Search

CN-122023953-A - Intelligent lightweight conversion and hierarchical loading method based on model feature recognition

CN122023953ACN 122023953 ACN122023953 ACN 122023953ACN-122023953-A

Abstract

The invention relates to an intelligent lightweight conversion and hierarchical loading method based on model feature recognition, and belongs to the technical field of computer graphic processing and digital twin. The method comprises the steps of analyzing an original three-dimensional model and configuration files thereof, classifying model structure units according to predefined motion constraint, data interfaces and interaction event relations, constructing a semantic feature tree, receiving real-time service data streams, calculating real-time service criticality of each semantic feature by comparing entity states with a preset threshold value, dynamically matching a light-weight strategy according to the real-time service criticality, performing differential simplification operation to generate a light-weight model data set, calculating comprehensive loading priorities by integrating view parameters, user interaction instructions and the real-time service criticality, scheduling and rendering corresponding feature components according to the comprehensive loading priorities, and triggering generation and transmission of data acquisition frequency regulation instructions of binding data sources when the rendering detail level changes. And realizing the model dynamic optimization and the resource closed-loop scheduling of service awareness.

Inventors

  • LIN JINSONG
  • ZENG XIANGYU

Assignees

  • 上海湃睿信息科技有限公司

Dates

Publication Date
20260512
Application Date
20260414

Claims (12)

  1. 1. An intelligent lightweight conversion and hierarchical loading method based on model feature recognition is characterized by comprising the following steps: The method comprises the steps of S1, responding to an original three-dimensional model file uploaded by a user, automatically analyzing the original three-dimensional model, extracting a geometric structure, analyzing a configuration file associated with the original three-dimensional model, wherein the configuration file defines a motion constraint relation, a data interface binding relation and a user interaction event binding relation of an internal structural unit of the original three-dimensional model; S2, receiving a real-time service data stream associated with the original three-dimensional model, analyzing entity states corresponding to semantic features according to the real-time service data stream, and comparing the entity states with preset thresholds of the corresponding features to calculate real-time service criticality; according to the lightweight strategy, corresponding geometric simplification, material simplification or special effect simplification operation is executed on the feature components, and a lightweight model dataset with multiple detail layers is generated; S3, when the three-dimensional scene operates, acquiring current view angle parameters, user interaction instructions and the real-time service criticality, carrying out comprehensive weighted calculation on the view angle parameters, the user interaction instructions and the real-time service criticality according to preset weighting rules, outputting comprehensive loading priorities of all semantic features, dispatching and rendering feature components of corresponding detail levels from the light-weight model dataset according to the comprehensive loading priorities, and generating and sending data acquisition frequency regulation instructions of data sources bound with the features if the rendering detail levels of specific semantic features change.
  2. 2. The method of claim 1, wherein the configuration file is generated or updated in accordance with user configuration instructions received via a graphical user interface, the user configuration instructions for customizing at least one of the following rules: assigning semantic feature types to different types of structural units; Setting a data out-of-limit alarm threshold value for the data monitoring point class characteristics; establishing a mapping relation between the real-time service criticality level and the lightweight strategy parameters; And setting various weight coefficients for calculating the comprehensive loading priority.
  3. 3. The method of claim 1, wherein the semantic features further comprise static structures, and wherein the classifying specifically comprises classifying structural elements in the configuration file that do not define motion constraints, data interfaces, and user interaction event bindings as static structures.
  4. 4. The method of claim 1, wherein the computing of the real-time business criticality in S2 comprises presetting a base criticality for each type of semantic feature, and increasing the criticality of the feature by a preset level on the base value when it is identified from the real-time business data stream that the data value associated with the data monitoring point feature exceeds an alert threshold defined in the configuration file or the event trigger frequency associated with the interactable interface feature exceeds a preset frequency threshold.
  5. 5. The method of claim 4, wherein the step of dynamically matching a lightweight policy in step S2 specifically includes querying a preset lightweight policy mapping table according to a preset level interval to which the real-time service criticality belongs to, to obtain a corresponding policy combination, where the policy combination at least includes a target geometric simplified strength, a target texture size, and a target special effect switch state.
  6. 6. The method of claim 5, wherein for a feature classified as a moving part, the geometric reduction operations in its lightweight policy combination are configured to exclude geometric surfaces associated with the feature in a motion axis or center of rotation defined in the profile.
  7. 7. The method of claim 1, wherein generating a lightweight model dataset with multiple levels of detail in S2 comprises generating at least two level of detail versions for the same feature component, a standard version that preserves both original material and major geometric details, and a schematic version that uses monochromatic material and simplifies geometry.
  8. 8. The method of claim 1, wherein the factors considered by the preset weighting rules in S3 include at least a distance from a current viewing angle, whether the user interaction focus is the user interaction focus, and the real-time business criticality.
  9. 9. The method according to claim 7, wherein the scheduling and rendering feature components of the corresponding detail level in S3 comprises, in particular, loading and rendering a standard version thereof immediately for features with integrated loading priority above a first set threshold, loading and rendering a schematic version thereof for features with priority below the first set threshold but above a second set threshold, and delaying loading or only preserving space placeholders for features with priority below the second set threshold.
  10. 10. The method according to claim 1, wherein the step of generating and transmitting the data collection frequency regulation command of the data source bound to the feature in the step S3 specifically includes generating and transmitting a command for increasing the data reporting frequency to a first target value when a feature of a certain data monitoring point is scheduled to be rendered in a standard version, generating and transmitting a command for decreasing the data reporting frequency to a second target value when the feature is degraded to be rendered in a schematic version, and the first target value and the second target value are defined in a configuration file.
  11. 11. The method of claim 10, wherein generating the data acquisition frequency regulation command for the data source bound to the feature further comprises comparing a target frequency value in the command with a current frequency value of the data source before transmitting the regulation command, and transmitting the command only when the target frequency value and the current frequency value are inconsistent.
  12. 12. The method of claim 1, further comprising, after S3, a step S4 of recording a loading history of the semantic feature tree, the applied lightweight policy, and the transmitted data collection frequency regulation instructions, generating an optimization process log.

Description

Intelligent lightweight conversion and hierarchical loading method based on model feature recognition Technical Field The invention belongs to the technical field of computer graphic processing and digital twin, and particularly relates to an intelligent lightweight conversion and hierarchical loading method based on model feature recognition. Background Along with the rapid development of technologies such as digital twinning, industrial Internet, smart cities and the like, the demands for real-time visualization, simulation analysis and interactive monitoring of high-fidelity and large-scale three-dimensional scenes are increasingly urgent. The core challenge of such applications is that the original three-dimensional model usually has extremely high geometric complexity and texture details, and there is a sharp contradiction between the massive data volume and the limited terminal computing resources, network bandwidth and real-time requirements. In order to solve the contradiction, the traditional technical path mainly focuses on two aspects, namely, light weight processing is carried out on the three-dimensional model, and hierarchical loading strategy is implemented. In terms of lightweight processing, the prior art relies mostly on generic geometry reduction algorithms (e.g. mesh reduction, vertex merging, texture compression, etc.) or pre-set level of detail (LOD) model generation. However, these methods are typically "one-shot" and lack an understanding of the inherent structure of the model and the business semantics. In terms of hierarchical loading strategies, existing methods mostly schedule models of different accuracy based on spatial distance of cameras or simple visibility decisions. This scheduling mechanism, which relies only on geometric and spatial information, is static and passive. In an actual digital twin application, the traffic state of the model (e.g., whether the device is alarming, whether the parameters are out of limit) and the user's interaction intent (e.g., being focused on a device to operate) are dynamically changing and have a decisive impact on rendering priority. The existing method cannot integrate the real-time business logic, so that at the key moment, equipment concerned by a user can only load a rough model due to a long distance to influence monitoring and decision, and on the contrary, a large number of models in a non-key state can consume precious rendering resources. Furthermore, the data flow of current three-dimensional visualization systems and underlying physical systems is typically unidirectional, i.e., a "data driven view". When the view reduces the rendering quality of a certain component due to performance consideration, the influence of the operation on the data acquisition end is not considered, a closed loop of 'view demand feedback to data acquisition' is not formed, and the opportunity of acquiring higher quality data in a high-fidelity rendering period may be missed, or resource waste is caused by continuously acquiring the data at high frequency when not necessary. In summary, in the prior art, when a three-dimensional model facing digital twin is processed, limitations of lack of service semantic guidance, disjointing of a hierarchical loading mechanism and a dynamic service state, visualization and data acquisition resource scheduling and splitting exist in a lightweight process. Therefore, a new processing method capable of intelligently identifying the service features of the model, dynamically optimizing according to the real-time service context, and realizing the integrated scheduling of visualization and data acquisition is needed in the industry to overcome the above defects, so that the efficient, intelligent and high-quality operation of the large-scale complex digital twin scene is truly realized. Disclosure of Invention In order to solve the problems in the prior art, the invention provides an intelligent lightweight conversion and hierarchical loading method based on model feature recognition, and the aim of the invention can be achieved by the following technical scheme: An intelligent lightweight conversion and hierarchical loading method based on model feature recognition comprises the following steps: The method comprises the steps of S1, responding to an original three-dimensional model file uploaded by a user, automatically analyzing the original three-dimensional model, extracting a geometric structure, analyzing a configuration file associated with the original three-dimensional model, wherein the configuration file defines a motion constraint relation, a data interface binding relation and a user interaction event binding relation of an internal structural unit of the model; S2, receiving a real-time service data stream associated with the original three-dimensional model, analyzing entity states corresponding to semantic features according to the real-time service data stream, and comparing the entity states with prese