A Method for Constructing 3D Wargaming Maps Based on DEM and Remote Sensing Land Feature Classification

Abstract

This invention proposes a method for constructing 3D wargaming maps based on DEM and remote sensing feature classification, comprising the following steps: acquiring Digital Elevation Model (DEM) data, performing block-level and layer-level processing using a two-level quadtree index encoding, and converting the data; constructing a regular grid model based on the transformed DEM data; using the regular grid model and hardware for terrain rendering to generate 3D terrain; performing terrain texture mapping on the obtained 3D terrain based on remote sensing feature classification to locate and map the textures of specific landforms on the terrain; performing texture blending processing on the 3D terrain after terrain texture mapping based on a multi-channel blended color map (Splatmap) to simulate the natural transition between terrain textures in a real scene; furthermore, this invention combines a multi-threading mechanism and an object pool strategy to achieve dynamic scheduling and real-time rendering. Compared with existing technologies, this invention has the advantages of high efficiency in 3D terrain generation and realistic texture mapping.

Description

Technical Field

[0001] This invention belongs to the field of computer application technology, and in particular relates to a method for constructing a three-dimensional wargaming map based on DEM and remote sensing feature classification. Background Technology

[0002] With the diversification of modern warfare requirements, the limitations of wargaming maps based on traditional battle scenarios in tactical analysis and military training are becoming increasingly apparent. Currently, most wargaming systems still use two-dimensional maps, primarily because they are relatively simple to create and easy to understand and operate. However, due to the excessive information elements in campaign maps, two-dimensional wargaming maps still fall short in terms of spatial representation and immersion. They cannot fully represent terrain elevation and slope, and their extraction and processing of key terrain information from actual battles are limited, making it difficult to simulate real battlefield environments. This results in low map accuracy and is detrimental to spatial warfare simulation. In contrast, three-dimensional maps can reproduce the three-dimensional structure of terrain in detail, providing richer perspectives and terrain features, thus more effectively simulating the battlefield environment, enhancing battlefield realism, and more accurately reflecting the impact of terrain on tactics and strategy. This allows trainees to more intuitively understand real-time battlefield changes and the complexity of decision-making. With the advancement of virtual reality and augmented reality technologies, three-dimensional wargaming systems will play an increasingly important role in the field of military simulation, providing richer and more realistic experiences for military training and education.

[0003] However, the current 3D terrain generation faces challenges such as difficulty in data acquisition, complexity in processing, and unrealistic texture mapping. There is currently little research on 3D wargame map generation, and the utilization of height information is insufficient. Most research focuses on 3D terrain generation and terrain data processing.

[0004] Commonly used 3D terrain generation methods mainly consist of terrain generation techniques based on data fitting, terrain generation techniques based on fractal technology, and terrain generation techniques based on real elevation data. Since this paper requires generating 3D terrain of classic battle examples, it necessitates the use of real geographic information. Therefore, terrain generation techniques based on elevation data are more suitable. Digital Elevation Models (DEMs), after digital terrain analysis, can intuitively display the undulations of the Earth's surface. Currently, DEMs have become important spatial information data in national basic geographic databases. Two commonly used representation models for DEMs are regular grid models and irregular triangular mesh models. Irregular triangular mesh models use irregularly distributed elevation points to form triangular meshes. Although this model more closely approximates real terrain, its structure is complex, and storage typically requires recording the topological relationships of points, edges, and triangles, which is detrimental to data processing efficiency and subsequent terrain management and application. In contrast, regular grid models are simpler and easier to use, and many researchers use regular grid models for 3D terrain generation. For example, some scholars combine low-resolution digital elevation models with aerial images to generate realistic terrain; others merge DEMs with satellite images to generate three-dimensional terrain.

[0005] Considering the increasing scale of terrain and the often wide scope of battles, researchers both domestically and internationally have begun to study terrain simplification techniques to ensure rendering quality and speed. For example, some researchers have proposed continuous Level of Depth (LOD) detail simplification algorithms, using variable screen thresholds to constrain the maximum geometric error of triangle vertices; others have improved this method by using quadtrees for node subdivision; still others have proposed the static LOD algorithm ChunkLOD, which processes data at each level during preprocessing and directly reads it during rendering, eliminating the need for dynamic calculations; in addition, there are methods such as using dual threads to implement quadtree-based LOD terrain simplification algorithms and utilizing quadtrees and fill curves to manage massive terrain data. These studies indicate that for the generation of realistic terrain, there is a growing research focus on using quadtree terrain organization combined with LOD terrain simplification algorithms to simplify large-scale terrain blocks and optimize scene rendering.

[0006] In the rendering of large-scale terrain scenes in 3D maps, terrain mesh models can display surface undulations, but are insufficient to showcase detailed landform features. In 1974, scholars proposed the concept of texture mapping, which greatly enhanced the visual richness of graphic objects and was of great significance for the generation of realistic maps. In 1976, two more scholars proposed the "reflection mapping" technique, considering the specular reflection parameters of texture images and lighting mapping techniques in the lighting model, thus significantly improving the image quality after texture mapping. In 1978, scholars proposed bump mapping, using a perturbation function to change the normal vector of the model surface to simulate the texture of a rough surface. Texture mapping is also used in 3D map modeling; mapping terrain textures onto terrain surfaces is an important application in 3D map generation. Texture mapping is a crucial technique for enhancing the realism of 3D terrain, enriching surface details by attaching texture images to the terrain mesh model. The core of texture mapping is establishing a match between terrain mesh points and texture coordinates, ensuring the stability of the texture mapping relationship when the terrain model deforms.

[0007] While significant progress has been made in research on wargaming systems, map element extraction and generation, and 3D terrain generation, research on generating 3D wargaming maps based on classic battle scenario maps remains limited due to the diversity and complexity of map information elements, as well as issues related to rendering quality and speed in optimizing 3D terrain generation. Therefore, an efficient and accurate method for generating 3D wargaming maps is urgently needed. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention proposes a method for constructing 3D wargaming maps based on DEM and remote sensing feature classification. Firstly, this invention creatively employs a DEM-based method to generate 3D terrain, providing a foundation for the quantification of terrain features in 3D maps. Secondly, it uses a texture blending method and feature generation to simulate the environment of a real battlefield. Finally, it proposes a data processing and terrain simplification method based on a two-level quadtree index, optimizing the speed and smoothness of large-scale terrain loading. This resolves the contradiction between high accuracy, realism, and immersion in 3D maps and the real-time efficiency and data requirements of the model, enabling rapid and accurate reconstruction of 3D simulation scenarios of classic battle situations.

[0009] To achieve the above-mentioned technical objectives, the present invention provides the following technical solution:

[0010] A method for constructing a 3D wargame map based on DEM and remote sensing feature classification, specifically including:

[0011] S1. Acquire digital elevation model (DEM) data, and perform block-level and layer-level processing and data transformation on it; the DEM data is a terrain data array containing terrain elevation and undulation information, stored in raster format;

[0012] S2. Construct a regular grid model based on the transformed DEM data; then use the regular grid model and hardware to perform terrain rendering to generate 3D terrain.

[0013] S3. Based on remote sensing land cover classification, perform terrain texture mapping on the three-dimensional terrain obtained in step S2 to locate and map the texture of specific landforms on the terrain.

[0014] S4. Based on the multi-channel blending color map Splatmap, the 3D terrain that has been mapped by terrain texture is processed to blend different textures, simulating the natural transition between terrain textures in real scenes.

[0015] S5. Test by adjusting the channel values ​​of Splatmap to affect the color representation of the final pixel o until the texture blending effect can be effectively achieved; otherwise, return to step S4.

[0016] Furthermore, the block-based and layered processing in step S1 involves dividing the DEM data into multiple blocks using a two-level quadtree index encoding method to ensure effective data management and improve processing efficiency. Specifically, this includes the following steps:

[0017] The block-based and layered processing in step S1 involves using a two-level quadtree index encoding method to divide the DEM data into layers and blocks to ensure effective data management and improve processing efficiency. Specifically, it includes the following steps:

[0018] 1) Large-scale terrain data in DEM data is segmented into blocks using vertical and horizontal partitioning methods to achieve the first-level encoding of the data block index: vertical-horizontal encoding. This involves setting the data block as the origin and dividing the terrain along the vertical and horizontal axes. An index code is then established for each segmented terrain data block, enabling rapid retrieval and access to the terrain data. The formula is expressed as:

[0019]

[0020]

[0021] Where (X0, Y0) are the position coordinates of the target region P; (C p ,R p ) represents the index coordinates of the data block within the entire terrain data. The size of the terrain block is l×l, where l is an odd number.

[0022] 2) Each data block independently constructs a two-level quadtree based on the traditional detail evaluation criteria, and only constructs it in real time when terrain rendering is required, so as to reduce the amount of computation; the data block starts with the origin as the local coordinate and is adjusted to the actual position during terrain rendering.

[0023] 3) Use a two-dimensional matrix to label the nodes in the two-level quadtree structure; depending on whether the nodes need to be subdivided, different labels are applied to form a hierarchical detail model with multi-resolution attributes, thereby realizing the layering and segmentation of terrain data.

[0024] Furthermore, in step S1, the data transformation process specifically includes:

[0025] S11. Using the Geospatial Data Abstraction Library (GDAL), the raster data in the DEM data is converted into latitude and longitude coordinates and elevations of the actual location on Earth in the WGS-84 system through geographic affine transformation coefficients.

[0026] S12. After projection calculation, the latitude and longitude coordinates and elevation are converted into three-dimensional rectangular coordinates in the Cartesian coordinate system; the Cartesian coordinate system, namely the geocentric rectangular coordinate system, has the Earth's center as its origin, with the positive direction of the X-axis pointing to the intersection of the equator and the prime meridian; the Z-axis is perpendicular to the equatorial plane, with its positive direction pointing to the North Pole; the formula for the conversion process is expressed as:

[0027]

[0028] Where (B, L, H) are the latitude, longitude and elevation of a point in the WGS-84 coordinate system; (X, Y, Z) are the corresponding three-dimensional coordinates of the point in the Cartesian coordinate system; N is the meridian radius and e is the first eccentricity of the Earth ellipsoid.

[0029] S13. Through translation and rotation transformations, the Cartesian coordinate system in step S12 is transformed into the scene coordinate system under the terrain rendering scene to obtain the three-dimensional coordinates under the scene coordinate system.

[0030] More specifically, the translation and rotation transformations in step S13 are as follows:

[0031] The translation transformation specifically refers to:

[0032] The formula for translating a point (X,Y,Z) in the original geocentric rectangular coordinate system to a point (X′,Y′,Z′) in the scene coordinate system is as follows:

[0033] X′=X+T x ;

[0034] Y′=Y+T y ;

[0035] Z′=Z+T z ;

[0036] Among them, T x ,T y ,T z The position of the starting point relative to the center of the earth in the terrain rendering scene;

[0037] The rotational transformation specifically refers to:

[0038] When the coordinate axes in the scene coordinate system do not completely coincide with the coordinate axes of the geocentric rectangular coordinate system, a rotation transformation is performed. Let the scene coordinate system be rotated by angles θ around the x, y, and z axes relative to the geocentric coordinate system. x ,θ y ,θ z Then the formula for the rotation transformation process is expressed as:

[0039] Rotation matrix around the x-axis:

[0040]

[0041] Rotation matrix around the y-axis:

[0042]

[0043] Rotation matrix around the z-axis:

[0044]

[0045] Transform the point (X′, Y′, Z′) sequentially using the three matrices described above:

[0046]

[0047] Where (X″,Y″,Z″) are the three-dimensional rectangular coordinates after translation and rotation transformations, i.e., mapped to the correct position.

[0048] Further, step S2 includes:

[0049] S21. Regular grid definition and grid vertex determination; Define an n×m DEM data regular grid, where each grid represents a region in the actual terrain; the three-dimensional coordinates of the grid vertices are (X... i,j ,Y i,j Z i,j );

[0050] X i,j =x origin +i·Δx

[0051] Y i,j =y origin +j·Δy;

[0052] Z i,j =DEM(i,j)

[0053] Where i and j represent the horizontal and vertical indices, respectively; x origin ,y origin These are the origin coordinates of the terrain region; Δx and Δy are the grid intervals; DEM(i,j) are the elevation values ​​in the DEM data;

[0054] For the entire terrain area, traverse all DEM data points and read out each grid vertex as described above. Each vertex corresponds to a point on the terrain surface.

[0055] S22. Assign texture coordinates to each vertex and map the texture onto the terrain surface; the texture coordinates are defined by the size of the DEM data regular grid; assuming the grid size is n×m, assign texture coordinates (u) to each vertex proportionally. i,j v i,j The formula is expressed as:

[0056]

[0057]

[0058] Where i and j represent the horizontal and vertical indices, respectively;

[0059] S23. Divide four adjacent grid vertices into two triangles and generate triangle indices; denote the four adjacent vertices as P. i,j ,P i+1,j ,P i,j+1 ,P i+1,j+1 Divide it into two triangles:

[0060] Triangle 1: P i,j ,P i+1,j ,P i,j+1 ;

[0061] Triangle 2: P i+1,j ,P i,j+1 ,P i+1,j+1 ;

[0062] The vertex indices of each triangle are:

[0063] Triangle 1=(i,j),(i+1,j),(i,j+1);

[0064] Triangle 2=(i+1,j),(i,j+1),(i+1,j+1);

[0065] This process involves traversing all vertex indices throughout the entire grid, ultimately generating triangle indices for all vertices.

[0066] S24. Define the connection relationship between vertices through triangle indices to ensure the correct connection and rendering of the mesh; the connection relationship is: each vertex is connected through a shared triangle edge, and the triangle indices of the mesh are sorted in counterclockwise order;

[0067] S25. Combine hardware rendering, apply basic texture mapping and lighting effects to generate 3D terrain.

[0068] More specifically, in step S1:

[0069] During the definition of the regular mesh, for the acquired DEM data, the sampling points and interpolation points are connected to form a terrain surface to approximate the original surface. The formula is as follows:

[0070] DEM = {M i =ξ(P j )|P j (x j ,y j H j )∈D,j=1,2,...,n; i=1,2,...,m}

[0071] Where n and m represent the size of the regular mesh model as n×m; P j M represents the sampling point and the interpolation point. i It is a connected surface, x j y j H j These are the points corresponding to the DEM, x j y j H represents the x and y coordinates of the grid vertices. j ξ represents the elevation value of the DEM data, and ξ represents the generation rule of the regular grid model.

[0072] Furthermore, step S3 specifically includes:

[0073] For the three-dimensional terrain obtained in step S2, different types of mask images obtained from the semantic segmentation results of remote sensing images are used to locate and map the texture of specific landforms on the terrain, thereby effectively restoring the landforms of historical battles in the three-dimensional terrain.

[0074] Furthermore, step S4 specifically involves:

[0075] A multi-channel blending color map (Splatmap) is used to achieve blending between different terrain textures, simulating the natural transition between textures in a real scene. The color map (Splatmap) is a four-channel RGBA color image that records the texture mapping of the terrain, with each channel representing a different texture map.

[0076] When blending two texture maps, the blended pixels are obtained through channel interpolation, expressed by the formula:

[0077] o = a * alpha+b * (1-alpha);

[0078] Where o is the color of the blended pixel, a and b are the colors of the corresponding pixels in the two texture maps respectively; alpha is the color value of a certain channel of the pixel in the color map Splatmap, which is used to adjust the blending ratio between the two texture maps.

[0079] Furthermore, a multi-threaded mechanism and object pooling strategy are employed to achieve dynamic scheduling of terrain tiles and real-time scene rendering, specifically including:

[0080] Multithreading is used to perform view volume clipping, while the main thread is responsible for drawing terrain tiles and loading and scheduling them.

[0081] During terrain rendering, the object pool is responsible for managing the storage, rendering, and destruction of terrain tile objects; it sets up a pre-storage table, a pending disposal table, and a destruction queue; the pre-storage table is used to store reactivated terrain tile objects; the pending disposal table stores unloaded terrain tile objects; and the destruction queue is responsible for handling terrain tile objects that need to be destroyed.

[0082] During each frame update, the system first checks the objects in the pre-storage table. If they meet the current rendering conditions, they are moved to the activation queue. Each frame, a batch of terrain tiles is activated according to a set number. For new terrain tiles that need to be rendered, if they are not in the pre-storage table, they are added to the rendering queue. At the same time, the pre-storage table is updated, including terrain tiles not in the rendering queue and terrain tiles with changes in level of detail. If the pre-storage table is full, the oldest objects are removed to the discard table. Finally, a certain number of objects in the queue are destroyed to maintain a smooth frame rate and avoid sudden drops in frame rate.

[0083] When the viewpoint moves, an improved AABB bounding box view volume clipping method is used for optimization; the clipping rules of the improved AABB bounding box are as follows:

[0084] (1) For child nodes of a terrain patch, if they fall entirely within the current view area, the node is considered visible and its visibility attribute is set to 1; based on the node evaluation criteria, determine whether the node needs to be further subdivided. If subdivision is required, the four child nodes of the node will inherit its visibility attribute, which is also set to true;

[0085] (2) For nodes that are completely outside the view volume, they are considered invisible nodes and will be excluded directly from the rendering process, and a clipping operation will be performed;

[0086] (3) If a node only partially overlaps with the view volume, meaning that part of the terrain block extends outside the view volume, it will still be set to visible; and it is assumed that the user pays less attention to these parts, so these nodes will not be further subdivided. Their previous state will remain unchanged.

[0087] Based on the above technical solution, the method proposed in this invention has the following beneficial effects:

[0088] 1. The method proposed in this invention uses DEM data containing terrain elevation and undulation information, which provides a basis for the quantification of three-dimensional terrain. Through multiple coordinate system transformations, the terrain information carried by the DEM data can adapt to the rendering requirements of three-dimensional scenes, while ensuring the accurate positioning of the data in three-dimensional space.

[0089] 2. The regular grid model used in this invention supports a hierarchical structure of a two-level quadtree, making adjacency relationships and topology clearer and easier to process. The combination of these two features also facilitates terrain segmentation and dynamic scheduling operations in the overall model. The regular grid model not only improves data processing efficiency but also optimizes subsequent terrain management and application processes; the two-level quadtree index encoding effectively realizes hierarchical management and display of terrain data, significantly improving indexing speed.

[0090] 3. By using a multi-threaded mechanism and object pool strategy, combined with two-level quadtree encoding, the overall loading speed and rendering smoothness are optimized, making real-time interaction possible, reducing algorithm time consumption, and improving the terrain loading speed and frame rate of 3D terrain generation.

[0091] 4. By using remote sensing land feature classification and color mapping Splatmap, the differences between different textures of the obtained 3D terrain are not too obvious, which is more in line with the natural transition in real nature; making the 3D scene more visually attractive and immersive, suitable for military training and simulation applications, and improving the realism of 3D terrain rendering. Attached Figure Description

[0092] Figure 1 This is the overall technical roadmap of this method;

[0093] Figure 2 This is a schematic diagram of a regular mesh model;

[0094] Figure 3 This is a flowchart of the 3D terrain generation process;

[0095] Figure 4 It is a 3D terrain generation result map;

[0096] Figure 5 These are the masked results for woodland, grassland, and cultivated land;

[0097] Figure 6It is a texture mapping result of a 3D scene;

[0098] Figure 7 It is a terrain splatmap;

[0099] Figure 8 This is a comparison image of the texture effects before and after blending;

[0100] Figure 9 This is a schematic diagram of terrain data block processing;

[0101] Figure 10 This is a schematic diagram of a two-level quadtree index encoding;

[0102] Figure 11 This is a schematic diagram of a two-dimensional matrix label;

[0103] Figure 12 This is a schematic diagram of a multi-threaded terrain data processing strategy;

[0104] Figure 13 This is a schematic diagram of DEM data. Detailed Implementation

[0105] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0106] Although the steps in this invention are arranged by reference numerals, this is not intended to limit the order of the steps. Unless the order of the steps is explicitly stated or the execution of a step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" as used herein refers to and covers any and all possible combinations of one or more of the associated listed items.

[0107] like Figure 1 As shown, this invention proposes a method for constructing a 3D wargaming map based on DEM and remote sensing feature classification, specifically including the following steps:

[0108] S1. Acquire digital elevation model (DEM) data, and perform block-level and layer-level processing and data transformation on it; the DEM data is a data array containing terrain elevation and undulation information and is stored in raster format.

[0109] In a preferred embodiment, the block-level processing in step S1 is specifically as follows:

[0110] 1) Large-scale terrain data in the DEM data is divided into blocks using vertical and horizontal segmentation methods to implement the first-level encoding of the data block index: vertical-horizontal encoding; such as... Figure 9As shown, a 257×257 database was chosen as the partitioning standard. If the elevation data value occupies 1 byte, a single data block is approximately 64KB; if it occupies two bytes, it is approximately 128KB. This data block size allows for real-time scheduling and caching strategies, enabling real-time terrain rendering and effectively reducing memory usage. The default starting coordinates of the data block are set at the lower left corner (0, 0), and the terrain is systematically divided along the vertical and horizontal axes. An index is created for each partitioned data block, enabling fast retrieval and access to terrain data and improving data processing efficiency; the formula is expressed as:

[0111]

[0112]

[0113] Where (X0, Y0) are the position coordinates of the target region P; (C p ,R p ) represents the index coordinates of the data block within the entire terrain block, where the size of the terrain block is l×l, and l is an odd number;

[0114] 2) Each terrain block independently constructs a two-level quadtree based on traditional detail evaluation criteria, and this construction is only performed in real-time when terrain rendering is required, thus reducing computational load; the data blocks start with the origin as their local coordinates, which are adjusted to their actual positions during terrain rendering; a schematic diagram of the two-level quadtree encoding is shown below. Figure 10 As shown, the data block partitioning and the quadtree structure of the partial blocks are illustrated.

[0115] In this embodiment, traditional detailed evaluation criteria include: region segmentation, storage efficiency, query performance, dynamic updates, and visualization. These will be elaborated upon below:

[0116] Region segmentation: When segmenting, choose a reasonable segmentation criterion, such as based on the distribution of data points or the characteristics of the region (such as density).

[0117] Storage efficiency: By storing node information reasonably and reducing the storage of unnecessary empty nodes, memory can be used effectively.

[0118] Query performance: Supports efficient query operations, such as range queries and nearest neighbor searches, and can return results with low time complexity.

[0119] Dynamic updates: Supports inserting and deleting nodes, and can adjust the structure without rebuilding the entire tree.

[0120] Visualization: It can support the visualization of regions, making the relationship between different regions and data more intuitive.

[0121] 3) Use a two-dimensional matrix to label the nodes in the two-dimensional quadtree structure, such as... Figure 11As shown, whether a node is subdivided depends on the calculation results of a specific evaluation criterion. If a node needs to be subdivided, its center point is marked as 1; otherwise, it is marked as 0. The question mark indicates that the node has not yet been traversed. After this process is completed, a hierarchical detail model with multi-resolution attributes is formed, effectively realizing the hierarchical management and display of terrain data.

[0122] With such a two-dimensional matrix for each terrain block, it is possible to quickly determine whether locally adjacent terrain blocks are at the same level based on the markings, which makes it easier to determine whether there are cracks between adjacent blocks, thereby saving search time and facilitating the elimination of cracks within blocks.

[0123] In this embodiment, the quadtree hierarchical algorithm is used to optimize data storage. That is, after acquiring the DEM data, in order to optimize computational efficiency and data processing, the DEM data is divided into layers and blocks using a two-level quadtree index encoding method, constructing a quadtree structure and storing it through a Boolean matrix. This can accelerate data retrieval, thereby effectively processing and rendering large-scale terrain data. In addition, in this embodiment, the terrain block is spatial, and it corresponds one-to-one with the data block. When dividing the terrain block, its corresponding terrain data is also divided. When processing the terrain data, it is processing the data block, which also corresponds to processing the terrain block in a spatial sense.

[0124] Furthermore, as an implementation method, such as Figure 3 As shown, the data transformation in step S1 further includes:

[0125] S11. Using the Geospatial Data Abstraction Library (GDAL), the raster data in the DEM data is converted into latitude and longitude coordinates and elevation of the actual location on Earth in the WGS-84 system, i.e., the geographic coordinate system, through geographic affine transformation coefficients.

[0126] In this embodiment, DEM data is processed by GDAL in the form of a dataset raster data. The information in the dataset mainly consists of raster bands, metadata, affine transformation coefficients, and a color table. The metadata includes information such as image time, resolution, sampling method, and color space, and is typically stored in GDAL as key-value pairs. The geographic affine transformation coefficients can convert raster coordinates into latitude, longitude, and altitude coordinates of the actual location on Earth in the WGS-84 system.

[0127] S12. After projection calculation, the latitude and longitude coordinates and elevation are converted into three-dimensional rectangular coordinates in the Cartesian coordinate system. This conversion process involves calculating parameters such as the meridian radius and the first eccentricity of the ellipsoid to ensure accurate positioning of the data in three-dimensional space. The Cartesian coordinate system, also known as the geocentric rectangular coordinate system, has the Earth's center as its origin. The positive direction of the X-axis points to the intersection of the equator and the primary meridian, representing the east-west direction on the Earth's surface. The Z-axis is perpendicular to the equatorial plane, with its positive direction pointing to the North Pole, symbolizing the north-south direction of the Earth. The Y-axis is orthogonal to the X-axis and Z-axis and lies in the equatorial plane, completing the construction of the three-dimensional coordinate system. The formula for the conversion process is expressed as:

[0128]

[0129] Where (B, L, H) are the latitude, longitude and elevation of a point in the WGS-84 coordinate system; (X, Y, Z) are the corresponding three-dimensional coordinates of the point in the Cartesian coordinate system; N is the meridian radius and e is the first eccentricity of the Earth ellipsoid.

[0130] S13. Through translation and rotation transformations, the Cartesian coordinate system in step S12 is transformed into the scene coordinate system under the terrain rendering scene to obtain the three-dimensional coordinates in the scene coordinate system. This process ensures that the data is correctly mapped to the grid points of the scene coordinate system.

[0131] The translation transformation specifically refers to:

[0132] The formula for translating a point (X,Y,Z) in the original geocentric rectangular coordinate system to a point (X′,Y′,Z′) in the scene coordinate system is as follows:

[0133] X′=X+T x ;

[0134] Y′=Y+T y ;

[0135] Z′=Z+T z ;

[0136] Among them, T x ,T y ,T z The position of the starting point relative to the center of the earth in the terrain rendering scene;

[0137] The rotational transformation specifically refers to:

[0138] When the coordinate axes in the scene coordinate system do not completely coincide with the coordinate axes of the geocentric rectangular coordinate system, a rotation transformation is performed. Let the scene coordinate system be rotated by angles θ around the x, y, and z axes relative to the geocentric coordinate system. x ,θ y ,θ zThen the formula for the rotation transformation process is expressed as:

[0139] Rotation matrix around the x-axis:

[0140]

[0141] Rotation matrix around the y-axis:

[0142]

[0143] Rotation matrix around the z-axis:

[0144]

[0145] Transform the point (X′, Y′, Z′) sequentially using the three matrices described above:

[0146]

[0147] Where (X″,Y″,Z″) are the three-dimensional rectangular coordinates after translation and rotation transformations, i.e., mapped to the correct position.

[0148] After the above data transformation and processing, the construction of a regular network model can ensure that each grid vertex is mapped to the correct position.

[0149] S2. Construct a regular grid model based on the transformed DEM data; then use the regular grid model and hardware to perform terrain rendering to generate 3D terrain.

[0150] As a preferred embodiment of step S2, such as Figure 3 As shown, step S2 specifically includes:

[0151] S21. Definition and Vertex Determination of Regular Mesh; A regular mesh refers to a two-dimensional or three-dimensional mesh with equal spacing. In three-dimensional terrain modeling, such as... Figure 2 As shown, a regular grid divides the terrain surface into small triangular or rectangular units to represent the details of the terrain.

[0152] Define an n×m DEM data regular grid, where each grid represents a region in the actual terrain; the three-dimensional coordinates of the grid vertices are (X, M, Z). i,j ,Y i,j Z i,j );

[0153] X i,j =x origin +i·Δx;

[0154] Y i,j =y origin +j·Δy

[0155] Zi,j =H j

[0156] Where i and j represent the horizontal and vertical indices, respectively; x origin ,y origin These are the origin coordinates of the terrain region; Δx and Δy are the grid spacing; H j These are elevation values ​​from the DEM data;

[0157] For the entire terrain area, traverse all DEM data points and read out each grid vertex as described above. Each vertex corresponds to a point on the terrain surface.

[0158] During the definition of the regular mesh, for the acquired DEM data, the sampling points and interpolation points are connected to form a terrain surface to approximate the original surface. The formula is as follows:

[0159] DEM = {M i =ξ(P j )|P j (x j ,y j H j )∈D,j=1,2,...,n; i=1,2,...,m}

[0160] Where n and m represent the size of the regular mesh model as n×m; P j M represents the sampling point and the interpolation point. i It is a connected surface, x j y j H j These are the points corresponding to the DEM, x j y j H represents the x and y coordinates of the grid vertices. j These are the elevation values ​​from the DEM data, and ξ represents the generation rule of the regular grid model. In three-dimensional space, DEM data appears as a data array containing information about the terrain's elevation and undulations, such as... Figure 13 As shown.

[0161] S22. Assign texture coordinates to each vertex and map the texture onto the terrain surface; the texture coordinates are defined by the size of the DEM data regular grid; assuming the grid size is n×m, assign texture coordinates (u) to each vertex proportionally. i,j v i,j The formula is expressed as:

[0162]

[0163]

[0164] Where i and j represent the horizontal and vertical indices, respectively;

[0165] S23. To render a regular mesh model, vertices need to be connected by triangles. A regular mesh typically consists of two triangles forming a rectangular cell. In an n×m mesh, rows and columns divide four adjacent vertices into two triangles, generating triangle indices; let P be the four adjacent vertices. i,j ,P i+1,j ,P i,j+1 ,P i+1,j+1 Then the two triangles formed are:

[0166] Triangle 1: P i,j ,P i+1,j ,P i,j+1 ;

[0167] Triangle 2: P i+1,j ,P i,j+1 ,P i+1,j+1 ;

[0168] The vertex indices of each triangle are:

[0169] Triangle 1=(i,j),(i+1,j),(i,j+1);

[0170] Triangle 2=(i+1,j),(i,j+1),(i+1,j+1);

[0171] For each matrix cell in the entire grid, iterate through all (i,j) indices and decompose them into two triangles as described above, ultimately generating the triangle indices of all vertices.

[0172] S24. Define the connection relationship between vertices through triangle indexes to ensure the correct connection and rendering of the mesh; the connection relationship is as follows: each vertex is connected through a shared triangle edge, the triangle indexes of the mesh are sorted in counterclockwise order to ensure consistent direction during the rendering process; each triangle consists of three vertices, and all triangles are combined to form the mesh surface of the entire terrain.

[0173] S25. Combining hardware rendering, basic texture mapping and lighting effects are applied to generate 3D terrain. Basic texture mapping and lighting effects are applied to enhance visual effects, ultimately resulting in... Figure 4 The results are shown in the figure.

[0174] Meanwhile, an effective terrain tile loading and scheduling method plays a crucial role in improving the speed of 3D terrain rendering. This embodiment also employs a multi-threading mechanism and an object pool strategy during terrain rendering to achieve dynamic scheduling of terrain data and real-time scene rendering. Specifically, multi-threading is used to perform view volume clipping, while the main thread is responsible for terrain tile rendering tasks and the loading and scheduling of terrain tiles, thereby optimizing overall rendering efficiency. Figure 12 As shown.

[0175] During terrain rendering, the object pool manages the storage, rendering, and destruction of terrain tile objects. This mechanism consists of three main parts: a pre-storage table, a deprecation table, and a destruction queue. The pre-storage table stores terrain tile objects that may be reused in the future; the deprecation table stores unloaded terrain tile objects; and the destruction queue handles terrain tile objects that need to be destroyed.

[0176] During each frame update, the system first checks objects in the pre-store table. If they meet the current rendering conditions, they are moved to the activation queue, and a batch of terrain tiles is activated each frame according to a set number. For newly rendered terrain tiles, if they are not in the pre-store table, they are added to the rendering queue. Simultaneously, the pre-store table is updated, including terrain tiles not in the rendering queue and terrain tiles with changes in level of detail. If the pre-store table is full, the oldest objects are removed and moved to the disposal table. Finally, a set number of objects in the queue are destroyed to maintain a smooth frame rate and avoid sudden frame rate drops.

[0177] Furthermore, as the viewpoint moves, it is necessary to continuously calculate whether each terrain block is within the view frustum, which imposes a significant computational burden. To reduce this computational load, a bounding box-based view frustum pruning method is employed for optimization. Since the terrain data in this paper is organized using a two-level quadtree, the terrain block model is relatively standardized, and AABB bounding boxes offer faster computation speeds; therefore, a view frustum pruning method based on AABB bounding boxes is adopted. In addition, the pruning rules are reasonably designed and modified as follows:

[0178] (1) For a child node of a terrain patch, if it falls entirely within the current view area, the node is considered visible, and its visibility attribute is set to 1. Based on the node evaluation criteria, determine whether the node needs to be further subdivided. If subdivision is required, the four child nodes of the node will inherit its visibility attribute, which is also set to true.

[0179] (2) For nodes that are completely outside the view volume, they are considered invisible nodes and will be excluded directly from the rendering process, and a clipping operation will be performed.

[0180] (3) If a node only partially overlaps with the view volume, meaning that part of the terrain block extends outside the view volume, the node is still set to visible. It can be assumed that the user pays less attention to these parts and will not further subdivide these nodes. Maintaining their previous state helps reduce the processing time for model simplification on these terrain blocks.

[0181] The optimized algorithm was compared with the original algorithm, which did not utilize quadtrees for hierarchical management and terrain tile preprocessing, nor did it employ multithreading or object pooling for scheduling. After determining the viewpoint position, the test involved sampling terrain at four different numbers of points. The average time and frame rate of the original and optimized algorithms were recorded, resulting in the comparison of algorithm time and frame rate shown in Table 1 below.

[0182] Table 1 Comparison of average time and frame rate

[0183]

[0184] As shown in the table, the larger the block size, the more obvious the optimization effect of the two-level quadtree index encoding combined with the multi-threading mechanism and object pool strategy is, which reduces the average loading time by up to 10 times and increases the frame rate by up to 9 times. This shows that the method proposed in this invention is more efficient than the traditional method.

[0185] Therefore, the preliminary 3D terrain map generated by this method still needs further texture mapping and blending.

[0186] S3. Based on remote sensing land cover classification, perform terrain texture mapping on the three-dimensional terrain obtained in step S2 to locate and map the texture of specific landforms on the terrain.

[0187] 3D wargame maps differ from 2D wargame maps in their visualization methods: 3D maps display continuous terrain, providing natural transitions and more accurately simulating real-world terrain; while 2D maps present discrete terrain. Despite the different visualization methods, the underlying decision-making logic of both uses a hexagonal grid rule to maintain consistency.

[0188] In rendering large-scale terrain scenes, terrain mesh models can display surface undulations, but their detail is limited. Texture mapping is a key technique for enhancing the realism of 3D terrain by attaching texture images to the terrain mesh model to enrich surface details. The key to texture mapping is ensuring correct matching between terrain mesh points and texture coordinates to guarantee the stability of the texture mapping relationship when the terrain model deforms. This process involves defining and controlling textures, as well as using filtering techniques, texture repetition, or scaling to optimize texture coordinate positioning.

[0189] Because town sizes and transportation routes can change significantly over decades, directly mapping remote sensing imagery onto 3D terrain mesh models may be inaccurate. Furthermore, remote sensing image textures are planar and cannot display the details of 3D models, such as rivers, buildings, and forests.

[0190] Therefore, in this embodiment, as a preferred implementation of step S3, different types of mask images obtained from the semantic segmentation results of remote sensing images are used to locate and map the textures of specific landforms such as forests, grasslands, and wastelands on the terrain. This method effectively reconstructs the landforms of historical battles in three-dimensional terrain. By traversing the mask result images of different landforms, such as... Figure 5 As shown, Figure 5 (a), (b), and (c) represent the masked results for woodland, grassland, and cultivated land, respectively. When the RGB value of the mask position corresponding to the terrain is not (0, 0, 0), that position is set as the corresponding texture, and the terrain texture mapping result is as follows: Figure 6 As shown.

[0191] While the texture mapping method described above can distinguish different terrain features, the boundaries between different textures are too obvious and do not match the natural transitions in the real world. Therefore, this invention further employs texture blending technology to simulate the natural transitions of textures in real-world scenes.

[0192] S4. Based on the multi-channel blending color map Splatmap, the 3D terrain that has been mapped by terrain texture is processed to blend different textures, simulating the natural transition between terrain textures in real scenes.

[0193] As a preferred embodiment of step S4, S4 specifically includes:

[0194] A multi-channel blending color map (Splatmap) is used to achieve blending between different terrain textures, simulating natural transitions between textures in a real scene. The Splatmap is a four-channel RGBA color image that records the terrain's texture mapping; each channel represents a different texture map, such as... Figure 7 As shown;

[0195] When blending two texture maps, the blended pixels are obtained through channel interpolation, expressed by the formula:

[0196] o = a * alpha+b * (1-alpha);

[0197] Where 'o' represents the color of the blended pixel, and 'a' and 'b' represent the colors of the corresponding pixels in the two texture maps, respectively; 'alpha' is the color value of a certain channel in the color map splatmap, used to adjust the blending ratio between the two texture maps. By adjusting the channel values ​​of the splatmap, the texture blending effect can be effectively achieved, thus affecting the final color representation of pixel 'o'. The blending results for sand and woodland are shown below. Figure 8 As shown, Figure 8 (a) is the 3D terrain before texture blending. Figure 8 (b) shows the 3D terrain after texture blending. It is clear that the blended terrain is closer to the natural transition of textures in real terrain.

[0198] After the above steps S3 and S4, the resulting three-dimensional terrain is more realistic.

[0199] S5. Test by adjusting the channel values ​​of Splatmap to affect the color representation of the final pixel o until the texture blending effect can be effectively achieved; otherwise, return to step S4.

[0200] In summary, this invention, based on DEM data combined with a regular grid model and a two-level quadtree index encoding, can efficiently generate 3D terrain. Furthermore, it utilizes remote sensing feature classification and color mapping to enhance texture mapping realism. The entire terrain rendering process also employs a multi-threading mechanism and an object pool strategy to streamline the generation process.

[0201] The method proposed in this invention resolves the contradiction between the high precision, realism, and immersiveness of 3D maps and the real-time efficiency and data requirements of models, enabling rapid and accurate reconstruction of 3D simulation scenarios of classic battle situations.

[0202] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0203] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for constructing a 3D wargaming map based on DEM and remote sensing feature classification, characterized in that, Specifically, the following steps are included: S1. Acquire digital elevation model (DEM) data, and perform block-level and layer-level processing and data transformation on it; the DEM data is a terrain data array containing terrain elevation and undulation information, stored in raster format; S2. Construct a regular grid model based on the transformed DEM data; then use the regular grid model and hardware to perform terrain rendering to generate 3D terrain. During terrain rendering, a multi-threading mechanism and object pooling strategy are employed to dynamically schedule terrain blocks and render the scene in real time. Specifically, this includes: Multithreading is used to perform view volume clipping, while the main thread is responsible for drawing terrain tiles and loading and scheduling them. During terrain rendering, the object pool is responsible for managing the storage, rendering, and destruction of terrain tile objects; it sets up a pre-storage table, a pending disposal table, and a destruction queue; the pre-storage table is used to store reactivated terrain tile objects; the pending disposal table stores unloaded terrain tile objects; and the destruction queue is responsible for processing terrain tile objects that need to be destroyed. During each frame update, the system first checks the objects in the pre-storage table. If they meet the current rendering conditions, they are moved to the activation queue. Each frame, a batch of terrain tiles is activated according to a set number. For new terrain tiles that need to be rendered, if they are not in the pre-storage table, they are added to the rendering queue. At the same time, the pre-storage table is updated, including terrain tiles not in the rendering queue and terrain tiles with changes in level of detail. If the pre-storage table is full, the oldest objects are removed from the discard table. Finally, a certain number of objects in the queue are destroyed to maintain a smooth frame rate and avoid sudden drops in frame rate. When the viewpoint moves, an improved AABB bounding box view volume clipping method is used for optimization; the clipping rules of the improved AABB bounding box are as follows: (1) For a child node of a terrain block, if it falls completely within the current field of view, the node is considered visible and its visibility attribute is set to 1. Based on the node evaluation criteria, determine whether the node needs to be further subdivided. If it needs to be subdivided, the four child nodes of the node will inherit its visibility attribute, that is, the visibility attribute is also set to true. (2) For nodes that are completely outside the view volume, they are considered invisible nodes and will be directly excluded from the rendering process, and a clipping operation will be performed; (3) If a node only partially overlaps with the view volume, i.e., some terrain blocks extend outside the view volume, it will still be set to visible; and it is assumed that the user pays less attention to these parts, so these nodes will not be further subdivided; their previous state will remain unchanged. S3. Based on remote sensing land cover classification, perform terrain texture mapping on the three-dimensional terrain obtained in step S2 to locate and map the texture of specific landforms on the terrain. S4. Based on the multi-channel blending color map Splatmap, the 3D terrain that has been mapped by terrain texture is processed to blend different textures, simulating the natural transition between terrain textures in real scenes. S5. By adjusting the channel values ​​of the Splatmap, the final pixel values ​​are affected. Test the color performance until the texture blending effect can be effectively achieved; otherwise, return to step S4.

2. The method for constructing a three-dimensional wargaming map based on DEM and remote sensing feature classification according to claim 1, characterized in that, The block-based and layered processing in step S1 involves using a two-level quadtree index encoding method to divide the DEM data into layers and blocks to ensure effective data management and improve processing efficiency. Specifically, it includes the following steps: 1) Large-scale terrain data in DEM data is segmented into blocks using vertical and horizontal partitioning methods to achieve the first-level encoding of the data block index: vertical-horizontal encoding. This involves setting the data block as the origin and dividing the terrain along the vertical and horizontal axes. An index code is then established for each segmented data block, enabling rapid retrieval and access to the terrain data. The formula is expressed as: ； ； in, The coordinates of the target region P; Here are the index coordinates of the data block within the entire terrain data; the size of the terrain block is... , It is an odd number; 2) Each data block independently constructs a two-level quadtree based on the traditional detail evaluation criteria, and only constructs it in real time when terrain rendering is required, so as to reduce the amount of computation; the data block starts with the origin as the local coordinate and is adjusted to the actual position during terrain rendering. 3) Use a two-dimensional matrix to label the nodes in the two-level quadtree structure; depending on whether the nodes need to be subdivided, different labels are applied to form a hierarchical detail model with multi-resolution attributes, thereby realizing the layering and segmentation of terrain data.

3. The method for constructing a three-dimensional wargaming map based on DEM and remote sensing feature classification according to claim 1, characterized in that, The data transformation process in step S1 specifically includes: S11. Using the Geospatial Data Abstraction Library (GDAL), the raster data in the DEM data is converted into latitude and longitude coordinates and elevations of the actual location on Earth in the WGS-84 system through geographic affine transformation coefficients. S12. After projection calculation, the latitude and longitude coordinates and elevation are converted into three-dimensional rectangular coordinates in the Cartesian coordinate system; the Cartesian coordinate system, namely the geocentric rectangular coordinate system, has the Earth's center as its origin, with the positive direction of the X-axis pointing to the intersection of the equator and the prime meridian; the Z-axis is perpendicular to the equatorial plane, with its positive direction pointing to the North Pole; the formula for the conversion process is expressed as: ； Where (B, L, H) are the latitude, longitude and elevation of a point in the WGS-84 coordinate system; (X, Y, Z) are the corresponding three-dimensional coordinates of the point in the Cartesian coordinate system; The radius of the meridian circle. This is the first eccentricity of the Earth's ellipsoid; S13. Through translation and rotation transformations, the Cartesian coordinate system in step S12 is transformed into the scene coordinate system under the terrain rendering scene to obtain the three-dimensional coordinates under the scene coordinate system.

4. The method for constructing a three-dimensional wargaming map based on DEM and remote sensing feature classification according to claim 3, characterized in that, In step S13: The translation transformation specifically refers to: Points in the original geocentric rectangular coordinate system Point translated to the scene coordinate system The formula is expressed as: ； ； ； in, The position of the starting point relative to the center of the earth in the terrain rendering scene; The rotational transformation specifically refers to: When the coordinate axes in the scene coordinate system do not completely coincide with the coordinate axes of the geocentric rectangular coordinate system, a rotation transformation is performed. Let the scene coordinate system be relative to the geocentric coordinate system and rotate around... The shaft has a rotation angle Then the formula for the rotation transformation process is expressed as: Rotation matrix around the x-axis: ； Rotation matrix around the y-axis: ； Rotation matrix around the z-axis: ； Point Transform the above three matrices in sequence: ； in, These are the three-dimensional rectangular coordinates after translation and rotation transformations, i.e., mapped to the correct position.

5. The method for constructing a three-dimensional wargaming map based on DEM and remote sensing feature classification according to claim 1, characterized in that, Step S2 includes: S21. Regular grid definition and grid vertex determination; Define an n×m DEM data regular grid, where each grid represents a region in the actual terrain; the three-dimensional coordinates of the grid vertices are... ; ； in, and These represent the horizontal and vertical indices, respectively; These are the origin coordinates of the terrain region; and It refers to the grid spacing; These are elevation values ​​from the DEM data; For the entire terrain area, traverse all DEM data points and read out each grid vertex as described above. Each vertex corresponds to a point on the terrain surface. S22. Assign texture coordinates to each vertex and map the texture onto the terrain surface; the texture coordinates are defined by the size of the DEM data regular grid; assuming the grid size is n×m, assign texture coordinates to each vertex proportionally. , The formula is expressed as: ； ； in, and These represent the horizontal and vertical indices, respectively; S23. Divide four adjacent grid vertices into two triangles and generate triangle indices; denote the four adjacent vertices as follows: Divide it into two triangles: Triangle 1: ; Triangle 2: ; The vertex indices of each triangle are: ； ； This process involves traversing all vertex indices throughout the entire grid, ultimately generating triangle indices for all vertices. S24. Define the connection relationship between vertices through triangle indices to ensure the correct connection and rendering of the mesh; the connection relationship is: each vertex is connected through a shared triangle edge, and the triangle indices of the mesh are sorted in counterclockwise order; S25. Combine hardware rendering, apply basic texture mapping and lighting effects to generate 3D terrain.

6. The method for constructing a three-dimensional wargaming map based on DEM and remote sensing feature classification according to claim 5, characterized in that, In step S21: During the definition of the regular mesh, for the acquired DEM data, the sampling points and interpolation points are connected to form a terrain surface to approximate the original surface. The formula is as follows: ； Where n and m represent the size of the regular grid model as n×m; Indicates the sampling point and the interpolation point. It is a sheet of dough connected together. , , These are the points corresponding to the DEM. , These are the x and y coordinates of the grid vertices. These are the elevation values ​​from the DEM data. This represents the generation rules for the regular mesh model.

7. The method for constructing a three-dimensional wargaming map based on DEM and remote sensing feature classification according to claim 1, characterized in that, Step S3 specifically involves: for the three-dimensional terrain obtained in step S2, using different types of mask images obtained from the semantic segmentation results of remote sensing images to locate and map the texture of specific landforms on the terrain, thereby effectively restoring the landforms of historical battles in the three-dimensional terrain.

8. The method for constructing a three-dimensional wargaming map based on DEM and remote sensing feature classification according to claim 1, characterized in that, Step S4 is as follows: A multi-channel blending color map (Splatmap) is used to achieve blending between different terrain textures, simulating the natural transition between textures in a real scene. The color map (Splatmap) is defined as a four-channel RGBA color image that records the texture mapping of the terrain, with each channel representing a different texture map. When blending two texture maps, the blended pixels are obtained through channel interpolation, expressed by the formula: ； Where o is the color of the blended pixel, a and b are the colors of the corresponding pixels in the two texture maps respectively; alpha is the color value of a certain channel of the pixel in the color map Splatmap, which is used to adjust the blending ratio between the two texture maps.

Images