Digital twinning scene construction method, system and device and medium

Abstract

The invention discloses a digital twinning scene construction method, system and device and a medium, and particularly relates to the technical field of scene construction, and the technical key points are as follows: constructing a mapping relation set of natural language description and semantic API call sequences, screening an optimal semantic API calling sequence corresponding to the natural language description in the mapping relation set through a pre-constructed target function, and adjusting the mapping relation by utilizing the optimal semantic API calling sequence to obtain an optimal mapping relation set; receiving a natural language demand description, querying a semantic API calling sequence corresponding to the natural language demand description by using the optimal mapping relationship set, and based on the digital twinning domain knowledge base and the semantic API calling sequence, performing matching in the model library to obtain an optimal digital twinning scene model, and generating an initial scene construction scheme; and performing verification adjustment on the initial scene construction scheme by using a pre-constructed model feasibility verification algorithm to obtain an adjusted scene construction scheme.

Description

Technical Field

[0001] The present invention relates to the field of scene construction technology, and in particular to a digital twin scene construction method, system, device and medium. Background Art

[0002] The digital twin model is based on physical entities. By collecting, organizing and modeling relevant data of the entities, and using mathematical modeling, physical simulation, data analysis and other technologies, it converts the characteristics, structure, working principle and performance of the physical entities into mathematical models and algorithms.

[0003] The current mainstream WebGL digital twin scene construction method mainly relies on manually written shader code (GLSL) or the use of WebGL frameworks (such as Three.js and Babylon.js), which directly develop lightweight digital twin scenes through JavaScript; however, the existing technology has the following disadvantages: First, manual development is adopted: developers need to write a large amount of JavaScript code and manually manage the WebGL state machine, which is a large workload and difficult to adapt to dynamically changing digital twin applications; second, there is a lack of direct mapping from natural language to WebGL API, resulting in a high threshold for developers; third, there is a lack of intelligence: it is unable to automatically identify business needs and generate adaptive three-dimensional scenes, and it is difficult to adapt to specific industry needs.

[0004] In view of this, the present invention aims to provide a digital twin scene construction method, system, device and medium to solve the above-mentioned related problems. Summary of the invention

[0005] The technical problem to be solved by the present invention is that the prior art lacks direct mapping from natural language to WebGL API, and cannot automatically identify business needs to generate adaptive three-dimensional scenes, resulting in low practicality. The purpose is to provide a digital twin scene construction method, system, device and medium, by constructing a mapping relationship set of natural language description and semantic API call sequence, and adjusting the mapping relationship through a pre-built objective function to obtain an optimal mapping relationship set, thereby completing the direct mapping from natural language description to WebGL API, realizing cross-modal instruction conversion, and achieving the technical effect of automatically identifying business needs to generate adaptive three-dimensional scenes; at the same time, the material mapping of the digital twin scene is optimized through the generative adversarial network model to improve the authenticity of the three-dimensional model of the digital twin scene; the combination of geometric structure information features and scene logic information features is optimized through the dual-stream attention mechanism, so as to improve the intelligence level of the digital twin scene construction system, so that it can be widely used in industrial simulation, medical training, smart cities and other fields, and has broad market application prospects.

[0006] The present invention is achieved through the following technical solutions:

[0007] A method for constructing a digital twin scenario, the method includes:

[0008] Using a natural language model to parse historical natural language descriptions to obtain corresponding semantic instructions, constructing a mapping relationship set between natural language descriptions and semantic API call sequences, screening the optimal semantic API call sequence corresponding to the natural language description in the mapping relationship set through a pre-constructed objective function, and using the optimal semantic API call sequence to adjust the mapping relationship to obtain an optimal mapping relationship set;

[0009] Receiving a natural language requirement description, querying the semantic API call sequence corresponding to the natural language requirement description using the optimal mapping relationship set, and based on the digital twin domain knowledge base and the semantic API call sequence, matching the optimal digital twin scenario model in the model library to generate an initial scenario construction plan;

[0010] Using a pre-constructed model feasibility verification algorithm to verify the feasibility of the initial scenario construction plan, and adjusting the initial scenario construction plan based on the verification result to obtain an adjusted scenario construction plan.

[0011] Further, the pre-constructed objective function is specifically: where F cmd represents the optimal semantic API call sequence; represents finding the parameter that maximizes the probability distribution P; P(T|S) represents the probability of obtaining the semantic API call sequence T given the natural language description S.

[0012] Further, the method further includes: constructing a generative adversarial network model, where the generator network generates different material mapping schemes by inputting the optimal digital twin scenario model, and the discriminator network judges the difference between the generated material mapping scheme and the real material mapping to obtain a difference result;

[0013] Based on the difference result, iteratively optimize the material mapping scheme of the digital twin scenario model until the loss function converges to obtain an optimal material mapping scheme.

[0014] Further, based on the digital twin domain knowledge base and the semantic API call sequence, matching the optimal digital twin scenario model in the model library to generate an initial scenario construction plan, specifically:

[0015] Extract the domain features in the digital twin domain knowledge base, input the domain features into the dual-stream attention mechanism model for correlation calculation, and combine the Transformer model to predict the best rendering parameters of the digital twin scenario;

[0016] Using the optimal rendering parameters and semantic API call sequences of the digital twin scenario, the optimal digital twin scenario model is matched in the model library to generate an initial scenario construction plan.

[0017] Furthermore, a pre-built model feasibility verification algorithm is used to verify the feasibility of the initial scenario construction plan, and the initial scenario construction plan is adjusted based on the verification results to obtain an adjusted scenario construction plan. The pre-built model feasibility verification algorithm includes a light sensitivity detection algorithm, a node loop detection algorithm, and a rendering performance prediction algorithm.

[0018] The present invention also provides a digital twin scenario construction system, which is used in any one of the above-mentioned digital twin scenario construction methods. The system includes:

[0019] A mapping relationship set construction module, which is used to parse the historical natural language description using a natural language model to obtain corresponding semantic instructions, construct a mapping relationship set between the natural language description and the semantic API call sequence, screen the optimal semantic API call sequence corresponding to the natural language description in the mapping relationship set through a pre-built objective function, and use the optimal semantic API call sequence to adjust the mapping relationship to obtain an optimal mapping relationship set;

[0020] A scenario construction plan generation module, which is used to receive the natural language requirement description, query the semantic API call sequence corresponding to the natural language requirement description using the optimal mapping relationship set, and match the optimal digital twin scenario model in the model library based on the digital twin domain knowledge base and the semantic API call sequence to generate an initial scenario construction plan;

[0021] The scenario construction plan generation module is also used to verify the feasibility of the initial scenario construction plan using a pre-built model feasibility verification algorithm, and adjust the initial scenario construction plan based on the verification results to obtain an adjusted scenario construction plan.

[0022] Furthermore, the system further includes: a construction of a generative adversarial network model, where the generator network inputs the optimal digital twin scenario model to generate different material mapping schemes, and the discriminator network judges the difference between the generated material mapping scheme and the real material mapping to obtain a difference result;

[0023] Iteratively optimize the material mapping scheme of the digital twin scenario model based on the difference result until the loss function converges to obtain an optimal material mapping scheme.

[0024] The present invention also provides a computer device, comprising a system memory and a processor, wherein the system memory stores a computer program, and the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0025] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any one of the methods described above are implemented.

[0026] The present invention also provides a computer program product comprising instructions, and when the instructions are executed by a computer device cluster, the computer device cluster executes any of the above methods.

[0027] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0028] In the present invention, by constructing a mapping relationship set between natural language description and semantic API call sequence, and adjusting the mapping relationship through a pre-built objective function to obtain the optimal mapping relationship set, the direct mapping from natural language description to WebGL API is completed, and cross-modal instruction conversion is realized to achieve the technical effect of automatically identifying business needs and generating adaptive three-dimensional scenes; at the same time, the material mapping of the digital twin scene is optimized through the generative adversarial network model to improve the authenticity of the three-dimensional model of the digital twin scene; the combination of geometric structure information features and scene logic information features is optimized through the dual-stream attention mechanism, so as to improve the intelligence level of the digital twin scene construction system, so that it can be widely used in industrial simulation, medical training, smart cities and other fields, and has broad market application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0030] Figure 1 This is a method flow chart of a digital twin scene construction method in this embodiment;

[0031] Figure 2 This is a module schematic diagram of a digital twin scene construction system in this embodiment;

[0032] Figure 3 It is a structural schematic diagram of a computer device in this embodiment. DETAILED DESCRIPTION

[0033] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0034] In the present disclosure, unless otherwise specified, the terms "first", "second", etc. are used to describe various elements and are not intended to limit the positional relationship, timing relationship, or importance relationship of these elements. Such terms are only used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of the element, and in certain cases, based on the context description, they may also refer to different instances.

[0035] The terms used in the description of various examples in the present disclosure are only for the purpose of describing specific examples and are not intended to be restrictive. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element may be one or more. In addition, the term "and / or" used in the present disclosure covers any one of the listed items and all possible combinations.

[0036] Embodiment 1

[0037] See Figure 1 As shown, a method flow chart of a digital twin scenario construction method is shown. The method includes:

[0038] S1: Use a natural language model to parse historical natural language descriptions to obtain corresponding semantic instructions, construct a mapping relationship set between natural language descriptions and semantic API call sequences, screen the optimal semantic API call sequence corresponding to the natural language description in the mapping relationship set through a pre-constructed objective function, and use the optimal semantic API call sequence to adjust the mapping relationship to obtain an optimal mapping relationship set;

[0039] It should be noted that in this embodiment, the natural language model uses the BERT model. The historical natural language description text is semantically parsed through the BERT model to obtain corresponding semantic instructions, and a semantic instruction set is constructed through multiple semantic instructions. At the same time, it should be noted that in this embodiment, the semantic instruction refers to a WebGL instruction, and the semantic API call sequence (WebGL API call sequence) refers to an API interface call combination formed by arranging a series of WebGL instructions in a specific order;

[0040] The pre-constructed objective function is specifically: Where Fcmd represents the optimal semantic API call sequence (or the optimal instruction conversion result), which is obtained by maximizing the conditional probability P(T|S); represents finding the parameters that maximize the probability distribution P; P(T|S) represents the probability of obtaining the semantic API call sequence T given the natural language description S. It should be noted that this is a conditional probability, where S is the natural language description, such as the natural language description "create a blue sphere"; T is the corresponding WebGL API call sequence, which is a combination of a series of instructions like gl.createSphere({color:'blue'});

[0041] Meanwhile, it should be noted that in this embodiment, the core idea of the objective function is to find a conversion method (corresponding to a probability distribution P) among all possible ways of converting the natural language instruction S into the WebGL API call sequence T, such that the probability P(T|S) of obtaining T given S is maximized;

[0042] Specifically, in this embodiment, first, models such as BERT are used to analyze the natural language input S to obtain a set of possible WebGL instructions. Then, in the structure mapping step, multiple possible mapping relationships from the natural language description to the WebGL API call are constructed, and these mapping relationships all correspond to different conditional probabilities P(T|S); finally, through operations, the optimal F cmd is selected from these possible mapping relationships, that is, the result that is most likely to correctly convert the natural language input into the WebGL API call sequence.

[0043] S2: Receive the natural language requirement description, query the semantic API call sequence corresponding to the natural language requirement description using the optimal mapping relationship set, and based on the digital twin domain knowledge base and the semantic API call sequence, match the optimal digital twin scenario model in the model library to generate an initial scenario construction plan;

[0044] It should be noted that in this embodiment, the digital twin domain knowledge base refers to the information features related to the digital twin domain, including geometric structure information features, scene logic information features, etc.; the model library refers to the existing three-dimensional scene model database, and an existing model library can be used for scene construction;

[0045] Specifically, in this embodiment, a natural language requirement description is first received, a semantic API call sequence corresponding to the natural language requirement description is found in the optimal mapping relationship set, domain features in the digital twin domain knowledge base are extracted, the domain features are input into the two-stream attention mechanism model for correlation calculation, and the optimal rendering parameters of the digital twin scene are obtained by combining the Transformer model prediction; the optimal rendering parameters of the digital twin scene and the semantic API call sequence are used to match the optimal digital twin scene model in the model library, and an initial scene construction plan is generated;

[0046] Among them, the correlation calculation formula of the dual-stream attention mechanism model is specifically as follows: Among them, A i，j represents the element in the i-th row and j-th column of the attention weight matrix, which measures the degree of association between the i-th position and the j-th position in the input sequence; Q i represents the query vector at the i-th position, which is used to find relevant information in the input sequence; represents the transpose of the key vector at the jth position. The key vector is used to store information for matching with the query vector. For all positions k The sum of the exponential values ​​plays a normalization role.

[0047] Specifically, in this embodiment, in the scenario of focusing on information features related to the digital twin field, the dual-stream attention mechanism can perform attention calculations on the geometric structure information features and the scene logic information features. For example, in 3D model rendering, geometric structure information features may include the vertex coordinates of the model, the connection relationship of the faces, etc.; scene logic information features may include rendering priority, properties of different materials, etc. Through the dual-stream attention mechanism, important correlation information within and between these two types of features can be captured respectively; the Transformer model has a powerful sequence modeling capability, which can process long sequence data and capture the global dependencies therein. The correlation information feature representation obtained by the dual-stream attention mechanism is input into the Transformer model. The Transformer model can learn the mapping relationship between different features and the optimal rendering parameters, thereby predicting the optimal rendering parameters for specific scenes and needs, such as light intensity, resolution, texture mapping method, etc.

[0048] S3: Use the pre-built model feasibility verification algorithm to verify the feasibility of the initial scenario construction plan, and adjust the initial scenario construction plan based on the verification result to obtain an adjusted scenario construction plan;

[0049] It should be noted that in this embodiment, the pre-built model feasibility verification algorithm includes a light sensitivity detection algorithm, a node loop detection algorithm, and a rendering performance prediction algorithm. The color difference offset of the material under different lighting conditions is calculated through the light sensitivity detection algorithm; the invalid connection path is identified based on the graph theory algorithm through the node loop detection mechanism; the rendering performance prediction model is used to estimate whether the scene frame rate meets the standard. At the same time, the light sensitivity detection algorithm, the node loop detection algorithm, and the rendering performance prediction algorithm are all conventional technical means in this field, and those skilled in the art are aware of their technical content and will not be elaborated here. In this embodiment, by combining the light sensitivity detection algorithm, the node loop detection algorithm, and the rendering performance prediction algorithm, an automated model feasibility verification system is formed for the dynamic adjustment and optimization of the digital twin model. Through the systematic combination of multiple modules, the overall efficiency is improved, and a more accurate and workable solution is provided.

[0050] In another embodiment, the method further includes: constructing a generative adversarial network model, wherein the generator network generates different material mapping schemes by inputting the optimal digital twin scene model, and the discriminator network judges the difference between the generated material mapping scheme and the real material mapping to obtain a difference result; based on the difference result, the material mapping scheme of the digital twin scene model is iteratively optimized until the loss function converges to obtain the optimal material mapping scheme.

[0051] Specifically, in this embodiment, the loss function of the generative adversarial network model is:

[0052]

[0053] where L GAN represents the loss function of the generative adversarial network, which is used to measure the performance of the generator and the discriminator; represents the expectation of the samples x in the real data distribution p data (x), where x represents the real sample data. For example, in the context of scene generation, it may be existing high-quality three-dimensional scene data; D(x) represents the output of the discriminator network for the real sample x, which is a value between 0 and 1, and logD(x) represents taking the logarithm of D(x), which is to map the probability value to a range that is more convenient for calculation and optimization; represents the expectation of the samples in the noise distribution p z(z) is the expectation of the sample z in the generator; z is the input noise vector of the generator, which is usually sampled from a known distribution (such as a normal distribution); G(z) represents the sample generated by the generator network based on the input noise z; D(G(z)) represents the output of the discriminator network for the generated sample G(z), that is, the probability that the discriminator network judges that the generated sample is a real sample; 1-D(G(z)) represents the probability that the discriminator network judges that the generated sample is a false sample, and log(1-D(G(z))) represents taking the logarithm of 1-D(G(z)) for ease of calculation and optimization;

[0054] For the discriminator network D, its goal is to encourage the discriminator to output high probability for real samples (because the logarithmic function is monotonically increasing within the domain of definition, so (the larger D(x) is, the larger logD(x) is); and to encourage the discriminator to output low probability for generated samples; for the generator G, its goal is to make the generated samples as likely to be mistaken by the discriminator as real samples, that is, to make D(G(z)) as large as possible.

[0055] Specifically, in this embodiment, when optimizing material mapping by using the generative adversarial network GAN model, the generator can generate different material mapping schemes, and the discriminator is responsible for judging the difference between the generated material mapping scheme and the real, high-quality material mapping. Through continuous adversarial training, the generator can generate material mappings that are closer to reality and more in line with rendering quality requirements, thereby improving the rendering quality of the three-dimensional scene; after parsing the natural language demand description and extracting relevant information and matching the optimal digital twin scene model, GAN further optimizes details such as materials. It can generate more realistic material mappings based on existing object types, materials, lighting and other information, so that the generated three-dimensional scene is more in line with expectations in terms of visual effects, and realizes the automatic generation of high-quality three-dimensional scenes based on existing template models.

[0056] Specifically, in this embodiment, by constructing a mapping relationship set between natural language description and semantic API call sequence, and adjusting the mapping relationship through a pre-built objective function to obtain the optimal mapping relationship set, the direct mapping from natural language description to WebGL API is completed, and cross-modal instruction conversion is realized to achieve the technical effect of automatically identifying business needs and generating adaptive three-dimensional scenes; at the same time, the material mapping of the digital twin scene is optimized through the generative adversarial network model to improve the authenticity of the three-dimensional model of the digital twin scene; the combination of geometric structure information features and scene logic information features is optimized through the dual-stream attention mechanism to improve the intelligence level of the digital twin scene construction system, so that it can be widely used in industrial simulation, medical training, smart cities and other fields, and has broad market application prospects.

[0057] Example 2

[0058] See Figure 2 As shown, the present invention also provides a digital twin scenario construction system, which is used in any one of the above-mentioned digital twin scenario construction methods. The system includes:

[0059] A mapping relationship set construction module 100, which is used to parse historical natural language descriptions by using a natural language model to obtain corresponding semantic instructions, construct a mapping relationship set between natural language descriptions and semantic API call sequences, screen the optimal semantic API call sequence corresponding to the natural language description in the mapping relationship set through a pre-constructed objective function, and use the optimal semantic API call sequence to adjust the mapping relationship to obtain an optimal mapping relationship set;

[0060] A scenario construction plan generation module 200, which is used to receive a natural language requirement description, query the semantic API call sequence corresponding to the natural language requirement description by using the optimal mapping relationship set, and match the optimal digital twin scenario model in the model library based on the digital twin domain knowledge base and the semantic API call sequence to generate an initial scenario construction plan;

[0061] The scenario construction plan generation module 200 is also used to verify the feasibility of the initial scenario construction plan by using a pre-constructed model feasibility verification algorithm, and adjust the initial scenario construction plan based on the verification result to obtain an adjusted scenario construction plan.

[0062] It should be noted that the modules in the system of Embodiment 2 correspond to the steps in the method of Embodiment 1. The steps in the method of Embodiment 1 have been elaborated in detail in Embodiment 1, and the content of the modules in the system will not be elaborated in detail in Embodiment 2 here.

[0063] Embodiment 3

[0064] See Figure 3 As shown, this embodiment also provides a computer device, including a system memory 1005 and a processor 1001. The system memory 1005 stores a computer program, and when the processor 1001 executes the computer program, it implements the steps of any one of the above methods.

[0065] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, when the processor 1001 executes the computer program, it implements the functions of each module / unit in the above system / device embodiments.

[0066] Specifically, in this embodiment, the computer program can be divided into one or more modules / units. One or more modules / units are stored in the system memory 1005 and executed by the processor 1001 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the terminal device.

[0067] The terminal device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, the processor 1001 and the system memory 1005. Those skilled in the art can understand that this does not limit the terminal device, and it may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0068] The processor 1001 can be a central processing unit (CPU), or can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or this processor can also be any conventional processor, etc.

[0069] The system memory 1005 can be an internal storage unit of the terminal device, such as the hard disk or memory of the terminal device. The system memory 1005 can also be a storage device 1004 of the terminal device, such as a plug-in hard disk equipped on the terminal device, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the system memory 1005 can also include both the internal storage unit of the terminal device and the storage device 1004. The system memory 1005 is used to store the computer program and other programs and data required by the terminal device. The system memory 1005 can also be used to temporarily store the data that has been output or will be output.

[0070] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described system, system, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0071] Example 4

[0072] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method according to any one of the above are implemented.

[0073] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device or component, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection with one or more wires, a portable computer disk, a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), registers, hard disks, optical fibers, portable Compact Disc Read-Only Memory (CD-ROM), optical storage devices, magnetic storage devices or any suitable combination of the above, or any other form of computer-readable storage medium known in the art.

[0074] An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an Application Specific Integrated Circuit (ASIC). In the embodiments of the present invention, the computer-readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device or component.

[0075] Example 5

[0076] This embodiment also provides a computer program product containing instructions. When the instructions are run by a computer device cluster, the computer device cluster is caused to execute the method described in Example 1.

[0077] The specific embodiments described above further elaborate on the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a digital twin scene, characterized in that: include: Use the natural language model to parse the historical natural language description to obtain the corresponding semantic instructions, build a mapping relationship set between the natural language description and the semantic API call sequence, select the optimal semantic API call sequence corresponding to the natural language description in the mapping relationship set through the pre-built objective function, and adjust the mapping relationship using the optimal semantic API call sequence to obtain the optimal mapping relationship set; Receive natural language requirement descriptions, use the optimal mapping relationship set to query the semantic API call sequence corresponding to the natural language requirement description, and match the optimal digital twin scene model in the model library based on the digital twin domain knowledge base and the semantic API call sequence to generate an initial scene construction plan; The feasibility of the initial scenario construction plan is verified using a pre-built model feasibility verification algorithm, and the initial scenario construction plan is adjusted based on the verification result to obtain an adjusted scenario construction plan.

2. A digital twin scene construction method according to claim 1, characterized in that: Pre-built objective functions, specifically: Among them, F cmd Represents the optimal semantic API call sequence; It means finding the parameters that maximize the probability distribution P; P(T|S) represents the probability of obtaining the semantic API call sequence T given the natural language description S.

3. A digital twin scene construction method according to claim 1, characterized in that: The method also includes: constructing a generative adversarial network model, wherein the generator network generates different material mapping schemes by inputting the optimal digital twin scene model, and the discriminator network judges the difference between the generated material mapping scheme and the real material mapping to obtain a difference result; Based on the difference results, the material mapping scheme of the digital twin scene model is iteratively optimized until the loss function converges to obtain the optimal material mapping scheme.

4. A digital twin scene construction method according to claim 1, characterized in that: Based on the digital twin domain knowledge base and semantic API call sequence, the optimal digital twin scene model is matched in the model library to generate the initial scene construction plan, specifically: Extract domain features from the digital twin domain knowledge base, input the domain features into the two-stream attention mechanism model for correlation calculation, and combine the Transformer model to predict the optimal rendering parameters of the digital twin scene; By using the optimal rendering parameters and semantic API call sequence of the digital twin scene, the optimal digital twin scene model is matched in the model library to generate the initial scene construction plan.

5. A digital twin scene construction method according to claim 1, characterized in that: The feasibility of the initial scene construction plan is verified by using a pre-built model feasibility verification algorithm, and the initial scene construction plan is adjusted based on the verification result to obtain an adjusted scene construction plan, wherein the pre-built model feasibility verification algorithm includes a light sensitivity detection algorithm, a node loop detection algorithm, and a rendering performance prediction algorithm.

6. A digital twin scene construction system, characterized in that: The system is used in a digital twin scene construction method according to any one of claims 1 to 5, and the system comprises: A mapping relationship set construction module is used to use a natural language model to parse historical natural language descriptions to obtain corresponding semantic instructions, construct a mapping relationship set between natural language descriptions and semantic API call sequences, select the optimal semantic API call sequence corresponding to the natural language description in the mapping relationship set through a pre-constructed objective function, and adjust the mapping relationship using the optimal semantic API call sequence to obtain the optimal mapping relationship set; The scenario construction scheme generation module is used to receive the natural language requirement description, use the optimal mapping relationship set to query the semantic API call sequence corresponding to the natural language requirement description, and match the optimal digital twin scenario model in the model library based on the digital twin domain knowledge base and the semantic API call sequence to generate the initial scenario construction scheme; The scenario construction plan generation module is also used to verify the feasibility of the initial scenario construction plan using a pre-built model feasibility verification algorithm, and adjust the initial scenario construction plan based on the verification result to obtain an adjusted scenario construction plan.

7. A digital twin scene construction system according to claim 6, characterized in that: The system also includes: constructing a generative adversarial network model, wherein the generator network generates different material mapping schemes by inputting the optimal digital twin scene model, and the discriminator network judges the difference between the generated material mapping scheme and the real material mapping to obtain a difference result; Based on the difference results, the material mapping scheme of the digital twin scene model is iteratively optimized until the loss function converges to obtain the optimal material mapping scheme.

8. A computer device comprising a system memory and a processor, wherein the system memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer program product comprising instructions, characterized in that When the instructions are executed by a computer device cluster, the computer device cluster executes the method according to any one of claims 1 to 5.

Images