Three-dimensional intelligent visualization method and device for multi-source data loading, medium and equipment

Abstract

The invention relates to the technical field of computer graphics, and particularly discloses a three-dimensional intelligent visualization method and device for multi-source data loading, a medium and equipment.The method comprises the steps that an oblique photography model is loaded and dynamically adjusted, and an oblique photography live-action base is constructed; generating semantic 3D tiles data on the basis of the base; identifying a key entity object, carrying out lightweight processing on a high-precision GLB / GLTF model of the key entity object, and dynamically implanting the high-precision GLB / GLTF model into a scene; performing dynamic analysis and spatial superposition on the KML data associated with the key object, and constructing a multi-level three-dimensional scene information system from entity annotation to macroscopic annotation; dynamically coupling the KML elements with terrain elevation data to form a high-precision digital elevation base supporting visualization; based on a video memory dynamic partitioning mechanism, optimized resource scheduling is performed on a substrate, and adaptive balance of terrain complexity and system load is realized. According to the invention, the precision and efficiency of multi-source heterogeneous data fusion and the system stability can be improved.

Description

Technical Field

[0001] This application belongs to the field of computer graphics technology, specifically relating to a three-dimensional intelligent visualization method, apparatus, medium, and device for loading multi-source data. Background Technology

[0002] In the deep application of smart cities, digital twins, geological exploration, and other fields, existing 3D visualization technologies face numerous challenges, including: poor compatibility with multi-source data (typically supporting only 3 to 5 mainstream data formats, with an overall format coverage of less than 50%), low dynamic fitting accuracy between models and terrain (fitting errors exceeding 2 to 3 meters), severe system resource contention (resource contention rates as high as 60%-80% in high-concurrency scenarios), and limited interactive functions (interactive response latency generally exceeding 200ms). Traditional visualization platforms often rely on a single data loading mode, making it difficult to achieve collaborative rendering and integrated management of cross-format and cross-coordinate system data. Furthermore, their inefficient algorithms often lead to issues such as memory overflow and rendering stuttering.

[0003] Although existing technologies have attempted to optimize the loading of specific data formats, their discrete processing architecture still suffers from problems such as format conflicts, inaccurate terrain fitting, and uneven resource scheduling. Furthermore, the lack of intelligent adaptation and scene-aware scheduling capabilities severely restricts the platform's practicality and scalability in complex engineering scenarios. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this application is to provide a three-dimensional intelligent visualization method, device, medium and equipment for multi-source data loading. This application aims to achieve high-precision and high-efficiency integrated loading and rendering of multi-source data, so as to improve the real-time performance and stability of the visualization platform.

[0005] To achieve the above objectives, this application provides the following technical solution: A multi-source data loading method for 3D intelligent visualization includes: loading an oblique photogrammetry model and dynamically adjusting it to construct an oblique photogrammetry real-world base; generating semantic 3D tile data based on the oblique photogrammetry real-world base using a GPU-accelerated vector conversion pipeline to achieve integrated rendering of the macro-geographic environment and micro-entity models; identifying key entity objects, performing lightweight processing on the high-precision GLB / GLTF refined models corresponding to the key entity objects, and dynamically embedding them into the integrated rendered macro-geographic environment and micro-entity models to achieve detail enhancement and visualization enhancement of the key entity objects; dynamically parsing and spatially overlaying the KML data associated with the enhanced key entity objects to construct a multi-level 3D scene information system from entity annotation to macro-spatial annotation; dynamically coupling the KML elements in the multi-level 3D scene information system with terrain elevation data to form a high-precision digital elevation base supporting visualization; and optimizing resource scheduling for the digital elevation base to achieve multi-source data 3D intelligent visualization rendering.

[0006] Optionally, loading the oblique photogrammetry model and dynamically adjusting it to construct the oblique photogrammetry real-world base includes: based on the inherent tiled storage structure of the oblique photogrammetry model, locating and loading model tiles within the visible area through R-tree spatial indexing; integrating a dynamic LOD mechanism to automatically switch data at different detail levels according to the real-time distance between the camera viewpoint and the tiles; and using vertex shaders to dynamically adjust the model height and terrain fit strength to complete the construction of the oblique photogrammetry real-world base.

[0007] Optionally, the step of generating semantic 3D tile data based on the oblique photogrammetry real-world base using a GPU-accelerated vector transformation pipeline to achieve integrated rendering of the macro-geographic environment and micro-entity models includes: preprocessing vector data in a geographic information system to match it with the oblique photogrammetry real-world base; generating 3D tiles based on the preprocessed vector data; performing dynamic stylized rendering on the generated 3D tiles; and deeply integrating and coupling the rendered 3D tiles with the oblique photogrammetry real-world base in a unified spatial coordinate system to achieve integrated rendering of the macro-geographic environment and micro-entity models.

[0008] Optionally, the identification of key entity objects involves lightweighting the high-precision GLB / GLTF refined model corresponding to the key entity object and dynamically embedding it into the integrated rendered macro-geographic environment and micro-entity model to achieve detail enhancement and visualization enhancement of the key entity object. This includes: identifying key entity objects based on external business data; lightweighting the GLB / GLTF refined model corresponding to the key entity object; and embedding the lightweighted GLB / GLTF refined model into the integrated rendered macro-geographic environment and micro-entity model to achieve detail enhancement and visualization enhancement of the key entity object.

[0009] Optionally, the dynamic parsing and spatial overlay of the KML data associated with the key entity objects after detail enhancement and visualization enhancement to construct a multi-level three-dimensional scene information system from entity annotation to macro-spatial annotation includes: extracting KML elements based on KML data; performing coordinate transformation on the extracted KML elements; and dynamically rendering and stylizing the coordinate-transformed KML elements. The rendered KML elements are deeply integrated with the oblique photogrammetry real-world base and semantic 3D tiles for rendering and interaction, in order to construct a multi-level three-dimensional scene information system from entity annotation to macro-spatial annotation.

[0010] Optionally, the step of dynamically coupling KML elements in the multi-layered three-dimensional scene information system with terrain elevation data to form a high-precision digital elevation base supporting visualization includes: constructing a multi-resolution seamless terrain base as a spatial carrying foundation to provide accurate geometric support and coordinate mapping framework for KML elements; performing dynamic elevation calibration on KML elements based on the multi-resolution seamless terrain base; performing real-time simulation and dynamic rendering of water level lines based on the calibrated KML elements; and maintaining the spatial consistency between terrain and KML elements to form a high-precision digital elevation base supporting visualization. Optionally, the resource scheduling optimization of the digital elevation base to achieve the rendering of multi-source data 3D intelligent visualization includes: constructing a memory dynamic partitioning structure based on a priority model; predicting the viewpoint movement trajectory based on machine learning based on the memory dynamic partitioning structure; scheduling and preloading resources based on the prediction results; and performing real-time load monitoring and adaptive adjustment to achieve the rendering of multi-source data 3D intelligent visualization. This application also provides a 3D intelligent visualization device for loading multi-source data. The device includes: a loading module for loading and dynamically adjusting an oblique photogrammetry model to construct an oblique photogrammetry real-world base; a conversion module for generating semantic 3D tile data based on the oblique photogrammetry real-world base using a GPU-accelerated vector conversion pipeline to achieve integrated rendering of macro-geographic environment and micro-entity model; an enhancement module for identifying key entity objects, performing lightweight processing on the high-precision GLB / GLTF refined model corresponding to the key entity objects, and dynamically embedding it into the integrated rendered macro-geographic environment and micro-entity model to achieve detail enhancement and visualization enhancement of the key entity objects; a parsing module for dynamically parsing and spatially overlaying the KML data associated with the enhanced key entity objects to construct a multi-level 3D scene information system from entity annotation to macro-spatial annotation; a calibration module for dynamically coupling the KML elements in the multi-level 3D scene information system with terrain elevation data to form a high-precision digital elevation base supporting visualization; and an optimization module for optimizing resource scheduling of the digital elevation base to achieve rendering of multi-source data 3D intelligent visualization.

[0011] This application also provides a storage medium including instructions that, when executed on a computer, cause the computer to perform the method as described in the preceding claim.

[0012] This application also provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any of the preceding claims.

[0013] Compared with the prior art, the beneficial effects of this application are as follows: This application establishes a complete multi-source data dynamic loading system by constructing an oblique photogrammetry real-world base, generating semantic 3D tiles using GPU acceleration, lightweight processing of refined models, dynamically parsing and overlaying KML data, fusing multi-resolution terrain, and performing real-time calibration. Finally, it relies on a dynamic memory partitioning mechanism for intelligent resource scheduling. This system effectively solves the core problems of traditional methods, such as poor format compatibility, inaccurate model-terrain fit, low rendering efficiency, and severe resource conflicts. It can achieve high-precision, high-fidelity integrated rendering and seamless fusion from macro-geographical environment to micro-entity elements, and can improve the real-time interactive performance, scene construction efficiency, and system stability of the 3D visualization platform. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating a three-dimensional intelligent visualization method for loading multi-source data according to an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a three-dimensional intelligent visualization device for multi-source data loading provided in another embodiment of this application; Figure 3 This is a schematic diagram of the structure of a storage medium provided in another embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0016] It should be noted that all directional indicators (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicator will also change accordingly.

[0017] In this application, unless otherwise expressly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0018] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the meaning of "and / or" throughout the text includes three parallel solutions; for example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0019] Figure 1 This application provides a method for multi-source data loading and three-dimensional intelligent visualization, as exemplified by one embodiment. Figure 1 As shown, the method includes the following steps: S100: Loads and dynamically adjusts the oblique photogrammetry model to construct an oblique photogrammetry real-world base. It should be noted that the loaded oblique photogrammetry model falls within the scope of existing technology. This model employs mature oblique photogrammetry technology, acquiring images from multiple angles (typically one vertical and four oblique) from the air, and then performing 3D reconstruction based on computer vision algorithms to obtain a realistic 3D model. It is important to clarify that this application does not involve any optimization or improvement of the generation algorithm or internal structure of this oblique photogrammetry model; rather, it uses it as a pre-set data source for loading and application.

[0020] S200: Based on the oblique photogrammetry real-world base, semantic 3D tiles data is generated through a GPU-accelerated vector conversion pipeline to achieve integrated rendering of macro-geographic environment and micro-entity model; S300: Identify key entity objects, perform lightweight processing on the high-precision GLB / GLTF refined models corresponding to the key entity objects, and dynamically embed them into the macro-geographic environment and micro-entity models after integrated rendering, so as to achieve detail enhancement and visualization enhancement of key entity objects. S400: Dynamically parses and spatially overlays the KML data associated with key entity objects after detail enhancement and visualization enhancement, in order to construct a multi-level three-dimensional scene information system from entity annotation to macro-spatial annotation; S500: Dynamically couple the KML elements in the multi-level three-dimensional scene information system with the terrain elevation data to form a high-precision digital elevation base that supports visualization. S600: Optimize resource scheduling for the digital elevation base to achieve three-dimensional intelligent visualization rendering of multi-source data.

[0021] This application achieves seamless integration of macroscopic geographical environment and microscopic entity models by constructing an integrated rendering framework of oblique photogrammetry real-world base and semantic 3D tiles. By leveraging the lightweight processing and dynamic implantation mechanism of refined models, it enhances the detail and visual realism of key entity objects. Through dynamic parsing of KML data and the construction of a multi-level spatial information system, it enables full-element 3D expression from entity-level annotation to macroscopic spatial annotation. Furthermore, by combining multi-source terrain data coupling and machine learning-based resource scheduling optimization, it effectively solves the performance bottleneck of loading massive heterogeneous data. Ultimately, while ensuring rendering accuracy, it achieves an adaptive balance between the load of 3D visualization devices and scene complexity, thereby comprehensively improving the system's real-time performance, stability, and visual immersion.

[0022] In another exemplary embodiment, step S100, loading the oblique photogrammetry model and dynamically adjusting it to construct the oblique photogrammetry real-world base, includes the following steps: S101: Based on the inherent tiled storage structure of the oblique photogrammetry model, the model tiles within the visible area are located and loaded through R-tree spatial indexing; This step begins by parsing the metadata file (e.g., metadata.xml) of the oblique photogrammetry model (OSGB / 3MX format) to obtain its spatial reference frame, geographic boundaries, and tile pyramid structure. Then, it iterates through all tile files (.osgb), calculating the minimum bounding rectangle (MBR) for each tile. Based on this, an R-tree spatial index is built in memory, stored as key-value pairs: the key is the tile's MBR, and the value is the path and hierarchy information of the tile file. Next, for each rendering frame, the system calculates the frustum based on the current camera parameters (position, orientation, field of view) and uses the constructed R-tree spatial index to perform fast collision detection between the frustum and the tile MBRs, retrieving all tiles located within the current visible area or intersecting with the frustum. Finally, the retrieved tile requests are sent to an asynchronous loading queue. This queue calculates loading priority based on the distance between the tile center and the camera viewpoint, prioritizing the loading of closer and more important tiles. The loading thread retrieves high-priority tasks from the queue, reads tile data (including geometry and texture) from the disk, and uploads it to the GPU memory.

[0023] This step, through R-tree spatial indexing and frustum culling, avoids the performance overhead of traversing all files, precisely concentrating data I / O and GPU upload resources in the user-visible area, greatly reducing loading time and laying the foundation for smooth visual interaction.

[0024] S102: Integrates a dynamic LOD mechanism, which automatically switches between different levels of detail based on the real-time distance between the camera viewpoint and the tile. In this step, firstly, for each tile retrieved in step S101, the Euclidean distance between its center point and the camera viewpoint is calculated in real time, and the level of detail (LOD) to be loaded for the tile is dynamically adjusted according to a preset distance-detail mapping strategy. The mapping strategy is expressed as follows: LOD_level=clamp(floor(log2(distance / baseLODDistance)),0,maxLOD); Here, `baseLODDistance` represents the configurable base distance parameter; `maxLOD` represents the lowest level of detail (LOD) of the tile; `distance` represents the Euclidean distance between the camera's viewpoint and the tile's center point; `log2()` represents a base-2 logarithmic function used to convert the linearly increasing distance (`distance / baseLODDistance`) into an exponentially changing LOD level; `floor()` represents a floor function used to discretize the result calculated by `log2()` and convert it into an integer value; `clamp()` represents a clamping function used to restrict the integer obtained by `floor()` to a minimum value of 0 and a maximum value of `maxLOD`; and `LOD_level` represents the calculated final level of detail, an integer in the range [0, maxLOD].

[0025] Based on the calculated LOD_level, the system intelligently loads or switches to tile model data of corresponding precision from the pre-generated pyramid-shaped LOD resources for rendering. Tiles closer to the viewpoint have lower calculated LOD_level values, thus receiving higher precision models to present rich geometric and texture details. Conversely, tiles farther away have higher LOD_level values, and the system assigns them coarser, less polygon-rich simplified models to save computational and memory resources. Furthermore, to avoid abrupt visual jumps when switching between different levels, the system uses vertex morphing or alpha fade-in / fade-out techniques in the shaders to achieve smooth visual transitions between different detail levels. This ensures rendering performance while maintaining the continuity, consistency, and immersion of the scene's visuals.

[0026] In summary, by integrating a dynamic LOD mechanism, the system can intelligently allocate computing and video memory resources, ensuring that distant areas do not consume excessive resources and that nearby areas maintain high precision, thereby maintaining a stable and high rendering frame rate from any perspective.

[0027] S103: The height of the oblique photogrammetry model and the terrain fit strength are dynamically adjusted through the vertex shader to complete the construction of the oblique photogrammetry real scene base.

[0028] In this step, the vertex shader dynamically adjusts the global offset of the entire oblique photogrammetry model along the Z-axis (elevation direction) based on the height adjustment slider set by the user in the interface (adjustment range ±50 meters, accuracy 0.1 meters). It also adjusts the adhesion strength between the oblique photogrammetry model and the terrain using a coefficient k (range 0~1) input through the adhesion strength knob. Furthermore, based on the dynamic adhesion formula—Adhesion Strength = k * (Terrain Elevation - Model Baseline Height)—the vertex shader performs real-time height synthesis for each vertex, enabling a smooth transition between "completely floating" and "closely attached" oblique photogrammetry model, overcoming the floating or embedding issues between the model and the terrain. Simultaneously, the vertex shader displays the viewpoint's ground clearance (based on high-precision ray collision detection, with an error controlled within ±0.5 meters) and the viewpoint calculated via Web Mercator projection in the lower right corner of its interface.

[0029] In another exemplary embodiment, step S200, which involves generating semantic 3D tile data based on the oblique photogrammetry real-world base using a GPU-accelerated vector transformation pipeline to achieve integrated rendering of the macroscopic geographical environment and the microscopic entity model, includes the following steps: S201: Preprocess the vector data in the geographic information system to match it with the oblique photogrammetric real-world base; This step first reads the vector data source (such as SHP file) from the Geographic Information System (GIS) database and parses its geometric information (points, lines, polygons) and attribute fields (such as name, type, and height value). Then, the read vector data is transformed into coordinates, that is, the vector data is transformed from the original coordinate system (such as the WGS84 geographic coordinate system) to the Web Mercator projection coordinate system consistent with the oblique photogrammetry real scene base, so as to achieve unified registration of spatial reference. At the same time, an attribute hash table associated with each vector feature is constructed to store its attribute information in a key-value pair structure, providing a semantic data foundation for subsequent rendering associated with real scene features.

[0030] S202: Generate 3D Tiles based on preprocessed vector data; This step requires GPU-accelerated 3D voxelization of the vector data (such as building outlines) after coordinate transformation. Specifically, this includes: vertically stretching the planar polygons based on the height field (such as height) in the feature attributes to generate a 3D volume model with real height; at the same time, using instantiation rendering technology, the semantic information (such as feature type and material parameters) in the attribute hash table is encoded into a low-dimensional numerical format (such as RGB three-channel color or single integer index), and the encoded data and geometric information are written together into a batch table in 3D Tiles format, thereby forming semantic 3D model data that combines geometric and attribute information and can be deeply integrated with the real scene base.

[0031] S203: Perform dynamic stylized rendering on the generated 3D Tiles; This step utilizes a programmable rendering pipeline to dynamically and diversely render the generated 3D Tiles. The specific implementation mechanism is as follows: First, in the fragment (pixel) coloring stage, the system reads the encoded data of each feature from a batch table in 3D Tiles format (e.g., a lookup table that maps feature types to specific RGB colors or integer indices). The shader then decodes these encoded values ​​in real time to restore the original semantic attributes of the feature, such as "building type is residential", "height exceeds 50 meters", or "belongs to a key facility".

[0032] Secondly, based on the parsed semantic attributes, the shader dynamically drives seamless switching and flexible combination of various rendering styles through a series of configurable mapping rules or style sheets. For example, it can achieve type-based coloring based on the "building type" attribute, rendering commercial, residential, and industrial buildings as blue, beige, and gray respectively; by recognizing element IDs or hierarchical attributes, it can overlay white or highlighted wireframe outlines on the entity model to clearly show the structural boundaries; for elements that meet specific conditions such as "key facilities" or "height > 100 meters", it can add a glowing post-processing effect in the fragment shader to make them stand out in the scene; at the same time, it supports color-changing the model according to its status using a semi-transparent fill method based on dynamic attributes such as "operational status", such as using green to represent normal operation and red to indicate alarm status, thereby realizing intuitive visualization of business data and multi-dimensional information fusion expression.

[0033] S204: Deeply integrate and couple the rendered 3D Tiles with the oblique photogrammetric real-world base in a unified spatial coordinate system to achieve integrated rendering of the macro-geographic environment and the micro-entity model.

[0034] In this step, this application constructs a multi-layered rendering pipeline to deeply integrate and couple semantic 3D tiles with oblique photogrammetric real-world bases in a unified spatial coordinate system. This includes a spatial registration stage and a visual fusion stage. In the spatial registration stage, a precise spatial mapping relationship is established based on a unified geographic reference system (such as the Web Mercator projection coordinate system and the WGS84 elevation datum). A coordinate transformation matrix is ​​used to transform the vertex data of the semantic 3D tiles to the corresponding 3D spatial coordinate system of the oblique photogrammetric real-world base in real time, ensuring that the positional deviation between vector elements such as building outline corners and road centerlines and the corresponding real-world model is less than 0.1 pixels. This process employs a four-parameter affine transformation and elevation interpolation algorithm to dynamically compensate for displacement deviations caused by coordinate system differences and data acquisition errors, achieving millimeter-level precise alignment between the microscopic entity model and the macroscopic geographic environment in 3D space.

[0035] In the visual fusion phase, a three-pronged approach is employed to achieve deep integration of geometry and lighting: First, a depth buffer synchronization management mechanism compares the depth values ​​of 3D Tiles and the oblique photogrammetry model in real time within the rendering pipeline. A strategy combining depth testing and stencil testing automatically corrects model interleaving caused by differences in data accuracy, ensuring seamless geometric integration between the building base and the terrain. Second, physically based rendering technology is used to calculate ambient occlusion, diffuse reflection, and specular reflection effects through a unified lighting model. The system dynamically generates hemispherical harmonic lighting parameters by collecting sky and environment maps from the real-world base and synchronizing them to the material shaders of the 3D Tiles, enabling the vector-generated building models to present ambient lighting images consistent with the real-world terrain. Finally, real-time shadow mapping technology uses the oblique photogrammetry model as a shadow projector, allowing the 3D Tiles buildings to project dynamic shadows on the ground that conform to the actual lighting angles, further enhancing the visual consistency of the two types of data in three-dimensional space.

[0036] Furthermore, this application innovatively introduces a semantically aware hybrid shader. This shader dynamically adjusts the material reflectivity and ambient occlusion coefficient based on the semantic attributes of 3D Tiles elements, highlighting the semantic features of key elements while maintaining the realism of the textures. Simultaneously, through screen-space reflection and global illumination calculations, it achieves natural light interaction between microscopic entities and the macroscopic environment, eliminating visual disjointness. Finally, under a GPU parallel architecture, a unified memory scheduling strategy enables synchronous rendering output of both types of data, forming a 3D scene that combines semantic information hierarchy with visual unity, thus achieving truly integrated intelligent visualization from the macroscopic geographical environment to the microscopic entity model.

[0037] In another exemplary embodiment, in step S300, identifying key entity objects, performing lightweight processing on the high-precision GLB / GLTF refined model corresponding to the key entity objects, and dynamically embedding it into the integrated rendered macro-geographic environment and micro-entity model to achieve detail enhancement and visualization enhancement of the key entity objects, includes the following steps: S301: Identify key entity objects based on external business data; In this step, this application achieves intelligent identification of key entity objects by accessing external business data sources (including KML / KMZ files, geographic information system databases, and business rule engines). First, it parses semantic attribute fields (such as facility type, management level, and operational status) and spatial geometric information in the data source. Then, it filters elements and selects targets through a pre-set business rule library (such as "electrical equipment with priority higher than Lv3" and "historical buildings listed in the protection list"). Subsequently, it binds entity elements that meet the rules with their spatial coordinates and attribute identifiers to generate a set of key entity objects carrying semantic identities. Finally, it associates and maps them with a unified resource identifier and a 3D resource library to provide target indexes and data foundations for subsequent refined model loading and spatial registration, forming a business demand-driven entity identification mechanism.

[0038] S302: Lightweighting of the GLB / GLTF refined model corresponding to the key entity object; In this embodiment, the lightweighting process for the GLB / GLTF refined model includes the following steps: Step 1: Simplify the geometric mesh of the GLB / GLTF refined model; In this step, the input GLB / GLTF refined model is first divided into an octree hierarchical mesh according to its spatial structure. Then, adjacent triangular faces with the same normal are merged layer by layer in a bottom-up manner. Next, the mesh is iteratively simplified by edge folding and vertex deletion algorithms. While maintaining the visual contour features, the number of model faces is compressed to about 20% of the original data to reduce the GPU rendering load.

[0039] Step 2: Compress the texture images in the simplified GLB / GLTF refined model; In this step, adaptive scalable texture compression (ASTC) is performed on the texture image in the GLB / GLTF refinement model. This involves dividing the texture into 4×4 pixel blocks and then performing lossy compression on each pixel block through bit-domain encoding and color space transformation. The compressed texture data is directly decoded by the GPU hardware, reducing texture memory usage by 60% to 80% while maintaining visual fidelity.

[0040] Step 3: Optimize the compressed texture image; In this step, for models containing skeletal animation or morphing animation, key frame nodes in the animation sequence are identified and extracted. During rendering, the intermediate animation state between adjacent key frames is restored in real time using a linear interpolation algorithm, reducing the amount of animation data by more than 70% while maintaining the smoothness of the animation.

[0041] S303: The lightweight GLB / GLTF refined model is embedded into the macro-geographic environment and micro-entity model after integrated rendering to achieve detailed enhancement and visualization of key entity objects.

[0042] This step first requires multi-level spatial adaptation of the lightweighted GLB / GLTF refined model, specifically including: 1. Based on the geographic coordinate system and scene matrix transformation, the model base is accurately matched with the real terrain.

[0043] Specifically, a final model-world transformation matrix is ​​calculated for each refined model that needs to be implanted. This matrix is ​​multiplied by the following translation, rotation, and scaling matrices: Model-world transformation matrix = translation matrix × rotation matrix × scaling matrix; The translation matrix determines the model's specific location in the global scene based on the high-precision geographic coordinates (such as latitude, longitude, elevation, or Web Mercator coordinates) of the model's base reference point; the rotation matrix performs rotation transformations based on the model's orientation angle, tilt angle, and other attitude parameters to ensure that its orientation is consistent with the real-world environment (for example, making the front of the building face the street); and the scaling matrix scales the model from its own defined unit size (such as 1 unit = 1 meter) to a scale that matches its real-world size.

[0044] Finally, in the vertex shader, the position of each model vertex is transformed through the following calculations: World space coordinates = Model - World transformation matrix × Model local space coordinates; Second, by reconstructing the normal field and recalculating the illumination, the implanted model maintains a continuous light influence response with the surrounding environment.

[0045] Specifically, the first step is to unify the lighting environment. The implanted model abandons its own isolated lighting parameters and instead adopts a global lighting environment extracted from the real-world base—including ambient irradiance defined in the form of a skybox or spherical harmonic function, and the main solar light source containing direction, intensity, and color information. Then, the standard Cook-Torrance BRDF shading model is executed in the fragment shader. By fusing world space normals, global lighting parameters, and model material properties, the diffuse and specular reflection components of direct lighting are accurately calculated, and the ambient occlusion effect is superimposed. Ultimately, the implanted model achieves a visually seamless integration with the surrounding real-world objects in terms of light and shadow relationships, specular reflection, and shadow representation, completely eliminating the feeling of pasting.

[0046] Secondly, a semantically aware instantiation rendering architecture is adopted. By parsing the semantic attributes of the model (such as type, size, and spatial hierarchy), it is possible to identify and dynamically aggregate entities of the same type in the scene (such as roadside trees and streetlights). During the rendering process, the entire group of objects is batch-drawn in a single draw call. Each instance dynamically adjusts its spatial pose and shape changes through a model matrix, and material parameters are distributed on demand (such as leaf color gradients and streetlight lighting status switching) by combining semantic information. At the shader level, the personalized feature calculation of each entity is realized through an instance ID indexing mechanism, which effectively reduces the frequency of draw calls (for example, it can be reduced to one-tenth of that in traditional rendering modes) while effectively maintaining visual differences. The above-described semantic clustering-based instantiation method can achieve efficient visualization of large-scale entity clusters while maintaining the complete presentation of micro-details, thus effectively supporting the ultra-large-scale deployment of micro-entities in macro-scenes.

[0047] In another exemplary embodiment, step S400, the dynamic parsing and spatial overlay of the KML data associated with the key entity objects after detail enhancement and visualization enhancement, to construct a multi-level three-dimensional scene information system from entity annotation to macro-spatial annotation, includes the following steps: S401: Extracting KML elements based on KML data; In this step, the KML / KMZ file is read and its XML structure is parsed. KML features (such as Placemark and GroundOverlay) and their contained KML features (including coordinate sequences of points, lines, and polygons), attribute data (such as name, description, and style attributes), and extended information are identified and extracted. At the same time, hierarchical nodes such as Document and Folder are parsed to construct the logical organization relationship and dependency structure between features, so as to form a structured feature set that combines geometric, attribute, and hierarchical information.

[0048] S402: Perform coordinate transformation on the extracted KML features; In this step, the extracted KML feature coordinates need to be converted in real time from the WGS84 geographic coordinate system (longitude, latitude, elevation) used by the KML standard to a projection coordinate system consistent with the 3D scene (such as Web Mercator projection), and the precise 3D spatial position of the feature is determined based on the elevation value or terrain sampling results, so as to provide an accurate spatial reference for overlay rendering.

[0049] S403: Dynamically render and style KML elements after coordinate transformation; In this step, the corresponding vector graphics need to be generated in real time on the GPU according to the style definition of KML elements (such as color, line width, icon, fill transparency) or user-defined rules. Point elements are rendered as icons that always face the camera using billboard technology; line elements are drawn using anti-aliased line segment shaders and support dynamic width adjustment (such as 0.1 meters to 10 meters); and polygon elements are achieved by triangulation and fragment shaders to achieve semi-transparent color fill or texture overlay.

[0050] S404: Deeply integrates and renders the rendered KML elements with the oblique photogrammetry real-world base and semantic 3D Tiles to construct a multi-layered 3D scene information system from entity annotation to macro-spatial annotation.

[0051] In this step, this application first achieves accurate occlusion processing of KML elements and oblique photogrammetry real-world base based on depth testing and template buffering technology. Adaptive transparent blending and screen space ambient light occlusion ensure visual coordination between annotation information and 3D scene. At the same time, this application establishes a dynamic visual hierarchy management mechanism. This mechanism adopts an intelligent rendering strategy based on observation perspective and spatial relationship. By calculating the distance and spatial distribution between the viewpoint and scene elements in real time, it can dynamically adjust the display details and density of KML elements. Specifically, the geometric details and text annotations of KML elements are fully displayed in the near range, adjacent annotations are automatically simplified and merged at medium distance, and regional aggregation rendering is enabled at far distance, retaining only key outlines. In addition, this mechanism integrates depth detection and semantic priority evaluation, automatically resolves visual occlusion conflicts, ensures that high-weight elements (such as alarm devices) are always presented first, and uses alpha gradient and shape interpolation to achieve smooth transition between different levels.

[0052] Furthermore, at the interaction level, this application binds a spatial event listener to each KML element, supporting interactive behaviors such as click highlighting, hover tooltips, and drag-and-drop editing. It also utilizes GPU-accelerated ray-picking technology to trigger attribute panel display and spatial analysis functions in real time. Ultimately, this constructs a multi-granularity three-dimensional spatial information system covering micro-level device positioning, meso-level facility distribution, and macro-level administrative divisions, achieving end-to-end interactive capabilities from basic visualization to intelligent spatial analysis.

[0053] In another exemplary embodiment, step S500, which involves dynamically coupling the KML elements in the multi-level three-dimensional scene information system with terrain elevation data to form a high-precision digital elevation base supporting visualization, includes the following steps: S501: Construct a multi-resolution seamless terrain base as the spatial foundation, providing precise geometric support and coordinate mapping framework for KML elements; This step aims to construct a high-precision, continuous, and scale-adaptive terrain geometry foundation. Specifically, the process begins by dynamically scheduling terrain tile data of appropriate precision from a pre-stored multi-level digital elevation model pyramid based on the real-time distance between the camera viewpoint and the ground surface. This scheduling process is efficiently managed using a GPU-accelerated quadtree index structure, ensuring that high-priority tiles (typically those in the near-field region) receive the fastest data loading and rendering. To address potential geometric cracks at the edges of tiles of different precision levels, the system implements a real-time crack detection and repair algorithm in the GPU shader. This algorithm generates a seamless, smoothly transitioning connection surface through geometric interpolation of adjacent tile boundary vertices and visual gradient fusion. Ultimately, this process constructs a visually and geometrically continuous, unbroken multi-resolution terrain model. This model not only provides a realistic surface undulation but, more importantly, establishes a unified and precise coordinate mapping framework and geometric bearing surface for the spatial positioning of all subsequent KML elements, ensuring that any element can find its accurate 3D spatial location on this foundation.

[0054] S502: Dynamic elevation calibration of KML elements based on multi-resolution seamless terrain substrate; In this step, dynamic elevation calibration is performed on KML features based on the constructed multi-resolution seamless terrain base. Specifically, this involves mapping the planar coordinates (X, Y) of each key vertex of the KML feature (anchor point for point features, a path node sequence for line features, and boundary contour vertices for area features) onto the elevation texture data of the multi-resolution terrain base in real time. Next, a bilinear interpolation algorithm is used to calculate the precise elevation value (Z) of the terrain surface corresponding to that point, based on the values ​​of four elevation sampling points surrounding that planar coordinate. Based on this calculated elevation, the system dynamically reconstructs the 3D coordinates of the KML feature vertices, ensuring they closely conform to the terrain surface. This calibration process enables precise spatial adaptation of three types of features: point markers achieve correct height for vertical positioning; line features (such as roads and rivers) form continuous paths following the terrain undulations through segmented elevation interpolation; and area features (such as administrative divisions and lakes) achieve complete alignment with the surface morphology through synchronous calibration of all vertices. Through this series of operations, the originally independent KML features and terrain elevation data are geometrically deeply integrated, thereby forming a terrain-feature composite structure that carries rich semantic information (from KML) and high-precision geometric information (from terrain). S503: Real-time simulation and dynamic rendering of water level lines based on calibrated KML elements; This step utilizes the high-precision terrain-feature composite structure generated in step S502 to simulate dynamic hydrological effects, thereby verifying and improving the realistic representation capabilities of the digital elevation matrix. The system receives preset or real-time water level parameters in the GPU's fragment shader. For each pixel (fragment) in the scene, the shader compares its corresponding world space coordinate elevation with the current water level. All areas with elevations below the waterline are identified as flooded areas. Subsequently, the shader applies dynamic shading effects to these flooded areas, typically by mixing a specific water color (such as blue) and adjusting its transparency to simulate water, possibly supplemented by ripple normal maps to enhance the dynamic details of the water surface. For KML features that cross waterline boundaries (such as partially submerged bridges or dams), the system performs local clipping and may add wetness, water stain, and other material effects. This real-time simulation process not only generates realistic water level visualization, but more importantly, it verifies the accuracy and reliability of the digital elevation base when simulating real-world physical phenomena (water cover), greatly enhancing the realism and dynamic expressiveness of the three-dimensional scene, enabling the base to support highly realistic visualization of complex hydrological scenarios such as flood evolution and tidal changes.

[0055] S504: Maintain the spatial consistency between terrain and KML elements to form a high-precision digital elevation base that supports visualization.

[0056] In this step, this embodiment constructs an automated state monitoring and response loop to continuously monitor two types of key events: first, the dynamic scheduling of terrain data itself (such as viewpoint movement leading to the loading of tiles with different precision); and second, editing operations on KML elements (such as moving, adding, deleting, or modifying attributes). Once such a change event is detected, the system immediately and seamlessly and automatically re-triggers the complete data processing pipeline, from terrain tile scheduling and KML element coordinate transformation to dynamic elevation calibration (i.e., steps S501 and S502 above), and even hydrological effect recalculation (step S503). This closed-loop feedback system ensures that regardless of whether the user is browsing, navigating, editing data, or adjusting scene parameters, the spatial position and geometry of the KML elements maintain real-time and strict consistency with the latest terrain surface. Through this dynamic maintenance mechanism, the high-precision digital elevation base transforms from a static data snapshot into a flexible and adaptive dynamic structure, continuously and reliably providing accurate spatial references and geometric foundations for 3D visualization rendering, ensuring the continuity of high precision and high realism in visualization effects under different interactive states.

[0057] In another exemplary embodiment, step S600, which optimizes resource scheduling for the digital elevation base to achieve rendering of multi-source data 3D intelligent visualization, includes the following steps: S601: Construct a dynamic memory partitioning structure based on a priority model; In this step, the priority-based dynamic memory partitioning structure includes a visual focus area, a prediction preload area, and a global context area. The visual focus area (P0 area), as the core layer of the memory architecture, allocates approximately 80% of the memory resources and is directly connected to the rendering pipeline via a high-frequency bandwidth channel. It is dedicated to hosting the highest-precision terrain tiles and key entity models within the current view frustum. This partition employs a material-aware caching strategy, prioritizing the residence and protection of high-performance resources such as specular reflection materials and dynamic lighting components to ensure the physical accuracy and material fidelity of the visual center area. The prediction preload area (P1 area), as a machine learning-driven intelligent buffer layer, allocates approximately 15% of the memory resources. It uses an asynchronous transmission mechanism and dynamic decompression technology to preload multi-precision data of surrounding areas based on the predicted viewpoint movement trajectory. This partition maintains the continuous distribution of adjacent tiles in the memory space through a topology-aware storage layout, effectively improving GPU cache hit rate and ensuring a seamless transition during viewpoint movement. The Global Context Area (P2 area), serving as the foundational support layer for scene coherence, allocates approximately 5% of video memory resources. It employs streaming loading and multi-level detail mapping techniques to maintain a low-precision topological representation of the global scene. This area utilizes a spatiotemporally consistent hash index to intelligently eliminate historical data and rapidly reconstruct the global context, providing a continuously stable spatial reference framework for large-scale scene browsing and effectively preventing frequent I / O operations.

[0058] The aforementioned three-partition structure achieves deep adaptation with the GPU architecture through a hardware abstraction layer, forming a gradient resource distribution from the visual focus to the global environment, which improves the utilization of video memory bandwidth by about 40%, thereby reducing rendering latency caused by data scheduling.

[0059] S602: Based on the dynamic partitioning structure of video memory, predicts the viewpoint movement trajectory based on machine learning; In this step, models such as Long Short-Term Memory (LSTM) networks can be used to analyze the user's historical movement trajectory (including movement direction, speed, and acceleration) to predict their most likely movement path and target area in the near future, thereby guiding the prediction of the preloaded content of the preload area data.

[0060] S603: Based on the prediction results, schedule and preload resources according to priority; In this step, the resource priority of the preloaded area can be dynamically calculated based on the prediction results of S602, and the terrain and entity model data in the prediction area can be actively scheduled to the P1 area (preloaded area) configured by S601 for asynchronous loading, so as to ensure that the data is ready before the user's perspective is switched, thereby achieving a seamless and smooth visual experience.

[0061] S604: Real-time load monitoring and adaptive adjustment.

[0062] In this step, the system monitors key performance indicators such as frame rate, video memory usage, and computational load in real time and compares them with preset thresholds. When excessive system load or resource conflicts are detected, a dynamic adjustment strategy is automatically triggered. This strategy can automatically adjust the capacity ratio of each video memory partition and reduce the precision level of preloaded data to achieve an adaptive balance between terrain rendering complexity and real-time system load, ensuring the overall stability of the system.

[0063] In summary, to address the issue of poor compatibility of multi-source data in 3D visualization technology, this application constructs a unified data loading and fusion pipeline, achieving dynamic integration and collaborative rendering of various heterogeneous data formats, including oblique photogrammetry models, vector data, GLB / GLTF refined models, and KML data. This application generates semantic 3D tiles using GPU-accelerated vector transformation technology and unifies data from different coordinate systems to the Web Mercator projection space through coordinate transformation and spatial registration algorithms. This improves data format compatibility and the ability to render cross-source data in an integrated manner, overcoming the limitations of traditional platforms in terms of limited format support and difficulties in collaborative rendering.

[0064] To address the issue of low dynamic model-terrain fitting accuracy in 3D visualization technology, this application innovatively adjusts the model height and terrain fitting intensity using vertex shaders, and integrates multi-resolution terrain fusion and real-time calibration algorithms to perform sub-meter-level dynamic elevation calibration of KML features. This method utilizes bilinear interpolation to calculate the ground-hugging height in real time, ensuring millimeter-level precise fitting of point, line, and surface features to the terrain surface. This solves the visual and accuracy problems of traditional methods, such as model floating, embedding, and large fitting errors.

[0065] To address the severe resource contention issue in 3D visualization technology, this application designs a dynamic memory partitioning mechanism based on machine learning prediction. This mechanism divides GPU memory into a visual focus area, a prediction preload area, and a global context area, optimizing data scheduling through viewpoint trajectory prediction and resource preloading strategies. This mechanism can improve memory bandwidth utilization by approximately 40%, significantly reducing resource contention and I / O latency in high-concurrency scenarios, effectively preventing memory overflow and rendering stutters, and achieving adaptive load balancing and stable system operation.

[0066] To address the issue of limited interactive functionality in 3D visualization technology, this application integrates a dynamic LOD mechanism with GPU-accelerated rendering technology to ensure high frame rate rendering from any viewpoint. Simultaneously, by binding spatial event listeners to KML elements and employing GPU ray picking technology to support click highlighting, drag-and-drop editing, and real-time attribute querying, a multi-layered interactive system from microscopic entity annotation to macroscopic spatial analysis is constructed. This enhances the system's real-time responsiveness and interactive experience, thereby helping to change the current situation of high latency and limited functionality in traditional platforms.

[0067] In another exemplary embodiment, this application also provides a three-dimensional intelligent visualization device for multi-source data loading, such as... Figure 2As shown, the device includes: a loading module 100, used to load and dynamically adjust an oblique photogrammetry model to construct an oblique photogrammetry real-world base; and a conversion module 200, used to generate semantic 3D based on the oblique photogrammetry real-world base using a GPU-accelerated vector conversion pipeline. Tiles data is used to achieve integrated rendering of macro-geographic environment and micro-entity model; Enhancement module 300 is used to identify key entity objects, perform lightweight processing on the high-precision GLB / GLTF refined model corresponding to the key entity objects, and dynamically embed it into the integrated rendered macro-geographic environment and micro-entity model to achieve detail enhancement and visualization enhancement of key entity objects; Parsing module 400 is used to dynamically parse and spatially overlay the KML data associated with the key entity objects after detail enhancement and visualization enhancement to construct a multi-level three-dimensional scene information system from entity annotation to macro-spatial annotation; Calibration module 500 is used to dynamically couple the KML elements in the multi-level three-dimensional scene information system with terrain elevation data to form a high-precision digital elevation base supporting visualization; Optimization module 600 is used to optimize resource scheduling of the digital elevation base to achieve rendering of multi-source data three-dimensional intelligent visualization.

[0068] Based on the above embodiments, refer to Figure 3 The computer-readable storage medium of exemplary embodiments of this application will be described below. Please refer to [link / reference]. Figure 3 The computer-readable storage medium shown is an optical disc 70, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it implements the steps described in the above method embodiments, such as loading and dynamically adjusting an oblique photogrammetry model to construct an oblique photogrammetry real-world base; and generating semantic 3D based on the oblique photogrammetry real-world base using a GPU-accelerated vector transformation pipeline. Tiles data is used to achieve integrated rendering of macroscopic geographic environment and microscopic entity models. Key entity objects are identified based on external business data. The high-precision GLB / GLTF refined models corresponding to these key entity objects are lightweighted and dynamically embedded into the integrated rendered macroscopic geographic environment and microscopic entity models to enhance the detail and visualization of key entity objects. The KML data associated with the enhanced key entity objects is dynamically parsed and spatially overlaid to construct a multi-layered 3D scene information system from entity annotation to macroscopic spatial annotation. Through multi-resolution terrain fusion and real-time calibration algorithms, the KML elements in the multi-layered 3D scene information system are dynamically coupled with terrain elevation data to form a high-precision digital elevation base supporting visualization. Based on a machine learning-predicted dynamic memory partitioning mechanism, resource scheduling optimization is performed on the digital elevation base to achieve an adaptive balance between terrain complexity and system load. The specific implementation methods of each step will not be repeated here.

[0069] It should be noted that the storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be described in detail here.

[0070] Based on the above embodiments, this application also proposes an electronic device, which is described below with reference to... Figure 4 An electronic device according to an exemplary embodiment of this application will be described.

[0071] Figure 4 A block diagram is shown of an exemplary electronic device 50 suitable for implementing embodiments of the present application. The electronic device 50 may be a computer system or a cloud server. Figure 4 The electronic device 50 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0072] like Figure 4 As shown, the components of electronic device 50 may include, but are not limited to: one or more processors or processing units 501, system memory 502, and bus 503 connecting different system components (including system memory 502 and processing unit 501).

[0073] Electronic device 50 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by electronic device 50, including volatile and non-volatile media, removable and non-removable media.

[0074] System memory 502 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 5021 and / or cache memory 5022. Electronic device 50 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, ROM 5023 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 4 (Not shown in the image, usually referred to as "hard drive"). Although not shown in... Figure 4As shown, but still available are disk drives for reading and writing to removable non-volatile disks (e.g., "floppy disks") and optical disc drives for reading and writing to removable non-volatile optical discs (e.g., CD-ROMs, DVD-ROMs, or other optical media). In these cases, each drive may be connected to bus 503 via one or more data media interfaces. System memory 502 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application.

[0075] A program / utility 5025 having a set (at least one) of program modules 5024 may be stored, for example, in system memory 502, and such program modules 5024 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment. Program modules 5024 typically perform the functions and / or methods described in the embodiments of this application.

[0076] Electronic device 50 can also communicate with one or more external devices 504 (such as a keyboard, pointing device, display, etc.). This communication can be performed through input / output (I / O) interface 505. Furthermore, electronic device 50 can also communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via network adapter 506. Figure 4 As shown, network adapter 506 communicates with other modules of electronic device 50 (such as processing unit 501) via bus 503. It should be understood that, although... Figure 4 As not shown, it can be used in conjunction with electronic device 50 with other hardware and / or software modules.

[0077] The processing unit 501 executes various functional applications and data processing by running programs stored in the system memory 502. For example, it loads and dynamically adjusts an oblique photogrammetry model to construct an oblique photogrammetry real-world base; based on the oblique photogrammetry real-world base, it generates semantic 3D data through a GPU-accelerated vector transformation pipeline. Tiles data is used to achieve integrated rendering of macroscopic geographic environment and microscopic entity model; key entity objects are identified based on external business data, and the high-precision GLB / GLTF refined models corresponding to the key entity objects are lightweighted and dynamically embedded into the integrated rendered macroscopic geographic environment and microscopic entity model to achieve detail enhancement and visualization enhancement of key entity objects; KML data associated with the key entity objects after detail enhancement and visualization enhancement is dynamically parsed and spatially overlaid to construct a multi-level 3D scene information system from entity annotation to macroscopic spatial annotation; through multi-resolution terrain fusion and real-time calibration algorithms, KML elements in the multi-level 3D scene information system are dynamically coupled with terrain elevation data to form a high-precision digital elevation base supporting visualization; based on a machine learning prediction-based dynamic memory partitioning mechanism, resource scheduling optimization is performed on the digital elevation base to achieve an adaptive balance between terrain complexity and system load. The specific implementation methods of the above steps will not be repeated here. It should be noted that although several units / modules or sub-units / sub-modules of the file concurrent download device are mentioned in the detailed description above, this division is only exemplary and not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules for embodiment.

[0078] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0079] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0080] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0081] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0082] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions that can execute all or part of the steps of the methods described in the various embodiments of this application based on a computer device (which may be a personal computer, a cloud server, or a network device, etc.). The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0083] The above embodiments are only for illustrating the technical concept and features of this application, and are intended to enable those skilled in the art to understand the content of this application and implement it accordingly. They should not be construed as limiting the scope of protection of this application. All equivalent changes or modifications made in accordance with the spirit and essence of this application should be included within the scope of protection of this application.

Claims

1. A three-dimensional intelligent visualization method for loading multi-source data, characterized in that, The method includes: Load the oblique photogrammetry model and make dynamic adjustments to construct an oblique photogrammetry real-world base; Based on the aforementioned oblique photogrammetry real-world base, semantic 3D tile data is generated through a GPU-accelerated vector transformation pipeline to achieve integrated rendering of the macro-geographic environment and micro-entity models. Identify key entity objects, perform lightweight processing on the high-precision GLB / GLTF refined models corresponding to the key entity objects, and dynamically embed them into the macro-geographic environment and micro-entity models after integrated rendering, so as to achieve detail enhancement and visualization enhancement of key entity objects. Dynamic parsing and spatial overlay of KML data associated with key entity objects after detail enhancement and visualization enhancement are performed to construct a multi-level three-dimensional scene information system from entity annotation to macro-spatial annotation. The KML elements in the multi-level three-dimensional scene information system are dynamically coupled with the terrain elevation data to form a high-precision digital elevation base that supports visualization. Resource scheduling optimization is performed on the digital elevation base to achieve three-dimensional intelligent visualization rendering of multi-source data.

2. The three-dimensional intelligent visualization method for multi-source data loading according to claim 1, characterized in that, The loading and dynamic adjustment of the oblique photogrammetry model to construct the oblique photogrammetry real-world base includes: Based on the inherent tiled storage structure of the oblique photogrammetry model, model tiles within the visible area are located and loaded using R-tree spatial indexing. It integrates a dynamic LOD mechanism, which automatically switches data at different detail levels based on the real-time distance between the camera viewpoint and the tile. The model height and terrain fit strength are dynamically adjusted by using vertex shaders to complete the construction of the oblique photogrammetry real-world base.

3. The three-dimensional intelligent visualization method for multi-source data loading according to claim 1, characterized in that, The process of generating semantically coded 3D tiles data based on the oblique photogrammetry real-world base using a GPU-accelerated vector transformation pipeline to achieve integrated rendering of the macroscopic geographical environment and microscopic entity models includes: The vector data in the geographic information system is preprocessed to match the oblique photogrammetric real-world base. 3D Tiles are generated based on preprocessed vector data; Dynamically stylize and render the generated 3D tiles; The rendered 3D tiles are deeply integrated and coupled with the oblique photogrammetric real-world base in a unified spatial coordinate system to achieve integrated rendering of the macro-geographic environment and the micro-entity model.

4. The three-dimensional intelligent visualization method for multi-source data loading according to claim 1, characterized in that, The process of identifying key entity objects involves lightweighting the high-precision GLB / GLTF refined models corresponding to these key entity objects and dynamically embedding them into the integrated rendered macro-geographic environment and micro-entity models to achieve detail enhancement and visualization enhancement of the key entity objects. This includes: Identify key entities based on external business data; Lightweighting is performed on the refined GLB / GLTF models corresponding to key entity objects; The lightweight GLB / GLTF refined model is embedded into the macro-geographic environment and micro-entity model after integrated rendering to achieve detailed enhancement and visualization of key entity objects.

5. The three-dimensional intelligent visualization method for multi-source data loading according to claim 1, characterized in that, The dynamic parsing and spatial overlay of KML data associated with key entity objects after detail enhancement and visualization enhancement are used to construct a multi-layered 3D scene information system from entity annotation to macro-spatial annotation, including: KML elements are extracted from KML data; Perform coordinate transformation on the extracted KML features; Dynamically render and style KML elements after coordinate transformation; The rendered KML elements are deeply integrated with the oblique photogrammetry real-world base and semantic 3D tiles for rendering and interaction, in order to construct a multi-level three-dimensional scene information system from entity annotation to macro-spatial annotation.

6. The three-dimensional intelligent visualization method for multi-source data loading according to claim 1, characterized in that, The step of dynamically coupling KML elements in the multi-level three-dimensional scene information system with terrain elevation data to form a high-precision digital elevation base supporting visualization includes: Construct a multi-resolution seamless terrain base as the spatial foundation, providing precise geometric support and coordinate mapping framework for KML elements; Dynamic elevation calibration of KML elements based on multi-resolution seamless terrain substrate; Based on the calibrated KML features, the water level line is simulated and dynamically rendered in real time. The spatial consistency between terrain and KML elements is maintained to form a high-precision digital elevation base that supports visualization.

7. The three-dimensional intelligent visualization method for multi-source data loading according to claim 1, characterized in that, The resource scheduling optimization of the digital elevation base to achieve 3D intelligent visualization rendering of multi-source data includes: Construct a dynamic memory partitioning structure based on a priority model; Based on the dynamic partitioning structure of video memory, the viewpoint movement trajectory is predicted using machine learning. Based on the prediction results, resources are prioritized and preloaded. Real-time load monitoring and adaptive adjustments are performed to achieve intelligent 3D visualization rendering of multi-source data.

8. A three-dimensional intelligent visualization device for loading multi-source data, characterized in that, The device includes: The loading module is used to load the oblique photogrammetry model and make dynamic adjustments to construct the oblique photogrammetry real-world base. The conversion module is used to generate semantic 3D tiles data based on the oblique photogrammetry real scene base through a GPU-accelerated vector conversion pipeline, so as to achieve integrated rendering of macro-geographic environment and micro-entity model; The enhancement module is used to identify key entity objects, perform lightweight processing on the high-precision GLB / GLTF refined model corresponding to the key entity objects, and dynamically embed it into the macro-geographic environment and micro-entity model after integrated rendering, so as to achieve detail enhancement and visualization enhancement of key entity objects. The parsing module is used to dynamically parse and spatially overlay the KML data associated with key entity objects after detail enhancement and visualization enhancement, so as to construct a multi-level three-dimensional scene information system from entity annotation to macro-spatial annotation. The calibration module is used to dynamically couple the KML elements in the multi-level three-dimensional scene information system with the terrain elevation data to form a high-precision digital elevation base that supports visualization. The optimization module is used to optimize resource scheduling for the digital elevation base to achieve rendering of multi-source data in three-dimensional intelligent visualization.

9. A storage medium, characterized in that, It includes instructions that, when executed on a computer, cause the computer to perform the method described in any one of claims 1 to 7.

10. An electronic device, characterized in that, The electronic device includes: Memory, processor, and computer programs stored in memory and executable on the processor, wherein, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.

Images