CN117934743A - A method and device for intelligently deriving multi-level structures of three-dimensional semantic models - Google Patents

Abstract

The invention provides a multi-level structure intelligent deriving method and device of a three-dimensional semantic model, wherein the method comprises the following steps: the method comprises the steps of firstly, obtaining original three-dimensional model data of a plurality of three-dimensional building models to determine semantic expressions corresponding to all the CityGML building components; extracting geometric information corresponding to each IFC building component in the IFC model based on the original three-dimensional model data to obtain geometric expression corresponding to each CityGML building component; thirdly, establishing a mapping body between semantic expressions and geometric expressions corresponding to each level model; and fourthly, intelligently reconstructing the three-dimensional building model based on the three-dimensional data structure diagram corresponding to each level model, thereby realizing the derivation of different level models. The invention realizes unified position relation and classified expression of building components on the space data, can facilitate management and classified storage of the space data, can carry out classified loading of ground objects under different scales according to different requirements during visualization, and improves the visualization and classification efficiency of mass data.

Description

A method and device for intelligently deriving multi-level structures of three-dimensional semantic models

Technical Field

The present invention relates to the technical field of three-dimensional modeling, and in particular to a method and device for intelligently deriving a multi-level structure of a three-dimensional semantic model.

Background technique

The current semantic expression ability of building 3D modeling is poor, resulting in low utilization of 3D geographic information system (GIS), which is difficult to meet the deep-level applications such as building component information query, energy consumption analysis and refined management. In addition, due to the lack of unified modeling standards, many 3D data formats are incompatible with each other, have poor reusability, and cannot reflect the mathematical relationship between spatial positions, resulting in difficulties in interoperability and data sharing of building 3D models, and lack of relationship between semantics and spatial positioning. Therefore, how to enhance the new mapping relationship of the semantic features of 3D models, reduce the cost of 3D model production and maintenance, and facilitate the retrieval, sharing and interoperability of its data has become a difficult problem that needs to be solved in the current smart city construction.

City Geography Markup Language (CityGML) is an international open standard for the storage and exchange of virtual 3D city models launched by the Open Geospatial Consortium (OGC). It is also a universal semantic information model in the field of 3D GIS. The CityGML model emphasizes the consistency of geometric, topological and semantic expressions. It makes up for the shortcomings of traditional 3D models in data sharing and interoperability, and enhances the reusability of 3D models. Therefore, it is widely used in many fields such as urban planning, building illumination estimation, energy demand analysis, shadow analysis, noise propagation estimation, 3D cadastral and facility management. In terms of 3D expression of buildings, CityGML not only defines the semantic information of various components such as roofs, walls, floors, doors, windows, rooms, etc., but also uses 5 levels of LoD for multi-scale expression from simple to complex. The emergence of CityGML (City Geography Markup Language) has brought opportunities for the widespread application and sharing of 3D geographic information. However, the current 5-level LoD is only classified from the logical structure, without reflecting the spatial position relationship between the logical layers, and is directly oriented to 3D reconstruction.

In the prior art, the positioning and positional relationship of building components are determined by a preset mathematical formula, which requires calculation through the coordinates of multiple fixed points, and the calculation is only performed for one building, so the efficiency is not high.

In fact, the coordinate ranges of floors, roofs, walls, structural parts such as columns and beams, doors, and windows in building components all have their own unique characteristics, and there are also unique range intersections between them. Therefore, using this feature for intelligent reconstruction becomes an optional solution.

Summary of the invention

In view of the above problems, the present invention proposes a method and device for deriving a multi-level structure of a three-dimensional semantic model that overcomes the above problems or at least partially solves the above problems. This patent aims at the problems that some existing three-dimensional modeling methods emphasize the construction of a single-scale geometric model, lack the characteristics of spatial position relationships and semantic information expression between multiple levels of detail (LoDs), and have weak interoperability. Based on the international open standard CityGML, this patent provides a method for deriving a multi-level structure of a three-dimensional semantic model.

According to one aspect of the present invention, a method for intelligently deriving a multi-level structure of a three-dimensional semantic model is provided, the method comprising:

Acquire original 3D model data of a plurality of 3D building models (i.e., original 3D model data of IFC models), establish a semantic mapping rule between the IFC model and the CityGML LoD4 model for each of the 3D building models, and determine a correspondence between the IFC building components and the CityGML building components based on the semantic mapping rule and the original 3D model data, so as to determine a semantic expression corresponding to each CityGML building component;

Extracting geometric information corresponding to each IFC building component in the IFC model based on the original three-dimensional model data, performing geometric transformation on the geometric information corresponding to each IFC building component to obtain a corresponding three-dimensional data structure diagram composed of spatial data points to represent the CityGML building component, and obtaining a geometric expression corresponding to each CityGML building component;

The geometric expression and semantic expression of different levels of models of the three-dimensional building model are defined by the OGC CityGML standard, a mapping is established between the semantic expression corresponding to the building component and the corresponding data point in the three-dimensional data structure diagram, and a mapping body between the semantic expression and the geometric expression corresponding to each level of model is established.

It is easy to understand that the three-dimensional data structure diagram establishes not a traditional two-dimensional mapping table, but a three-dimensional mapping body, which reflects the relationship between the semantics and spatial positions of each building component, enriches the expression of CityGML, and locates the position of building components more clearly and conveniently.

The three-dimensional building model is intelligently reconstructed based on the three-dimensional data structure diagram corresponding to each level model, thereby realizing the derivation of different level models.

Optionally, the method of geometric transformation includes:

S1 establishes a three-dimensional rectangular coordinate system XYZ, defines a coordinate origin, and in a plurality of original three-dimensional model data, a preset point coincides with the origin, a preset direction is parallel to an X-axis or a Y-axis in the three-dimensional rectangular coordinate system, and marks the building components corresponding to each level model;

S2 marks the intersections between different building components, obtains the coordinates of any point on the intersection, obtains the coordinate value range on the X-axis, Y-axis, and Z-axis, and defines a positioning point of any intersection so that the positioning point is closest or second closest to the origin;

S3 uses the normalized span of the coordinate value range as a grayscale value or an RGB color value, thereby obtaining a positioning point and a grayscale value or an RGB color value of any intersection line; the positioning point and the grayscale value or the RGB color value form a data point, and the data point is positioned in the coordinate system according to the position of the positioning point in the coordinate system, forming a three-dimensional data structure diagram, and the data point is represented by a symbol with the grayscale value or the RGB color value.

The so-called span refers to the length of the coordinate value range. For example, if the x-coordinate range is [1,3], the span is 2.

Optionally, the preset point is a vertex of the bottom surface of the building, and the preset direction is selected so that the plane where one of the facades is located is parallel to the X-axis or the Y-axis.

Optionally, the data points that have been located use different symbols to represent intersection lines of different orientations.

The method for intelligently reconstructing the three-dimensional building model based on the three-dimensional data structure graph corresponding to each level model specifically includes:

S4: For any data point in the three-dimensional data structure diagram, starting from the positioning point, move in the positive direction of the X-axis to form a first track, and then starting from the positioning point, move in the negative direction of the X-axis to form a second track, so that the first union length of the first and second tracks coincides with the span in the X-axis direction (that is, the span of the intersection line corresponding to the positioning point in the X-axis direction, that is, the projection length of the intersection line on the X-axis, or the range of the coordinate values projected on the X-axis); starting from the positioning point, move in the positive direction of the Y-axis to form a third track, and then starting from the positioning point, move in the negative direction of the Y-axis to form a fourth track, so that the second union length of the third and fourth tracks coincides with the span in the Y-axis direction; starting from the positioning point, move in the positive direction of the Z-axis to form a fifth track, and then starting from the positioning point, move in the negative direction of the Z-axis to form a sixth track, so that the third union length of the fifth and sixth tracks coincides with the span in the Z-axis direction, thereby obtaining a total union of the first union to the third union;

S5 establishes an oblique line generation model, the oblique line generation model includes forming an oblique line by automatically connecting the corresponding calculation method by clicking or identifying the plane orthogonal line segment or the three-dimensional orthogonal line segment when the total union displayed in the three-dimensional data structure diagram is a plane orthogonal line segment or a three-dimensional orthogonal line segment, the calculation method is that for the plane orthogonal line segment, a rectangle is made in two positive directions and two negative directions of the orthogonal line, and the two diagonals of the two faces where the positioning point is located are connected to generate an oblique line; for the three-dimensional orthogonal line segment, a parallel hexagonal right prism is made in three positive directions and three negative directions of the orthogonal line, and the two body diagonals where the positioning point is located are connected to generate an oblique line;

The identification of plane orthogonal line segments or stereo orthogonal line segments can be performed by performing a difference operation on the lines in the original three-dimensional model data before reconstruction and the total union after reconstruction. If the difference result is zero, it means that the total union after reconstruction is not an orthogonal line segment, otherwise it is an orthogonal line segment.

It can be understood that if it is not a diagonal line, the total union after reconstruction coincides with the intersection line of the reconstructed signature, so the difference is zero, otherwise the difference result is a subset of the total union, that is, there are residual lines after the difference, which are displayed as grayscale values or color values that are not zero on the difference image. In this way, the intelligent recognition and connection calculation method of the total union of orthogonal line segments can be realized to reconstruct diagonal lines.

The terrain tangent curves and building shell curves belong to the overall appearance of the building. These lines are the interfaces between building components and external non-building components (the earth and the air), not the intersection lines between building components. Therefore, they are not considered in the scope of trajectory and oblique line generation.

S6 establishes a building component classification and recognition model, obtains the classification of building components through analysis, and saves it to the corresponding data points.

Optionally, the recognition model includes completing classification by using an SVM algorithm on the plane where the total union is located, and/or completing classification by performing a three-dimensional clustering analysis based on the projection coordinates and grayscale values of the positioning points on the XY plane.

Preferably, the SVM algorithm includes using a Gaussian kernel function to map data points on the plane where the total union is located into a three-dimensional Hilbert space, and using the first hyperplane for classification.

Optionally, a second hyperplane is used to further perform linear SVM classification in the same classification obtained after classification by the first hyperplane.

Therefore, the geometric and spatial position relationship of different building components is given through the three-dimensional data structure diagram as the three-dimensional relationship of the geometry and semantic mapping of all building components.

Preferably, obtaining the coordinates of any point on the intersection line is replaced by obtaining the coordinates of points at a preset interval on the intersection line. It is easy to understand that this can reduce the amount of calculation of the closest distance.

Preferably, the points of the preset distance are the endpoints and the equally divided points on the intersection line.

Optionally, for a building complex within a preselected geographic area, the geometric transformation method is as follows: select a reference building, construct a three-dimensional rectangular coordinate system XYZ in the reference building, and define a coordinate origin, in the original three-dimensional model data corresponding to the reference building, a preset point coincides with the origin, a preset direction is parallel to the X-axis or Y-axis in the three-dimensional rectangular coordinate system, and the building components corresponding to each level model of the reference building and all other buildings within the preselected geographic area are marked; then steps S2 and S3 are executed; the method for intelligently reconstructing the three-dimensional building model based on the three-dimensional data structure diagram corresponding to each level model also includes steps S4-S5, and then S6' is performed, including establishing building components A group classification and recognition model is provided, which firstly makes the preset points corresponding to the reference buildings in all non-reference buildings within the pre-selected geographical area coincide with the origin, and the corresponding preset direction is parallel to the X-axis or Y-axis in the three-dimensional rectangular coordinate system, thereby translating the three-dimensional data structure diagrams corresponding to the hierarchical models of all non-reference buildings into the three-dimensional rectangular coordinate system of the reference building, and establishing a building component group classification and recognition model, and obtaining the classification of the building component group through analysis, and saving it into the data points of the three-dimensional data structure diagrams corresponding to the hierarchical models of the corresponding non-reference buildings, and finally reversely translating the three-dimensional data structure diagrams corresponding to the hierarchical models of the non-reference buildings to the positions before the translation in the three-dimensional rectangular coordinate system.

The building component group classification and recognition model includes classifying each building using an SVM algorithm on the plane corresponding to the total union, and/or performing three-dimensional clustering analysis on the projection coordinates of all positioning points of each building on the XY plane and the corresponding grayscale values to complete the classification.

Preferably, the SVM algorithm includes constructing a Gaussian kernel function, mapping the data points on the plane where the total union is located into a three-dimensional Hilbert space, and using the first hyperplane for classification.

Optionally, a second hyperplane is used to further perform linear SVM classification in the same classification obtained after classification by the first hyperplane.

Thus, through a unified coordinate system, the intersection lines of the buildings in the selected geographical area are collectively obtained, the three-dimensional data structure diagram is produced, and the overall reconstruction of the building complex is carried out. Since the building types in a selected area are similar, the spatial distribution of each building component is similar, so the collective construction classification and recognition can be obtained through SVM or cluster analysis, and then the recognition is completed through reverse translation and relocation, which improves the efficiency of classification and recognition as a whole.

Optionally, defining the geometric and semantic expressions of different hierarchical models of the three-dimensional building model by the OGC CityGML standard, establishing a mapping between the semantic expressions corresponding to the building components and the corresponding data points in the three-dimensional data structure diagram, and forming a mapping body between the semantic expressions and geometric expressions corresponding to the hierarchical models includes:

Defining the three-dimensional data structure diagram and semantic expression of different level models of the three-dimensional building model through the OGC CityGML standard, and determining multiple three-dimensional data structure diagram semantic topics (titles) corresponding to the three-dimensional building model;

Establish associations between the semantic expressions of the different hierarchical models and the semantic themes of each of the three-dimensional data structure graphs to form a mapping body.

Optionally, the semantic topic includes a building component classification name in a data point corresponding to a positioning point selected in the corresponding three-dimensional data structure diagram.

Optionally, the method further comprises:

Associating LoD tags according to the classification of each of the CityGML building components stored in the mapping body, wherein the LoD tags are used to represent different hierarchical models;

The same classification of CityGML building components corresponds to at least one geometric semantic theme, and further corresponds to at least one LoD tag. The same CityGML building component has the same or different display modes in the spatial rectangular coordinate system XYZ under different LoD tags. The derivation of different hierarchical models specifically includes: determining the geometric semantic theme and multiple CityGML building components corresponding to any hierarchical model based on the mapping body corresponding to the hierarchical model;

Constructing a hierarchical display model corresponding to the hierarchical model according to the CityGML building components and their display modes;

The three-dimensional building model is reconstructed by integrating the hierarchical models corresponding to the hierarchical display model, thereby achieving the derivation of models at different hierarchical levels.

Optionally, OGC CityGML defines five hierarchical models including LoD0, LoD1, LoD2, LoD3, and LoD4 to achieve multi-scale expression of building appearances, building components, and ancillary facilities;

Among them, LoD0 expresses the bottom or roof outline plane of the building, which is a 2.5D polygon; LoD1 simply represents the three-dimensional model of the building with blocks; LoD2 adds a description of the auxiliary structure and roof of the house on the basis of LoD1; LoD3 describes the detailed appearance structure of the building, including but not limited to doors, windows, and balconies; LoD4 adds the expression of objects such as stairs, rooms, and furniture inside the building, with detailed geometric and semantic information.

According to another aspect of the present invention, a multi-level structure derivation device for a three-dimensional semantic model is provided to implement the multi-level structure intelligent derivation method for the three-dimensional semantic model, the device comprising:

A data parsing module is used to obtain original three-dimensional model data of the three-dimensional building model, and construct geometric expression and semantic expression of the original three-dimensional model data;

A mapping body establishment module is used to define the geometric expression and semantic expression of different level models of the three-dimensional building model through the OGC CityGML standard, and establish a mapping body between the semantic expression and the geometric expression corresponding to each level model;

The model intelligent reconstruction module is used to intelligently reconstruct the three-dimensional building model based on the mapping bodies corresponding to the models at each level, so as to realize the derivation of models at different levels.

Optionally, the data parsing module is further used to: establish semantic mapping rules between the IFC model and the CityGML LoD4 model, and determine the correspondence between the IFC building components and the CityGML building components based on the semantic mapping rules and the original three-dimensional model data to determine the semantic expression corresponding to each CityGML building component; and extract the geometric information corresponding to each IFC building component in the IFC model based on the original three-dimensional model data, perform geometric transformation on the geometric information corresponding to each IFC building component to obtain a corresponding three-dimensional data structure diagram composed of spatial data points to represent the CityGML building component, and obtain the geometric expression corresponding to each CityGML building component.

The mapping establishment module can also be used to: define the three-dimensional data structure diagram and semantic expression of different level models of the three-dimensional building model through the OGC CityGML standard, and determine multiple three-dimensional data structure diagram semantic themes (titles) corresponding to the three-dimensional building model;

Establishing associations between the semantic expressions of the different hierarchical models and the semantic themes of the three-dimensional data structure graphs to form a mapping body;

The semantic topic includes the building component classification name in the data point corresponding to a positioning point selected in the corresponding three-dimensional data structure diagram;

Associating LoD tags according to the classification of each of the CityGML building components stored in the mapping body, wherein the LoD tags are used to represent different hierarchical models;

The same classification of CityGML building components corresponds to at least one geometric semantic theme and further corresponds to at least one LoD tag. The same CityGML building component may be displayed in the same or different manners in the spatial rectangular coordinate system XYZ under different LoD tags.

The model intelligent reconstruction module is also used to: determine the geometric semantic theme and multiple CityGML building components corresponding to any hierarchical model based on the mapping body corresponding to the hierarchical model;

Constructing a hierarchical display model corresponding to the hierarchical model according to the CityGML building components and their display modes;

The three-dimensional building model is reconstructed by integrating the hierarchical models corresponding to the hierarchical display model, thereby achieving the derivation of models at different hierarchical levels.

According to a third aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the multi-level structure intelligent derivation method of the three-dimensional semantic model.

According to a fourth aspect of the present invention, there is further provided a computing device, the computing device comprising a processor and a memory:

The memory is used to store program code and transmit the program code to the processor;

The processor is used to execute the multi-level structure intelligent derivation method of the three-dimensional semantic model according to the instructions in the program code.

The present invention provides a multi-level structure intelligent derivation method for a three-dimensional semantic model, which performs geometric and semantic analysis on original three-dimensional model data of a three-dimensional building model, defines geometric and semantic expressions of models at different levels through the OGC CityGML standard, establishes geometric and semantic mapping bodies at each level, and then reflects the spatial position relationship of different building components based on the geometric and semantic mapping relationship, thereby intelligently reconstructing the model through a trajectory forming method and an established building component recognition model to achieve derivation of models at different levels, that is, the derivation of the present invention is different from traditional surface generation, and linearly generates a three-dimensional model by taking the intersection line of the building component as the starting point, and integrates the building model, building component classification, and the three-dimensional spatial position of the building component in the form of a three-dimensional mapping body in semantic association, thereby enriching the information of the model.

In the solution of the present invention, each CityGML building component is associated with a LoD tag, that is, the same object can have model expressions corresponding to different LoDs, which are all associated with the same entity, and the corresponding display mode is selected at different detail levels to achieve significant expression of the entity at the corresponding LoD.

The present invention realizes unified expression of spatial data based on semantic information extraction, geometric mapping and semantic mapping based on geometric features and rules, which can facilitate the management and hierarchical storage of spatial data. During visualization, hierarchical loading of objects at different scales can be performed according to different needs, thereby improving the visualization efficiency of massive data.

For building complexes within a geographical range, classification and identification are performed by superimposing them on a unified reference building coordinate system, which improves the efficiency of classification processing.

The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below.

Based on the following detailed description of specific embodiments of the present invention in conjunction with the accompanying drawings, those skilled in the art will become more aware of the above and other objects, advantages and features of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

FIG1 is a four-step flow chart of a multi-level structure intelligent derivation method of a three-dimensional semantic model according to Embodiment 1 of the present invention;

FIG2 is a partial IFC model of two buildings in a residential area according to Embodiment 1 of the present invention, wherein (a) is a reference building without a slope, and (b) is a non-reference building with a slope, wherein the top floor is shown as an example to describe the geometric transformation;

FIG3 is a schematic diagram of intelligent reconstruction of horizontal intersection lines, taking the roof surface corresponding to the H’ and M positioning points in FIG2b as an example.

FIG4 is a schematic diagram of the SVM wall and window classification algorithm process for the facade with a single window in FIG2a.

Figure 5 is a schematic diagram of the SVM wall, window 1 and window 2 classification algorithm process for a facade with two windows.

FIG6 is a schematic diagram of the cluster analysis results of the horizontal and vertical intersection lines of the top and wall, and the window intersection line in FIG2b.

FIG. 7 is a diagram showing the association relationship between the semantic expressions of different hierarchical models in the mapping body and the semantic themes of the three-dimensional data structure diagram.

FIG8 is a schematic diagram of a multi-level structure derivation device configuration of a three-dimensional semantic model and data interaction between modules in Embodiment 2 of the present invention.

Detailed ways

The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

Example 1

An embodiment of the present invention provides a multi-level structure derivation method for a three-dimensional semantic model, as shown in Figure 1. The multi-level structure intelligent derivation method for a three-dimensional semantic model of this embodiment includes: four steps as shown in Figure 1, namely, the first step, the semantic expression corresponding to each CityGML building component; the second step, the geometric expression corresponding to each CityGML building component; the third step, forming a mapping body between the semantic expression and the geometric expression corresponding to each level model; the fourth step, intelligently reconstructing the three-dimensional building model to realize the derivation of models of different levels.

Among them, the first step, as shown in FIG2, takes two buildings within a selected community as an example, obtains the original three-dimensional model data corresponding to the two three-dimensional building models, and for each of them, establishes a semantic mapping rule between the two corresponding IFC models and the CityGML LoD4 model, and determines the correspondence between the IFC building components and the CityGML building components based on the semantic mapping rule and the original three-dimensional model data, so as to determine the semantic expression corresponding to each CityGML building component;

The second step is to extract geometric information corresponding to each IFC building component in the IFC model (in this embodiment, the outlines of walls, windows, roofs, etc.) based on the original three-dimensional model data, perform geometric transformation on the geometric information corresponding to each IFC building component to obtain a corresponding three-dimensional data structure diagram composed of spatial data points to represent the CityGML building component, and obtain the geometric expression corresponding to each CityGML building component;

The geometric transformation includes, as shown in Figure 2a and Figure 2b, establishing an XYZ three-dimensional coordinate system for each of the two buildings, coinciding the origin O with a vertex of the bottom surface, and making the plane containing the unexpected window in Figure 2a and Figure 2b parallel to the X-axis.

Mark the intersections between walls, windows, and roofs of different building components, and obtain the coordinates of the endpoints and trisection points on the intersections. As shown in Figure 2a and Figure 2b, points A-I respectively show the endpoints (A, D, G, J) and trisection points (B, C, E, F, H, I, K, L, A', F', H', M) on the intersections between three walls and the roof, as well as the walls. K and L are the positioning points on the intersection of the window and the wall where the window is installed (hereinafter, K' and L' are the intersections between the top surface of the floor slab and the wall).

Among them, Figure 2a uses the point closest to the origin among these points as the positioning point. It can be seen that the use of the closest distance solution results in the overlap of positioning points on A and G (the positioning symbols overlap), which are the positioning points of the intersection line p, and the intersection line l and the vertical intersection line DG below. In addition, there is also an overlap of positioning points at the intersection of the intersection line m and the vertical intersection line orthogonal to the intersection line p, which is caused by the overlap of the positioning points of the intersection line m and the vertical intersection line of the cover. Among them, the intersection line l and the intersection line m are the intersection lines of the top floor slab top surface and the wall (the following intersection lines m' and intersection line l' are similarly the intersection lines of the top floor slab top surface and the wall).

Therefore, the closest distance can only partially describe the spatial position relationship of the intersection line and the resolution is not high.

However, when the second closest distance scheme is selected, as shown in Figure 2b, taking the top floor of a building with a slope as an example, the trisection points A', F', H', K', L', the trisection point M on the slope line where the slope and the top intersect, the trisection points of the intersection lines n', p', m', and l', the trisection points of another slope line just below M, and the trisection points of the two perpendicular intersection lines between the intersection lines n' and m' and the trisection points of the two perpendicular intersection lines between the two slope lines form the visible positioning points in the perspective of Figure 2b. The spatial position relationship of each intersection line is clearly shown.

Then, the span of the coordinate value range is normalized as the grayscale value (RGB color value can also be used to represent it), thereby obtaining the positioning point and grayscale value of any intersection line; the positioning point and grayscale value form a data point, and the data point is positioned in the coordinate system by the position of the positioning point in the coordinate system, forming a three-dimensional data structure diagram, and the data point is represented by the symbol √ and ○ with the grayscale value to mark the horizontal intersection line and the vertical intersection line respectively, and the positioning point marked with the symbol √ and ○ in Figure 2a and Figure 2b can be used as the position of the positioning data point. Therefore, these positioning points in the figure form a point group reflecting the spatial structure of the three-dimensional data structure diagram.

Therefore, the mapping body between the semantic expression and geometric expression corresponding to each level model formed in the third step also presents the spatial structure of the three-dimensional data structure diagram, and at this time, the semantic expression corresponding to each level model is associated with the data point.

In the fourth step, the three-dimensional data structure diagram corresponding to each level model is intelligently reconstructed. The following steps are included:

S4 is shown in FIG3 . For the data points of the three-dimensional data structure diagram corresponding to the positioning point in FIG2b , taking the reconstruction of the intersection line with the positioning point H' as an example, starting from the positioning point H', moving in the positive direction of the X-axis to form a first track, and then starting from the positioning point H', moving in the negative direction of the X-axis to form a second track, so that the length of the first union of the first and second tracks, that is, the length of the intersection line with the positioning point H', coincides with the span in the X-axis direction; starting from the positioning point H', moving in the positive direction of the Y-axis to form a third track, moving in the negative direction of the Y-axis to form a fourth track, and moving in the positive direction of the Z-axis to form a fifth track, and moving in the negative direction of the Z-axis to form a sixth track, are all empty, respectively represented by null2 and null3, representing the second union and the third union, respectively, thereby obtaining the first union ∪null2∪null3=total union, that is, reconstructing the intersection line with the positioning point H' in FIG3 ;

Taking the reconstruction of the oblique line in FIG. 2 b as an example, starting from the positioning point M, the first to fourth trajectories are formed respectively to form a cross-shaped total union. Thus, S5 is performed to establish an oblique line generation model.

The oblique line generation model includes the cross-shaped total union plane orthogonal line segments shown in the three-dimensional data structure diagram of Figure 3, and the oblique line is formed by clicking the cross-shaped to automatically connect the corresponding calculation method. The calculation method is, as shown in the lower part of Figure 3, rectangles are drawn in the two positive directions and two negative directions of the orthogonal X-axis and Y-axis, and the two diagonal lines of the two faces where the positioning point M is located are connected to generate the oblique line.

Finally, S6 is performed to establish a building component classification and recognition model, and the classification of building components is obtained through analysis and saved in the corresponding data points.

Taking a facade with a window in FIG. 2a as an example (FIG. 2b can be analyzed similarly), the recognition model includes using the SVM algorithm to complete classification on the plane where the reconstructed facade, ie, the total union of all intersections therein, is located.

As shown in Figure 4, the three-dimensional data structure diagram corresponding to the facade cannot be analyzed linearly, because there is no hyperplane between the data points representing the two intersection lines of the window and the data points representing the four intersection lines of the wall, and at least only a circle can be used as the classification boundary. However, by constructing a Gaussian kernel function to map each data point to the three-dimensional Hilbert space, the first hyperplane is obtained to classify the two types of building components, windows and walls.

For other apartment types in the community, the facade has two windows, namely window 1 and window 2. As shown in FIG5 , after mapping to the three-dimensional Hilbert space, the first hyperplane is obtained for classification. Then, in the same window classification obtained (the runway-shaped area in the figure), the second hyperplane is used for linear SVM classification to complete the classification of window 1 and window 2.

For cluster analysis, still taking the top floor of Figure 2b as an example, as shown in Figure 6, with the XY plane as the projection positioning surface of the data points, and the normalized grayscale value as the third dimension, the data points represented by the trisection points on the vertical intersection line and the trisection points on the top intersection line are expressed in this coordinate system. It can be clearly seen that the two different planes formed by the vertical intersection lines and the parallel intersection lines, so that the classification of vertical intersection lines and parallel intersection lines can be obtained through cluster analysis. Since the vertical intersection lines are of equal length, the corresponding calibration points are of equal height. Therefore, in the cluster analysis, the first cluster is presented, which is clustered on the horizontal plane parallel to the XY plane, and the positioning points on the parallel intersection lines have different normalized grayscale values due to the different corresponding spans of the positioning points, resulting in a second cluster of three-dimensional kink surfaces interspersed up and down in the plane.

For the window intersection line, since the span is the smallest and the normalized gray value is the highest, it is located above the plane where the vertical intersection line is located, forming the third cluster.

The reconstruction of the building complex in the community is carried out in the same manner, still taking the two buildings in the community of FIG2 as an example, the difference is that in this embodiment, the building in FIG2a is taken as the reference building, and the building in FIG2b is taken as the non-reference building, and the three-dimensional rectangular coordinate system is shown in FIG2a, while the coordinate system of the non-reference building in FIG2b is deleted. The aforementioned first to fourth steps are carried out for each building, but in step S6', it is different from step S6.

S6' includes establishing a classification and recognition model for a building component group. First, the corresponding bottom vertices in the non-reference building and the reference building in the community are made to coincide with the origin, while the corresponding window facades are parallel to the X-axis in the three-dimensional rectangular coordinate system. This allows the three-dimensional data structure diagram corresponding to each level model of the non-reference building to be translated into the XYZ three-dimensional rectangular coordinate system of the reference building, and establishes a classification and recognition model for a building component group. The classification of the building component group is obtained through analysis and saved in the data points of the three-dimensional data structure diagram corresponding to each level model of the corresponding non-reference building. Finally, the three-dimensional data structure diagram corresponding to each level model of the non-reference building is reversely translated to the position before the translation in the three-dimensional rectangular coordinate system, that is, the position in Figure 2b.

The building component group classification and recognition model includes the aforementioned SVM and cluster analysis methods as used in FIG. 4 and FIG. 6 .

OGC CityGML defines five hierarchical models including LoD0, LoD1, LoD2, LoD3, and LoD4 (LoD0, LoD1, LoD2, LoD3, and LoD4 are labels representing models at each level) to achieve multi-scale expression of building exteriors, building components, and ancillary facilities; among them, LoD0 expresses the bottom or roof outline plane of the building, which is a 2.5D polygon; LoD1 simply represents the three-dimensional model of the building in blocks; LoD2 adds a description of the auxiliary structures and roof of the house on the basis of LoD1; LoD3 describes the detailed exterior structure of the building, including but not limited to doors, windows, and balconies; LoD4 adds the expression of objects such as stairs, rooms, and furniture inside the building, with detailed geometric and semantic information.

The third step is to establish the mapping between the semantic expression and geometric expression corresponding to each level model, including:

As shown in FIG7 , the three-dimensional data structure diagrams and corresponding semantic expressions of the different levels of LoD0, LoD1, LoD2, LoD3, and LoD4 of the three-dimensional building model are defined by the OGC CityGML standard, and multiple three-dimensional data structure diagram semantic topics LoD0-semantic topic LoD4 corresponding to the three-dimensional building model are determined;

Establish associations between the semantic expressions of the different hierarchical models and the semantic themes of each of the three-dimensional data structure graphs to form a mapping body.

The semantic topic includes the building component classification name in the data point corresponding to a positioning point selected in the corresponding three-dimensional data structure diagram in FIG. 2 .

Optionally, the method further comprises:

Associating the classification of each of the CityGML building components stored in the mapping body with a LoD tag, wherein the LoD tag is used to represent different hierarchical models;

The same classification of CityGML building components corresponds to at least one geometric semantic theme and further corresponds to at least one LoD tag. For example, when the data points at the trisection point on the vertical intersection line in Figure 2b represent geometric semantic themes, they can be data points in the building shell block components of LoD1 level, or data points in the building shell surface components of LoD2-LoD4 level.

The same CityGML building component is displayed differently in the spatial rectangular coordinate system XYZ under different LoD tags to show the difference when it is represented in different levels. It can also be displayed the same in several levels and different in the remaining levels to compare the difference when it is represented between different level groups.

Table 1 Relationship between different semantic themes and different LoD levels of CityGML

As can be seen from Table 1, each hierarchical model has different display effects, and the architectural components and display methods of the architectural components are also different. Among them, different architectural components can belong to different geometric semantic themes. Table 1 schematically shows the association between different geometric semantic themes and hierarchical models, where "√" indicates that this type of geometric semantic theme and the corresponding hierarchical model have an association relationship. In this embodiment, each geometric semantic theme may contain multiple architectural components, and each architectural component can be associated with a LoD tag, that is, the same object can have model expressions corresponding to different LoDs, which are all associated with the same entity, and the corresponding display methods are selected at different levels of detail to achieve significant expression of the entity at the corresponding LoD.

The LoD surface model generation algorithm is used to construct each level model in the CityGML model, and each level model is integrated to form the final CityGML model, so as to realize the derivation of different level models. It should also be noted that the same CityGML building component may be displayed differently in different level models. For example, since the LoD0 level model is displayed roughly, smaller building components may not be displayed, or only a line or a point may be displayed. Specifically, it can be customized according to the size and type of the three-dimensional building model, which is not limited in the embodiment of the present invention.

Example 2

As shown in FIG8 , this embodiment provides a multi-level structure derivation device for a three-dimensional semantic model to implement the multi-level structure intelligent derivation method for the three-dimensional semantic model. The device includes:

The data parsing module 201 is used to obtain the original 3D model data of the 3D building model, and construct geometric expression and semantic expression for the original 3D model data;

A mapping body establishment module 202 is used to define the geometric expression and semantic expression of different hierarchical models of the three-dimensional building model through the OGC CityGML standard, and establish a mapping body between the semantic expression and geometric expression corresponding to each hierarchical model;

The model intelligent reconstruction module 203 is used to intelligently reconstruct the three-dimensional building model based on the mapping bodies corresponding to the models at each level, so as to achieve the derivation of models at different levels.

In specific operation, after the data parsing module 201 completes the construction of the geometric expression and the semantic expression, the mapping body establishment module 202 makes a request for the parsing result. After receiving the request, the data parsing module 201 transmits the parsing result to the mapping body establishment module 202, and completes the establishment of the mapping body in the mapping body establishment module 202. Through the request of the model intelligent reconstruction module 203 to the mapping body establishment module 202, the mapping body establishment module 202 transmits the established mapping body to the model intelligent reconstruction module 203, and completes the model intelligent reconstruction in the model intelligent reconstruction module 203.

The data analysis module 201 can also be used to

It is also used to establish semantic mapping rules between the IFC model and the CityGML LoD4 model, and determine the corresponding relationship between the IFC building components and the CityGML building components based on the semantic mapping rules and the original three-dimensional model data to determine the semantic expression corresponding to each CityGML building component; and extract the geometric information corresponding to each IFC building component in the IFC model based on the original three-dimensional model data, perform geometric transformation on the geometric information corresponding to each IFC building component to obtain a corresponding three-dimensional data structure diagram composed of spatial data points to represent the CityGML building component, and obtain the geometric expression corresponding to each CityGML building component.

In an optional embodiment of the present invention, the mapping establishment module 202 may also be used for:

Defining the three-dimensional data structure diagram and semantic expression of different level models of the three-dimensional building model through the OGC CityGML standard, and determining multiple three-dimensional data structure diagram semantic topics (titles) corresponding to the three-dimensional building model;

Establishing associations between the semantic expressions of the different hierarchical models and the semantic themes of the three-dimensional data structure graphs to form a mapping body;

The semantic topic includes the building component classification name in the data point corresponding to a positioning point selected in the corresponding three-dimensional data structure diagram;

Associating LoD tags according to the classification of each of the CityGML building components stored in the mapping body, wherein the LoD tags are used to represent different hierarchical models;

The same classification of CityGML building components corresponds to at least one geometric semantic theme and further corresponds to at least one LoD tag. The same CityGML building component may be displayed in the same or different manners in the spatial rectangular coordinate system XYZ under different LoD tags.

In an optional embodiment of the present invention, the model reconstruction module 330 may also be used to:

Determine the geometric semantic theme and multiple CityGML building components corresponding to any hierarchical model based on the mapping body corresponding to the hierarchical model;

Constructing a hierarchical display model corresponding to the hierarchical model according to the CityGML building components and their display modes;

The three-dimensional building model is reconstructed by integrating the hierarchical models corresponding to the hierarchical display model, thereby achieving the derivation of models at different hierarchical levels.

An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program codes, and the program codes are used to execute the methods described in the above embodiments.

An embodiment of the present invention further provides a computing device, which includes a processor and a memory: the memory is used to store program code and transmit the program code to the processor; the processor is used to execute the method described in the above embodiment according to the instructions in the program code.

Those skilled in the art can clearly understand that the specific working processes of the systems, devices, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and for the sake of brevity, they are not further described here.

In addition, the functional units in various embodiments of the present invention may be physically independent of each other, or two or more functional units may be integrated together, or all functional units may be integrated into one processing unit. The above integrated functional units may be implemented in the form of hardware, or in the form of software or firmware.

Those skilled in the art can understand that if the integrated functional unit is implemented in the form of software and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can essentially or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, which includes a number of instructions to enable a computing device (such as a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention when running the instructions. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program codes.

Alternatively, all or part of the steps of implementing the aforementioned method embodiments may be accomplished by hardware associated with program instructions (such as a computing device such as a personal computer, a server, or a network device), and the program instructions may be stored in a computer-readable storage medium. When the program instructions are executed by a processor of a computing device, the computing device executes all or part of the steps of the methods described in the embodiments of the present invention.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that within the spirit and principles of the present invention, the technical solutions described in the aforementioned embodiments can still be modified, or some or all of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate from the protection scope of the present invention.

Claims (10)

1. The multi-level structure intelligent deriving method of the three-dimensional semantic model is characterized by comprising the following steps:

Firstly, acquiring original three-dimensional model data of a plurality of three-dimensional building models, establishing a semantic mapping rule between an IFC model and a CityGML LoD4 model for each of the three-dimensional building models, and determining a corresponding relation between an IFC building component and a CityGML building component based on the semantic mapping rule and the original three-dimensional model data so as to determine semantic expressions corresponding to the CityGML building components;

Extracting geometric information corresponding to each IFC building component in the IFC model based on the original three-dimensional model data, performing geometric transformation on the geometric information corresponding to each IFC building component to obtain a corresponding three-dimensional data structure diagram formed by space data points, representing the CityGML building component, and obtaining a geometric expression corresponding to each CityGML building component;

Defining geometric expressions and semantic expressions of different level models of the three-dimensional building model through OGC CityGML standard, establishing mapping between the semantic expressions corresponding to the building components and corresponding data points in the three-dimensional data structure diagram, and establishing a mapping body between the semantic expressions and the geometric expressions corresponding to the level models;

And fourthly, intelligently reconstructing the three-dimensional building model based on the three-dimensional data structure diagram corresponding to each level model, thereby realizing the derivation of different level models.

2. The method of claim 1, wherein the method of geometric transformation of the second step comprises:

S1, establishing a three-dimensional rectangular coordinate system XYZ, defining a coordinate origin, overlapping a preset point with the origin in a plurality of original three-dimensional model data, enabling a preset direction to be parallel to an X axis or a Y axis in the three-dimensional rectangular coordinate system, and marking out building components corresponding to each level model;

S2, marking intersecting lines among different building components, acquiring coordinates of any point on the intersecting lines, obtaining coordinate value ranges on an X axis, a Y axis and a Z axis, and defining a positioning point of any intersecting line at the same time, so that the positioning point is closest or next closest to the origin;

S3, taking the value after span normalization of the coordinate value range as a gray value or an RGB color value, thereby obtaining a positioning point of any intersection line and the gray value or the RGB color value; forming a data point by the positioning point and the gray value or the RGB color value, positioning the data point into a coordinate system by the position of the positioning point in the coordinate system, forming a three-dimensional data structure diagram, and representing the data point by a symbol with the gray value or the RGB color value.

3. The method of claim 2, wherein the predetermined point is an apex of the building floor, and the predetermined direction is selected such that a plane in which one of the facades lies is parallel to the X-axis or the Y-axis;

the data points for which positioning is complete take different symbols to represent intersecting lines of different orientations.

4. The method according to claim 1, wherein in the fourth step, the method for intelligently reconstructing the three-dimensional building model based on the three-dimensional data structure map corresponding to each hierarchical model specifically comprises:

s4, for data points in any three-dimensional data structure diagram, moving from a positioning point to an X-axis positive direction to form a first track, and then moving from the positioning point to an X-axis negative direction to form a second track, so that the first aggregation length of the first track and the second aggregation length of the first track are overlapped with the span in the X-axis direction; moving from the positioning point to the Y-axis positive direction to form a third track, and then moving from the positioning point to the Y-axis negative direction to form a fourth track, so that the second union length of the third and fourth tracks coincides with the span in the Y-axis direction; moving from the positioning point to the positive Z-axis direction to form a fifth track, and then moving from the positioning point to the negative Z-axis direction to form a sixth track, so that the third union length of the fifth sixth track coincides with the span in the Z-axis direction, and a total union of the first union to the third union is obtained;

S5, establishing a diagonal generation model, wherein the diagonal generation model comprises the steps of automatically connecting corresponding calculation methods by clicking or identifying a plane orthogonal line segment or a three-dimensional orthogonal line segment to form diagonal when the total union set displayed in the three-dimensional data structure diagram is the plane orthogonal line segment or the three-dimensional orthogonal line segment, and the calculation methods are that for the plane orthogonal line segment, two orthogonal positive directions and two orthogonal negative directions are respectively rectangular, and two facing corner lines where positioning points are connected to generate diagonal; for a three-dimensional orthogonal line segment, respectively making parallelepiped straight prisms in three orthogonal positive directions and three orthogonal negative directions, connecting two body diagonal lines where positioning points are positioned, and generating oblique lines;

s6, building a building component classification recognition model, obtaining the classification of the building components through analysis, and storing the classification into corresponding data points.

5. The method according to claim 4, wherein the identifying a planar orthogonal line segment or a stereoscopic orthogonal line segment uses a difference operation between a line in the original three-dimensional model data before reconstruction and a total union after reconstruction, and if the difference result is zero, it indicates that the total union after reconstruction is not an orthogonal intersecting line segment, otherwise is an orthogonal line segment; the identification model comprises the steps of completing classification on a plane where the total union is located by adopting an SVM algorithm, and/or completing classification by carrying out three-dimensional clustering analysis on projection coordinates and gray values of positioning points on an XY plane;

The SVM algorithm comprises the steps of mapping data points on a plane where a total union set is located into a three-dimensional Hilbert space by adopting a Gaussian kernel function, and classifying by adopting a first hyperplane; and further performing linear SVM classification by adopting a second hyperplane in the same classification obtained after the classification by adopting the first hyperplane.

6. The method according to any one of claims 1-5, wherein obtaining coordinates of any point on the intersection line is replaced by obtaining coordinates of a point on the intersection line at a predetermined separation distance;

the points of the preset distance are the end points and the equal dividing points on the intersecting line.

7. The method of any one of claims 1-5, wherein for a group of buildings within a preselected geographic area, the method of geometric transformation is: selecting a reference building, constructing a three-dimensional rectangular coordinate system XYZ in the reference building, defining a coordinate origin, overlapping a preset point with the origin in original three-dimensional model data corresponding to the reference building, enabling a preset direction to be parallel to an X axis or a Y axis in the three-dimensional rectangular coordinate system, and marking out building components corresponding to all other building level models in a preselected geographical area range; steps S2 and S3 are then performed; the method for intelligently reconstructing the three-dimensional building model based on the three-dimensional data structure diagram corresponding to each level model also comprises the steps of S4-S5, and then S6', wherein the method comprises the steps of establishing a building component group classification model, overlapping preset points corresponding to the reference building in all non-reference buildings within a preselected geographical area range with the origin, enabling corresponding preset directions to be parallel to X-axis or Y-axis in a three-dimensional rectangular coordinate system, enabling the three-dimensional data structure diagram corresponding to each level model corresponding to all non-reference buildings to translate into the three-dimensional rectangular coordinate system of the reference building, establishing a building component group classification model, obtaining the classification of the building component group through analysis, storing the classification into data points of the three-dimensional data structure diagram corresponding to each level model corresponding to the corresponding non-reference building, and finally reversely translating the three-dimensional data structure diagram corresponding to each level model corresponding to each non-reference building to the position before translation in the three-dimensional rectangular coordinate system.

8. The method according to claim 7, wherein the building component group classification and identification model comprises the steps of classifying each building on a plane corresponding to a total union by adopting an SVM algorithm, and/or performing three-dimensional clustering analysis on projection coordinates of all positioning points of each building on an XY plane and corresponding gray values;

The SVM algorithm comprises the steps of constructing a Gaussian kernel function, mapping data points on a plane where a total union set is located into a three-dimensional Hilbert space, and classifying by adopting a first hyperplane; and further performing linear SVM classification by adopting a second hyperplane in the same classification obtained after the classification by adopting the first hyperplane.

9. The method according to any one of claims 10-13, characterized in that the third step comprises in particular:

Defining three-dimensional data structure diagrams and semantic expressions of different level models of the three-dimensional building model through OGC (open g language) CityGML standards, and determining semantic subjects of a plurality of three-dimensional data structure diagrams corresponding to the three-dimensional building model; establishing association relations between semantic expressions of the different-level models and semantic topics of the three-dimensional data structure diagram to form a mapping body; the semantic theme comprises a building component classification name in a data point corresponding to a positioning point selected from the corresponding three-dimensional data structure diagram;

the deriving of the model with different levels in the fourth step specifically comprises: determining a geometric semantic topic corresponding to any hierarchical model based on a mapping body corresponding to the hierarchical model and a plurality of CityGML building components; constructing a hierarchical display model corresponding to the hierarchical model according to the CityGML building component and the display mode thereof; and synthesizing the corresponding hierarchical display models of the hierarchical models to reconstruct the three-dimensional building model and realize the derivation of different hierarchical models.

10. A multi-level structure deriving device for three-dimensional semantic modeling to implement a method according to any one of claims 1-18, the device comprising:

the data analysis module is used for acquiring original three-dimensional model data of the three-dimensional building model and constructing geometric expression and semantic expression on the original three-dimensional model data;

The mapping body building module is used for defining geometric expressions and semantic expressions of different level models of the three-dimensional building model through OGC CityGML standard and building mapping bodies between the semantic expressions and the geometric expressions corresponding to the level models;

the model intelligent reconstruction module is used for intelligently reconstructing the three-dimensional building model based on the mapping body corresponding to each level model so as to realize the derivation of different levels of models;

The data parsing module is further configured to: establishing a semantic mapping rule between an IFC model and a CityGML LoD4 model, and determining a corresponding relation between an IFC building component and a CityGML building component based on the semantic mapping rule and the original three-dimensional model data so as to determine a semantic expression corresponding to each CityGML building component; extracting geometric information corresponding to each IFC building component in the IFC model based on the original three-dimensional model data, performing geometric transformation on the geometric information corresponding to each IFC building component to obtain a corresponding three-dimensional data structure diagram formed by space data points, representing the CityGML building component, and obtaining geometric expression corresponding to each CityGML building component;

The map creation module may also be configured to: defining three-dimensional data structure diagrams and semantic expressions of different level models of the three-dimensional building model through OGC (open g language) standards, and determining a semantic theme (title) of a plurality of three-dimensional data structure diagrams corresponding to the three-dimensional building model;

Establishing association relations between semantic expressions of the different-level models and semantic topics of the three-dimensional data structure diagram to form a mapping body;

The semantic theme comprises a building component classification name in a data point corresponding to a positioning point selected from the corresponding three-dimensional data structure diagram;

Associating LoD tags according to the classification of each of the citysml building elements stored in the mapping body, the LoD tags being used to characterize different hierarchical models;

wherein, the same classification of the CityGML building components corresponds to at least one geometric semantic theme, further corresponds to at least one LoD label, and the same CiyGML building components are displayed in the same or different modes in a space rectangular coordinate system XYZ under different LoD labels;

The model intelligent reconstruction module is also used for: determining a geometric semantic topic corresponding to any hierarchical model based on a mapping body corresponding to the hierarchical model and a plurality of CityGML building components;

constructing a hierarchical display model corresponding to the hierarchical model according to the CityGML building component and the display mode thereof;

And synthesizing the corresponding hierarchical display models of the hierarchical models to reconstruct the three-dimensional building model and realize the derivation of different hierarchical models.