A high-fidelity three-dimensional model compression and multi-resolution dynamic loading method and device

Abstract

The present application relates to the field of three-dimensional model loading, and relates to a high-fidelity three-dimensional model compression and multi-resolution dynamic loading method and device, the method comprising: obtaining a three-dimensional grid model; calculating according to the three-dimensional grid model to obtain a geometric saliency index; performing adaptive quantization decision processing according to the geometric saliency index to obtain a globally optimal quantization bit number; performing attribute collaborative scaling and encoding mode scheduling processing according to the globally optimal quantization bit number to obtain a compression encoding mode adapted to the current scene; and performing predictive dynamic loading and rendering processing according to the compression encoding mode and a preset multi-resolution level to obtain a loaded three-dimensional model. The present application as a whole solves the technical problems of detail distortion or data redundancy caused by fixed quantization parameters in the prior art, difficulty in balancing compression rate and decoding speed caused by a single encoding mode, visual jump caused by LOD hard switching, and perception delay caused by passive loading.

Description

Technical Field

[0001] This invention relates to the field of 3D model loading, and more specifically, to a method and apparatus for high-fidelity 3D model compression and multi-resolution dynamic loading. Background Technology

[0002] With the widespread application of 3D digitization technology in the field of digital cultural heritage preservation, high-precision 3D scanning can generate massive model data containing millions of polygons, enabling high-fidelity archiving of the geometric details of cultural relics. However, how to efficiently and smoothly display these large-scale models on the web under limited network bandwidth and terminal computing power has become a key technical challenge. Currently, the industry generally adopts 3D geometric compression algorithms to reduce transmission volume, relies on standard formats such as glTF for data exchange, and uses multi-level of detail (LOD) technology to switch between models of different precision based on the viewing distance to balance performance. However, existing solutions still have significant limitations: First, the compression process typically relies on manual experience to preset fixed parameters, making it impossible to automatically adjust precision based on the richness of detail on different model surfaces (such as inscriptions and patterns), resulting in a difficulty in achieving an intelligent balance between compression rate and detail fidelity. Second, the encoding and decoding strategies are simplistic, making it difficult to dynamically adapt between low-bandwidth scenarios requiring extreme compression and high-interaction scenarios seeking ultra-fast decoding. Third, switching between LOD levels can cause visual jumps due to instantaneous changes in the model's geometric structure, disrupting the immersive viewing experience. Fourth, data loading often employs a passive response mode, which is prone to perceptible latency under network fluctuations. These shortcomings limit the potential for high-quality digital cultural relics to be displayed efficiently, smoothly, and with high fidelity on the web. Summary of the Invention

[0003] The purpose of this invention is to provide a method and apparatus for high-fidelity 3D model compression and multi-resolution dynamic loading, so as to improve the above-mentioned problems.

[0004] To achieve the above objectives, the embodiments of this application provide the following technical solutions:

[0005] On one hand, embodiments of this application provide a method for high-fidelity 3D model compression and multi-resolution dynamic loading, the method comprising:

[0006] Obtain a 3D mesh model;

[0007] The geometric significance index is obtained by calculation based on the three-dimensional mesh model.

[0008] Adaptive quantization decision processing is performed based on the geometric significance index to obtain the globally optimal quantization bit depth;

[0009] Based on the globally optimal quantization bit depth, attribute co-scaling and encoding mode scheduling are performed to obtain a compression encoding mode adapted to the current scenario;

[0010] Predictive dynamic loading and rendering are performed based on the compression encoding mode and preset multi-resolution levels to obtain a loaded 3D model, which is then used for display on the Web.

[0011] Secondly, embodiments of this application provide a high-fidelity 3D model compression and multi-resolution dynamic loading device, the device comprising:

[0012] The acquisition module is used to acquire 3D mesh models;

[0013] The first processing module is used to calculate the geometric saliency index based on the three-dimensional mesh model;

[0014] The second processing module is used to perform adaptive quantization decision processing based on the geometric significance index to obtain the globally optimal quantization bit depth.

[0015] The third processing module is used to perform attribute collaborative scaling and encoding mode scheduling processing based on the global optimal quantization bit depth to obtain a compression encoding mode that adapts to the current scenario.

[0016] The fourth processing module is used to perform predictive dynamic loading and rendering processing based on the compression encoding mode and the preset multi-resolution level to obtain the loaded 3D model, which is used for display on the Web.

[0017] Thirdly, embodiments of this application provide an apparatus comprising a memory and a processor. The memory stores a computer program; the processor executes the computer program to implement the steps of the above-described high-fidelity 3D model compression and multi-resolution dynamic loading method.

[0018] Fourthly, embodiments of this application provide a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described high-fidelity 3D model compression and multi-resolution dynamic loading method.

[0019] The beneficial effects of this invention are as follows:

[0020] This invention identifies and quantifies the importance of surface details in a 3D mesh model by calculating its geometric saliency index. Based on this index, a rate-distortion cost function is constructed for adaptive quantization decision-making, intelligently determining the globally optimal quantization bit width for models of different complexities, thus achieving on-demand allocation of compression accuracy. Attribute co-scaling is performed on vertex and texture coordinates according to this optimal bit width, ensuring the consistency of spatial mapping between geometry and texture after compression. Furthermore, high compression rate or ultra-fast decoding modes are dynamically scheduled based on terminal environment parameters to adapt to diverse network and computing power conditions. Finally, by constructing a multi-resolution hierarchy coupled with quantization accuracy and utilizing predictive data preloading and gradient blending techniques in vertex shaders, efficient model transmission, fast decoding, and visually seamless smooth loading are achieved, thus avoiding visual jumps caused by LOD hard switching while balancing compression rate and decoding speed.

[0021] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the high-fidelity 3D model compression and multi-resolution dynamic loading method described in this embodiment of the invention.

[0024] Figure 2 This is a schematic diagram of the high-fidelity 3D model compression and multi-resolution dynamic loading device described in an embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of the camera motion prediction and resource preloading mechanism.

[0026] The diagram is labeled as follows: 800, High-fidelity 3D model compression and multi-resolution dynamic loading device; 801, Processor; 802, Memory; 803, Multimedia component; 804, I / O interface; 805, Communication component. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0028] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0029] Example 1:

[0030] This embodiment provides a method for high-fidelity 3D model compression and multi-resolution dynamic loading. It can be understood that this embodiment can be used to illustrate a scenario, such as in museums, online education, or digital exhibitions where high-fidelity 3D cultural relic models need to be presented to the public via web browsers. Due to significant differences in the network environment (from high-speed broadband to mobile weak networks) and the terminal devices used (from high-performance computers to ordinary mobile phones), traditional fixed compression and loading schemes can lead to a systemic contradiction: high-intensity compression to ensure accessibility in weak network environments will lose key model details (such as inscriptions and patterns), rendering the cultural relic display academically worthless; while lightweight compression to preserve details will cause unbearable loading delays and interactive stutters on weak network or low-computing-power devices. Furthermore, switching between near and far distances during model viewing will produce noticeable visual jumps, disrupting the immersive experience.

[0031] See Figure 1 The figure shows that the method includes steps S1, S2, S3, S4 and S5.

[0032] Step S1: Obtain the 3D mesh model;

[0033] In this step, an initial three-dimensional mesh model is obtained by collecting and reconstructing the physical cultural relic using three-dimensional digitization technology.

[0034] Step S2: Calculate the geometric saliency index based on the three-dimensional mesh model;

[0035] Step S2 further includes steps S21-S23, which specifically include:

[0036] Step S21: Preprocess the three-dimensional mesh model to obtain a denoised three-dimensional mesh model;

[0037] The preprocessing in this step includes manifold restoration and bilateral filtering. During the scanning and reconstruction process, non-manifold edges, isolated vertices, or mesh cracks may be generated. These topological defects can seriously interfere with subsequent differential geometry-based calculations. Manifold restoration can identify and repair such defects, resulting in a repaired mesh. Bilateral filtering is then applied to the repaired mesh. This algorithm can effectively filter out high-frequency random noise introduced by the scanning, while preserving the sharp geometric features of inscriptions, patterns, and other areas characterized by drastic changes in normal direction to the greatest extent possible, ensuring that the key high-frequency details of the cultural relic are not blurred.

[0038] Step S22: Calculate the average curvature and Gaussian curvature of the surface of the denoised 3D mesh model using the discrete differential geometry algorithm;

[0039] The formula for calculating the mean curvature in this step is:

[0040]

[0041] in, Represents the global average curvature. Let V be the area of ​​the Voronoi region. This represents the total number of vertices in the denoised 3D mesh model. As vertex The set of adjacent vertices in a ring; and Representation and shared edges In two adjacent triangular facets, the two interior angles corresponding to the shared edge; This represents the three-dimensional spatial coordinate vector of the vertex to be calculated. Represents the vertex The first one connected directly by the grid edge The coordinate vectors of the adjacent vertices; Represents vertices The unit normal vector at that location, through comparison with... The result is obtained by weighting and normalizing the normal vectors of all adjacent facets.

[0042] The formula for calculating Gaussian curvature is:

[0043]

[0044] In the above formula, Indicates the global Gaussian curvature; Represents the vertex The set of all associated triangular faces; Indicates the first Adjacent triangular facets at the vertex The interior angle at the location (in radians); Represents vertices The absolute value of the discrete Gaussian curvature at a point (i.e., the modulus of the angular deficit) reflects the intensity of the local topological change at that point.

[0045] Step S23: Perform a weighted calculation based on the average curvature and the Gaussian curvature to obtain the geometric significance index.

[0046] In this step, the formula for calculating the geometric significance index is:

[0047]

[0048] In the above formula, Indicates the geometric significance index; and These represent the mean curvature and Gaussian curvature, respectively. and This represents the weighting coefficient.

[0049] Step S3: Perform adaptive quantization decision processing based on the geometric significance index to obtain the globally optimal quantization bit depth;

[0050] Step S3 further includes steps S31-S34, which specifically include:

[0051] Step S31: Construct a geometric distortion term based on the geometric saliency index;

[0052] In this step, the geometric distortion term uses the weighted root mean square error as the evaluation index. The core of this step is that it is not simply taking the square root of the average of the squared errors of all vertices, but rather calculating the geometric significance index for each vertex. As a local weighting factor for this vertex, the error in the highly salient region (the edge of the inscription) is amplified in the overall distortion calculation, while the error contribution of the low salient region (the flat body of the vessel) is relatively reduced, thus protecting the key geometric features that carry historical and artistic information.

[0053] Step S32: Obtain the encoding bitrate item;

[0054] Step S33: Construct the rate-distortion cost function based on the geometric distortion term and the coding rate term to obtain the rate-distortion cost function;

[0055] In this step, the rate-distortion cost function is specifically as follows:

[0056]

[0057] In the above formula, Represents the rate-distortion cost function; Represents the geometric distortion term; Indicates the encoding rate; It represents the Lagrange multiplier.

[0058] Step S34: Perform an optimal solution search based on the rate-distortion cost function to obtain the globally optimal quantization bit depth that minimizes the total cost.

[0059] In this step, The algorithm iterates through the set of candidate bit widths. For each candidate bit width in the set, it calculates its corresponding total cost and selects the candidate bit width with the minimum total cost as the globally optimal quantization bit width. This step automatically obtains the globally optimal solution that balances detail preservation and compression efficiency without performing physical partitioning.

[0060] Step S4: Perform attribute co-scaling and encoding mode scheduling processing based on the globally optimal quantization bit depth to obtain a compression encoding mode adapted to the current scene;

[0061] Step S4 further includes steps S41-S44, which specifically include:

[0062] Step S41: Perform attribute co-scaling processing based on the globally optimal quantization bit depth to obtain quantized grid data that maintains consistent spatial proportions;

[0063] Understandably, after determining the global quantization bit depth, the attribute space is scaled synchronously for both mesh vertex coordinates and texture coordinates. Because a globally uniform quantization step size is used, geometric vertices and UV sampling points maintain a strict linear ratio during dequantization, thus physically eliminating texture stretching and misalignment that might result from heterogeneous compression.

[0064] Step S42: Obtain the environmental parameters of the terminal device;

[0065] In this step, the environmental parameters of the terminal device include bandwidth and computing power.

[0066] Step S43: Compare the environmental parameters of the terminal device with the preset threshold information to obtain a judgment result;

[0067] Step S44: Determine the compression encoding mode suitable for the current scenario based on the judgment result.

[0068] In this step, when the network bandwidth is <10Mbps and the model file size is >20MB, the high compression mode is forced; when the network bandwidth is >50Mbps and the device is identified as a mobile device (computing power <100GFLOPS), the high-speed decoding mode is triggered.

[0069] In one specific implementation, the original unprocessed model was used as a benchmark, and its performance was compared with the two decoding modes of the present invention under different network environments (10Mbps weak network vs. 200Mbps high bandwidth). The results are shown in the table below:

[0070] Table 1 Comparison of Actual Measurements

[0071]

[0072] Based on the table above, it can be seen that in a weak network environment, transmission bandwidth is the main bottleneck. Although the decoding time of the high compression mode (1.14s) is much longer than that of the original model (0.086s), the transmission benefit brought by its 90.1% size reduction (saving about 55s) far outweighs the decoding cost, that is, the total loading time is reduced from 62.24s to 7.18s, and the efficiency is improved by 8.6 times. In a high bandwidth environment, transmission is no longer the only bottleneck, and the importance of decoding speed increases significantly. The high compression mode takes 1.45s, of which the decoding time of 1.14s accounts for as much as 79%, becoming a new performance drag. Although the ultra-fast decoding mode has a larger size (14.17MB), its decoding speed (68ms) is about 16 times faster than that of the high compression mode. In a 200Mbps environment, its total time is only 0.66s, which is about 2.2 times faster than the high compression mode and about 4.8 times faster than the original model. Therefore, under excellent network conditions, the system should automatically switch to the ultra-fast decoding mode to obtain the ultimate millisecond-level loading experience. Experimental data strongly demonstrates that a single compression strategy cannot cover all scenarios. The dynamic scheduling mechanism based on bandwidth / computing power proposed in this invention can ensure optimal loading experience in any network environment.

[0073] Step S5: Perform predictive dynamic loading and rendering processing according to the compression encoding mode and preset multi-resolution levels to obtain the loaded 3D model, which is used for display on the Web.

[0074] Step S5 further includes steps S51-S56, which specifically include:

[0075] Step S51: Compress and encode the quantized mesh data according to the compression encoding mode to obtain a compressed three-dimensional mesh model;

[0076] Step S52: Construct a multi-resolution hierarchy based on the compressed 3D mesh model to obtain the LOD hierarchy sequence;

[0077] In this step, based on the compressed 3D mesh model (LOD0), a quadratic error metric simplification algorithm is used to generate LOD1 (reduced surface area ratio). ) and LOD2 (reduced face ratio) ).

[0078] LOD0 (close-up) ): Retain 100% of the surface area to ensure fine features.

[0079] LOD1 (Medium shot) ): 50% dough pieces, It is suitable for medium-distance observation with high precision.

[0080] LOD2 (distant view) ): 10% dough pieces, Only the basic outline is retained, prioritizing extremely low bandwidth usage.

[0081] Step S53: Determine the display level based on the current relative distance between the camera and the model to obtain the target LOD level that should be rendered.

[0082] This step is a decision point in the real-time rendering loop to determine which LOD level of the model should be used for rendering. Specifically, in the web-based 3D rendering environment, the distance between the virtual camera position and the center of the model's bounding box needs to be calculated every frame. Figure 3 As shown, a switching threshold is set for every two adjacent levels in the LOD level sequence. By comparing the distance calculated in the current frame with the switching threshold, the target LOD level to be rendered can be determined. Specifically, this includes: Displaying LOD0; Display LOD1; The LOD2 display indicates that 'd' represents the camera's current distance. The decision-making mechanism employed in this step enables on-demand allocation of visual resources. When the user zooms in to examine the inscription more closely, the system automatically provides the highest-precision model; when the user zooms out to view the entire device, the system seamlessly switches to a lighter version. This intelligently concentrates limited graphics computing resources on the area of ​​greatest interest to the user while maintaining overall visual consistency.

[0083] Step S54: Perform motion trend prediction based on camera historical position data to obtain prediction results, including the target level to be reached at the next moment;

[0084] The purpose of this step is to anticipate the user's interaction intent and eliminate network latency by preloading data. Specifically, the camera position of the most recent few frames is continuously recorded in the rendering loop, and the instantaneous motion vector of the camera is calculated. The instantaneous motion vector includes the camera's motion direction and speed. At this time, based on the current position of the camera and the instantaneous motion vector, the distance range that the camera may enter after a short time window in the future can be predicted.

[0085] Step S55: Preload data based on the prediction results;

[0086] In this step, the current camera distance is d=2.5 (in the LOD1 region), its instantaneous motion vector points to the model and its speed is 0.5m / s. It is predicted that it will enter the LOD0 region within 60 frames. At this time, the asynchronous network request for LOD0 is immediately triggered to ensure that the LOD0 data is ready when it reaches 2.0m.

[0087] Step S56: Perform a smooth transition between layers based on the continuous change of camera distance to obtain the loaded 3D model.

[0088] Step S56 further includes steps S561-S565, which specifically include:

[0089] Step S561: Use the edge collapse path of the mesh simplification algorithm to establish vertex mapping and obtain the vertex topology mapping table between levels;

[0090] When using the mesh simplification algorithm, the algorithm records a series of edge collapse operations, which merge two vertices of an edge into a new vertex. In this step, based on the history of collapse operations, the corresponding vertex in the low-resolution model can be found for each vertex in the high-resolution model. Specifically, for a vertex removed during the simplification process, its mapping target is the new vertex generated after the collapse operation; for a vertex that is retained, it is mapped to its corresponding vertex in the low-resolution model, thereby generating the final vertex topology mapping table. This ensures that the topological logic of geometric simplification is strictly followed during deformation, avoiding model overlap.

[0091] Step S562: Perform geometric difference calculation based on the vertex topology mapping table to obtain the offset describing the shape change between layers;

[0092] After obtaining the vertex mapping relationship, this step calculates the specific geometric differences between corresponding vertices of the two levels. For each pair of mapping relationships in the mapping table, the position offset is calculated. The position offset can intuitively express how far each point on the high-poly model needs to move to be completely aligned with the corresponding shape of the low-poly model.

[0093] Step S563: Compress and store the data according to the offset to obtain geometric deformation information;

[0094] In this step, since the geometric differences between adjacent levels are small, the offset is locally quantized using an 8-bit signed integer and stored as a normalized auxiliary attribute a_DeltaPosition. Compared with storing absolute coordinates, this can keep the data redundancy within 20% of the original geometric data.

[0095] Step S564: Perform residual compensation calculation based on the quantization bit difference to obtain the residual compensation term;

[0096] Since different LOD levels may use different global quantization bits, direct interpolation may introduce jump jitter. In this step, the quantization residual is pre-calculated:

[0097]

[0098] In the above formula, Indicates the residual compensation term; Represents the geometric coordinates in actual storage; Represents the geometric coordinates under ideal conditions.

[0099] Step S565: Perform real-time interpolation calculation in the vertex shader based on the time-varying blending factor. By linearly blending the current coordinates, the decompressed geometric deformation information, and the residual compensation term, the interpolated vertex position is obtained.

[0100] In this step, when the camera distance is in the transition range between two LOD levels, the CPU calculates and inputs a blending factor that linearly changes from 0 to 1. The shader first reads the current coordinates (i.e., the high-resolution vertex coordinates of the currently rendered image), the decompressed geometric deformation information (i.e., the offset), and the residual compensation term, and then performs interpolation calculations, specifically:

[0101]

[0102] In the above formula, Indicates the position of the vertex after interpolation; Represents the coordinates of the currently rendered high-resolution vertex; Indicates the mixing factor; This indicates the geometric deformation information after decompression; This represents the residual compensation term. Through this step, when users interact with and browse cultural relics, the switching between different levels of detail in the model becomes a continuous and gradual process, eliminating abrupt visual transitions and improving the user's viewing experience.

[0103] Example 2:

[0104] This embodiment provides a high-fidelity 3D model compression and multi-resolution dynamic loading device. The device includes an acquisition module, a first processing module, a second processing module, a third processing module, and a fourth processing module, specifically including:

[0105] The acquisition module is used to acquire 3D mesh models;

[0106] The first processing module is used to calculate the geometric saliency index based on the three-dimensional mesh model;

[0107] The second processing module is used to perform adaptive quantization decision processing based on the geometric significance index to obtain the globally optimal quantization bit depth.

[0108] The third processing module is used to perform attribute collaborative scaling and encoding mode scheduling processing based on the global optimal quantization bit depth to obtain a compression encoding mode that adapts to the current scenario.

[0109] The fourth processing module is used to perform predictive dynamic loading and rendering processing based on the compression encoding mode and the preset multi-resolution level to obtain the loaded 3D model, which is used for display on the Web.

[0110] In one specific embodiment of this disclosure, the first processing module further includes a first processing unit, a second processing unit, and a third processing unit, specifically:

[0111] The first processing unit is used to preprocess the three-dimensional mesh model to obtain a denoised three-dimensional mesh model.

[0112] The second processing unit is used to calculate the average curvature and Gaussian curvature of the surface of the denoised three-dimensional mesh model using a discrete differential geometry algorithm.

[0113] The third processing unit is used to perform a weighted calculation based on the average curvature and the Gaussian curvature to obtain a geometric significance index.

[0114] In one specific embodiment of this disclosure, the second processing module further includes a fourth processing unit, a first acquisition unit, a fifth processing unit, and a sixth processing unit, specifically:

[0115] The fourth processing unit is used to construct a geometric distortion term based on the geometric saliency index;

[0116] The first acquisition unit is used to acquire the coding rate item;

[0117] The fifth processing unit is used to construct a rate-distortion cost function based on the geometric distortion term and the coding rate term, thereby obtaining the rate-distortion cost function;

[0118] The sixth processing unit is used to perform an optimal solution search based on the rate-distortion cost function to obtain the globally optimal quantization bit depth that minimizes the total cost.

[0119] In one specific embodiment of this disclosure, the third processing module further includes a seventh processing unit, a second acquisition unit, an eighth processing unit, and a ninth processing unit, specifically:

[0120] The seventh processing unit is used to perform attribute co-scaling processing based on the globally optimal quantization bit depth to obtain quantized grid data that maintains a consistent spatial ratio.

[0121] The second acquisition unit is used to acquire environmental parameters of the terminal device;

[0122] The eighth processing unit is used to compare the environmental parameters of the terminal device with preset threshold information to obtain a judgment result;

[0123] The ninth processing unit is used to determine a compression encoding mode suitable for the current scenario based on the judgment result.

[0124] In one specific embodiment of this disclosure, the fourth processing module further includes a tenth processing unit, an eleventh processing unit, a twelfth processing unit, a thirteenth processing unit, a fourteenth processing unit, and a fifteenth processing unit, specifically as follows:

[0125] The tenth processing unit is used to compress and encode the quantized mesh data according to the compression encoding mode to obtain a compressed three-dimensional mesh model.

[0126] The eleventh processing unit is used to construct a multi-resolution hierarchy based on the compressed 3D mesh model to obtain a LOD hierarchy sequence.

[0127] The twelfth processing unit is used to determine the display level based on the current relative distance between the camera and the model, and to obtain the target LOD level that should be rendered.

[0128] The thirteenth processing unit is used to predict motion trends based on historical camera position data and obtain prediction results, including the target level to be reached at the next moment.

[0129] The fourteenth processing unit is used to preload data based on the prediction results;

[0130] The fifteenth processing unit is used to perform smooth transitions between layers based on continuous changes in camera distance, resulting in the loaded 3D model.

[0131] In one specific embodiment of this disclosure, the fifteenth processing unit further includes a sixteenth processing unit, a seventeenth processing unit, an eighteenth processing unit, a nineteenth processing unit, and a twentieth processing unit, specifically as follows:

[0132] The sixteenth processing unit is used to establish vertex mappings using the edge collapse path of the mesh simplification algorithm, and obtain the vertex topology mapping table between levels.

[0133] The seventeenth processing unit is used to perform geometric difference calculations based on the vertex topology mapping table to obtain the offset describing the shape changes between layers;

[0134] The eighteenth processing unit is used to compress and store data according to the offset to obtain geometric deformation information;

[0135] The nineteenth processing unit is used to perform residual compensation calculation based on the difference in the number of quantization bits to obtain the residual compensation term;

[0136] The twentieth processing unit is used to perform real-time interpolation calculations in the vertex shader based on the time-varying blending factor. It obtains the interpolated vertex position by linearly blending the current coordinates, the decompressed geometric deformation information, and the residual compensation term.

[0137] It should be noted that the specific manner in which each module performs its operation in the apparatus described in the above embodiments has been described in detail in the embodiments of the method, and will not be elaborated here.

[0138] Example 3:

[0139] Corresponding to the above method embodiments, this embodiment also provides a high-fidelity 3D model compression and multi-resolution dynamic loading device. The high-fidelity 3D model compression and multi-resolution dynamic loading device described below and the high-fidelity 3D model compression and multi-resolution dynamic loading method described above can be referred to in correspondence.

[0140] Figure 2 This is a block diagram illustrating a high-fidelity 3D model compression and multi-resolution dynamic loading device 800 according to an exemplary embodiment. Figure 2 As shown, the high-fidelity 3D model compression and multi-resolution dynamic loading device 800 may include: a processor 801 and a memory 802. The high-fidelity 3D model compression and multi-resolution dynamic loading device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0141] The processor 801 controls the overall operation of the high-fidelity 3D model compression and multi-resolution dynamic loading device 800 to complete all or part of the steps in the aforementioned high-fidelity 3D model compression and multi-resolution dynamic loading method. The memory 802 stores various types of data to support the operation of the high-fidelity 3D model compression and multi-resolution dynamic loading device 800. This data may include, for example, instructions for any application or method operating on the high-fidelity 3D model compression and multi-resolution dynamic loading device 800, as well as application-related data, such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the high-fidelity 3D model compression and multi-resolution dynamic loading device 800 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0142] In an exemplary embodiment, the high-fidelity 3D model compression and multi-resolution dynamic loading device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described high-fidelity 3D model compression and multi-resolution dynamic loading method.

[0143] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the high-fidelity 3D model compression and multi-resolution dynamic loading method described above. For example, the computer-readable storage medium may be the memory 802 including the program instructions described above. These program instructions may be executed by the processor 801 of the high-fidelity 3D model compression and multi-resolution dynamic loading device 800 to complete the high-fidelity 3D model compression and multi-resolution dynamic loading method described above.

[0144] Example 4:

[0145] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below and the high-fidelity 3D model compression and multi-resolution dynamic loading method described above can be referred to in relation to each other.

[0146] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the high-fidelity 3D model compression and multi-resolution dynamic loading method described in the above method embodiments.

[0147] The readable storage medium can specifically be a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, or any other readable storage medium capable of storing program code.

[0148] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0149] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for high-fidelity 3D model compression and multi-resolution dynamic loading, characterized in that, include: Obtain a 3D mesh model; The geometric significance index is obtained by calculation based on the three-dimensional mesh model. Adaptive quantization decision processing is performed based on the geometric significance index to obtain the globally optimal quantization bit depth; Based on the globally optimal quantization bit depth, attribute co-scaling and encoding mode scheduling are performed to obtain a compression encoding mode adapted to the current scenario; Predictive dynamic loading and rendering are performed based on the compression encoding mode and preset multi-resolution levels to obtain a loaded 3D model, which is then used for display on the Web.

2. The high-fidelity 3D model compression and multi-resolution dynamic loading method according to claim 1, characterized in that, Based on the aforementioned three-dimensional mesh model, geometric saliency indices are calculated, including: The three-dimensional mesh model is preprocessed to obtain a denoised three-dimensional mesh model; The average curvature and Gaussian curvature of the surface of the denoised 3D mesh model are calculated using a discrete differential geometry algorithm. The geometric significance index is obtained by weighting the mean curvature and the Gaussian curvature.

3. The high-fidelity 3D model compression and multi-resolution dynamic loading method according to claim 1, characterized in that, Adaptive quantitative decision processing based on the geometric significance index includes: Construct a geometric distortion term based on the geometric saliency index; Get the encoding bitrate item; The rate-distortion cost function is constructed based on the geometric distortion term and the coding rate term to obtain the rate-distortion cost function. The optimal solution is searched based on the rate-distortion cost function to obtain the globally optimal quantization bit depth that minimizes the total cost.

4. The high-fidelity 3D model compression and multi-resolution dynamic loading method according to claim 1, characterized in that, Based on the globally optimal quantization bit depth, attribute co-scaling and encoding mode scheduling processing are performed, including: Based on the globally optimal quantization bit depth, attribute co-scaling processing is performed to obtain quantized grid data that maintains consistent spatial proportions. Obtain environmental parameters of the terminal device; The judgment result is obtained by comparing the environmental parameters of the terminal device with preset threshold information; Based on the judgment result, a compression encoding mode suitable for the current scenario is determined.

5. The high-fidelity 3D model compression and multi-resolution dynamic loading method according to claim 1, characterized in that, Predictive dynamic loading and rendering processing is performed based on the compression encoding mode and preset multi-resolution levels, including: The quantized mesh data is compressed and encoded according to the compression encoding mode to obtain a compressed three-dimensional mesh model. Based on the compressed 3D mesh model, a multi-resolution hierarchy is constructed to obtain the LOD hierarchy sequence; The display level is determined based on the relative distance between the current camera and the model, thus obtaining the target LOD level that should be rendered. Motion trend prediction is performed based on historical camera location data to obtain prediction results, which include the target level to be reached at the next moment. Data preloading is performed based on the prediction results; By performing a smooth transition between layers based on the continuous change in camera distance, a loaded 3D model is obtained.

6. A high-fidelity 3D model compression and multi-resolution dynamic loading device, characterized in that, include: The acquisition module is used to acquire 3D mesh models; The first processing module is used to calculate the geometric saliency index based on the three-dimensional mesh model; The second processing module is used to perform adaptive quantization decision processing based on the geometric significance index to obtain the globally optimal quantization bit depth. The third processing module is used to perform attribute collaborative scaling and encoding mode scheduling processing based on the global optimal quantization bit depth to obtain a compression encoding mode that adapts to the current scenario. The fourth processing module is used to perform predictive dynamic loading and rendering processing based on the compression encoding mode and the preset multi-resolution level to obtain the loaded 3D model, which is used for display on the Web.

7. The high-fidelity 3D model compression and multi-resolution dynamic loading device according to claim 6, characterized in that, The first processing module includes: The first processing unit is used to preprocess the three-dimensional mesh model to obtain a denoised three-dimensional mesh model. The second processing unit is used to calculate the average curvature and Gaussian curvature of the surface of the denoised three-dimensional mesh model using a discrete differential geometry algorithm. The third processing unit is used to perform a weighted calculation based on the average curvature and the Gaussian curvature to obtain a geometric significance index.

8. The high-fidelity 3D model compression and multi-resolution dynamic loading device according to claim 6, characterized in that, The second processing module includes: The fourth processing unit is used to construct a geometric distortion term based on the geometric saliency index; The first acquisition unit is used to acquire the coding rate item; The fifth processing unit is used to construct a rate-distortion cost function based on the geometric distortion term and the coding rate term, thereby obtaining the rate-distortion cost function; The sixth processing unit is used to perform an optimal solution search based on the rate-distortion cost function to obtain the globally optimal quantization bit depth that minimizes the total cost.

9. The high-fidelity 3D model compression and multi-resolution dynamic loading device according to claim 6, characterized in that, The third processing module includes: The seventh processing unit is used to perform attribute co-scaling processing based on the globally optimal quantization bit depth to obtain quantized grid data that maintains a consistent spatial ratio. The second acquisition unit is used to acquire environmental parameters of the terminal device; The eighth processing unit is used to compare the environmental parameters of the terminal device with preset threshold information to obtain a judgment result; The ninth processing unit is used to determine a compression encoding mode suitable for the current scenario based on the judgment result.

10. The high-fidelity 3D model compression and multi-resolution dynamic loading device according to claim 6, characterized in that, The fourth processing module includes: The tenth processing unit is used to compress and encode the quantized mesh data according to the compression encoding mode to obtain a compressed three-dimensional mesh model. The eleventh processing unit is used to construct a multi-resolution hierarchy based on the compressed 3D mesh model to obtain a LOD hierarchy sequence. The twelfth processing unit is used to determine the display level based on the current relative distance between the camera and the model, and to obtain the target LOD level that should be rendered. The thirteenth processing unit is used to predict motion trends based on historical camera position data and obtain prediction results, including the target level to be reached at the next moment. The fourteenth processing unit is used to preload data based on the prediction results; The fifteenth processing unit is used to perform smooth transitions between layers based on continuous changes in camera distance, resulting in the loaded 3D model.

Images