A lightweight method for ship 3D models based on multimodal collaborative simplification

By employing a multimodal collaborative simplification method, combined with semantic feature weighting and dynamic scaling mesh simplification, a multi-level LOD system is constructed. This solves the problems of large data volume and low rendering and interaction efficiency of ship 3D models in existing technologies, and enables real-time rendering and interaction that stably preserves key structural and functional features on low-configuration devices.

CN122695145BActive Publication Date: 2026-11-10CHINA SHIP DEV & DESIGN CENT
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202611163783.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-08-03
Publication Date
2026-11-10
Estimated Expiration
2046-08-03

AI Technical Summary

Technical Problem

Existing technologies struggle to reduce the amount of data in a ship's 3D model while stably preserving key structural and functional features, thus meeting the real-time rendering and interaction requirements of low-configuration devices.

Method used

By using a multimodal collaborative simplification method, geometric data, texture data, semantic data, and metadata are extracted to establish a multi-dimensional data association map of geometry, texture, and semantics. Combined with semantic feature weighting and dynamic scaling mesh simplification, topology-aware vertex clustering and texture block compression are adopted to construct a multi-level LOD system and perform cross-platform adaptation to achieve efficient model loading and interaction.

Benefits of technology

While significantly reducing the amount of model data, it stably retains key structural and functional features, meets the real-time rendering and interaction needs of low-configuration devices, and provides an accurate and efficient lightweight solution for 3D models.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122695145B_ABST
    Figure CN122695145B_ABST
Patent Text Reader

Abstract

This invention discloses a lightweight method for ship 3D models based on multimodal collaborative simplification, belonging to the field of equipment digital model technology. The method includes: deep analysis of the original whole-ship model to establish a multi-dimensional data association map of geometry, texture, and semantics; integrating component semantic information into the weight calculation and edge folding cost function of the QEM simplification algorithm based on semantic feature weighting and dynamic scaling, and using dynamic threshold control to control the simplification process, achieving feature-aware adaptive simplification; setting differentiated optimization parameters for geometry, texture, and LOD respectively, constructing a three-level LOD system of global, local, and detailed models to achieve a balance between volume reduction and feature preservation; and finally, cross-platform adaptation to support real-time rendering and interaction on low-configuration terminals. This invention, through a semantic and geometric collaborative simplification mechanism, significantly reduces the amount of model data while stably preserving key structural and functional features, making it suitable for scenarios such as ship design, operation and maintenance, and training.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of equipment digital modeling technology, specifically to a lightweight method for ship 3D models based on multimodal collaborative simplification. Background Technology

[0002] Against the backdrop of the shipbuilding industry's transformation towards intelligence and digitalization, 3D models have evolved from traditional "design aids" to "core digital carriers" for the entire lifecycle management of ships. From virtual simulation in ship R&D to process verification during construction, from equipment maintenance in operation and maintenance to crew training, 3D models are deeply integrated into the entire chain of ship equipment design, manufacturing, operation and maintenance, and services. However, the contradiction between the large size of ship 3D visualization models and the limited computing power of low-configuration computers (such as office PCs, embedded terminals, and domestically produced display and control consoles) has become a key bottleneck restricting the implementation of digital applications in shipbuilding.

[0003] Chinese Patent Publication No. CN120655826A discloses a method for generating a lightweight 3D data format suitable for ship model annotation. This method involves exporting a 3D model file from ship design software and extracting its model attribute information file. The method parses the 3D model file and the model attribute information file to calculate the attribute values ​​of the geometric models within the model. These attribute values ​​are then integrated and converted into a lightweight data exchange format. The lightweight data exchange format file is then converted to binary format, and after transcoding and compression, a lightweight 3D data format file is obtained. Finally, the lightweight 3D data format file is stored in a database after matching names based on the model attribute information. This invention produces a small-sized lightweight 3D data format file, retaining only the key data for reconstructing the model, and performing binary encoding conversion and compression. Therefore, compared to general-purpose model file formats, it is smaller in size, which is beneficial for storage and network transmission.

[0004] Chinese Patent Publication No. CN120180581A discloses a method for lightweight integration of 3D models of ship products, including the following steps: inputting model data from a 3D CAD model into a Product Data Management System (PDM); in the PDM, representing, modeling, parsing, and storing the 3D model data structure using a data integration framework; converting the 3D CAD model data into lightweight files using a 3D model lightweight conversion interface; mounting the lightweight 3D CAD model files in the PDM's model structure tree and embedding a 3D model lightweight browser to achieve lightweight browsing of the 3D CAD model in the PDM. This invention standardizes the data modeling standards and 3D model lightweight file standards for 3D CAD models, providing convenience for the management and browsing of 3D models in PDM.

[0005] Chinese patent CN111210521B discloses a method, system, terminal, and medium for lightweighting ship megadata models for VR. The method includes: preprocessing the ship megadata model with unstructured data during export; converting the preprocessed ship megadata model into a meshed model that can be displayed in VR, and removing the meshes; adjusting the rendering configuration parameters of the meshed model after mesh removal to obtain a lightweight model that meets the display frame rate requirements. This invention solves the problem that existing ship megadata models are not sufficiently lightweight, resulting in extremely low VR rendering frame rates that are insufficient to meet the requirements for lightweight VR rendering of ship megadata models. This invention improves the average rendering frame rate after importing the ship megadata model into a VR platform, ensuring a smooth VR experience, reducing dizziness, and providing technical support for long-term VR full-ship roaming, cabin design reviews, and maintenance training.

[0006] In summary, existing technologies (such as CN120655826A, CN120180581A, and CN111210521B) primarily achieve data reduction and visualization applications for ship 3D models through lightweight format conversion and binary encoding compression, lightweight file mounting and browsing for management systems, or mesh culling and rendering parameter adjustment for VR display frame rates. These technologies generally rely on global static conversion / compression or single-dimensional (geometry or rendering) optimization, lacking a collaborative modeling and differentiated processing mechanism for the ship model's "geometric structure-material texture-semantic / task elements." Therefore, it is difficult to significantly reduce data volume while stably preserving key structural and functional features and meeting the real-time rendering and interaction requirements of low-configuration devices. Summary of the Invention

[0007] This invention provides a lightweight method for ship 3D models based on multimodal collaborative simplification, aiming to achieve a balance between model volume compression and key feature preservation, and to adapt to efficient loading and interaction on low-configuration terminals.

[0008] This invention provides a lightweight method for ship 3D models based on multimodal collaborative simplification, comprising: The original ship model was analyzed in depth to extract geometric data, texture data, semantic data and metadata, and a multi-dimensional data association map of geometry-texture-semantics was established. Based on the multi-dimensional data association map, a core simplification driven by multi-modal features is carried out. A grid simplification method based on semantic feature weighting and dynamic ratio matching is introduced, and component semantic information is integrated into the simplification weight calculation to achieve adaptive simplification of feature perception. Based on the obtained structural importance region division results, texture type division results and their importance weights, optimization parameters and constraints are set differently for geometric simplification, texture simplification and level of detail (LOD) to achieve a balance between volume reduction and feature preservation. The processed model is adapted for cross-platform use, including format conversion, shader optimization, and progressive loading configuration, to support real-time rendering and interaction on the target configuration terminal.

[0009] In some instances, the process involves deep analysis of the original ship model to extract geometric data, texture data, semantic data, and metadata, and to establish a multi-dimensional data association map of geometry, texture, and semantics, including: The parsing engine extracts geometric data, texture data, semantic data containing component names and functional attributes, and metadata containing units and precision from the whole ship model. The whole ship model includes hull structure and cross-disciplinary data, including power system, energy supply system, cooling and fluid transport system, auxiliary support system and general equipment components.

[0010] In some instances, the core simplification based on the multi-dimensional data association map, driven by multi-modal features, introduces a mesh simplification method based on semantic feature weighting and dynamic scaling. This method integrates component semantic information into the simplification weight calculation, achieving adaptive simplification for feature perception, including: The first and second target regions in the model are identified by a geometric feature extraction algorithm, and dynamic weights are assigned to different features by combining prior knowledge in the field of shipbuilding. The first target region is more important than the second target region. Based on the simplified algorithm of the quadratic error metric QEM, a feature-aware collapse cost threshold mechanism is introduced, and semantic weights are incorporated into the error metric. Users can customize simplification strategies through the configuration interface, and the system dynamically matches the simplification ratio based on the semantic information of the components; Analyze the texture maps to distinguish between functional textures and decorative textures; Each feature is assigned an importance weight based on prior knowledge in the field of shipbuilding; By fusing and analyzing multimodal features, we can identify the information that needs to be retained and the redundant information.

[0011] In some instances, incorporating semantic weights into the error metric includes: vertex semantic weight Take the maximum value of the weight of its associated facet, for the edge Its semantic cost Defined as semantic weights of the two endpoints and The product of the product and the geometric cost of fusion With semantic cost The final feature-perception collapse cost is obtained. : , Represents vertices The second-order error measure Represents vertices The second-order error measure This represents the new vertex generated after the edge is folded; And it uses an iterative dynamic threshold. Controlling the edges that can be collapsed in a single iteration: ,in, This represents the current iteration number. It is the base scaling factor. It is a constant. It is a growth index; During the iteration process, when If the edge is not allowed to be collapsed, it should be retained for subsequent iterations, and after each iteration... Increase, It will increase accordingly.

[0012] In some instances, based on the obtained structural importance region segmentation results, texture type segmentation results, and their importance weights, optimization parameters and constraints are set differently for geometric simplification, texture simplification, and Level of Detail (LOD), including: In terms of geometric simplification, a topology-aware vertex clustering simplification strategy is adopted: for adjacent vertices of the second target region, clustering and merging are performed under the condition of satisfying the preset curvature threshold and topology consistency constraints; for vertices of the first target region, clustering range parameters are set according to importance weights and clustering and merging operations are restricted, while the simplification error is constrained by a preset geometric error threshold. In terms of texture simplification, the texture data is compressed in blocks: the first texture compression parameter is used for the pixel blocks corresponding to the functional texture, and the second texture compression parameter is used for the pixel blocks corresponding to the decorative texture. The structural similarity index (SSIM) is constrained by a preset structural similarity threshold. The precision of the first texture compression parameter is greater than that of the second texture compression parameter. In terms of LOD layering, a three-level LOD system of "global-local-detail" is constructed: the global layer loads the first LOD model, which only retains the outline of the ship and the position of the target equipment; the local layer dynamically switches to the second LOD model, which includes the approximate shape of the equipment and the position of the interface; the detail layer loads the third LOD model, which retains the fine structure of the equipment. The level of detail of the first LOD model, the second LOD model and the third LOD model increases sequentially.

[0013] In some instances, the cross-platform adaptation of the processed model includes format conversion, shader optimization, and progressive loading configuration, including: Convert the model to the universal GLB or GLTF format; optimize the shader code for the GPU characteristics of the target device; develop a progressive loading module that prioritizes displaying the first LOD model outline when the model is loaded, and gradually replaces it with the third LOD model details as the data is downloaded.

[0014] In some instances, the method further includes: A lightweight model system with multimodal collaborative simplification is constructed. A geometry-texture-mesh local reconstruction simplification algorithm is developed for the whole ship model. The target features of the hull surface are preserved through mesh parameterization technology. The details of the first target region are preserved by combining the mesh reconstruction algorithm, while the second target region is simplified.

[0015] In some instances, the method further includes: For the application of the accompanying model of the Interactive Electronic Technical Manual (IETM), a mapping relationship of "task semantics - component importance weight - LOD / loading strategy" is established by utilizing the standardized description of component fault handling priority, maintenance task process and repair time limit in IETM. Based on this, a three-level LOD strategy of "component - scenario - task" is constructed: the first LOD model is used for the global scenario; the third LOD model is loaded as needed when the user focuses on the target component in the task process; and standard parts are instantiated and reused in maintenance operation scenarios.

[0016] In some instances, the method further includes preserving component identifiers and semantic mapping relationships during simplification and LOD processes.

[0017] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: This invention creatively transforms prior knowledge and task semantics in the maritime domain into computable mathematical constraints by embedding semantic weights into the cost function of the QEM in a product form and using dynamic threshold control to simplify the process. This lightweight method not only approximates the original model geometrically but also preserves the key features of the model at the functional level, realizing a paradigm shift from "geometry-driven" to "semantic and geometric co-driven".

[0018] Meanwhile, by constructing a three-level LOD strategy of "component-scene-task", a cross-platform universal intermediate representation, texture block compression and progressive loading mechanism, this invention can stably retain the key structural and functional features of ship models while significantly reducing the amount of model data, meet the needs of low-configuration devices for real-time rendering and interactive response, and provide a precise and efficient solution for lightweighting of 3D models in the field of shipbuilding engineering. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a method for lightweighting ship 3D models based on multimodal collaborative simplification provided in an embodiment of the present invention; Figure 2 This is a simplified schematic diagram of the core "multimodal feature-driven" method provided in the embodiments of the present invention; Figure 3 This is a simplified process diagram provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the iterative process provided in an embodiment of the present invention. Detailed Implementation

[0021] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] In the following description, specific embodiments of the invention will be illustrated with reference to steps and symbols performed by one or more computers, unless otherwise stated. Therefore, these steps and operations will be referred to several times as being performed by a computer, and computer execution as referred to herein includes operations by a computer processing unit representing electronic signals of data in a structured format. This operation transforms the data or maintains it at a location in the computer's memory system, which can be reconfigured or otherwise alter the operation of the computer in a manner well known to those skilled in the art. The data structure maintained by the data is the physical location of the memory, which has specific characteristics defined by the data format. However, the principles of the invention described above are not intended to be limiting, and those skilled in the art will understand that many of the following steps and operations can also be implemented in hardware.

[0023] The terms "module" or "unit" as used herein can be considered as software objects executing on the computing system. Different components, modules, engines, and services described herein can be considered as implementations on the computing system. The apparatus and methods described herein are preferably implemented in software, but can also be implemented in hardware, both of which are within the scope of this invention.

[0024] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0025] In this embodiment of the invention, a flowchart of a method for lightweighting a ship 3D model based on multimodal collaborative simplification is provided, as follows: Figure 1 As shown, the lightweighting method for ship 3D models based on multimodal collaborative simplification includes: S1. In-depth analysis of the original ship model, including multi-disciplinary data such as hull structure, fuel system, chilled water system, and other equipment components. A dedicated analysis engine extracts geometric data, texture data, semantic data (component names, functional attributes, etc.), and metadata (units, precision, etc.), and establishes a multi-dimensional data association map of "geometry-texture-semantics." The in-depth analysis of the original ship model specifically includes: Ship models typically contain cross-disciplinary data on hull structure and various equipment / piping / electromechanical systems (including but not limited to power, energy supply, cooling and fluid transport, auxiliary support, and general equipment components). Their high geometric complexity and diverse texture types easily lead to model size expansion. The data analysis phase uses a dedicated analysis engine to extract geometric data, texture data, semantic data (component names, functional attributes, etc.), and metadata (units, precision, etc.), establishing a multi-dimensional data association map of "geometry-texture-semantics."

[0026] Geometric data, including vertex coordinates, normals, and patch topological relationships, is used to describe the spatial shape and structure of the model. Texture data includes texture images, UV mapping relationships, material properties, etc., which are used to describe the appearance and visual effects of the model. Semantic data includes component names, functional attributes, assembly relationships, etc., which are used to identify the type and purpose of each component. Metadata, including units, precision, and version information, is used to ensure consistency in data parsing and transformation.

[0027] Based on this, a multi-dimensional data association graph of "geometry-texture-semantics" is established. This graph uses components as nodes and geometric dependencies, texture attachment relationships, and semantic inclusion relationships as edges to form a structured data association network, providing a decision-making basis for identifying important regions and assigning simplified weights in subsequent steps.

[0028] S2. Based on the data analysis results, a core simplification method driven by "multimodal features" is implemented. This involves introducing a mesh simplification method based on semantic feature weighting and dynamic ratio matching. Component semantic information (such as naming rules and functional types) is integrated into the simplification weight calculation to achieve adaptive simplification based on feature perception. The process is as follows: Figure 2 As shown, it specifically includes: 1) Identify the first target region in the model through geometric feature extraction algorithms (such as curvature-based and boundary detection). The first target region corresponds to the structurally important region (such as key hull components and connection parts) and the second target region. The second target region corresponds to the low-importance region (such as flat surfaces). Combine prior knowledge in the shipbuilding field (such as component naming rules and functional classification) to assign dynamic weights to different features. For example, the weight of key power components named "propulsion shaft" and "steering gear" is set to a higher value, and the weight of secondary components named "decorative panel" and "non-load-bearing bulkhead" is set to a lower value. 2) Based on the simplified Quadric Error Metrics (QEM) algorithm, a "feature-aware collapse cost threshold" mechanism is introduced. In the standard QEM algorithm, the error for a vertex v is defined as the sum of squared distances to its associated set of planes. This embodiment combines this geometric cost with semantic cost to form the final feature-aware collapse cost. By dynamically adjusting the collapse cost calculation formula, semantic weights are incorporated into the error metric, increasing the edge folding cost of high-weight regions, thereby reducing their simplification ratio. 3) Supports users to customize simplification strategies through the configuration interface, dynamically matching the simplification ratio based on the semantic information of the components, and achieving personalized simplification of "one model, one strategy"; 4) Analyze the texture maps to distinguish between "functional textures" and "decorative textures"; 5) Based on prior knowledge in the field of shipbuilding, assign an "importance weight" to each feature; Through the fusion analysis of multimodal features, the model can accurately locate the "core information that needs to be retained" and the "redundant information that can be simplified", providing a basis for decision-making for subsequent differentiated simplification strategies.

[0029] The semantic weights are incorporated into the cost function of QEM in a product form, and a dynamic threshold control is used to simplify the process, as follows: Figure 3 As shown. Specifically includes: 1) In the standard QEM algorithm, for a vertex The error is defined as the sum of squared distances to its associated set of planes; 2) For an edge The folding cost is usually calculated by folding the new vertex. The error; 3) Assign a semantic weight factor to each triangle or vertex in the model. ; 4) Semantic weights of vertices Take the maximum or average weight of its associated facets to ensure that key features are not diluted. To strengthen protection, the maximum value is used. ; 5) For the edge The semantic cost is defined as a function of the weights of the two endpoints. To significantly improve the folding cost of important edges, a product form is used: ; 6) Fusion Cost Function: Combining geometric and semantic costs to form the final feature-aware collapse cost:

[0030] 7) Employ an iterative dynamic threshold. This controls the edges that can be folded during a single iteration.

[0031]

[0032] in, It is the current iteration number. It is the base scaling factor (e.g.) ), It is a constant (such as 1) to ensure that the early threshold is not zero. It is a growth index (e.g., 1.5).

[0033] Iterative process as follows Figure 4 As shown, the details are as follows: 1) in the At the start of the next iteration, set the current threshold. ; 2) Traverse all valid edges: For each edge Calculate its ; 3) Selective folding: If If the condition is met, then the edge can be collapsed; otherwise, it is retained for subsequent iterations. 4) Iteration and Update: After completing one round of traversal and folding, Increase, threshold This increases the cost of edges, allowing more expensive edges to be collapsed in subsequent iterations.

[0034] S3. Based on the obtained structural importance region segmentation results, texture type segmentation results, and their importance weights, optimize parameters and constraints are set for differences in geometry, texture, and LOD (Level of Detail) to achieve a balance between "volume reduction" and "feature preservation." The optimization parameters include at least curvature threshold, geometric error threshold, clustering range parameter, texture compression accuracy parameter, and structural similarity threshold, and these thresholds and parameters are configurable. 1) In terms of geometric simplification, an improved "topology-aware vertex clustering algorithm" is adopted: For adjacent vertices in the second target region, i.e., the low-importance region, clustering and merging are performed based on a curvature threshold to reduce the number of faces; for vertices in the first target region, i.e., the high-importance region, the clustering range is limited by constraints to ensure the accuracy requirements for visualizing the geometric errors of key structures. Specifically, for adjacent vertices in the low-importance region, clustering and merging are performed under the condition of satisfying a preset curvature threshold and topological consistency constraints to reduce the number of faces; for vertices in the high-importance region, clustering range parameters are set according to importance weights and clustering and merging operations are limited, while a preset geometric error threshold is used to constrain the simplification error to ensure that the key structures meet the visualization accuracy requirements. 2) In terms of texture simplification, the texture data is compressed in blocks. The first texture compression precision parameter is used for the pixel blocks corresponding to the functional texture, and the second texture compression precision parameter is used for the pixel blocks corresponding to the decorative texture. The Structural Similarity Index (SSIM) is constrained by a preset structural similarity threshold to control the visual distortion caused by compression within an acceptable range. The first texture compression precision parameter corresponds to a higher texture compression precision parameter, and the second texture compression precision parameter corresponds to a lower texture compression precision parameter.

[0035] 3) Regarding LOD layering, a three-level LOD system of "global-local-detail" is constructed: the global layer loads the first LOD model, retaining only the hull outline and target device positions (such as the main device positions); the local layer dynamically switches to the second LOD model, including the shape of the device and interface positions; the detail layer loads the third LOD model, retaining the detailed structure of the device. Through layered LOD management, the model only needs to load the minimum LOD level required for the current viewpoint on the target configuration device (such as a low-configuration device), significantly reducing memory and video memory usage. The first LOD model corresponds to the low-poly model, the second LOD model corresponds to the mid-poly model, and the third LOD model corresponds to the high-poly model, with the level of detail increasing sequentially from the first LOD model to the third LOD model.

[0036] S4. Finally, perform cross-platform adaptation, as follows: Convert the model to the universal GLB / GLTF format to avoid the high redundancy of text data in traditional formats; Optimize shader code to reduce shader compilation time and runtime computation for the GPU characteristics of target devices (such as low-configuration devices); A "progressive loading module" was developed, which prioritizes displaying the low-poly outline when the model is loading, and gradually replaces it with high-poly details as the data is downloaded, thus avoiding long waiting times for users to wait for the complete model to load.

[0037] Furthermore, when constructing a lightweight model system of "multimodal collaborative simplification", a geometry-texture-mesh local reconstruction simplification algorithm is developed for the whole ship model. The target features of the hull surface (such as the key features of the hull surface) are preserved through mesh parameterization technology. The mesh reconstruction algorithm is combined with the preservation of the details of the first target region, i.e. the high-importance region, while the second target region, i.e. the low-importance region, is greatly simplified.

[0038] Furthermore, for the application of the IETM (Interactive Electronic Technical Manual) supporting model, this embodiment of the invention utilizes the standardized description of the guarantee semantics in IETM, such as the priority of component fault handling, maintenance task process, and repair time limit, to establish a mapping relationship of "task semantics - component importance weight - LOD / loading strategy", and constructs a three-level LOD strategy of "component - scene - task": the global scene adopts the first LOD model, i.e., low-poly model, to ensure smooth roaming; when the user focuses on the target component (such as a high-weight key component) in the task process, the third LOD model, i.e., high-poly model, is loaded as needed; and the maintenance operation scene uses instantiation and reuse of standard parts to reduce the overall computing and rendering load.

[0039] Furthermore, a "cross-platform universal intermediate representation" format was developed to streamline the entire data flow. A universal intermediate format incorporating geometry, materials, assembly relationships, and business information was designed. This format supports seamless conversion between the design end (browser) and the cloud (lightweight server), automatically preserving key metadata during the conversion process.

[0040] Furthermore, this invention promotes the deep integration of models and knowledge to empower intelligent operation and maintenance services. By simplifying the LOD (Level of Detail) process while preserving component identifiers and semantic mapping relationships, this lightweight model can be directly used in IETM (Intelligent Operation and Maintenance Technology) interaction scenarios, supporting the display and push of maintenance knowledge based on component selection.

[0041] The above provides a detailed description of a lightweight method for ship 3D models based on multimodal collaborative simplification provided by the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A lightweight method for ship 3D models based on multimodal collaborative simplification, characterized in that, include: The original ship model was analyzed in depth to extract geometric data, texture data, semantic data and metadata, and a multi-dimensional data association map of geometry-texture-semantics was established. Based on the aforementioned multi-dimensional data association map, a core simplification driven by multi-modal features is implemented. A mesh simplification method based on semantic feature weighting and dynamic ratio matching is introduced, integrating component semantic information into the simplification weight calculation to achieve adaptive simplification based on feature perception. Specifically, this includes: identifying the first and second target regions in the model through a geometric feature extraction algorithm, and assigning dynamic weights to different features based on prior knowledge of the shipbuilding domain, where the first target region is more important than the second target region; introducing a feature-perceived collapse cost threshold mechanism based on the quadratic error metric (QEM) simplification algorithm, integrating semantic weights into the error metric; allowing users to customize simplification strategies through a configuration interface, with the system dynamically matching simplification ratios based on component semantic information; analyzing texture maps to distinguish between functional and decorative textures; assigning importance weights to each feature based on prior knowledge of the shipbuilding domain; and locating information to be retained and redundant information through multi-modal feature fusion analysis. The integration of semantic weights into the error metric includes: vertex... semantic weight Take the maximum value of the weight of its associated facet, for the edge Its semantic cost Defined as semantic weights of the two endpoints and The product of the product and the geometric cost of fusion With semantic cost The final feature-perception collapse cost is obtained. : , Represents vertices The second-order error measure Represents vertices The second-order error measure This represents the new vertex generated after edge folding; and employs an iterative dynamic threshold. Controlling the edges that can be collapsed in a single iteration: ,in, This represents the current iteration number. It is the base scaling factor. It is a constant. It is a growth index; during the iteration process, when If the edge is not allowed to be collapsed, it should be retained for subsequent iterations, and after each iteration... Increase, Consequently, it increases; Based on the obtained structural importance region division results, texture type division results and their importance weights, optimization parameters and constraints are set differently for geometric simplification, texture simplification and level of detail (LOD) to achieve a balance between volume reduction and feature preservation. The processed model is adapted for cross-platform use, including format conversion, shader optimization, and progressive loading configuration, to support real-time rendering and interaction on the target configuration terminal.

2. The method according to claim 1, characterized in that, The process involves deep analysis of the original ship model to extract geometric data, texture data, semantic data, and metadata, and to establish a multi-dimensional data association map of geometry, texture, and semantics, including: The parsing engine extracts geometric data, texture data, semantic data containing component names and functional attributes, and metadata containing units and precision from the whole ship model. The whole ship model includes hull structure and cross-disciplinary data, including power system, energy supply system, cooling and fluid transport system, auxiliary support system and general equipment components.

3. The method according to claim 2, characterized in that, Based on the obtained structural importance region segmentation results, texture type segmentation results, and their importance weights, optimization parameters and constraints are set differently for geometric simplification, texture simplification, and Level of Detail (LOD), including: In terms of geometric simplification, a topology-aware vertex clustering simplification strategy is adopted: for adjacent vertices of the second target region, clustering and merging are performed under the condition of satisfying the preset curvature threshold and topology consistency constraints; for vertices of the first target region, clustering range parameters are set according to importance weights and clustering and merging operations are restricted, while the simplification error is constrained by a preset geometric error threshold. In terms of texture simplification, the texture data is compressed in blocks: the first texture compression parameter is used for the pixel blocks corresponding to the functional texture, and the second texture compression parameter is used for the pixel blocks corresponding to the decorative texture. The structural similarity index (SSIM) is constrained by a preset structural similarity threshold. The precision of the first texture compression parameter is greater than that of the second texture compression parameter. In terms of LOD layering, a three-level LOD system of global-local-detail is constructed: the global layer loads the first LOD model, which only retains the outline of the ship and the position of the target equipment; the local layer dynamically switches to the second LOD model, which includes the shape of the equipment and the position of the interface; the detail layer loads the third LOD model, which retains the fine structure of the equipment. The level of detail of the first LOD model, the second LOD model and the third LOD model increases sequentially.

4. The method according to claim 3, characterized in that, The cross-platform adaptation of the processed model includes format conversion, shader optimization, and progressive loading configuration, including: Convert the model to the universal GLB or GLTF format; optimize the shader code for the GPU characteristics of the target device; develop a progressive loading module that prioritizes displaying the first LOD model outline when the model is loaded, and gradually replaces it with the third LOD model details as the data is downloaded.

5. The method according to any one of claims 1 to 4, characterized in that, The method further includes: A lightweight model system with multimodal collaborative simplification is constructed. A geometry-texture-mesh local reconstruction simplification algorithm is developed for the whole ship model. The target features of the hull surface are preserved through mesh parameterization technology. The details of the first target region are preserved by combining the mesh reconstruction algorithm, while the second target region is simplified.

6. The method according to any one of claims 1 to 4, characterized in that, The method further includes: For the application of the accompanying model of the Interactive Electronic Technical Manual (IETM), the standardized descriptions of component fault handling priorities, maintenance task processes, and repair time limits in IETM are used to establish a mapping relationship between task semantics, component importance weight, and LOD / loading strategy. Based on this, a three-level LOD strategy of component-scenario-task is constructed: the first LOD model is used for global scenarios; the third LOD model is loaded as needed when the user focuses on the target component in the task process; and standard parts are instantiated and reused in maintenance operation scenarios.

7. The method according to any one of claims 1 to 4, further comprising: Component identifiers and semantic mapping relationships are preserved during the simplification and LOD process.

Citation Information

Patent Citations

  • Lightweight methods, systems, terminals, and media for VR-oriented ship big data models

    CN111210521B

  • Lightweight integration method for three-dimensional model of ship product

    CN120180581A

  • Lightweight three-dimensional data format generation method suitable for ship model labeling

    CN120655826A

  • Three-dimensional model lightweight reconstruction and rendering method based on multi-strategy collaboration

    CN121280602A