Virtual geographic scene fusion modeling method, device and computer equipment

Abstract

The present application relates to a method, device and computer equipment for fusion modeling of virtual geographic scenes. The method includes: obtaining training sample optimization data; using an irregular triangulated network model to process terrain optimization data terrain scene; reconstructing the three-dimensional model and processing the texture coordinate mapping of the object optimization data according to the three-dimensional modeling algorithm to obtain the object model; fusing the terrain scene and the object model to obtain the geographic scene fusion model; obtaining the spatial position parameters of the VR device according to the scale and scaling of the geographic scene fusion model, and sending the geographic scene fusion model and the spatial position parameters to the VR device. The present application adopts machine learning to efficiently classify, simplify and extract features of massive three-dimensional data, thereby extracting effective scene modeling data, greatly reducing data redundancy, improving modeling efficiency, and using spatial constraint relationships to fuse and model geographic scenes to break through the existing modeling methods.

Description

Technical Field

[0001] The present application relates to the field of virtual reality technology, and in particular to a virtual geographic scene fusion modeling method, device and computer equipment. Background Art

[0002] Three-dimensional geographic scene modeling and visualization refers to the process of obtaining geographic spatial data by various data collection methods, and constructing a three-dimensional geographic scene model that can truly reflect the characteristics of the geographic environment through specific algorithms and technologies, and presenting the constructed three-dimensional geographic scene model in an intuitive visual form, so that users can intuitively understand geographic spatial information, including the form, position, spatial relationship of geographic objects, and dynamic changes of various geographic phenomena. Among them, geographic environmental characteristics include topography (such as mountains, rivers, plains, etc.), landforms (such as buildings, roads, bridges, etc.) and the spatial relationship between them. However, most of the existing three-dimensional geographic scene modeling and visualization adopts the method of manual modeling. For example, data collection and preliminary processing are first carried out, and then the landforms in the scene are modeled through relevant modeling software (such as 3dsmax, SketchUp, etc.), and finally the modeling of the geographic scene is realized. However, the existing modeling method has a limited scene range, and does not take into account the constraints such as semantics, position and attributes in various geographic scenes. In addition, in terms of visualization, it is often a traditional three-dimensional visualization technology, lacking scene immersion effect, and the user experience needs to be improved. Summary of the invention

[0003] Based on this, it is necessary to provide a virtual geographic scene fusion modeling method, device and computer equipment that can improve modeling efficiency and visualization experience in response to the above technical problems.

[0004] A virtual geographic scene fusion modeling method comprises the following steps:

[0005] Obtain a training sample data set, optimize the training sample data set based on a support vector machine algorithm, and obtain training sample optimized data; the training sample optimized data at least includes terrain optimized data and ground feature optimized data;

[0006] The terrain optimization data is processed using the irregular triangulated network model, and the terrain optimization data is constructed into a triangulated network data structure to obtain the terrain scene;

[0007] Reconstruct the three-dimensional model and process the texture coordinate mapping of the optimized data of the ground object according to the three-dimensional modeling algorithm to obtain the ground object model;

[0008] Based on the preset fusion rule mapping, the terrain scene and the object model are fused to obtain a geographic scene fusion model; the preset fusion rule mapping includes spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints;

[0009] According to the scale and scaling ratio of the geographic scene fusion model, the spatial position parameters of the VR device are obtained, and the geographic scene fusion model and the spatial position parameters are sent to the VR device, so that the VR device displays the geographic scene fusion model according to the spatial position parameters and realizes scene interaction according to a preset interaction method, which includes a bird's-eye view interaction method, an automatic roaming interaction method, and a moving walking interaction method.

[0010] In one embodiment, a training sample data set is obtained, and the training sample data set is optimized based on a support vector machine algorithm. The step of obtaining the training sample optimized data includes the following steps:

[0011] Collecting raw data and extracting training sample data sets from the raw data; the raw data include manually surveyed terrain raster data, high-resolution satellite image data and ground object model data; the training sample data sets include at least terrain data and ground object data;

[0012] Based on the preset feature extraction rules, feature extraction is performed on the training sample data set to obtain feature extraction results, and a feature space is constructed based on the feature extraction results; the feature extraction results include texture features, structural features, elevation features, scale features, semantic features, location features, and attribute features;

[0013] According to the training sample data set and feature space, a hyperplane is constructed on the feature space;

[0014] A hyperplane is used to test the training sample data set. If the data in the training sample data set meets the classification rules of the hyperplane, it is retained; otherwise, the support vector machine algorithm is used to reassign the data in the training sample data set that does not meet the classification rules of the hyperplane, and features are extracted from the data in the reassigned training sample data set. The corresponding hyperplane is constructed and data testing is performed until all data in the training sample data set meets the classification rules of the hyperplane to obtain the optimized training sample data.

[0015] In one embodiment, the steps are also included:

[0016] According to the feature extraction results, the training sample optimization data is classified to obtain the data classification results; the data classification results include spatial location information, attribute data, spatial semantic data, structural morphology, texture coordinates and elevation values;

[0017] Based on the feature extraction results and data classification results, index rules are constructed for the training sample optimization data, and based on the index rules, the training sample optimization data is stored in a topological relationship.

[0018] In one embodiment, based on the feature extraction results and the data classification results, index rules are constructed for the training sample optimization data, and based on the index rules, the steps of topologically storing the training sample optimization data include the steps of:

[0019] Configure corresponding indexes for feature extraction results and data classification results respectively;

[0020] The index is tested in the training sample optimization data. If the content pointed to is unique, the preliminary index is retained; if the content pointed to is not unique, the index is changed and the changed index is tested until the content pointed to by the index is unique, so as to construct the index rule.

[0021] In one embodiment, the step of processing terrain optimization data using an irregular triangulated network model and constructing the terrain optimization data into a triangulated network data structure to obtain a terrain scene further includes the steps of:

[0022] The coordinate and elevation information of the irregular triangulated network of the triangulated network data structure is obtained, and the coordinate and elevation information are stored in a topological relationship using index rules.

[0023] In one embodiment, the step of performing 3D model reconstruction and texture coordinate mapping processing on the object optimization data according to the 3D modeling algorithm to obtain the object model includes the following steps:

[0024] Using index rules, retrieve and obtain the data of building elements of features from the feature optimization data;

[0025] Reconstruct the three-dimensional model of the ground feature construction element data according to the three-dimensional modeling algorithm to obtain a preliminary ground feature model;

[0026] The preliminary object model is processed by texture coordinate mapping to obtain the object model.

[0027] In one embodiment, based on the preset fusion rule mapping, the terrain scene and the object model are fused to obtain the geographical scene fusion model, including the steps of:

[0028] Extract parameters of the object model to obtain the spatial coordinates, spatial posture and scaling ratio of the object model;

[0029] According to spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints, the spatial coordinates, spatial posture and scaling ratio are mapped respectively, and the object model is integrated into the terrain scene.

[0030] In one embodiment, the steps are also included:

[0031] Calculate the data level of the feature optimization data and the viewing angle distance of the VR device, and determine the display effect corresponding to the data level according to the viewing angle distance;

[0032] Obtain the scene camera position of the data VR device, calculate the viewing angle parameters of the VR device according to the scene camera position, and send the viewing angle parameters to the VR device;

[0033] According to the geographic scene fusion model, spatial position parameters, display effect and viewing angle parameters, configure the bird's-eye view interaction mode, automatic roaming interaction mode and mobile walking interaction mode.

[0034] A virtual geographic scene fusion modeling device, the device comprising:

[0035] A data processing module is used to obtain a training sample data set, optimize the training sample data set based on a support vector machine algorithm, and obtain training sample optimized data; the training sample optimized data at least includes terrain optimized data and ground feature optimized data;

[0036] A terrain scene acquisition module, used for processing the terrain optimization data using an irregular triangulated network model, and constructing the terrain optimization data into a triangulated network data structure to acquire a terrain scene;

[0037] A ground object model acquisition module is used to perform three-dimensional model reconstruction and texture coordinate mapping processing on the ground object optimization data according to the three-dimensional modeling algorithm to obtain the ground object model;

[0038] A fusion module is used to fuse the terrain scene and the object model based on a preset fusion rule mapping to obtain a geographic scene fusion model; the preset fusion rule mapping includes spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints;

[0039] The parameter configuration module is used to obtain the spatial position parameters of the VR device according to the scale and scaling ratio of the geographic scene fusion model, and send the geographic scene fusion model and the spatial position parameters to the VR device so that the VR device displays the geographic scene fusion model according to the spatial position parameters.

[0040] A computer device comprises a memory and a processor, wherein the memory stores a computer program and the processor implements the steps of the above method when executing the computer program.

[0041] One of the above technical solutions has the following advantages and beneficial effects:

[0042] The virtual geographic scene fusion modeling method of the present application is carried out through the following steps: obtaining a training sample data set, optimizing the training sample data set based on a support vector machine algorithm, and obtaining training sample optimization data; using an irregular triangulated network model to process terrain optimization data, and constructing the terrain optimization data into a triangulated network data structure to obtain a terrain scene; reconstructing a three-dimensional model and mapping texture coordinates on the object optimization data according to a three-dimensional modeling algorithm to obtain an object model; based on a preset fusion rule mapping, fusion of the terrain scene and the object model to obtain a geographic scene fusion model; according to the scale and scaling of the geographic scene fusion model, obtaining the spatial position parameters of the VR device, and sending the geographic scene fusion model and the spatial position parameters to the VR device, so that the VR device displays the geographic scene fusion model according to the spatial position parameters, and realizes scene interaction according to a preset interaction method. The present application adopts a machine learning method to efficiently classify, simplify and extract features of massive three-dimensional data, thereby extracting effective scene modeling data, greatly reducing data redundancy, and improving modeling efficiency. The spatial constraint relationship is used to fuse and model the geographic scene to break through the existing modeling method, and under the support of VR technology, a diversified interaction mode of the geographic scene and a visual expression at multiple scales are realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is an application environment diagram of the virtual geographic scene fusion modeling method in the embodiment of the present application.

[0044] Figure 2 It is a flowchart of the virtual geographic scene fusion modeling method in the embodiment of the present application.

[0045] Figure 3 A schematic diagram of the process of obtaining optimized data in an embodiment of the present application.

[0046] Figure 4 A flowchart of the steps for constructing index rules in an embodiment of the present application.

[0047] Figure 5 A flowchart of the interactive configuration steps in an embodiment of the present application.

[0048] Figure 6 Schematic diagram of the internal structure of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0050] The virtual geographic scene fusion modeling method provided in this application can be applied to Figure 1 In the application environment shown. Among them, the VR device 102 communicates with the server 104 through the network. The server 104 optimizes the training sample data set based on the support vector machine algorithm to obtain the training sample optimization data; the server 104 uses the irregular triangulated network model to process the coordinates and elevation information in the terrain optimization data, and constructs the terrain optimization data into a triangulated network data structure to obtain the terrain scene; the server 104 performs three-dimensional model reconstruction and texture coordinate mapping processing on the object optimization data according to the three-dimensional modeling algorithm to obtain the object model; the server 104 maps based on the preset fusion rule, fuses the terrain scene and the object model, and obtains the geographic scene fusion model; the server 104 obtains the spatial position parameters of the VR device 102 according to the scale and scaling of the geographic scene fusion model, and sends the geographic scene fusion model and the spatial position parameters to the VR device 102, so that the VR device 102 displays the geographic scene fusion model according to the spatial position parameters, and realizes scene interaction according to the preset interaction method. Among them, the VR device 102 refers to a device that uses computer technology to create and simulate a virtual environment so that users can immerse themselves in and interact with it. It provides users with an immersive experience through various technical means, such as stereoscopic display, head tracking, handle interaction, etc. The server 104 can be implemented as an independent server or a server cluster composed of multiple servers.

[0051] In one embodiment, Figure 2 As shown in the figure, a virtual geographic scene fusion modeling method is provided, which is applied to Figure 1 The server in the example is used to illustrate the following steps:

[0052] Step S210, obtain a training sample data set, optimize the training sample data set based on the support vector machine algorithm, and obtain training sample optimized data. The training sample data is extracted from the original data. The original data is a sample data set with the characteristics of large data volume, high redundancy, deep complexity between data and weak correlation. It is necessary to perform rough extraction on the original data to obtain the training sample data set. The obtained training sample data can be divided into terrain data, high-resolution image data (belonging to landform data), building model (belonging to land object data), oblique photography (belonging to land object data)), attribute data (belonging to attribute expression) and metadata (belonging to attribute expression) according to the terrain, landform, land object and attribute expression of the geographical scene. Among them, the terrain data is to process the terrain (Digital Elevation Model, DEM) raster data into a triangulated network data structure based on the irregular triangulated network model (Triangulated Irregular Network, TIN). Specifically, the terrain surface triangulated network model is constructed through the coordinates and elevation information of the terrain raster grid, so as to realize the triangulation of a single grid. The grid coordinates and elevation can be calculated by the following formula:

[0053]

[0054] in and Respectively represent the starting point coordinates of the terrain raster data, represents the size of the grid, and Represent the column and row number of the current grid respectively, Indicates the total number of rows.

[0055] High-resolution image data is image data obtained by high-resolution earth observation satellites or aerial remote sensing platforms. Building models are a physical or virtual form of representation that intuitively displays, analyzes, studies, assists in design, or presents results of buildings or building groups in three-dimensional form. Oblique photography is a type of data obtained using aerial photography technology. Specifically, by carrying multiple sensors (cameras) on an aircraft (such as a drone or a manned aircraft), ground targets are photographed from multiple angles such as vertical and oblique to obtain more comprehensive and richer image information of the ground objects. Both building models and oblique photography belong to ground object data. Ground object data modeling is the most important link in virtual geographic scene modeling. This application uses modeling software to perform refined modeling processing on key buildings (such as houses, roads, bridges, vegetation, etc.). The specific method is to construct a refined three-dimensional building model in the target scene area in the three-dimensional modeling software; then import the constructed building model into the established real terrain scene; and finally obtain a refined three-dimensional scene model of the ground object. Attribute data is data that describes the characteristics or properties of geographic entities (such as ground objects, landforms, etc.). Metadata, also known as intermediary data or relay data, is data that describes data. It is mainly information that describes the attributes of data and is used to support functions such as indicating storage location, historical data, resource search, and file records.

[0056] The original data includes manually surveyed terrain raster data, high-resolution satellite image data and terrain model data. Terrain raster data is a type of raster data used to represent terrain information. It divides the earth's surface into a series of regular grid cells (pixels), each of which has a corresponding value to represent the terrain features or related attributes at the location of the cell. High-resolution satellite image data refers to data formed by earth observation images acquired by high-resolution satellites. Terrain model data can be terrain model data such as buildings, roads, transportation, security, etc. acquired through refined modeling software or drone aerial photography, as well as oblique photography acquired through aerial photography technology.

[0057] After obtaining the training sample data set, the training sample data set is optimized based on the support vector machine algorithm to obtain the training sample optimized data. The training sample optimized data includes at least terrain optimization data and feature optimization data. Among them, the support vector machine algorithm (SVM) is a supervised learning model for classification and regression analysis. The purpose is to establish a decision boundary (hyperplane) based on the maximized interval of the training sample data set to achieve classification and simulation of the training samples. It should be noted that the optimization of the training sample data set can be used to process missing values, duplicate values, correct erroneous values, merge multi-source data or standardize data.

[0058] In a specific example, if Figure 3As shown, a training sample data set is obtained, and the training sample data set is optimized based on a support vector machine algorithm. The step of obtaining the training sample optimized data includes the following steps:

[0059] Step S310, collect raw data, and extract training sample data set from the raw data. The raw data includes manually surveyed terrain raster data, high-resolution satellite image data, and object model data. The training sample data set includes at least terrain data and object data. In another example, the training sample data set also includes landform data and attribute description.

[0060] Step S320, based on the preset feature extraction rules, feature extraction is performed on the training sample data set, feature extraction results are obtained, and a feature space is constructed according to the feature extraction results. Among them, the feature extraction results include texture features, structural features, elevation features, scale features, semantic features, location features and attribute features. The preset feature extraction rules are a set of predetermined criteria and methods for identifying, selecting and extracting meaningful features (or attributes) from a data set. Taking the texture feature extraction rules as an example, other feature extraction rules are similar. Texture features need to cover image recognition channels and high-resolution spectra of all ground scenes such as buildings, water systems, roads, and transportation facilities to ensure that the extraction results can be quickly and accurately recognized by the machine learning model. The feature space is a space composed of the features (attributes) of the data, and the feature space is specifically constructed by texture features, structural features, elevation features, scale features, semantic features, location features, and attribute features. In an example, part of the training sample data can be randomly selected from the training sample data set, features can be extracted from part of the training sample data, feature extraction results can be obtained, and a feature space can be constructed according to the feature extraction results. It should be noted that part of the training sample data should cover three or more features.

[0061] Step S330, construct a hyperplane on the feature space according to the training sample data set and the feature space. Since the support vector machine algorithm is a generalized linear classifier that performs binary classification of data in a supervised learning manner, a classification hyperplane is constructed on the constructed feature space by selecting a kernel function, and the binary classification of samples is achieved by using a regularization factor to balance the interval maximization and the training error minimization. Therefore, this step is to establish a hyperplane based on the training sample data and the feature space. In the support vector machine algorithm, for linearly separable binary classification problems, a hyperplane is used to separate data points of different categories. The goal is to find an optimal hyperplane so that the distance (called the interval) between the two types of data points and this hyperplane is maximized, and this hyperplane can classify new data points well.

[0062] Step S340, use a hyperplane to test the training sample data set. If the data in the training sample data set meets the classification rules of the hyperplane, it is retained; otherwise, the support vector machine algorithm is used to reassign the data in the training sample data set that does not meet the classification rules of the hyperplane, and features are extracted from the data in the reassigned training sample data set. The corresponding hyperplane is constructed and data verification is performed until all data in the training sample data set meets the classification rules of the hyperplane to obtain the optimized training sample data.

[0063] The constructed hyperplane is used to check the data in the training sample data set. The data that meets the hyperplane is judged to meet the requirements, and the data that does not meet the hyperplane is reassigned. Based on the reassigned data, a new hyperplane is reconstructed, and the new hyperplane is used to check the reassigned data, and so on until all the data meet the corresponding hyperplane.

[0064] Combined with machine learning related technologies, the massive geographic scene data is preliminarily identified and features extracted. The decision boundary is established through the support vector machine algorithm, and based on this, the data is classified and extracted and features are identified to achieve efficient data induction and retrieval. Then, unified spatial constraint rules are established for the summarized data to achieve the purpose of rapid fusion modeling of geographic scenes.

[0065] In one example, data is revalued based on the following function:

[0066]

[0067] Among them, among them, represents the initial assignment of geographic scene training samples, Represents coefficient-based The hyperplane structure of is defined by the vector, represents the transmission coefficient of the sample parameter to be classified, represents the eigenvalue of the training sample based on the improved support vector machine theory, Represents the information sample assignment parameter.

[0068] In order to enrich the classification rules of feature types, in an example, Figure 4 As shown, the virtual geographic scene fusion modeling method of the present application also includes the steps of:

[0069] Step S410, classify the training sample optimization data according to the feature extraction result to obtain the data classification result. Among them, the feature extraction result divides the features in the training sample data set into texture features, structural features, elevation features, scale features, semantic features, position features and attribute features, and classifies the data in the training sample data set based on the feature classification. Specifically, the data classification result includes spatial position information, attribute data, spatial semantic data, structural morphology, texture coordinates and elevation values. Among them, feature classification is to classify the features (attributes) of the data, and it mainly focuses on dividing the various features in the training sample data set into different categories according to certain rules or standards. Data classification is to classify the entire data sample (instance), and its goal is to divide each data point in the training sample data set into different categories.

[0070] Step S420, based on the feature extraction results and data classification results, index rules are constructed for the training sample optimization data, and based on the index rules, the training sample optimization data is topologically stored. An efficient data retrieval mechanism is provided for the next step of geographic scene fusion modeling. In this step, index rules are established for the feature extraction results and data classification results to complete the establishment of index rules for the training sample optimization data. After the index rules are established, the training sample optimization data is topologically stored according to the index rules.

[0071] In a specific example, based on the feature extraction results and the data classification results, index rules are constructed for the training sample optimization data, and based on the index rules, the steps of storing the topological relationship of the training sample optimization data include the following steps: configuring corresponding indexes for the feature extraction results and the data classification results, respectively; checking the index in the training sample optimization data, and if the content pointed to is unique, retaining the preliminary index; if the content pointed to is not unique, changing the index and checking the changed index until the content pointed to by the index is unique, so as to construct the index rule.

[0072] Step S220, use the irregular triangulated network model to process the terrain optimization data, and construct the terrain optimization data into a triangulated network data structure to obtain the terrain scene. The irregular triangulated network model (Triangulated Irregular Network, TIN) is a data structure used to describe the surface model. It can accurately reflect the specific ups and downs and detailed features of the surface model. It is often used to depict geographical scenes such as terrain and landforms. The terrain scene obtained is a comprehensive presentation of the terrain and landform characteristics of the earth's surface, including natural terrain elements such as mountains, hills, plains, rivers, lakes, and traces left by human activities on the terrain (such as terraces, roads, urban buildings and other artificial landforms). It is a three-dimensional space concept that covers the geometric features of the terrain such as shape, height, slope, and slope aspect, as well as the appearance features of the surface such as material, texture, and color.

[0073] In one example, the step of processing terrain optimization data using an irregular triangulated network model and constructing the terrain optimization data into a triangulated network data structure to obtain a terrain scene also includes the steps of obtaining coordinates and elevation information of the irregular triangulated network of the triangulated network data structure, and using index rules to store the coordinates and elevation information in a topological relationship. After the triangulated network data structure is established, in order to improve data retrieval and management efficiency, the coordinates and elevation information of the irregular triangulated network of the triangulated network data structure are stored using the index rules that have been constructed.

[0074] Step S230, based on the three-dimensional modeling algorithm, the optimized data of the object is reconstructed into a three-dimensional model and the texture coordinate mapping is processed to obtain the object model. For example, the three-dimensional modeling algorithm can be 3dsmax, sketchUp or Rhino. Among them, the three-dimensional model reconstruction refers to the process of reconstructing or generating a three-dimensional model based on the existing data information. The texture coordinate mapping process is the process of mapping the two-dimensional texture image to the surface of the three-dimensional model. By establishing the correspondence between the points on the three-dimensional model and the points on the two-dimensional texture image, texture details are added to the model to enhance its sense of reality. Among them, the acquired object model is a digital representation of various objects on the earth's surface (i.e., various objects on the ground, such as buildings, roads, bridges, vegetation, water bodies, etc.).

[0075] In one example, the step of reconstructing a 3D model and performing texture coordinate mapping processing on the optimized data of the feature according to the 3D modeling algorithm to obtain the feature model includes the following steps: using index rules to retrieve and obtain feature construction element data from the optimized data of the feature; reconstructing a 3D model on the feature construction element data according to the 3D modeling algorithm to obtain a preliminary feature model; and performing texture coordinate mapping processing on the preliminary feature model to obtain the feature model. The feature construction element data includes all feature information such as buildings, roads, transportation, and river systems.

[0076] Step S240, based on the preset fusion rule mapping, the terrain scene and the object model are fused to obtain the geographical scene fusion model. The preset fusion rule mapping includes spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints. Spatial position constraints refer to the establishment of spatial position constraint rules through spatial relationships (such as adjacent, separated, etc.), absolute positioning and coordinate matching between terrain and terrain, terrain and objects, and objects and objects. Attribute information constraints refer to the use of texture coordinates to establish mapping rules and spatial processing methods to achieve the fusion of model data and texture maps, the combination of non-spatial data and models, and the combination of terrain data and image data. Semantic relationship constraints are to construct constraint rules for the spatial semantic relationship between local object objects from the perspectives of spatial attitude angle, spatial topological relationship and spatial layout, so that the object objects and terrain scenes can be accurately expressed. The optimal scale constraint is to achieve the accurate fusion of objects of different sizes and terrain by scaling different building models and scene models, as well as terrain scheduling methods. The fusion of terrain scenes and object models includes steps such as data preparation, coordinate system unification, geometry fusion, texture fusion and interactive processing.

[0077] In one example, based on the preset fusion rule mapping, the terrain scene and the object model are fused to obtain the step of the fusion model of the geographic scene, including the steps of: extracting parameters of the object model to obtain the spatial coordinates, spatial posture and scaling ratio of the object model; mapping the spatial coordinates, spatial posture and scaling ratio according to the spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints, and integrating the object model into the terrain scene. By mapping the fusion modeling rules such as spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints, the object model is positioned, loaded, accurately matched, scaled and other operations are performed on the object model to achieve accurate fusion matching between the object model and the terrain scene.

[0078] Step S250, according to the scale and scaling ratio of the geographic scene fusion model, obtain the spatial position parameters of the VR device, and send the geographic scene fusion model and the spatial position parameters to the VR device, so that the VR device displays the geographic scene fusion model according to the spatial position parameters, and realizes scene interaction according to a preset interaction method, and the preset interaction method includes a bird's-eye view interaction method, an automatic roaming interaction method and a moving walking interaction method.

[0079] The completed geographic scene fusion model is imported into the unity3D software that supports the development of VR devices, and the spatial position parameters of the VR device are adjusted according to the scale of the scene and the specific scaling ratio. The spatial position parameters of the VR device can include position coordinates, direction parameters (pitch angle, yaw angle, roll angle), displacement speed and rotation speed, etc. The VR device adaptively displays the geographic scene fusion model according to the spatial position parameters. Among them, the core operation of the bird's-eye view interaction method first needs to extract the camera's coordinates, attitude angle and spatial position and other parameters, and calculate the size of the visual field when the camera range changes in real time. Then, combined with the gravity sensor of the VR device, the obtained related parameters are dynamically calculated; finally, the calculated data is transmitted in real time to ensure that the user can move and transform freely in the scene while ensuring the real-time update of the camera's orientation, attitude angle and other parameters. The specific steps of the automatic roaming interaction method are to first plan the roaming route in the unity3D software, and set the camera's spatial position and attitude angle and other parameters at the corners and non-corners according to the route, so that the camera can smoothly change its position and attitude angle during the roaming process. The mobile walking interaction method first needs to set up collision detection between objects and terrain surfaces, and at the same time realize the walking effect according to the real-time changes of camera position parameters.

[0080] In one example, in order for a VR device to better display a geographic scene fusion model and interact with a user, Figure 5 As shown, the virtual geographic scene fusion modeling method of the present application also includes the steps of:

[0081] Step S510, calculating the data level of the feature optimization data and the viewing angle distance of the VR device, and determining the display effect corresponding to the data level according to the viewing angle distance.

[0082] The data levels of the ground feature optimization data are recorded as: LOD1, LOD2, LOD3, 、LOD n . According to the actual scene requirements, that is, the LOD data levels corresponding to the two perspectives of bird's-eye view and scene placement, the data required for the LOD data level of bird's-eye view does not need to be too precise, and the appearance of various objects can be displayed; the LOD data level of the scene placement needs to be displayed in a refined manner. Then, the range of each level of LOD data level is determined according to the calculated viewing angle distance, so as to display different levels of data. The specific calculation method is as follows:

[0083] ① Analyze the geometric characteristics of the building to determine the length, width, height and other geometric parameters of the model;

[0084] ② Determine the display mode of the building model, such as points, cuboids and fine models;

[0085] ③Determine the distance of the building model, that is, determine the resolution of the terrain in a certain range;

[0086] ④Determine the distance range of buildings under different resolution conditions;

[0087] ⑤ Determine the resolution of the current terrain visualization and determine the visualization range of its corresponding state;

[0088] ⑥ Determine the geometric display mode of the current object and establish the LOD data level of the building model.

[0089] Step S520, obtain the scene camera position of the data VR device, calculate the viewing angle parameters of the VR device according to the scene camera position, and send the viewing angle parameters to the VR device. By using the real-time change of the scene camera position of the VR device and setting the viewing angle parameters, the geographic scene data within the visual range of the VR device is dynamically loaded and displayed, and the data outside the viewing angle of the VR device is ignored and not rendered, so as to reduce the performance problems caused by massive data and improve the efficiency of scene loading.

[0090] Step S530 configures the bird's-eye view interaction mode, the automatic roaming interaction mode and the mobile walking interaction mode according to the geographic scene fusion model, spatial position parameters, display effect and viewing angle parameters. VR technology performs a full range of multi-sensory visualization of geographic scenes, while taking into account the immersive characteristics of VR visualization, and through technologies such as real-time tracking of targets, scene LOD grading and camera cone clipping, it improves the user's immersive experience effect and enriches the visualization methods and means of geographic scenes.

[0091] The virtual geographic scene fusion modeling method of the present application is carried out through the following steps: obtaining a training sample data set, optimizing the training sample data set based on a support vector machine algorithm, and obtaining training sample optimization data; using an irregular triangulated network model to process terrain optimization data, and constructing the terrain optimization data into a triangulated network data structure to obtain a terrain scene; reconstructing a three-dimensional model and mapping texture coordinates on the object optimization data according to a three-dimensional modeling algorithm to obtain an object model; based on a preset fusion rule mapping, fusion of the terrain scene and the object model to obtain a geographic scene fusion model; according to the scale and scaling of the geographic scene fusion model, obtaining the spatial position parameters of the VR device, and sending the geographic scene fusion model and the spatial position parameters to the VR device, so that the VR device displays the geographic scene fusion model according to the spatial position parameters, and realizes scene interaction according to a preset interaction method. The present application adopts a machine learning method to efficiently classify, simplify and extract features of massive three-dimensional data, thereby extracting effective scene modeling data, greatly reducing data redundancy, and improving modeling efficiency. The spatial constraint relationship is used to fuse and model the geographic scene to break through the existing modeling method, and under the support of VR technology, a diversified interaction mode of the geographic scene and a visual expression at multiple scales are realized.

[0092] With the support of machine learning technology, this application is supplemented by relevant technologies and theories in the field of virtual reality to quickly model and visualize massive three-dimensional geographic scenes, achieving the advantages of fast modeling speed, high algorithm fault tolerance, and visualization efficiency. Through machine learning, massive geographic scenes are automatically modeled while taking into account multi-scale, multi-dimensional and multi-constraint conditions, and visualized through virtual reality technology to enhance the immersive effect of the scene and improve the user experience. The accuracy and fault tolerance of the classification of various types of landforms in geographical scenes are improved through machine learning. Through VR visualization technology, the appearance is improved from the traditional three-dimensional scene perspective to an all-round multi-perceptual immersive virtual visualization perspective, which is richer in appearance, more layered, and the user experience has been greatly improved.

[0093] It should be understood that although Figures 2 to 5 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figures 2 to 5 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0094] In one embodiment, a virtual geographic scene fusion modeling device is provided, comprising:

[0095] A data processing module is used to obtain a training sample data set, optimize the training sample data set based on a support vector machine algorithm, and obtain training sample optimized data; the training sample optimized data at least includes terrain optimized data and ground feature optimized data;

[0096] A terrain scene acquisition module, used for processing the terrain optimization data using an irregular triangulated network model, and constructing the terrain optimization data into a triangulated network data structure to acquire a terrain scene;

[0097] A ground object model acquisition module is used to perform three-dimensional model reconstruction and texture coordinate mapping processing on the ground object optimization data according to the three-dimensional modeling algorithm to obtain the ground object model;

[0098] A fusion module is used to fuse the terrain scene and the object model based on a preset fusion rule mapping to obtain a geographic scene fusion model; the preset fusion rule mapping includes spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints;

[0099] The parameter configuration module is used to obtain the spatial position parameters of the VR device according to the scale and scaling ratio of the geographic scene fusion model, and send the geographic scene fusion model and the spatial position parameters to the VR device so that the VR device displays the geographic scene fusion model according to the spatial position parameters.

[0100] For the specific definition of the virtual geographic scene fusion modeling device, please refer to the definition of the virtual geographic scene fusion modeling method above, which will not be repeated here. The various modules in the above-mentioned virtual geographic scene fusion modeling device can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0101] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 6As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as original data, training sample data sets, and training sample optimization data. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement a virtual geographic scene fusion modeling method.

[0102] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0103] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0104] Obtain a training sample data set, optimize the training sample data set based on a support vector machine algorithm, and obtain training sample optimized data; the training sample optimized data at least includes terrain optimized data and ground feature optimized data;

[0105] The terrain optimization data is processed using the irregular triangulated network model, and the terrain optimization data is constructed into a triangulated network data structure to obtain the terrain scene;

[0106] Reconstruct the three-dimensional model and process the texture coordinate mapping of the optimized data of the ground object according to the three-dimensional modeling algorithm to obtain the ground object model;

[0107] Based on the preset fusion rule mapping, the terrain scene and the object model are fused to obtain a geographic scene fusion model; the preset fusion rule mapping includes spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints;

[0108] According to the scale and scaling ratio of the geographic scene fusion model, the spatial position parameters of the VR device are obtained, and the geographic scene fusion model and the spatial position parameters are sent to the VR device, so that the VR device displays the geographic scene fusion model according to the spatial position parameters and realizes scene interaction according to a preset interaction method, which includes a bird's-eye view interaction method, an automatic roaming interaction method, and a moving walking interaction method.

[0109] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0110] Obtain a training sample data set, optimize the training sample data set based on a support vector machine algorithm, and obtain training sample optimized data; the training sample optimized data at least includes terrain optimized data and ground feature optimized data;

[0111] The terrain optimization data is processed using the irregular triangulated network model, and the terrain optimization data is constructed into a triangulated network data structure to obtain the terrain scene;

[0112] Reconstruct the three-dimensional model and process the texture coordinate mapping of the optimized data of the ground object according to the three-dimensional modeling algorithm to obtain the ground object model;

[0113] Based on the preset fusion rule mapping, the terrain scene and the object model are fused to obtain a geographic scene fusion model; the preset fusion rule mapping includes spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints;

[0114] According to the scale and scaling ratio of the geographic scene fusion model, the spatial position parameters of the VR device are obtained, and the geographic scene fusion model and the spatial position parameters are sent to the VR device, so that the VR device displays the geographic scene fusion model according to the spatial position parameters, and realizes scene interaction according to the preset interaction mode, and the preset interaction mode includes a bird's-eye view interaction mode, an automatic roaming interaction mode, and a mobile walking interaction mode. A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0115] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0116] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent application shall be subject to the attached claims.

Claims

1. A virtual geographic scene fusion modeling method, characterized in that: The following steps are involved: Obtaining a training sample data set, and optimizing the training sample data set based on a support vector machine algorithm to obtain training sample optimized data; The training sample optimization data at least includes terrain optimization data and ground feature optimization data; Processing the terrain optimization data using an irregular triangulated network model, and constructing the terrain optimization data into a triangulated network data structure to obtain a terrain scene; Reconstructing the three-dimensional model and performing texture coordinate mapping processing on the optimized ground object data according to the three-dimensional modeling algorithm to obtain a ground object model; Based on a preset fusion rule mapping, the terrain scene and the object model are fused to obtain a geographic scene fusion model; the preset fusion rule mapping includes spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints; According to the scale and scaling ratio of the geographic scene fusion model, the spatial position parameters of the VR device are obtained, and the geographic scene fusion model and the spatial position parameters are sent to the VR device, so that the VR device displays the geographic scene fusion model according to the spatial position parameters, and realizes scene interaction according to a preset interaction mode, wherein the preset interaction mode includes a bird's-eye view interaction mode, an automatic roaming interaction mode, and a moving walking interaction mode; The step of obtaining a training sample data set and optimizing the training sample data set based on a support vector machine algorithm includes the following steps: Collecting raw data, and extracting the training sample data set from the raw data; the raw data includes manually surveyed terrain raster data, high-resolution satellite image data and ground object model data; the training sample data set includes at least terrain data and ground object data; Based on preset feature extraction rules, feature extraction is performed on the training sample data set to obtain feature extraction results, and a feature space is constructed according to the feature extraction results; the feature extraction results include texture features, structural features, elevation features, scale features, semantic features, location features and attribute features; Constructing a hyperplane on the feature space according to the training sample data set and the feature space; The hyperplane is used to test the training sample data set. If the data in the training sample data set meets the classification rules of the hyperplane, it is retained; otherwise, the support vector machine algorithm is used to reassign the data in the training sample data set that does not meet the classification rules of the hyperplane, and features are extracted from the data in the reassigned training sample data set. The corresponding hyperplane is constructed and data testing is performed until all data in the training sample data set meets the classification rules of the hyperplane to obtain the training sample optimized data.

2. The virtual geographic scene fusion modeling method according to claim 1 is characterized in that: Also includes the steps: According to the feature extraction result, the training sample optimization data is classified to obtain a data classification result; the data classification result includes spatial position information, attribute data, spatial semantic data, structural morphology, texture coordinates and elevation value; Based on the feature extraction result and the data classification result, an index rule is constructed for the training sample optimization data, and based on the index rule, the training sample optimization data is topologically stored.

3. The virtual geographic scene fusion modeling method according to claim 2 is characterized in that: The step of constructing index rules for the training sample optimization data based on the feature extraction results and the data classification results, and storing the training sample optimization data in a topological relationship based on the index rules, comprises the steps of: Configuring corresponding indexes for the feature extraction result and the data classification result respectively; The index is checked in the training sample optimization data. If the content pointed to is unique, the preliminary index is retained. If the content pointed to is not unique, the index is changed and the changed index is checked until the content pointed to by the index is unique, so as to construct the index rule.

4. The virtual geographic scene fusion modeling method according to claim 2 is characterized in that: The step of processing the terrain optimization data using an irregular triangulated network model and constructing the terrain optimization data into a triangulated network data structure to obtain a terrain scene also includes the following steps: The coordinates and elevation information of the irregular triangulated network of the triangulated network data structure are obtained, and the coordinates and the elevation information are stored in a topological relationship using the index rule.

5. The virtual geographic scene fusion modeling method according to claim 2 is characterized in that: The step of performing three-dimensional model reconstruction and texture coordinate mapping processing on the optimized ground object data according to a three-dimensional modeling algorithm to obtain a ground object model comprises the following steps: Using the index rule, the ground feature construction element data is retrieved from the ground feature optimization data; Reconstructing the three-dimensional model of the ground object construction element data according to the three-dimensional modeling algorithm to obtain a preliminary ground object model; The preliminary ground object model is subjected to texture coordinate mapping processing to obtain the ground object model.

6. The virtual geographic scene fusion modeling method according to claim 1, characterized in that: Based on the preset fusion rule mapping, the terrain scene and the object model are fused to obtain a geographical scene fusion model, including the steps of: Extracting parameters of the object model to obtain the spatial coordinates, spatial posture and scaling ratio of the object model; According to the spatial position constraint, the attribute information constraint, the semantic relationship constraint and the optimal scale constraint, the spatial coordinates, the spatial posture and the scaling ratio are mapped respectively, and the terrain model is integrated into the terrain scene.

7. The virtual geographic scene fusion modeling method according to claim 1 is characterized in that: Also includes the steps: Calculating the data level of the feature optimization data and the viewing angle distance of the VR device, and determining the display effect corresponding to the data level according to the viewing angle distance; Acquire the scene camera position of the VR device of the data, calculate the viewing angle parameter of the VR device according to the scene camera position, and send the viewing angle parameter to the VR device; The bird's-eye view interaction mode, the automatic roaming interaction mode and the mobile walking interaction mode are configured according to the geographic scene fusion model, the spatial position parameters, the display effect and the viewing angle parameters.

8. A virtual geographic scene fusion modeling device, characterized in that: The device comprises: A data processing module is used to obtain a training sample data set, optimize the training sample data set based on a support vector machine algorithm, and obtain training sample optimized data; the training sample optimized data at least includes terrain optimized data and ground feature optimized data; A terrain scene acquisition module, used to process the coordinate and elevation information in the terrain optimization data using an irregular triangulated network model, and construct the terrain optimization data into a triangulated network data structure to acquire a terrain scene; A ground object model acquisition module is used to perform three-dimensional model reconstruction and texture coordinate mapping processing on the ground object optimization data according to a three-dimensional modeling algorithm to obtain a ground object model; A fusion module, used to fuse the terrain scene and the object model based on a preset fusion rule mapping to obtain a geographic scene fusion model; the preset fusion rule mapping includes spatial position constraints, attribute information constraints, semantic relationship constraints and optimal scale constraints; A parameter configuration module is used to obtain the spatial position parameters of the VR device according to the scale and scaling ratio of the geographic scene fusion model, and send the geographic scene fusion model and the spatial position parameters to the VR device, so that the VR device displays the geographic scene fusion model according to the spatial position parameters.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images