vector are combined into a complete transformation matrix, which is not limited here.

[0127] In some specific embodiments, a quaternion characteristic equation is constructed based on matching point pairs, the eigenvalues ​​and eigenvectors of the equation are solved to obtain a unit quaternion representing the rotation, the quaternion is converted into a rotation matrix form, and the translation component is optimized by the least squares method to finally obtain a complete transformation matrix, which is not limited here.

[0128] S1076, applying the coordinate transformation matrix data to the point cloud data to calculate distance error data between the transformed feature points;

[0129] The distance error data represents the spatial distance between the corresponding points after the transformation. The transformed feature point refers to the position of the point cloud feature point after the coordinate transformation.

[0130] In some embodiments, the calculated transformation matrix is ​​applied to the source point cloud, and the Euclidean distance between the transformed feature points and the corresponding feature points in the target point cloud is calculated. These distance values ​​reflect the accuracy of the current transformation result and provide an evaluation basis for iterative optimization.

[0131] S1077, determining whether the distance error data is less than a preset distance threshold;

[0132] If it is less than the preset distance threshold, then S1078, the coordinate transformation matrix data is determined as the optimal transformation matrix;

[0133] If it is not less than the preset distance threshold, execute step S1075.

[0134] In some embodiments, the calculated distance error is compared with a preset threshold. If the error is less than the threshold, it is considered that the current transformation matrix has met the requirements and is determined as the final result; otherwise, the transformation matrix is ​​returned to be recalculated and iteratively optimized until the accuracy requirements are met or the maximum number of iterations is reached.

[0135] It can be seen that by extracting corner points and contour feature points, the key features of point cloud data are obtained. The scale-invariant feature transformation algorithm is used for feature description to ensure the stability of feature matching. The use of the random sampling consistency algorithm effectively removes the wrong matching point pairs and improves the accuracy of subsequent transformation matrix calculation. Through the iterative optimization process, continuous adjustment is made until the preset threshold requirements are met, and finally a high-precision spatial transformation matrix is ​​obtained, achieving accurate alignment of point cloud data.

[0136] It can be seen that by acquiring multi-source heterogeneous data and unifying them into the coordinate system of the geographic base layer, a benchmark reference for data fusion is established. The geographic base layer serves as a benchmark and provides a stable spatial reference for subsequent data fusion. The surface smoothing algorithm is used to preprocess the real-life 3D model and the monomer model data, reducing the impact of data noise on fusion accuracy. The precise alignment of different types of 3D data is achieved through the combined use of feature point extraction and iterative nearest point algorithm. This feature-based iterative optimization method avoids the error accumulation problem caused by direct overall fusion, improves fusion accuracy, and ensures the geometric consistency of the fused scene.

[0137] In actual use, during the data collection process of the real scene model, due to factors such as occlusion and viewing angle limitation, data loss may occur due to blind spots. This data loss is manifested as holes on the surface of the model, affecting the integrity and visual effect of the model.

[0138] See also Figure 5 , Figure 5 This is another flow chart of the method for seamless and rapid fusion of multivariate three-dimensional data in an embodiment of the present application;

[0139] Therefore, in some embodiments, after step S107, the method further includes:

[0140] S201, identifying missing regions of the real-scene 3D model data, and generating feature entropy data of the missing regions;

[0141] Among them, the missing area indicates that there are data holes or incomplete areas on the model surface, which are usually caused by acquisition blind spots or occlusions. Feature entropy data refers to a numerical indicator used to quantify the complexity of the missing area. The larger the value, the richer the regional features.

[0142] In some embodiments, the model is first subjected to grid analysis to detect the connectivity and boundary features of the grid and identify incomplete or empty areas. The geometric feature distribution and local complexity of these areas are then calculated to generate entropy data reflecting the richness of regional features, providing a reference for subsequent feature propagation.

[0143] It should be noted that the feature entropy data is actually an indicator that describes the complexity of the missing region boundary. It does not directly calculate the features of the missing region (because it is indeed empty), but is determined based on analyzing the geometric features of the edge of the missing region, such as the curvature change of the boundary curve, the normal vector distribution of the boundary points, etc.

[0144] S202, extracting geometric features of the complete area of ​​the real-scene three-dimensional model data to generate characteristic flow field data, wherein the characteristic flow field data includes characteristic propagation direction data and characteristic propagation intensity data;

[0145] Among them, the complete area refers to the part of the model with complete data and good quality. Geometric features refer to the numerical description of the local morphology of the model, including curvature, normal vector, etc. Feature flow field data represents the propagation trend of features in space and is used to guide feature completion. Feature propagation direction data refers to the spatial direction of feature diffusion. Feature propagation intensity data is used to indicate the magnitude of feature influence.

[0146] In some embodiments, a grid analysis is first performed on the complete area to extract local geometric features including curvature, normal vectors, boundary features, etc.; then a characteristic flow field is constructed based on these features, and the characteristic propagation direction and intensity of each position are calculated; finally, the flow field data is stored in the form of a vector field for subsequent feature propagation processes.

[0147] S203, propagating the geometric features of the complete area to the missing area according to the preset propagation rules; the preset propagation rules include: the feature propagation intensity data decays nonlinearly with the increase of the propagation distance; the feature propagation intensity data increases with the increase of the feature entropy data; the propagation process of the geometric features is constrained by the feature propagation direction data.

[0148] Propagation intensity attenuation: The characteristic propagation intensity will decay nonlinearly with the increase of propagation distance, avoiding characteristic distortion caused by long-distance propagation

[0149] Entropy-guided propagation: The propagation intensity will be dynamically adjusted according to the previously calculated feature entropy data. When the feature entropy value is large, it indicates that the area is highly complex and requires stronger feature propagation to ensure completion quality.

[0150] Direction constraint: The feature propagation process is constrained by the direction of the feature flow field, ensuring that the feature propagates in a reasonable direction and maintains the continuity of the geometric structure.

[0151] It can be seen that the geometric complexity of the missing area is quantified by the feature entropy data, which provides an accurate reference for the subsequent feature propagation. Combined with the propagation direction and intensity information in the feature flow field data, an adaptive feature propagation mechanism is established. The feature propagation intensity will decay nonlinearly with the increase of propagation distance. This attenuation characteristic is consistent with the continuity characteristics of the actual scene. At the same time, the propagation intensity will increase with the increase of feature entropy, ensuring that more propagation supplementary information can be obtained in areas with complex geometric features, and finally achieving the natural restoration of the missing area and maintaining the visual continuity of the scene.

[0152] In some embodiments, after step S107, the method further includes: S204, setting an initial granularity parameter, wherein the initial granularity parameter is a preset maximum grid side length;

[0153] The initial granularity parameter represents the initial accuracy control value of the model meshing. The mesh edge length refers to the boundary length of the mesh unit on the surface of the three-dimensional model. The preset maximum mesh edge length represents the maximum mesh size threshold allowed, which is used to control the initial accuracy of feature propagation.

[0154] In some embodiments, the appropriate maximum grid edge length is first determined based on the overall size of the model and the feature complexity as an initial granularity parameter; this parameter determines the grid accuracy of the first round of feature propagation, and a larger initial granularity helps to quickly build a rough feature framework.

[0155] S205, executing steps S201 to S203 based on the initial granularity parameters to obtain propagated real-scene three-dimensional model data;

[0156] S206, reducing the initial granularity parameter by a preset step size; treating the propagated real-scene 3D model data as the real-scene 3D model data and executing steps S201 to S203 to obtain the propagated real-scene 3D model data until the initial granularity parameter reaches a preset minimum grid side length.

[0157] In some embodiments, the granularity parameter is first decreased according to a preset step size, and then the model obtained in the previous round is used as input to repeat the feature propagation process at the new granularity; multi-scale feature completion from coarse to fine is achieved by gradually refining the granularity until the preset minimum grid edge length is reached.

[0158] It can be seen that a multi-scale feature propagation mechanism from coarse to fine is established. By setting the initial maximum grid edge length, feature propagation is completed at a larger scale, ensuring the correctness of the overall morphology. As the grid edge length gradually decreases, feature propagation gradually focuses on more detailed geometric features. This progressive processing method not only improves computational efficiency, but also maintains the continuity of geometric features at different scales, ultimately achieving high-quality missing area repair effects.

[0159] The following introduces an exemplary multivariate three-dimensional data seamless rapid fusion system 600 provided in an embodiment of the present application. Figure 6 It is a schematic diagram of an exemplary hardware structure of a multivariate three-dimensional data seamless rapid fusion system 600 provided in an embodiment of the present application.

[0160] In some embodiments, the multivariate three-dimensional data seamless rapid fusion system 600 is a computer device or the multivariate three-dimensional data seamless rapid fusion system 600 includes a computer device. The computer device includes a processor, a memory and a network interface connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with other external terminals or servers through a network connection. In some embodiments, the network interface can be a wired network interface, and in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, the method in the embodiment of the present application is implemented.

[0161] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0162] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0163] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0164] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk), etc.

[0165] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A method for seamless and rapid fusion of multivariate three-dimensional data, characterized in that: include: Acquire real-scene 3D model data, monomerized model data, digital elevation model data and digital orthophoto data, wherein the real-scene 3D model data is in osgb format and the monomerized model data is in obj format; Performing registration processing on the digital elevation model data and the digital orthophoto data to generate a geographic base layer of the three-dimensional scene; Converting the real-scene 3D model data and the individualized model data to the coordinate system of the geographic base layer, so that the real-scene 3D model data and the individualized model data are aligned with the spatial position of the geographic base layer; Using a surface smoothing algorithm to process the real-scene three-dimensional model data and the monomerized model data in the three-dimensional scene; Receive BIM data and CAD data; Converting the BIM data and the CAD data to the coordinate system of the geographic base layer; The feature points in the real-scene three-dimensional model data, the individualized model data, the BIM data and the CAD data are extracted as point cloud data, and the optimal transformation matrix between the point cloud data is calculated using an iterative closest point algorithm. The three-dimensional data is spatially transformed according to the optimal transformation matrix until the distance error between the feature points is less than a preset distance threshold, thereby generating fused three-dimensional scene data.

2. The method according to claim 1, characterized in that After the step of generating fused three-dimensional scene data, the method further includes: Identifying missing regions of the real-scene three-dimensional model data and generating characteristic entropy data of the missing regions; Extracting geometric features of a complete area of ​​the real-scene three-dimensional model data to generate characteristic flow field data, wherein the characteristic flow field data includes characteristic propagation direction data and characteristic propagation intensity data; Propagate the geometric features of the complete area to the missing area according to a preset propagation rule; The preset propagation rule includes: the characteristic propagation intensity data decays nonlinearly with the increase of propagation distance; The characteristic propagation intensity data increases as the characteristic entropy data increases; The propagation process of the geometric feature is constrained by the feature propagation direction data.

3. The method according to claim 2, characterized in that Before the step of identifying missing regions of the real-scene three-dimensional model data and generating characteristic entropy data of the missing regions, the method further includes: Setting an initial granularity parameter, wherein the initial granularity parameter is a preset maximum grid side length; Based on the initial granularity parameter, the step of identifying missing regions of the real-scene 3D model data and generating characteristic entropy data of the missing regions is performed to obtain propagated real-scene 3D model data; The initial granularity parameter is reduced by a preset step size; the real-scene 3D model data after propagation is regarded as the real-scene 3D model data to perform the steps of identifying missing areas of the real-scene 3D model data and generating characteristic entropy data of the missing areas to obtain the real-scene 3D model data after propagation, until the initial granularity parameter reaches a preset minimum grid side length.

4. The method according to claim 1, characterized in that The step of using a surface smoothing algorithm to process the real-scene 3D model data and the monomerized model data in the 3D scene specifically includes: Establishing a neighbor index for each grid point in the surface grid of the real-scene three-dimensional model data and the monomerized model data, and generating neighborhood index data of the grid point; Determine the neighboring points of the grid point according to the neighborhood index data; Calculating the local curvature of the grid point according to the grid point and the corresponding adjacent points; Determine the grid points whose local curvature is greater than a preset curvature threshold as elevation mutation points, and determine the grid points whose local curvature is not greater than the preset curvature threshold as non-elevation mutation points; Calculating the regional average elevation data of the grid point and the adjacent points to obtain initial smoothed elevation data; Determining weight coefficient data of the elevation mutation point according to the local curvature, wherein the weight coefficient data approaches 1 as the local curvature increases; For the non-elevation mutation point, the initial smoothed elevation data is used for smoothing; For the elevation mutation point, the weight coefficient data is used for smoothing.

5. The method according to claim 4, characterized in that After the step of smoothing the elevation mutation point using the weight coefficient data, the method further comprises: Extracting elevation reference values ​​of positions corresponding to smoothed grid points from the geographic base layer to obtain reference elevation data; Comparing the elevation data of the smoothed grid points with the reference elevation data to generate elevation deviation data; Setting an elevation constraint weight according to the elevation deviation data to generate constraint weight data, wherein the constraint weight data decreases as the elevation deviation data increases; The elevation data of the grid points are corrected based on the constraint weight data to generate corrected grid data.

6. The method according to claim 1, characterized in that The step of using the iterative closest point algorithm to calculate the optimal transformation matrix between point cloud data specifically includes: Extracting corner point features and contour feature points from the point cloud data to generate feature point set data; Using a scale-invariant feature transformation algorithm to perform feature description on the feature point set data to generate feature description sub-data; Performing feature point matching on the feature descriptor data based on a nearest neighbor search algorithm to generate initial matching point pair data; The initial matching point pair data is optimized using a random sampling consensus algorithm to generate optimized matching point pair data; Calculate coordinate transformation matrix data based on the optimized matching point pair data; Applying the coordinate transformation matrix data to the point cloud data to calculate distance error data between the transformed feature points; Determining whether the distance error data is less than the preset distance threshold; If it is less than the preset distance threshold, determining the coordinate transformation matrix data as the optimal transformation matrix; If it is not less than the preset distance threshold, return to the step of calculating the coordinate transformation matrix data based on the optimized matching point pair data.

7. The method according to claim 1, characterized in that The step of converting the real-scene 3D model data and the individualized model data into the coordinate system of the geographic base layer specifically includes: Acquire the digital elevation model data and the digital orthophoto data; Resampling the digital elevation model data to obtain resampled elevation data, so that the resampled elevation data is consistent with the spatial resolution data of the digital orthophoto data; Using a control point matching algorithm to perform geo-registration processing on the resampled elevation data and the digital orthophoto data to generate registered geographic data; constructing triangular mesh model data based on the resampled elevation data in the registered geographic data; The digital orthophoto data in the registered geographic data is mapped to the triangular mesh model data as texture data.

8. A multi-dimensional three-dimensional data seamless rapid fusion system, characterized in that: The multivariate three-dimensional data seamless and rapid fusion system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the multivariate three-dimensional data seamless and rapid fusion system to execute the method described in any one of claims 1-7.

9. A computer program product comprising instructions, characterized in that When the computer program product runs on a system for seamless and rapid fusion of multivariate three-dimensional data, the system for seamless and rapid fusion of multivariate three-dimensional data executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on a system for seamless and rapid fusion of multivariate three-dimensional data, the system for seamless and rapid fusion of multivariate three-dimensional data executes the method as described in any one of claims 1 to 7.

Images