CN-121708228-B - Lightweight hierarchical digital twin modeling and rendering method for large-scale scene

CN121708228BCN 121708228 BCN121708228 BCN 121708228BCN-121708228-B

Abstract

The application relates to the technical field of three-dimensional modeling, and provides a lightweight hierarchical digital twin modeling and rendering method for a large-scale scene. The method comprises the steps of analyzing surface geometry of an original high-precision three-dimensional model, calculating and extracting curvature characteristics of each local area corresponding to the high-precision three-dimensional model, constructing curvature response distribution through the curvature characteristics, screening and judging each model area corresponding to the original high-precision three-dimensional model based on the obtained curvature response distribution and a dynamic curvature threshold value, obtaining a second key detail area, generating independent multi-resolution detail enhancement data blocks for the second key detail area, generating a plurality of data versions with decreasing detail degrees for the same second key detail area, forming a multi-resolution detail library capable of being directly indexed, and completing seamless fusion rendering according to the detail data blocks and a basic main body model in a real-time rendering stage, so that the overall scene is light and the local detail is unified with high fidelity.

Inventors

LIU BIN
ZHENG YULONG
LI WENBIN
LI HONGYAN

Assignees

深圳市鼎粤科技有限公司
深圳华制智能制造技术有限公司

Dates

Publication Date: 20260505
Application Date: 20260213

Claims (9)

1. A lightweight hierarchical digital twin modeling and rendering method for a large-scale scene is characterized by comprising the following steps: Analyzing the surface geometry of the original high-precision three-dimensional model, calculating and extracting curvature characteristics of each local area corresponding to the high-precision three-dimensional model, distinguishing a flat area and a first key detail area of the original high-precision three-dimensional model through the curvature characteristics, and constructing curvature response distribution for subsequent area screening; Based on the obtained curvature response distribution, generating a dynamic curvature threshold by combining global curvature statistics corresponding to the original high-precision three-dimensional model with detail fidelity adjustment, wherein the dynamic curvature threshold comprises the steps of extracting and calculating normalized curvature average values and global maximum curvature response values of all areas of the original high-precision three-dimensional model from the constructed curvature response distribution to obtain model global curvature statistics; the method comprises the steps of obtaining a preset detail fidelity adjusting coefficient, wherein the detail fidelity adjusting coefficient is used for manually setting or automatically adapting a system according to specific requirements of different industries on detail fidelity to control the strictness degree of a key detail area identification standard, generating a dynamic curvature threshold value under a current scene by carrying out weighted calculation on a normalized curvature mean value, a global maximum curvature response value and the detail fidelity adjusting coefficient; Generating independent multi-resolution detail enhancement data blocks for the second key detail areas by adopting a detail level technology, generating a plurality of data versions with decreasing detail levels for the same second key detail areas, and forming a multi-resolution detail library capable of being directly indexed according to the data versions; In the real-time rendering stage, the distance between the observation viewpoint and the space center of each key detail area is calculated, detail data blocks corresponding to resolution levels in a multi-resolution detail library are dynamically scheduled according to the distance and the line-of-sight sensitivity coefficient, and in a graph drawing pipeline, seamless fusion rendering is completed according to the detail data blocks and a basic main body model, so that the light weight of the whole scene and the high-fidelity unification of local details are achieved.
2. The method of claim 1, wherein the first critical detail region is a region including a complex structure, sharp edges or precise lines, the curvature feature is used for quantifying the surface bending degree and detail richness of a local region of the characterization model, and the curvature feature of all the first critical detail regions is normalized, and the analyzing the surface geometry of the original high-precision three-dimensional model, and calculating and extracting the curvature feature of the high-precision three-dimensional model corresponding to each local region comprises: Acquiring triangular mesh data of an original high-precision three-dimensional model, performing traversal analysis on each vertex and each surface patch in the triangular mesh data one by one, and acquiring a set of directly adjacent vertices of each vertex and a unit normal vector corresponding to each vertex; obtaining a discrete curvature value of each vertex by carrying out average calculation on a unit normal vector included angle in the neighborhood of each vertex; In order to inhibit the interference of single-point noise on a subsequent judgment result, introducing an area weighting-based curvature aggregation mechanism at a patch level, and carrying out aggregation calculation on vertex discrete curvature values according to the geometric area weight of the patch of each vertex in a corresponding area to obtain a regional level geometric complexity measure; And mapping the geometric complexity measurement of all the regions to a unified normalized feature space, and completing the normalization processing of curvature features of all the first key detail regions, thereby eliminating absolute scale differences among different models.
3. The method of claim 2, wherein said constructing a curvature response profile for subsequent region screening comprises: After normalization processing of curvature characteristics of all areas is completed, counting global maximum values and global minimum values of geometric complexity metrics of all areas in the current original high-precision three-dimensional model; And constructing a distribution relation of detail response intensity based on the curvature characteristic, the global maximum value and the global minimum value of each region after normalization, and forming curvature response distribution for subsequent region screening.
4. The method according to claim 1, wherein the screening and determining each model region corresponding to the original high-precision three-dimensional model according to the dynamic curvature threshold, identifying a region with a curvature response exceeding the dynamic curvature threshold as a second critical detail region that cannot be lost in a simplified manner, and generating a unique semantic tag for each identified second critical detail region, includes: comparing the normalized curvature response value of each model region with the generated dynamic curvature threshold value one by one, and marking the corresponding model region as a second key detail region when the normalized curvature response value is larger than the dynamic curvature threshold value; acquiring space center coordinates, region geometric feature vectors and local geometric descriptors of each marked region; And generating unique reproducible semantic tags for each second key detail region by carrying out weighted encoding processing on the space center coordinates, the region geometric feature vectors and the local geometric descriptors, wherein the semantic tags are used for realizing accurate positioning and management of the key detail region in light weight, storage and rendering.
5. The method of claim 1, wherein generating independent multi-resolution detail enhancement data blocks for a second critical detail region using a level of detail technique, generating multiple data versions of decreasing detail for the same second critical detail region, comprises: Acquiring all area data identified as second key detail areas, and presetting a plurality of detail resolution levels for each second key detail area by adopting a detail level technology, wherein the detail resolution levels comprise the highest detail level and the simplest detail level; acquiring preset level attenuation factors for controlling the detail compression degree of different resolution levels, and ensuring that key morphological characteristics of a high-curvature region can still be reserved in a low-resolution level; generating a plurality of data versions with decreasing detail levels for each critical detail region based on the original high-precision geometric features of each second critical detail region in combination with the level attenuation factors; In the data version generation process, a curvature-guided local grid resampling algorithm is adopted, the data vertex density of each resolution level is adaptively adjusted according to regional curvature gradients, so that a high curvature region still keeps geometric outlines in a low resolution level, the low curvature region is subjected to remarkable downsampling processing, detail balance of data blocks of each resolution level is achieved, and independent multi-resolution detail enhancement data blocks corresponding to each second key detail region are generated.
6. The method of claim 5, wherein forming a directly indexable multi-resolution detail library from the version of the data comprises: Classifying and sorting a plurality of data versions with decreasing detail degrees generated in each second key detail area, and associating corresponding key detail area identifiers, resolution level identifiers and detail characteristic information corresponding to the data versions for each data version; Constructing a unified index mechanism, carrying out association mapping on each second key detail area and all data versions corresponding to the area, and determining the corresponding relation between each key detail area and the data versions of different resolution levels; And integrating the data versions of all the second key detail areas based on the incidence mapping relation to form a multi-resolution detail library capable of being directly indexed, and retrieving and calling the corresponding data versions through the key detail area identification and the resolution level identification.
7. The method of claim 1, wherein all data versions are stored in data block form independent of the lightweight base body model, and the spatial location, curvature characteristics and semantic tags of the corresponding key detail regions are recorded in data block metadata, the directly indexable multi-resolution detail library is formed from the data versions, comprising: The data versions corresponding to each second key detail area are stored independently in the form of independent data blocks, and all the data blocks do not depend on the basic main body model after light weight; In the metadata of each data block, recording the space position information, curvature characteristic information and generated semantic marks of the corresponding key detail area, and ensuring that each data block can trace back to the corresponding key detail area; Based on metadata information of all data blocks, constructing a unified index system, wherein the index system associates key information in metadata of the data blocks with the data blocks to form a multi-resolution detail library capable of being directly indexed; and locating and extracting the corresponding data blocks with different resolution levels by searching the space position, curvature characteristics or semantic marks of the key detail areas, so that the efficient calling of the data blocks is realized.
8. The method according to claim 1, wherein in the real-time rendering stage, a distance between the observation viewpoint and a spatial center of each key detail region is calculated, and the detail data blocks of the corresponding resolution level in the multi-resolution detail library are dynamically scheduled according to the distance and the line-of-sight sensitivity coefficient, and the method comprises: in the real-time rendering process, acquiring the space coordinates of the current observation viewpoint and the space center coordinates of each second key detail area in real time; Obtaining the space proximity degree of the observation viewpoint and each key detail area by calculating the linear distance between the space coordinate of the observation viewpoint and the space center coordinate of each second key detail area; Acquiring a preset sight distance sensitivity coefficient and a maximum sight distance threshold value, wherein the sight distance sensitivity coefficient is used for controlling the sensitivity of switching of detail level, and the maximum sight distance threshold value is used for defining the scheduling range of a detail data block; Determining a detail level index corresponding to each second key detail area according to the calculated space distance, the apparent distance sensitivity coefficient and the maximum apparent distance threshold, and when the space distance is smaller than the maximum apparent distance threshold, observing that the closer the distance between the view point and the key detail area is, scheduling the detail data blocks of the higher resolution level of the area, and the farther the distance is, scheduling the detail data blocks of the lower resolution level; And introducing a prediction caching mechanism, acquiring a motion vector and a motion amplitude of an observation viewpoint in real time, predicting a second key detail area which is possibly accessed by a next frame according to the motion vector and the motion amplitude, and loading detail data blocks corresponding to the resolution level in advance to avoid loading delay.
9. The method of claim 1, wherein the performing seamless fusion rendering with the underlying body model based on the detail data blocks comprises: And carrying out real-time space alignment processing on the detail data block and the basic main body model, ensuring accurate superposition of the detail data block and the basic main body model in a three-dimensional space, dynamically mixing color information of the detail data block and textures of the basic main body model through a texture fusion algorithm, introducing an illumination correction mechanism to eliminate illumination and shadow deviation between the detail data block and the basic main body model, realizing seamless fusion rendering of the detail data block and the basic main body model, and achieving light weight of the whole scene and high-fidelity unification of local details.

Description

Lightweight hierarchical digital twin modeling and rendering method for large-scale scene Technical Field The application relates to the technical field of three-dimensional modeling, in particular to a lightweight hierarchical digital twin modeling and rendering method for a large-scale scene. Background In the field of three-dimensional digital twin modeling and real-time rendering, the balance of light weight and high fidelity of local detail of a large-scale scene is a core technical problem. In the prior art, although the improvement of the overall modeling efficiency and the optimization of rendering resources are realized, the key detail processing has obvious defects: firstly, a quantitative analysis means for model geometric features is lacking, a high-curvature key region cannot be automatically identified, so that details such as a local complex structure, sharp edges and the like are lost in the light-weight process, virtual-real synchronization precision is affected, secondly, a hierarchical storage and dynamic scheduling mechanism is not established, calculation resources cannot be dynamically allocated according to user viewpoints, loading delay and interaction blocking are prone to occurring in a very large scale scene, thirdly, illumination and texture consistency processing are lacking in fusion of high-resolution details and a light-weight basic model, and vision splitting is caused. Accordingly, a need exists for a method that addresses at least one of the problems described above. Disclosure of Invention The application provides a lightweight hierarchical digital twin modeling and rendering method for a large-scale scene, and aims to solve the core technical problem of balance between lightweight and local detail high-fidelity of the large-scale scene in the field of three-dimensional digital twin modeling and real-time rendering. In a first aspect, an embodiment of the present application provides a method for lightweight hierarchical digital twin modeling and rendering for a large-scale scene, where the method includes: Analyzing the surface geometry of the original high-precision three-dimensional model, calculating and extracting curvature characteristics of each local area corresponding to the high-precision three-dimensional model, distinguishing a flat area and a first key detail area of the original high-precision three-dimensional model through the curvature characteristics, and constructing curvature response distribution for subsequent area screening; Based on the obtained curvature response distribution, generating a dynamic curvature threshold value by combining global curvature statistics corresponding to the original high-precision three-dimensional model and detail fidelity adjustment, screening and judging each model area corresponding to the original high-precision three-dimensional model according to the dynamic curvature threshold value, identifying areas with curvature responses exceeding the dynamic curvature threshold value as second key detail areas which cannot simplify loss, and generating unique semantic tags for each identified second key detail area; Generating independent multi-resolution detail enhancement data blocks for the second key detail areas by adopting a detail level technology, generating a plurality of data versions with decreasing detail levels for the same second key detail areas, and forming a multi-resolution detail library capable of being directly indexed according to the data versions; In the real-time rendering stage, the distance between the observation viewpoint and the space center of each key detail area is calculated, detail data blocks corresponding to resolution levels in a multi-resolution detail library are dynamically scheduled according to the distance and the line-of-sight sensitivity coefficient, and in a graph drawing pipeline, seamless fusion rendering is completed according to the detail data blocks and a basic main body model, so that the light weight of the whole scene and the high-fidelity unification of local details are achieved. In some embodiments, the first critical detail region is a region including a complex structure, sharp edges or precise lines, the curvature features are used for quantifying the surface bending degree and detail richness of a local region of a characterization model, curvature features of all first critical detail regions are normalized, the surface geometry of an original high-precision three-dimensional model is analyzed, curvature features of each local region corresponding to the high-precision three-dimensional model are calculated and extracted, the method comprises the steps of acquiring triangular mesh data of the original high-precision three-dimensional model, performing traversal analysis on each vertex and each patch in the triangular mesh data one by one, acquiring a set of directly adjacent vertices of each vertex and a unit normal vector corresponding to each vertex, obtainin