A multi-level tile joint large-scale digital twin scene construction method

Abstract

The present application relates to the technical field of digital twin, three-dimensional tile and its processing, in particular to a multi-level tile joint large-scale digital twin scene construction method, the method comprises the following steps: obtaining various basic models and material files to generate glTF files, and realizing coordinate system transformation to unify the coordinate system; generating I3DM tiles by combining the obtained glTF files with orientation and custom attributes; determining the tile index information such as boundingVolume and GeometricError of each tile; constructing R-tree spatial index according to the tile bounding box and hierarchical relationship; generating Tileset.json file and linking I3DM tiles to form 3DTiles file, and rendering and loading through Cesium platform. The present application can meet the needs of large-scale scene rendering, and can shorten the scene construction period by using open source models, thereby improving the scene construction efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital twin, three-dimensional tile and its processing, and particularly relates to a multi-level tile joint large-scale digital twin scene construction method. BACKGROUND

[0002] Three-dimensional tiles (3DTiles) are widely used in large-scale three-dimensional geographic data loading and display. Currently, there are mainly four types of large-scale digital twin scene construction methods: traditional surveying and mapping method, panoramic oblique photography, artificial modeling, and laser radar scanning. These methods use quadtree and octree to index the source data to generate tile models through a single data source (source model, oblique photography image, grid data), and finally load them on a three-dimensional GIS (Geographic Information System) platform. However, these methods have three shortcomings:

[0003] First, they have high dependence on original data and require self-construction of source data files for loading. For example, to construct a building digital model, professional modeling personnel need to first construct a BIM (Building Information Modeling) building model and then convert it into a 3DTiles file, which limits the efficiency of digital twin scene construction.

[0004] Second, they cannot meet the LOD (Level of Detail) level relationship of multiple data sources. LOD is an important concept in three-dimensional rendering, which divides model details into levels to load under different rendering conditions, thereby saving GPU resources. Traditional LOD strategies can reduce model meshes through mesh simplification methods such as triangle collapse to achieve LOD level relationships for single models. However, in the case of multiple data sources, it is obviously inappropriate to use the same LOD strategy for models of different sizes.

[0005] Third, they are not friendly to non-uniform atomic data. The digital twin scene construction strategy divides and indexes source data in a uniform manner using quadtree, which is difficult to adapt to non-divisible and non-uniformly distributed data sets, and can easily cause index performance degradation due to multiple queries. SUMMARY

[0006] The summary section is used to briefly introduce the concepts, which will be described in detail in the specific embodiments section. The summary section is not intended to identify key or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0007] To simplify the digital twin scene modeling process using open source models, the present application proposes a multi-level tile joint large-scale digital twin scene construction method.

[0008] The application provides a multi-level tile joint large-scale digital twin scene construction method, which comprises the following steps:

[0009] Obtain various basic models and material files to generate glTF files, and realize coordinate system transformation and unified coordinate system.

[0010] The obtained glTF files are combined with the orientation and custom attributes to generate I3DM tiles.

[0011] Determine the boundingVolume, GeometricError tile index information of each tile.

[0012] According to the tile bounding box and the hierarchical relationship, an R-tree spatial index is constructed.

[0013] According to the R-tree spatial index and the tile index information, a Tileset.json file is generated, the I3DM tiles are linked, a 3DTiles file is formed, and rendering and loading are performed through the Cesium platform.

[0014] Optionally, the glTF file is generated, and the unified coordinate system is realized by coordinate system transformation, which comprises the following steps:

[0015] The private coordinate system of the basic model is processed into the right-hand coordinate system, and the material file in the basic model is extracted and embedded into the glTF file.

[0016] Optionally, the glTF file is combined with the orientation and custom attributes to generate I3DM tiles, which comprises the following steps:

[0017] A FileHead file header is generated, glTF file object IDs are extracted, and BatchTable is generated by combining interactive request information. The rotation matrix of each glTF file is calculated by orientation, and custom attribute information is stored in FeatureTable. The glTF file is embedded in a linked manner to generate a single I3DM tile.

[0018] Optionally, the boundingVolume, GeometricError tile index information of each tile is determined, which comprises the following steps:

[0019] The boundary body of each glTF model mesh is extracted, the position information is superimposed, the coordinate conversion is performed to obtain the boundingVolume bounding box, and the GeometricError geometric error is calculated according to the bounding box.

[0020] Optionally, the R-tree spatial index is constructed according to the tile bounding box and the hierarchical relationship, which comprises the following steps:

[0021] According to the tile bounding box and the hierarchical relationship, the R-tree spatial index is constructed through the insertion, selection, adjustment and splitting processes, wherein the R-tree index construction is completed by selectively inserting into the index tree through the self-defined hierarchical relationship of nodes in the insertion process.

[0022] Optionally, the Tileset.json file is generated according to the R-tree spatial index and the tile index information, the I3DM tiles are linked, the 3DTiles file is formed, and the rendering loading is performed through the Cesium platform, including:

[0023] According to the constructed R-tree spatial index and the determined tile index information, the node Tileset.json is recursively calculated and merged into the total Tileset.json file to generate the 3DTiles file, and the tile set file is called through CesiumJS to realize loading.

[0024] The present application has the following beneficial effects:

[0025] The present application uses the general models contained in the Internet as public resources, and combines the self-defined index information to construct a digital twin scene, so that part of the work of obtaining source data is avoided. While using the open source model, the scene construction period can be shortened, and the need for large-scale scene rendering can be met. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0027] Figure 1 A flowchart of a multi-level tile joint large-scale digital twin scene construction method of the present application;

[0028] Figure 2 An I3DM tile data structure diagram of the present application;

[0029] Figure 3 A tile index node data structure diagram of the present application;

[0030] Figure 4 An R-tree index construction flowchart of the present application;

[0031] Figure 5 Another flowchart of the present application. DETAILED DESCRIPTION

[0032] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined inventive objectives, the specific implementation, structure, features and effects of the technical solutions proposed according to the present application are described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0034] The present application provides a multi-level tile joint large-scale digital twin scene construction method, which comprises the following steps:

[0035] Obtain various basic models and their material files to generate glTF files, and realize coordinate system transformation and unified coordinate system;

[0036] The obtained glTF files are combined with the orientation and custom attributes to generate I3DM tiles;

[0037] Determine the boundingVolume, GeometricError tile index information of each tile;

[0038] According to the tile bounding box and hierarchical relationship, an R-tree spatial index is constructed;

[0039] According to the R-tree spatial index and the tile index information, a Tileset.json file is generated and linked to the I3DM tiles to form a 3DTiles file, and the rendering and loading are performed through the Cesium platform.

[0040] The above steps are described in detail as follows:

[0041] Reference Figure 1 , shows the flow of some embodiments of a multi-level tile joint large-scale digital twin scene construction method according to the present application. The multi-level tile joint large-scale digital twin scene construction method comprises the following steps:

[0042] Step S1, obtain various basic models and their material files to generate glTF files, and realize coordinate system transformation and unified coordinate system.

[0043] In some embodiments, various basic models and their material files can be collected according to the required scene to generate glTF files, and coordinate system transformation and unified coordinate system can be realized.

[0044] The basic model can be a general model publicly available on the Internet. For example, the basic model can be, but is not limited to, an obj model. The categories can be basic categories such as trees, cars, and buildings that constitute a digital twin scene. Each basic model in the embodiment can be a plurality of basic models required to build a digital twin scene according to actual conditions.

[0045] As an example, each basic model and its material file can be obtained, the basic model private coordinate system is processed into a right-handed coordinate system, and the material file in the basic model is extracted and embedded into a glTF file. For example, if the basic model is an obj model, in the glTF file generation process, first, the model vertex coordinates and the triangular face normal vector are extracted, and the y-axis and z-axis are changed to the right-handed coordinate system. In addition, the model texture data is extracted and saved to the glTF file bufferView, and finally the mesh data and texture data are saved to generate a glTF file.

[0046] It should be noted that most general model files have their own private coordinate system, and the glTF file conforms to the right-handed Cartesian coordinate system, which requires rotation transformation of the mesh data. Secondly, we assume that the obj model is a loadable, real-size model, and additional spatial position information, orientation information, attribute information, and LOD level information of a plurality of basic models are known, so as to be used in subsequent work.

[0047] Step S2, combining the obtained glTF file with the orientation and custom attributes to generate an I3DM tile.

[0048] In some embodiments, the glTF file obtained in step S1 can be combined with the orientation and custom attributes to generate an I3DM tile.

[0049] As an example, first, a FileHead file header is generated, the glTF file object ID is extracted and combined with the interactive request and other information to generate a BatchTable, the rotation matrix of each glTF file is calculated through the orientation, and the custom attribute information is stored in a FeatureTable, and finally the glTF file is embedded in a linked manner to generate a single I3DM tile.

[0050] As Figure 2As shown, the I3DM file includes Filehead, FeatureTable, BatchTable, glb four parts, which mainly include glb model mesh data, user-defined attributes and other information. In the glTF to I3DM conversion process, the file version number and the size of the I3DM tile data are set to the Filehead, the glTF object ID, the category, and the tile API interaction request are stored in the BatchTable. The model orientation is converted into a rotation matrix and combined with the custom attribute information in the FeatureTable. Finally, the glTF file is embedded in the glb in a linked manner to generate the I3DM tile.

[0051] Step S3, determine the boundingVolume, GeometricError tile index information of each tile.

[0052] In some embodiments, the glTF model mesh can be extracted, the boundingVolume is obtained, and the position information is superimposed for coordinate conversion to obtain the boundingVolume bounding box, and the GeometricError geometric error is calculated according to the bounding box.

[0053] As an example, as shown in Figure 3 The spatial index information of each tile needs to be calculated, which is used to enable the 3DTiles file to be displayed correctly during the loading process. First, the glTF file mesh boundary value is obtained and its own position information is superimposed to obtain the boundingVolume, second, the diagonal distance of the boundingVolume is multiplied by the adjustment coefficient to obtain the geometricError, and finally the I3DM tile is linked in the content. The Children array will be linked after the tile index tree is successfully constructed.

[0054] Step S4, according to the tile bounding box and hierarchical relationship, construct R-tree spatial index.

[0055] In some embodiments, the R-tree spatial index can be constructed according to the tile bounding box and hierarchical relationship.

[0056] As an example, the R-tree spatial index can be constructed according to the tile bounding box and hierarchical relationship through the insertion, selection, adjustment and split process. In the insertion process, the R-tree index is constructed by selectively inserting the node into the index tree according to the custom hierarchical relationship.

[0057] It is to be noted that the R-tree index is widely used for GIS data indexing, and has good indexing performance on non-uniform data sets. The R-tree uses MBR (Minimum Bounding Rectangle) to build tile index, so that it does not need to split sub-data, but each node has no hierarchical distinction on data. In order to adapt to the LOD hierarchy of different models, the creation process of the traditional R-tree index is improved. It includes the following processes: insertion, selection, adjustment, and split. In the insertion process, the nodes are selectively inserted into the index tree through the self-defined hierarchical relationship, so as to complete the R-tree index construction.

[0058] According to the boundingVolume of each tile and the hierarchical relationship, the improved R-tree spatial index is constructed. First, each tile is classified according to the level of the hierarchy, and each class is stored in the form of a queue. The MBR node is gradually inserted into the R-tree in the order from the high LOD level to the bottom LOD level. In addition, the tile boundary volume boundingVolume is used as the MBR node to construct the index tree. In the index construction process, the parent MBR node is always the minimum node that encloses the child MBR node, and the parent node is always smaller than the child node in the LOD level.

[0059] Here it is assumed that there are three kinds of nodes: A\LOD0, B\LOD1, and C\LOD2.

[0060] The index construction process includes the following sub-processes:

[0061] The insertion (insert) process: first, chooseNode selects node B to insert C on the index tree, which may have two unexpected situations. One situation is that the B node is overflowed, and then the adjust operation is performed. The other situation is that there is no insertion node B that meets the hierarchical relationship. At this time, a node needs to be popped out from the LOD1 queue to perform the insert operation.

[0062] The split process: for the overflowed node B1, first chooseNode selects the split node B2. There may be two situations: the first situation is that there is no split node, and then the A node is inserted into the B2 node. The second situation is that there is a split node, and then all the nodes under B1 are divided into B1 and B2 nodes according to the minimum intersection principle of P. Finally, the A node is adjusted.

[0063] The adjustment (adjust) process: if the node is overflowed, the split operation is performed, otherwise the process is ended.

[0064] The selection (chooseNode) process: the node is mainly selected through the p index to ensure that the intersection area of each MBR region is minimum. The formula corresponding to the p index is:

[0065]

[0066] Wherein, k represents the number of possible intersection sibling MBR regions caused by inserting the node, MBR represents the node range, Si represents each node, and P represents the intersection area of each MBR.

[0067] The R-tree index construction process is not limited to the above node, and each process is recursively called, and inserting a node may need to insert multiple nodes due to the absence of a suitable node and the like. Figure 4 The construction process of the binary R-tree index is exemplified, wherein only three tiles a, b and c are exemplified, and the former is smaller than the latter on the LOD level, and the following will introduce the flow of the example:

[0068] State 1: It is assumed that the R-tree index has two c nodes, one b node at this time, and b is the parent node of c and is at the bottom level on the LOD level.

[0069] State 2: When a new c node is inserted, the b node is overflowed and is ready to be adjusted.

[0070] State 3: It is found that the a node is missing, so the a node is selected and inserted first.

[0071] State 4: It is found that the b node is missing, so the b node is selected and inserted to prepare to start splitting.

[0072] State 5: The original node is split to the b node through the p index, and the splitting is completed.

[0073] The worst case is exemplified in the entire R-tree index construction process, and the insertion of the remaining nodes meets the above process, until all nodes are inserted into the index, and the R-tree index construction is completed.

[0074] Step S5: generating a Tileset.json file according to the R-tree spatial index and tile index information of each tile, linking I3DM tiles, forming a 3DTiles file, and rendering and loading through the Cesium platform.

[0075] In some embodiments, according to the R-tree spatial index constructed in step S4 and the tile index information determined in step S3, each node Tileset.json can be recursively calculated and merged into a total Tileset.json file to generate a 3DTiles file, and the tileset file can be requested and called through CesiumJS to realize loading.

[0076] As an example, as Figure 3As shown, each tile node includes a children array that stores all high-LOD level nodes under the node. The children array of each node can be filled according to the R-tree index created in step S4, and the Tileset link of the child node is recorded in children. Finally, the Tileset.json file of the R-tree root node is returned to complete the construction of 3DTiles. Finally, the 3DTiles tileset is implemented on the Cesium platform through a CesiumJS request call to realize scene rendering and loading.

[0077] If the basic model is an obj model, another flowchart of the present application can be as shown in Figure 5 As shown, the flow takes the obj model and its orientation level information as input, and the target constructs a 3DTiles three-dimensional tile model file based on the basic model and loads it. It meets the LOD level detail requirements at different heights through the hierarchical mechanism of the basic model. It includes steps S1-S5. Among them, Figure 5 S1-S5 in the above are steps S1-S5, respectively.

[0078] In summary, the present application collects public models on the Internet, uses these public resources in combination with custom index information to construct a digital twin scene, so as to avoid part of the work of obtaining source data, and while using open source models, it can shorten the construction period of the digital twin scene to a certain extent and meet the needs of large-scale scene rendering.

[0079] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for constructing a large-scale digital twin scene with multi-level tile federation, characterized in that, The method comprises the following steps: obtaining various basic models and material files to generate a glTF file, and realizing coordinate system transformation and unifying a coordinate system; generating I3DM tiles by combining the obtained glTF file with orientation and custom attributes; determining the boundingVolume and GeometricError tile index information of each tile; constructing an R-tree spatial index according to the tile bounding box and hierarchical relationship; generating a Tileset.json file and linking I3DM tiles according to the R-tree spatial index and tile index information, forming a 3DTiles file, and rendering and loading through the Cesium platform; the R-tree spatial index is constructed according to the tile bounding box and hierarchical relationship, comprising: the R-tree spatial index is constructed by inserting, selecting, adjusting and splitting according to the tile bounding box and hierarchical relationship, wherein the R-tree index is constructed by selectively inserting the nodes into the index tree according to the custom hierarchical relationship in the insertion process; an improved R-tree spatial index is constructed according to the tile boundingVolume and hierarchical relationship; the tiles are classified according to the hierarchical level, each class is stored in the form of a queue, and the R-tree insertion operation is gradually performed from the high LOD level to the bottom LOD level; in addition, the tile boundary body boundingVolume is used as the MBR node to construct the index tree, and in the index construction process, the parent MBR node always encloses the smallest node of the child MBR node, and the parent node is always smaller than the child node in the LOD level.

2. The multi-level tile federated large-scale digital twin scene construction method according to claim 1, characterized in that, the glTF file is generated, and the coordinate system transformation and unification of the coordinate system are realized, comprising: processing the private coordinate system of the basic model to the right-hand coordinate system, and extracting the material file in the basic model and embedding it into the glTF file.

3. The multi-level tile federated large-scale digital twin scene construction method according to claim 1, characterized in that, the obtained glTF file is combined with the orientation and custom attributes to generate I3DM tiles, comprising: generating a FileHead file header, extracting glTF file object ID and generating a BatchTable by combining interactive request information, calculating the rotation matrix of each glTF file through orientation and storing custom attribute information in FeatureTable, and embedding glTF files in a linked manner to generate a single I3DM tile.

4. The multi-level tile federated large-scale digital twin scene construction method according to claim 1, characterized in that, the boundingVolume and GeometricError tile index information of each tile are determined, comprising: extracting the boundary body of each glTF model mesh to obtain the boundingVolume bounding box by superimposing the position information and performing coordinate conversion, and calculating the GeometricError geometric error according to the bounding box.

5. The multi-level tile federated large-scale digital twin scene construction method according to claim 1, characterized in that, the R-tree spatial index and tile index information are used to generate a Tileset.json file and link I3DM tiles, form a 3DTiles file, and render and load through the Cesium platform, comprising: According to the constructed R-tree space index and the determined tile index information, each node Tileset.json is calculated recursively and merged into a total Tileset.json file to generate a 3DTiles file, and the tile set file is called to realize loading through CesiumJS request.

Images