Lightweight transmission and loading optimization method for high-precision three-dimensional model cloud rendering

Abstract

The invention discloses a lightweight transmission and loading optimization method for cloud rendering of a high-precision three-dimensional model, belongs to the technical field of network transmission, and aims to solve the problems of large data volume, high transmission delay and slow loading of cloud rendering of the high-precision three-dimensional model. Carrying out importance evaluation on the rendering primitives, constructing a feature matrix and a weight vector, and determining important data points through matrix multiplication; important data points are subjected to lightweight coding and preferential transmission, and non-important data points are subjected to high-strength compression or delayed transmission. At the client side, client importance re-evaluation is carried out on received rendering data blocks, a client feature matrix and a weight vector are constructed, high-priority loading data are determined through matrix multiplication, and decoding, loading and rendering are preferentially carried out. According to the invention, lightweight transmission and rapid loading of the high-precision three-dimensional model are realized, and the user experience and the resource utilization efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of network transmission, and in particular to a lightweight transmission and loading optimization method for high-precision three-dimensional model cloud rendering. Background Art

[0002] With the rapid development of 3D modeling technology and cloud computing, cloud rendering of high-precision 3D models is becoming more and more widely used, such as online design, virtual reality, augmented reality and digital twins. However, high-precision 3D models usually contain massive amounts of geometric data, texture data and material data. After cloud rendering is completed, the rendering results need to be transmitted to the client through the network. Traditional transmission methods face bandwidth limitations and network latency challenges, which may cause the client to receive data stuck and load slowly, seriously affecting the user experience. In the existing technology, although there are some compression and streaming transmission methods, it is often difficult to achieve the best balance between high precision and low latency. In particular, there is a lack of refined processing strategies for the different important parts of the rendering data, resulting in room for improvement in transmission efficiency and loading speed. Summary of the Invention

[0003] In response to the above-mentioned problems existing in the prior art, the present invention aims to provide a lightweight transmission and loading optimization method for high-precision 3D model cloud rendering, including a cloud rendering lightweight transmission method, which is executed by a cloud rendering server and includes: Step S1, obtaining the original rendering scene data of the high-precision three-dimensional model and preprocessing the original rendering scene data, wherein the preprocessing includes constructing a hierarchical detail structure of the rendering scene data and generating a spatial division structure of the rendering scene data, and preliminarily compressing the common data in the rendering scene data.

[0004] Step S2: performing an importance evaluation on the rendering primitives in the rendering scene data according to the viewpoint data transmitted in real time by the client to generate an importance score for the rendering primitives, wherein the importance evaluation is based on at least one of the distance from the rendering primitive to the client viewpoint, the projection area on the client screen, the geometric curvature, the visibility, the user interaction, and the attention factor.

[0005] Step S3: construct a feature matrix and a weight vector of the rendering primitive, calculate the importance score vector of the rendering primitive by performing a matrix multiplication operation on the feature matrix and the weight vector, and determine the important data points in the rendering primitive that need to be transmitted first based on the importance score vector.

[0006] Step S4: Lightweight encoding is performed on the important data points and they are transmitted preferentially, and high-intensity compression encoding or delayed transmission is performed on the unimportant data points.

[0007] Furthermore, the client loading step, performed by the client, includes: Step Y1: Receive the lightweight transmitted rendering data stream.

[0008] Step Y2: re-evaluate the client importance of the rendering data blocks in the rendering data stream according to at least one of the current rendering state, display resolution, and user interaction of the client.

[0009] Step Y3: constructing a client feature matrix and a client weight vector of the rendering data block, and determining high-priority loading data in the rendering data block that needs to be loaded and rendered first through matrix multiplication operation.

[0010] Step Y4, preferentially decode, decompress and load the high-priority loading data for rendering, and gradually decode, decompress and load the low-priority loading data for detail supplementary rendering, gradually improving the fineness of the rendering picture.

[0011] Furthermore, the hierarchical detail structure includes multiple geometric representations of the three-dimensional model at different levels of geometric detail, and the geometric representations are polygon numbers or texture resolutions from coarse to fine, so as to dynamically select and adjust the display complexity of the three-dimensional model according to the client viewpoint distance or the importance of the rendering primitive; and perform preliminary lossless or light lossy compression on the geometric data, texture data and material data in the rendering scene data to reduce the data volume and prepare for subsequent transmission.

[0012] Furthermore, in step S2, the calculation formula for the importance score of the rendering primitive is: ,in Indicates the distance from the rendered primitive to the client viewpoint. Represents the projected area of ​​the rendered primitive on the client screen. Indicates the geometric curvature of the area where the rendering primitive is located. Represents the visibility factor of the rendered primitive, Indicates the attention factor of the rendered primitive, Represents the corresponding weighting coefficient.

[0013] Furthermore, in the S3 step, the feature matrix and weight vector of the rendering primitive are constructed by performing matrix multiplication operation on the feature matrix and the weight vector: the distance, projected area, geometric curvature, visibility, and attention factor contained in the rendering primitive are standardized to ensure the dimensional consistency of each feature in the feature matrix; the feature matrix is ​​a matrix of , wherein is the total number of the rendering primitives, is the number of features corresponding to each rendering primitive, each row of the feature matrix represents a feature vector of the rendering primitive, and each component of the feature vector corresponds to at least one of the distance, projected area, geometric curvature, visibility, and attention factor of the rendering primitive; the column vector of the weight vector, wherein each element corresponds to the weighting coefficient of the corresponding feature in the feature matrix.

[0014] Furthermore, in the S3 step, the important data points in the rendering primitives that need to be transmitted first are determined based on the importance score vector, a transmission importance threshold is set, and the rendering primitives in the importance score vector whose scores are greater than or equal to the transmission importance threshold are marked as the important data points. The transmission importance threshold is dynamically adjusted based on the current network bandwidth, client device performance or user tolerance to transmission delays.

[0015] Furthermore, in the S4 step, the transmission method of the non-important data points includes high-intensity compression coding or delayed transmission, and the high-intensity compression coding adopts a lossy compression algorithm, including low-resolution compression of textures in non-critical areas; the delayed transmission refers to transmitting the non-important data points after the network bandwidth allows or the transmission of the important data points is completed.

[0016] Furthermore, the client loading step also includes, after step Y1, the client dynamically adjusts the buffering strategy or loading order of the rendering data block according to the priority of the data block in the rendering data stream, the system resource load of the client, and the current display performance, so as to ensure that the high-priority loaded data can preferentially occupy and utilize the client resources for rapid decoding, decompression or uploading to the graphics processor, while the low-priority loaded data is delayed or dynamically unloaded, thereby optimizing the resource utilization efficiency of the client.

[0017] Furthermore, in step Y3, high-priority loading data in the rendering data blocks that need to be loaded and rendered first is determined, and the client feature matrix is ​​a matrix of , wherein is the total number of the rendering data blocks received or to be loaded, and is the number of client-side features corresponding to each rendering data block, and each row of the client feature matrix represents a feature vector of the rendering data block, and each component of the feature vector corresponds to at least one of the geometric detail level, rendering readiness, or user interaction area correlation of the rendering data block; the client weight vector is a column vector of , wherein each element corresponds to the weighting coefficient of the corresponding feature in the client feature matrix; a loading importance threshold is set and the rendering data blocks with scores greater than or equal to the loading importance threshold in the client importance score vector are marked as high-priority loading data.

[0018] Furthermore, in step Y4, the client uses high-priority loaded data to quickly generate and display a rough skeleton of the three-dimensional model during the first rendering; after the high-priority loaded data is rendered, the client gradually decodes, decompresses and loads the low-priority loaded data, and performs detail supplement rendering on it. The detail supplement rendering includes at least one of the following: loading a higher-level detail model of the three-dimensional model to improve geometric accuracy; loading higher-resolution texture data of the three-dimensional model to improve texture clarity; applying more complex materials or lighting effects to improve visual realism; the client dynamically updates the rendering pipeline or pixel-by-pixel shader to apply detail supplement rendering to the displayed three-dimensional model in real time, thereby gradually improving the fineness of the rendered image.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The present invention introduces a viewpoint-aware importance assessment model to achieve refined grading and differentiated transmission of rendering data, significantly improving transmission efficiency. The present invention can accurately identify the most important data points that are most critical to the user's visual experience based on factors such as the client's real-time viewpoint, the projected area of ​​the model on the screen, and the complexity of geometric details. By prioritizing and efficiently transmitting important data points while highly compressing or delaying the transmission of non-important data points, the bandwidth waste and transmission bottlenecks caused by the traditional method of treating all data equally are avoided, ensuring the rapid delivery of key information, thereby achieving rapid response and display of high-precision three-dimensional models under limited bandwidth.

[0020] The present invention implements an importance re-evaluation and hierarchical loading mechanism based on matrix calculation on the client side, which greatly optimizes the local loading and rendering experience of high-precision three-dimensional models. The client can dynamically re-evaluate the loading priority of the received data according to its own device performance, rendering status and user interaction behavior. By utilizing matrix calculation to efficiently process data in batches, high-priority data is decoded, loaded and rendered first, allowing users to quickly obtain a rough overview of the model and interact. Subsequently, low-priority data is gradually added and loaded, and the picture fineness is gradually improved, effectively alleviating the problems of long waiting time for loading large models and initial rendering jams, providing users with a smoother and real-time three-dimensional model browsing experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The figure is a flowchart of an exemplary step of the lightweight transmission method of the present invention.

[0022] Figure 2 An exemplary flowchart of the optimization steps of the present invention is shown below. DETAILED DESCRIPTION

[0023] The present invention will be further described below with reference to specific embodiments.

[0024] This application discloses a lightweight transmission and loading optimization method for high-precision 3D model cloud rendering. This method aims to significantly improve the rendering efficiency, responsiveness, and user experience of complex 3D models across various devices and network conditions by intelligently managing and optimizing the transmission and client loading of 3D rendering data. This method is particularly suitable for cloud rendering scenarios. Through server-side data preprocessing, importance assessment, and encoded transmission, and client-side reception, reassessment, and hierarchical loading and rendering, it achieves a balance between high-precision visual effects and a smooth interactive experience.

[0025] In one embodiment, the method includes two main parts: a cloud-based rendering lightweight transmission method performed by a cloud-based rendering server, and a client-side loading step performed by a client. These two parts work together to optimize the rendering process of the 3D model.

[0026] like Figure 1 The following is a lightweight transmission method for cloud rendering in accordance with this embodiment, which is executed by a cloud rendering server and includes: Step S1, obtaining the original rendering scene data of the high-precision three-dimensional model and preprocessing the original rendering scene data, the preprocessing includes constructing a hierarchical detail structure of the rendering scene data and generating a spatial division structure of the rendering scene data, and preliminarily compressing the common data in the rendering scene data.

[0027] In one embodiment, step S1 is a key step in server-side data preparation.

[0028] Obtaining raw rendering scene data: The server first retrieves complete, high-precision 3D model data from a data storage system or asset library. This data may include the model's geometry, topology, texture data, material properties, animation data, lighting information, and other entities and attributes in the scene. This raw data is often very large and unsuitable for direct transmission. Preprocessing: Preprocessing optimizes the data structure and size for efficient subsequent transmission and rendering. Constructing a level-of-detail structure: This structure contains multiple geometric representations of the 3D model at different levels of detail. For example, polygon simplification can be performed on the original model to generate a series of model versions with varying vertex and face counts. These versions can be dynamically selected and adjusted for display complexity based on the client's viewpoint distance. Furthermore, texture images of varying resolutions can be generated for texture data. Generating a spatial partitioning structure: To efficiently perform frustum culling and level-of-detail structure selection, the server constructs spatial partitioning structures, such as octrees, quadtrees, and bounding box hierarchies. These structures divide the entire 3D scene into several subregions, each containing a portion of rendering primitives, so that the client only loads and renders data for the viewpoint's visible area. Initial compression: Before transmission, the common data in the rendered scene data is initially compressed. For example, a lossless compression algorithm or a light lossy compression algorithm can be used to reduce the data volume while maintaining good data quality to prepare for subsequent transmission.

[0029] Step S2: Based on the viewpoint data transmitted in real time by the client, an importance evaluation is performed on the rendering primitives in the rendering scene data to generate an importance score for the rendering primitives. The importance evaluation is based on at least one of the distance from the rendering primitive to the client viewpoint, the projected area on the client screen, the geometric curvature, the visibility, the user interaction, and the attention factor.

[0030] In step S2, the calculation formula for the importance score of the rendering primitive is: ,in Indicates the distance from the rendered primitive to the client's viewpoint. Indicates the projection area of ​​the rendered primitive on the client screen. Indicates the geometric curvature of the area where the rendering primitive is located. Represents the visibility factor of the rendered primitive, Indicates the attention factor of the rendered primitive, Represents the corresponding weighting coefficient.

[0031] In one embodiment, the parameters and weights of the formula are specifically defined as follows: Indicates the comprehensive importance score of a single rendered primitive. The higher the score, the more critical the primitive is to the user's visual experience or interactive behavior at the current viewpoint, and should be prioritized for transmission and rendering.

[0032] Indicates the Euclidean distance from the rendered primitive to the client's viewpoint. The term means that the closer the distance is, the greater the contribution of the term is, which is consistent with the visual perception that "near objects are more important".

[0033] Represents the projected area of ​​the rendered primitive on the client screen. This value is calculated based on the primitive's actual size and distance from the viewpoint, reflecting the visual proportion of the primitive on the screen. The larger the projected area, the greater the contribution.

[0034] Indicates the geometric curvature of the region where the primitive is rendered. For example, for a mesh, this could be the average of the rate of change of vertex normals or the Gaussian curvature. The greater the curvature, the greater the contribution of this term.

[0035] Indicates the visibility factor of the rendered primitive. This factor is typically 0 or 1, indicating whether the primitive is within the current view frustum and not completely obscured by other objects. It is 1 if visible and 0 if not, ensuring that only visible data is transmitted. In more complex implementations, it can be a value between 0 and 1 to indicate partial occlusion.

[0036] Indicates the attention factor of the rendered primitive. This is an indicator that reflects the user's intention. For example, if the user is pointing or clicking an object with the mouse, or if the user is gazing at a certain area through eye tracking, the primitives in that area will be more visible. The value will increase significantly.

[0037] These weight coefficients are configurable and are used to adjust the relative contribution of each importance assessment factor in the total score. In one embodiment, these weights can be adjusted according to the application type. For example, for applications that require high interactive accuracy, the weights can be appropriately increased. 、 For applications that need to emphasize visual details, the weights of and can be increased. These weights are usually optimized through expert experience, user research data, or machine learning algorithms.

[0038] Step S3, constructing the feature matrix and weight vector of the rendering primitive, calculating the importance score vector of the rendering primitive by performing matrix multiplication operation on the feature matrix and the weight vector, and determining the important data points in the rendering primitive that need to be transmitted first based on the importance score vector.

[0039] In step S3, the feature matrix and weight vector of the rendering primitive are constructed by performing matrix multiplication operation on the feature matrix and the weight vector: the distance, projected area, geometric curvature, visibility, and attention factor contained in the rendering primitive are standardized to ensure the dimensional consistency of each feature in the feature matrix; the feature matrix is ​​a matrix of , where is the total number of rendering primitives and is the number of features corresponding to each rendering primitive. Each row of the feature matrix represents a feature vector of a rendering primitive, and each component of the feature vector corresponds to at least one of the distance, projected area, geometric curvature, visibility, and attention factor of the rendering primitive; the weight vector is a column vector, where each element corresponds to the weighting coefficient of the corresponding feature in the feature matrix.

[0040] In step S3, the important data points in the rendering primitives that need to be transmitted first are determined based on the importance score vector, a transmission importance threshold is set, and the rendering primitives in the importance score vector whose scores are greater than or equal to the transmission importance threshold are marked as important data points. The transmission importance threshold is dynamically adjusted based on the current network bandwidth, client device performance, or the user's tolerance for transmission delay.

[0041] Step S4: Lightweight encoding is performed on important data points and they are transmitted first, while high-intensity compression encoding or delayed transmission is performed on non-important data points.

[0042] In one embodiment, step S4 is the execution link of the server-side transmission strategy, which embodies the essence of differentiated transmission.

[0043] Important data point processing: Lightweight encoding of the important data points. This means using an efficient and fast decoding encoding algorithm and maintaining high data quality as much as possible. For example, progressive grid coding, efficient geometric data compression algorithms, or high-quality texture compression formats can be used. These encoded data will be transmitted to the client via the network first. Non-important data point processing: For the non-important data points, the system will adopt a more aggressive transmission strategy to maximize bandwidth and server resource savings. High-intensity compression encoding: A lossy compression algorithm is used to achieve a higher compression ratio within an acceptable visual loss range. For example, low-resolution compression can be performed for textures in non-critical areas; more aggressive simplification can be performed for geometric details. Delayed transmission: Delayed transmission means transmitting the non-important data points after the network bandwidth allows or the transmission of the important data points is completed. This means that non-important data will only be sent when the network is idle or the client has completed rendering the main content, thereby avoiding blocking the transmission channel for critical data and ensuring the rapid presentation of core visual content.

[0044] In step S4, the transmission methods of non-important data points include high-intensity compression coding or delayed transmission. High-intensity compression coding uses a lossy compression algorithm, including low-resolution compression of textures in non-critical areas; delayed transmission refers to transmitting non-important data points after network bandwidth allows or the transmission of important data points is completed.

[0045] like Figure 2 The following are the client loading steps of this embodiment, which are executed by the client: Step Y1: Receive the lightweight rendered data stream. In one embodiment, step Y1 involves the client receiving data sent by the server. The client continuously monitors the rendered data stream from the server and caches it for processing. Because the server performs differentiated transmission and encoding, the data streams received by the client exhibit different priorities, with important data arriving first.

[0046] Step Y2: re-evaluate the client importance of the rendering data blocks in the rendering data stream according to at least one of the current rendering state, display resolution, and user interaction of the client.

[0047] In one embodiment, step Y2 is a secondary intelligent evaluation performed by the client based on its own real-time situation. Although the server has already performed an importance evaluation, the client may have its own unique real-time status, and these factors require the client to adjust the importance by itself.

[0048] Client rendering status: for example, whether the current frame rate is smooth, video memory usage, CPU load, etc. If the client resources are tight, it may be necessary to lower the requirements for data accuracy and prioritize loading rough models. Display resolution: If the client screen resolution is low, high-resolution textures or extremely fine geometric details may not need to be loaded immediately. User interaction: Similar to the server side, if the user performs a new zoom, translation, or rotation operation on the client, or clicks on an object, the client will re-evaluate the importance of the data blocks that have been received or are being received based on these new interactive behaviors, so as to prioritize parsing and rendering the area or object that the user is currently focusing on. For example, when the user zooms in on an area, the priority of the data blocks in that area will be immediately increased, even if its priority is not high when it is first evaluated on the server side. Client feature matrix and weight vector : The client feature matrix for The matrix of is the total number of rendering data blocks received or to be loaded, The number of client-side features corresponding to each rendering data block. Each row of the feature matrix represents a feature vector of the rendering data block, and each component thereof corresponds to at least one of the geometric detail level, rendering readiness, or user interaction area relevance of the rendering data block. The client weight vector for A column vector where each element corresponds to the weighted coefficient of the corresponding feature in the client feature matrix. These coefficients can be dynamically adjusted based on client performance, network conditions, and user preferences. Loading importance threshold: A loading importance threshold is set, and rendering data blocks with a score greater than or equal to the loading importance threshold in the client importance score vector are marked as high-priority loading data. This threshold can also be dynamically adjusted based on the client's real-time resource status and the user's requirements for rendering smoothness.

[0049] Step Y3: construct a client feature matrix and a client weight vector of the rendering data block, and determine the high-priority loading data in the rendering data block that needs to be loaded and rendered first through matrix multiplication operation.

[0050] Step Y4, prioritize decoding, decompressing and loading high-priority loading data for rendering, and gradually decode, decompressing and loading low-priority loading data for detail supplementary rendering, gradually improving the fineness of the rendered image.

[0051] In one embodiment, step Y4 is the final step of the client performing data loading and rendering, which implements progressive rendering.

[0052] Coarse Skeleton Fast Rendering: During the initial rendering, the client utilizes the high-priority data identified in Y3 to prioritize decoding, decompression, and uploading it to the GPU. This quickly generates and displays a coarse skeleton or low-precision view of the 3D model. This ensures that users see a preliminary, interactive screen in the shortest possible time, even if details are still incomplete. This improves the user experience and avoids long blank screen waits. Detailed Rendering: After rendering the high-priority data, the client gradually decodes, decompresses, and loads the lower-priority data. This lower-priority data is used to render detailed supplementary data for the displayed coarse model, gradually increasing the resolution of the rendered image. Higher-Level Detail Model: Loads a higher-level detailed model of the 3D model to improve geometric accuracy and create smoother and more realistic surfaces. Higher-Resolution Texture Data: Loads higher-resolution texture data of the 3D model to enhance texture clarity and enrich surface details such as scratches and wear. Complex Materials or Lighting Effects: Apply more complex materials or lighting effects to enhance visual realism and create a more immersive experience. Dynamically updating the rendering pipeline: The client dynamically updates the rendering pipeline or pixel-by-pixel shaders, applying these additional details to the displayed 3D model in real time. This means the client doesn't need to re-render the entire scene, but instead incrementally adds details to the existing rendering, effectively improving the rendering's sophistication and achieving a smooth transition from blur to clarity, and from rough to fine.

[0053] In this embodiment, the hierarchical detail structure includes multiple geometric representations of the three-dimensional model at different levels of geometric detail. The geometric representations are the number of polygons or texture resolutions from coarse to fine, so as to dynamically select and adjust the display complexity of the three-dimensional model according to the client viewpoint distance or the importance of the rendered primitives; the geometric data, texture data and material data in the rendered scene data are preliminarily compressed without loss or with light loss to reduce the amount of data and prepare for subsequent transmission.

[0054] The client loading step of this embodiment also includes, after step Y1, the client dynamically adjusts the buffering strategy or loading order of the rendering data blocks according to the priority of the data blocks in the rendering data stream, the system resource load of the client, and the current display performance, so as to ensure that high-priority loaded data can preferentially occupy and utilize client resources for rapid decoding, decompression or uploading to the graphics processor, while low-priority loaded data is delayed or dynamically unloaded, thereby optimizing the resource utilization efficiency of the client.

[0055] In step Y3, high-priority loading data in the rendering data blocks that need to be loaded and rendered first is determined, and the client feature matrix is ​​a matrix of , where is the total number of rendering data blocks received or to be loaded, and is the number of client-side features corresponding to each rendering data block. Each row of the client feature matrix represents a feature vector of a rendering data block, and each component of the feature vector corresponds to at least one of the geometric detail level, rendering readiness, or user interaction area correlation of the rendering data block; the client weight vector is a column vector of , where each element corresponds to the weighting coefficient of the corresponding feature in the client feature matrix; a loading importance threshold is set and the rendering data blocks with scores greater than or equal to the loading importance threshold in the client importance score vector are marked as high-priority loading data.

[0056] In step Y4, when rendering for the first time, the client uses the high-priority loaded data to quickly generate and display the rough skeleton of the three-dimensional model; after the high-priority loaded data is rendered, the client gradually decodes, decompresses and loads the low-priority loaded data, and performs detail supplement rendering on it. The detail supplement rendering includes at least one of the following: loading a higher-level detail model of the three-dimensional model to improve geometric accuracy; loading higher-resolution texture data of the three-dimensional model to improve texture clarity; applying more complex materials or lighting effects to improve visual realism; the client dynamically updates the rendering pipeline or pixel-by-pixel shader to apply detail supplement rendering to the displayed three-dimensional model in real time, thereby gradually improving the fineness of the rendered image.

[0057] Those skilled in the art will appreciate that the above embodiments are merely illustrative and that various modifications and equivalent substitutions may be made without departing from the spirit and scope of the present invention. For example, the specific feature point algorithm, optimizer selection, and distortion model details can all be adjusted based on actual needs. Therefore, the scope of protection of the present invention shall be defined by the appended claims.

Claims

1. A lightweight transmission and loading optimization method for high-precision 3D model cloud rendering, characterized by: This includes a lightweight transmission method for cloud rendering, which is executed by a cloud rendering server and includes: Step S1, obtaining original rendered scene data of the high-precision three-dimensional model and preprocessing the original rendered scene data, wherein the preprocessing includes constructing a hierarchical detail structure of the rendered scene data and generating a spatial partitioning structure of the rendered scene data, and preliminarily compressing common data in the rendered scene data; Step S2: performing an importance evaluation on the rendering primitives in the rendering scene data according to the viewpoint data transmitted in real time by the client to generate an importance score for the rendering primitive, wherein the importance evaluation is based on at least one of a distance from the rendering primitive to the client viewpoint, a projected area on the client screen, geometric curvature, visibility, user interaction, and an attention factor; Step S3: constructing a feature matrix and a weight vector of the rendering primitive, calculating an importance score vector of the rendering primitive by performing a matrix multiplication operation on the feature matrix and the weight vector, and determining important data points in the rendering primitive that need to be transmitted first based on the importance score vector; Step S4: Lightweight encoding is performed on the important data points and they are transmitted preferentially, and high-intensity compression encoding or delayed transmission is performed on the unimportant data points.

2. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 1 is characterized by: Includes client loading steps, executed by the client, including: Step Y1, receiving the rendering data stream after the lightweight transmission; Step Y2, re-evaluating the client importance of the rendering data blocks in the rendering data stream according to at least one of the current rendering state, display resolution, and user interaction of the client; Step Y3: constructing a client feature matrix and a client weight vector of the rendering data block, and determining high-priority loading data in the rendering data block that needs to be loaded and rendered first through matrix multiplication operation; Step Y4, preferentially decode, decompress and load the high-priority loading data for rendering, and gradually decode, decompress and load the low-priority loading data for detail supplementary rendering, gradually improving the fineness of the rendering picture.

3. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 1 is characterized by: The level of detail structure includes multiple geometric representations of the three-dimensional model at different levels of geometric detail, wherein the geometric representations are polygon counts or texture resolutions from coarse to fine, so as to dynamically select and adjust the display complexity of the three-dimensional model based on the client viewpoint distance or the importance of the rendered primitives; Perform preliminary lossless or light lossy compression on the geometric data, texture data, and material data in the rendered scene data to reduce the data volume and prepare for subsequent transmission.

4. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 1 is characterized by: In step S2, the calculation formula for the importance score of the rendering primitive is: ,in Indicates the distance from the rendered primitive to the client viewpoint. Represents the projected area of ​​the rendered primitive on the client screen. Indicates the geometric curvature of the area where the rendering primitive is located. Represents the visibility factor of the rendered primitive, Indicates the attention factor of the rendered primitive, Represents the corresponding weighting coefficient.

5. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 1 is characterized by: In the S3 step, the feature matrix and weight vector of the rendering primitive are constructed by performing matrix multiplication operation on the feature matrix and the weight vector: the distance, projected area, geometric curvature, visibility, and attention factor contained in the rendering primitive are standardized to ensure the dimensional consistency of each feature in the feature matrix; the feature matrix is ​​a matrix of , wherein is the total number of the rendering primitives, is the number of features corresponding to each rendering primitive, each row of the feature matrix represents a feature vector of the rendering primitive, and each component of the feature vector corresponds to at least one of the distance, projected area, geometric curvature, visibility, and attention factor of the rendering primitive; the column vector of the weight vector, wherein each element corresponds to the weighting coefficient of the corresponding feature in the feature matrix.

6. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 1 is characterized by: In the S3 step, the important data points in the rendering primitives that need to be transmitted first are determined based on the importance score vector, a transmission importance threshold is set, and the rendering primitives in the importance score vector whose scores are greater than or equal to the transmission importance threshold are marked as the important data points. The transmission importance threshold is dynamically adjusted based on the current network bandwidth, client device performance or user tolerance for transmission delays.

7. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 1 is characterized by: In the S4 step, the transmission method of the non-important data points includes high-intensity compression coding or delayed transmission. The high-intensity compression coding adopts a lossy compression algorithm, including low-resolution compression of textures in non-critical areas; the delayed transmission refers to transmitting the non-important data points after the network bandwidth allows or the transmission of the important data points is completed.

8. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 2 is characterized by: The client loading step also includes, after step Y1, the client dynamically adjusts the buffering strategy or loading order of the rendering data block according to the priority of the data block in the rendering data stream, the system resource load of the client, and the current display performance, so as to ensure that the high-priority loaded data can preferentially occupy and utilize the client resources for rapid decoding, decompression or uploading to the graphics processor, while the low-priority loaded data is delayed or dynamically unloaded, thereby optimizing the resource utilization efficiency of the client.

9. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 2 is characterized by: In step Y3, high-priority loading data in the rendering data blocks that need to be loaded and rendered first is determined, and the client feature matrix is ​​a matrix of , wherein is the total number of the rendering data blocks received or to be loaded, and is the number of client-side features corresponding to each rendering data block, and each row of the client feature matrix represents a feature vector of the rendering data block, and each component of the feature vector corresponds to at least one of the geometric detail level, rendering readiness, or user interaction area association of the rendering data block; the client weight vector is a column vector of , wherein each element corresponds to a weighting coefficient of a corresponding feature in the client feature matrix; A loading importance threshold is set and rendering data blocks with scores in the client importance score vector greater than or equal to the loading importance threshold are marked as high-priority loading data.

10. The lightweight transmission and loading optimization method for high-precision 3D model cloud rendering according to claim 2 is characterized by: In step Y4, during the first rendering, the client uses the high-priority loaded data to quickly generate and display a rough skeleton of the three-dimensional model; after the high-priority loaded data is rendered, the client gradually decodes, decompresses, and loads the low-priority loaded data, and performs detailed supplementary rendering on the data, wherein the detailed supplementary rendering includes at least one of the following: loading a higher-level detailed model of the three-dimensional model to improve geometric accuracy; Load higher-resolution texture data of 3D models to improve texture clarity; apply more complex materials or lighting effects to enhance visual realism; The client dynamically updates the rendering pipeline or pixel-by-pixel shader to apply detailed supplementary rendering to the displayed three-dimensional model in real time, thereby gradually improving the fineness of the rendered image.

Images