Car seeing and selecting platform based on 3D modeling and AI technology

Abstract

The invention provides a vehicle viewing and selecting platform based on 3D modeling and AI technology, which comprises a 3D vehicle model demonstration unit, an AI model unit and a real-time communication unit, and is characterized in that the 3D vehicle model demonstration unit and the AI model unit are connected through the real-time communication unit; an interaction instruction of the AI model unit is sent to the 3D car model demonstration unit through the real-time communication unit, and the 3D car model demonstration unit executes the interaction instruction. According to the invention, the appearance, interior decoration and technical details of the vehicle are truly restored through the 3D vehicle model demonstration unit, various characteristics and technical parameters of the vehicle are interpreted for a user through the AI model unit in a professional, accurate, popular and easy-to-understand manner, the user is helped to better understand vehicle information, the difficulty in understanding is eliminated, meanwhile, the vehicle viewing and selecting become more efficient and convenient, and the user experience is improved. A user can access the 3D car model through the network anytime and anywhere and communicate with the AI conversation robot in real time to obtain required information, so that time and energy are greatly saved, and the efficiency of seeing and selecting a car is improved.

Description

Technical Field

[0001] The present invention relates to the field of 3D modeling and AI technology, and in particular to a car viewing and selection platform based on 3D modeling and AI technology. Background Art

[0002] A car is a vehicle that has its own power to drive and can be driven without tracks or electricity. In a broad sense, vehicles that are driven by their own mechanical energy are generally called cars. There are many cars in the world, and cars in each country have their own characteristics, such as safety, speed, comfort, and endurance. So when buying a car, you should choose a model based on your own situation.

[0003] When users are currently looking for a car, they can only learn about the vehicle through flat pictures, text materials and the sales staff's explanations. However, these methods have limitations. Flat pictures cannot show the three-dimensional appearance and details of the vehicle. Text materials may be too professional or abstract for users to understand intuitively. The sales staff's explanations may contain subjective biases or inaccurate information. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the defect that the way of understanding vehicle information when viewing and selecting a car in the prior art is relatively limited, and to provide a car viewing and selection platform based on 3D modeling and AI technology.

[0005] The present invention solves the above technical problems through the following technical solutions:

[0006] The present invention provides a car viewing and selection platform based on 3D modeling and AI technology, comprising a 3D car model demonstration unit, an AI model unit and a real-time communication unit, wherein the 3D car model demonstration unit and the AI ​​model unit are connected via the real-time communication unit;

[0007] The interaction instructions of the AI ​​model unit are sent to the 3D car model demonstration unit through the real-time communication unit, and the 3D car model demonstration unit executes the interaction instructions.

[0008] In this technical solution, the vehicle's appearance, interior and technical details are truly restored through the 3D car model demonstration unit, and the AI ​​model unit interprets the various characteristics and technical parameters of the vehicle to users in a professional, accurate and easy-to-understand manner, helping users to better understand vehicle information and eliminate difficulties in understanding. At the same time, it makes car viewing and selection more efficient and convenient. Users can access the 3D car model through the Internet anytime and anywhere, and communicate with the AI ​​dialogue robot in real time to obtain the required information, which greatly saves time and energy and improves the efficiency of car viewing and selection.

[0009] Preferably, the 3D car model demonstration unit and the AI ​​model unit are respectively connected to the central control management unit.

[0010] In this technical solution, a central control management unit is used to perform overall coordination and control.

[0011] Preferably, the 3D car model demonstration unit includes a data acquisition module, a 3D modeling module, a rendering module, a car model simulation plug-in module, a static display module and a dynamic demonstration module.

[0012] In this technical solution, the realistic model effect in the 3D car model demonstration unit is used to support users to observe the vehicle's exterior design and interior layout in an all-round way.

[0013] Preferably, the data acquisition module acquires data by laser scanning or photogrammetry to accurately acquire the appearance data of the vehicle.

[0014] In this technical solution, the data acquisition module is used to accurately obtain the vehicle's appearance data through various methods such as laser scanning and photogrammetry, and then professional modeling software is used for fine processing to create a highly realistic 3D car model, achieving a one-to-one restoration of the real car model information.

[0015] Preferably, the static display module displays the exterior design, interior layout and technical details of the vehicle. The interior layout is combined with a virtual dummy and a luggage set to help users evaluate the comfort of the passenger space and the size requirements of the storage space. The technical details include the layout of the engine and the structure of the suspension system to help users who have a certain understanding of automobile technology to evaluate the performance and reliability of the vehicle.

[0016] The dynamic demonstration module demonstrates contents including simulating the vehicle's form and state, the performance of the acceleration and braking systems, and demonstrating the functions of intelligent driving.

[0017] Preferably, the AI ​​model unit includes an AI explanation robot, an AI dialogue robot, a tool module, a plug-in module, a business knowledge module and a basic model module.

[0018] In this technical solution, the AI ​​explanation robot uses the content generation model (AIGC) to accurately create expert-level vehicle model introductions covering six core areas, including appearance, interior, and power; the AI ​​dialogue robot's voice and text generation question-and-answer functions are very unique.

[0019] Preferably, the AI ​​dialogue robot includes a text dialogue robot and a voice dialogue robot;

[0020] The tool module includes an AIGC service submodule, an intelligent voice service submodule, a digital human service submodule and a robot service submodule;

[0021] The plug-in module includes a vertical plug-in service submodule, a blockage plug-in service submodule and a 3D voice digital human plug-in service submodule;

[0022] The business knowledge module includes a vertical knowledge submodule, a risk control knowledge submodule and a model basic knowledge submodule;

[0023] The basic model module includes a commercial general large model submodule and an open source self-built general large model submodule.

[0024] In this technical solution, the vertical plug-in service sub-modules include appearance plug-in, interior plug-in, intelligence and entertainment plug-in, configuration plug-in, inquiry plug-in, comparison plug-in, maintenance and care plug-in, and safety and comfort plug-in.

[0025] Preferably, the implementation of the AI ​​model in the AI ​​model unit includes the construction of a knowledge base, and the construction of the knowledge base includes the following steps:

[0026] S1. Data collection: collect vehicle model data, news information, external business data and external vehicle model information;

[0027] S2, data preprocessing, cleaning, text segmentation and annotation of the data collected in step S1;

[0028] S3, knowledge extraction and representation, extracting entity relationships and learning knowledge representation;

[0029] S4. Knowledge base storage and management, database storage, and updating of the knowledge base.

[0030] In this technical solution, the construction of the knowledge base is a key link in an AI model. The knowledge graph of the automotive industry is constructed using automobile vertical knowledge, and a knowledge update mechanism is established to regularly collect new automobile data and update the knowledge base. At the same time, the content and structure of the knowledge base are continuously optimized based on user feedback and questions.

[0031] Preferably, the explanation model of the AI ​​explanation robot creates a prompt pool for vehicle model explanations according to the energy type and highlights of the vehicle model. During model training, the information in the prompt pool is optimized according to the subjective feeling of generating the copy. After the target vehicle model is input, a suitable prompt is selected according to the characteristics of the target vehicle model. The vehicle model explanation information is explained from six aspects: appearance, interior, power, smart cockpit, smart driving, and safety performance. The output copy needs to be corrected by the "parameter performance" module. The corrected content copy is displayed in the intelligent explanation visual content editing background. The operator manually edits the copy and finally outputs it to the 3D car model demonstration unit for display.

[0032] In this technical solution, the output copy needs to be corrected by the "parameter performance" module to prevent the large model from outputting wrong information, such as horsepower and power values ​​may be confused, and the greater the power, the better is common sense.

[0033] Preferably, during the model training process, the dialogue model of the AI ​​dialogue robot utilizes large-scale computing resources and follows the scaling law to train a large model to improve the training speed and model performance;

[0034] Through model self-testing and manual evaluation, the accuracy, recall, and F1 value indicators of the large model are used to measure the quality of the model's answers, and manual labeling is used to improve the model's supervised learning capabilities.

[0035] In this technical solution, during the dialogue stage, the dialogue control module is responsible for coordinating the work of each subsystem and managing the flow and status of the dialogue; the agent distribution technology can assign different tasks to different subsystems or modules for processing; during the dialogue process, the relevant knowledge and information are searched from the knowledge base through hybrid retrieval technology.

[0036] On the basis of being in accordance with the common sense in the art, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present invention.

[0037] The positive and progressive effects of the present invention are:

[0038] The present invention utilizes 3D high-precision car models to truly restore the vehicle's appearance, interior and technical details. The AI ​​model unit interprets the various characteristics and technical parameters of the vehicle for users in a professional, accurate and easy-to-understand manner, helping users to better understand vehicle information and eliminate difficulties in understanding. At the same time, it makes car viewing and selection more efficient and convenient. Users can access the 3D car model through the Internet anytime and anywhere, and communicate with the AI ​​dialogue robot in real time to obtain the required information, which greatly saves time and energy and improves the efficiency of car viewing and selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a schematic diagram of the overall principle of a car viewing and selection platform based on 3D modeling and AI technology in an embodiment of the present invention.

[0040] Figure 2 for Figure 1 The schematic diagram of the AI ​​model unit principle of the car viewing and selection platform based on 3D modeling and AI technology is shown.

[0041] Figure 3 for Figure 1 The figure shows a schematic diagram of the construction process of the knowledge base of the car viewing and selection platform based on 3D modeling and AI technology.

[0042] Figure 4 for Figure 1 The diagram shown is a schematic diagram of the 3D-AI car model interaction process of the car viewing and selection platform based on 3D modeling and AI technology.

[0043] Figure 5 for Figure 1 The diagram shown is a schematic diagram of the AI ​​dialogue model structure of the car viewing and selection platform based on 3D modeling and AI technology.

[0044] Figure 6 for Figure 1 The figure shows the flow chart of the AI ​​explanation model of the car viewing and selection platform based on 3D modeling and AI technology.

[0045] Description of Reference Numerals

[0046] 1. 3D car model demonstration unit;

[0047] 2. AI model unit;

[0048] 3. Real-time communication unit;

[0049] 4. Central control management unit;

[0050] 5. Data acquisition module;

[0051] 6. 3D modeling module;

[0052] 7. Rendering module;

[0053] 8. Car model effect plug-in module;

[0054] 9. Static display module;

[0055] 10. Dynamic demonstration module;

[0056] 11. AI explanation robot;

[0057] 12. AI conversation robot; 121. Text conversation robot; 122. Voice conversation robot;

[0058] 13. Tool module; 131. AIGC service submodule; 132. Intelligent voice service submodule; 133. Digital human service submodule; 134. Robot service submodule;

[0059] 14. Plug-in module; 141. Vertical plug-in service submodule; 142. Blockage plug-in service submodule; 143. 3D voice digital human plug-in service submodule;

[0060] 15. Business knowledge module; 151. Vertical knowledge submodule; 152. Risk control knowledge submodule; 153. Model basic knowledge submodule;

[0061] 16. Basic model module; 161. Commercial general large model sub-module; 162. Open source self-built general large model sub-module. DETAILED DESCRIPTION

[0062] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples.

[0063] Figures 1 to 6 The structure diagram of an embodiment of the vehicle viewing and selection platform based on 3D modeling and AI technology of the present invention is shown. The vehicle viewing and selection platform based on 3D modeling and AI technology comprises a 3D vehicle model demonstration unit 1, an AI model unit 2 and a real-time communication unit 3, wherein the 3D vehicle model demonstration unit 1 and the AI ​​model unit 2 are connected via the real-time communication unit 3;

[0064] The interaction instruction of the AI ​​model unit 2 is sent to the 3D car model demonstration unit 1 through the real-time communication unit 3, and the 3D car model demonstration unit 1 executes the interaction instruction.

[0065] The AI ​​robot in the present invention can match the generated text content in real time, drive the dynamic display of the 3D car model, and realize the seamless interaction between the digital tour guide and the virtual vehicle, allowing users to "see what they hear" and enjoy an immersive car viewing experience.

[0066] 3D car model demonstration unit 1 has a realistic model effect, which allows users to observe the exterior design and interior layout of the vehicle in all directions. The interior space is combined with virtual dummies and luggage sets to help users evaluate the comfort of the passenger space and the size of the storage space. Users can view technical details including the layout of the engine and the structure of the suspension system, helping users who have a certain understanding of automotive technology to evaluate the performance and reliability of the vehicle;

[0067] In addition, the 3D car model also provides dynamic display capabilities, which can simulate the vehicle's driving status, show the performance of the acceleration and braking systems, and demonstrate the functions of intelligent driving, allowing users to have a more intuitive understanding of the vehicle's dynamic performance.

[0068] The voice and text-generated question-and-answer function of the AI ​​dialogue robot 12 is very unique. Users only need to easily speak out their questions or input them through text, and the robot can quickly understand and give accurate answers. Whether it is about the vehicle's performance parameters, configuration details, or the comparative advantages of different models, detailed answers can be obtained immediately. At the same time, AI will combine 3D car models for simultaneous display when answering, making the information easier to understand.

[0069] The AI ​​explanation robot 11 uses the content generation model (AIGC) to accurately create expert-level vehicle model introductions covering six core areas, including appearance, interior, power, etc. It can also match the generated text content in real time, drive the dynamic display of 3D car models, and achieve seamless interaction between digital tour guides and virtual vehicles, allowing users to "see what they hear" and enjoy an immersive car viewing experience.

[0070] The present invention fully restores the actual experience of users viewing and selecting cars, with "cars" (3D car models) and "sales" (AI robots), allowing users to complete the preliminary exploration of car purchase decisions in a relaxed and pleasant atmosphere, opening up a new car purchasing mode for users.

[0071] In this technical solution, the 3D car model demonstration unit 1 truly restores the vehicle's appearance, interior and technical details, and the AI ​​model unit 2 interprets the various characteristics and technical parameters of the vehicle for users in a professional, accurate and easy-to-understand manner, helping users to better understand vehicle information and eliminate difficulties in understanding. At the same time, it makes car viewing and selection more efficient and convenient. Users can access the 3D car model through the Internet anytime and anywhere, and communicate with the AI ​​dialogue robot in real time to obtain the required information, which greatly saves time and energy and improves the efficiency of car viewing and selection.

[0072] The 3D car model demonstration unit 1 and the AI ​​model unit 2 are respectively connected to the central control management unit 4.

[0073] In this technical solution, the central control management unit 4 is used to perform overall coordination and control.

[0074] The 3D car model demonstration unit 1 includes a data acquisition module 5, a 3D modeling module 6, a rendering module 7, a car model simulation plug-in module 8, a static display module 9 and a dynamic demonstration module 10.

[0075] During 3D modeling, the 3D modeling module 6 combines high-precision measuring equipment and data processing algorithms to ensure that the car model is highly consistent with the real vehicle in appearance and size. By collecting data and comparing and verifying multiple real vehicles, the key dimensional errors of the car model are controlled within a very small range, and the real shape and details of the vehicle, such as body lines, wheel style, interior layout, etc., can be accurately presented.

[0076] In terms of materials and light and shadow effects, the rendering module 7 uses the physically based rendering technology PBR to simulate real light reflection and refraction, so that the performance of the car model under different lighting conditions is consistent with the actual vehicle. This not only improves the visual authenticity, but also helps users to understand the appearance characteristics of the vehicle more accurately.

[0077] The data acquisition module 5 uses data collection and annotation technology to obtain rich vehicle structured data including vehicle configuration, comments, information, etc. For image and video content, AI image processing technology is used to slice and segment the image content and classify and store it. At the same time, there is also a closed-loop data annotation operation process to promote model supervision learning tasks.

[0078] The 3D modeling module 6 and the rendering module 7 utilize 3D modeling technology and rendering technology. In the field of computer graphics, a wealth of 3D modeling technology has been accumulated over the years. For example, polygonal modeling technology can accurately construct the appearance structure of the vehicle by dividing and editing the polygonal mesh of the object surface. Powerful rendering engines such as Unity and UnrealEngine provide support for high-quality rendering of 3D car models. These engines have advanced lighting, material and texture processing capabilities, and can simulate real light and shadow effects, so that the paint, interior materials, etc. of the car model present a high degree of realism.

[0079] In this technical solution, the realistic model effect in the 3D car model demonstration unit 1 is used to support users to observe the exterior design and interior layout of the vehicle in an all-round manner.

[0080] The data acquisition module 5 acquires data in a manner including laser scanning and photogrammetry to accurately acquire the vehicle's appearance data.

[0081] The data acquisition module 5 and the 3D modeling module 6 are used to accurately acquire the vehicle's appearance data through various methods such as laser scanning and photogrammetry, and then refined by professional modeling software to create a highly realistic 3D vehicle model, achieving a one-to-one restoration of the real vehicle model information, whether it is the complex curves or subtle textures of the vehicle, can be accurately restored;

[0082] Using the car model effect plug-in module 8, a car model effect plug-in (unity) was created based on the 3D high-precision car model to support 167 model actions;

[0083] The rendering module 7 enables the 3D car model to present high-quality visual effects on various platforms. Whether on a web page or a mobile device, users can enjoy a smooth and clear 3D display.

[0084] The static display module 9 displays the exterior design, interior layout and technical details of the vehicle. The interior layout is combined with a virtual dummy and a luggage set to help users evaluate the comfort of the passenger space and the size of the storage space. The technical details include the layout of the engine and the structure of the suspension system to help users who have a certain understanding of automobile technology to evaluate the performance and reliability of the vehicle.

[0085] The dynamic demonstration module 10 demonstrates contents including simulating the vehicle's form and state, the performance of the acceleration and braking systems, and demonstrating the functions of intelligent driving.

[0086] The interactive implementation of the 3D car model in 3D car model demonstration unit 1 needs to go through three paths: modeling, animation, model loading and interaction, and also needs to pay attention to performance issues;

[0087] Fine modeling, use professional 3D modeling software (3dsMax) to model the car in high detail. During the modeling process, the movable parts such as the doors are modeled separately, and ensure that their connection with the car body is accurate. For example, for the doors, its hinge structure, handles, windows and other details must be accurately portrayed so that its movement can be realistically simulated in the animation.

[0088] Animation production, through the keyframe animation method, set the position and rotation angle of the car door in different states. For example, define the closed state of the car door as the initial keyframe, and then set the position and rotation of the car door opened to a certain angle in another keyframe, and then automatically generate the intermediate transition animation through the interpolation algorithm; in order to make the animation more natural and smooth, you can add easing effects (easing) to make the movement of the car door have acceleration and deceleration to avoid stiff linear motion.

[0089] Model loading and interaction: Use the Unity-3D engine to load the pre-made 3D car model file (FBX) into the front-end page. During the loading process, optimize and adjust the model, such as adjusting the size, position, and rotation angle of the model to adapt to the page layout and user perspective; add event listeners in the front-end code to enable user interaction with the 3D car model; for example, when the user clicks on the door area, the corresponding event is triggered, calling the animation control interface provided by the 3D engine to play the door opening animation.

[0090] Performance optimization: In order to ensure the smooth operation of the 3D car model on the front end, performance optimization is required, which can be achieved by reducing the number of model faces, splitting the model, etc.

[0091] In addition, the central control management unit 4 manages the car model and the action data supported by the car model. The main menus are action management and action configuration. Action management is full action management, which includes information such as action category (one-time action or cyclic action), part (door or trunk, etc.), scene (inside or outside the car, etc.). Currently, 167 unity car model actions have been implemented; the main function of action configuration is to match the car model with the action so that other systems can map to the corresponding relationship between the car model and the action when calling.

[0092] The AI ​​model unit (2) includes an AI explanation robot (11), an AI dialogue robot (12), a tool module (13), a plug-in module (14), a business knowledge module (15) and a basic model module (16).

[0093] like Figure 4As shown, the AI ​​model in the AI ​​model unit (2) uses a semantic analysis algorithm in natural language processing technology to deeply understand the text content generated by the AI ​​robot. For example, a neural network model based on deep learning, such as a variant of a convolutional neural network (CNN) or a recurrent neural network (RNN), is used to classify and semantically annotate the text, so that the description of different vehicle components, functions, and operations in the text can be accurately identified;

[0094] By building a word vector model, the words in the text are converted into numerical vectors, so that the computer can better understand the semantic relationship between words and parse text instructions more accurately, such as distinguishing different operation instructions such as "open the door" and "close the door";

[0095] Develop a special command extraction module to extract specific instructions related to 3D car model operations from the semantically understood text, and filter out valid operation instructions from the text based on pre-defined command templates, such as "rotate the car body" and "turn on the flash".

[0096] The central control management unit 4 receives the interactive instructions from the AI ​​model unit 2, judges the validity of the car model's actions, and pushes the results to the 3D car model demonstration unit 1. Through the real-time communication unit 3, when the AI ​​model unit 2 generates new text content and instructions, the instructions are immediately sent to the 3D car model demonstration unit 1 to ensure that the instructions can be transmitted and executed in time.

[0097] It is worth noting that in order to synchronize the explanation content with the car model demonstration, it is necessary to accurately match and synchronize the text and car model movements on the timeline. By setting timestamps in the programs of the AI ​​robot and 3D car model, it is ensured that the car model's movements are consistent with the rhythm of the text explanation to achieve a good demonstration effect.

[0098] For the AI ​​vehicle model explanation model and AI robot question-and-answer model, a large amount of real car data and professional knowledge texts are used in the training of the models. Through the carefully designed machine learning algorithm, the model can accurately understand and answer various questions about the user's vehicle, including technical parameters, configuration information, performance characteristics, and comparison between two vehicles.

[0099] At present, the model’s self-test accuracy rate for answering vehicle dialogues has reached over 90%. It can accurately describe the configuration differences of different models and provide users with detailed explanations.

[0100] An efficient data communication and synchronization mechanism is designed between the 3D car model demonstration unit 1 and the AI ​​model unit 2 to avoid data conflicts and delays. Through rigorous testing and optimization, it can stably handle a large number of user requests and ensure performance in high-concurrency scenarios.

[0101] In this technical solution, the AI ​​explanation robot 11 uses the content generation model (AIGC) to accurately create an expert-level vehicle model introduction covering six core aspects such as appearance, interior, and power; the AI ​​dialogue robot 12 has a unique voice and text generation question and answer function.

[0102] The AI ​​conversational robot 12 uses its capabilities of intent recognition, sentiment analysis, and multi-round conversations to resolve references, complete omissions, and integrate contexts to correctly understand users' complex questions.

[0103] The knowledge graph in the automotive field, which is built with rich automotive vertical knowledge, stores and manages vehicle-related knowledge in a structured manner, so that the AI ​​model unit 2 can quickly retrieve the information required by the user based on the knowledge graph;

[0104] With the help of the natural language processing technology (NLP) and text content generation capabilities of the AIGC service model, the content is integrated and output;

[0105] By combining intelligent voice services and digital human services, the AI ​​robot’s voice commentary capabilities can ultimately be realized.

[0106] The AI ​​dialogue robot 12 includes a text dialogue robot 121 and a voice dialogue robot 122;

[0107] The tool module 13 includes an AIGC service submodule 131, an intelligent voice service submodule 132, a digital human service submodule 133 and a robot service submodule 134;

[0108] The AIGC service submodule 131 includes Vincent text, Vincent card, Vincent voice, and Vincent video;

[0109] The intelligent voice service submodule 132 includes TTS (text-to-speech, exclusive voice) and ASR (speech-to-text, exclusive error correction);

[0110] The digital human service submodule 133 includes digital human image (exclusive training) and lip shape calibration.

[0111] The plug-in module 14 includes a vertical plug-in service submodule 141, a blockage plug-in service submodule 142 and a 3D voice digital human plug-in service submodule 143;

[0112] The vertical plug-in service submodule 141 includes an appearance plug-in, an interior plug-in, an intelligence and entertainment plug-in, a configuration plug-in, an inquiry plug-in, a comparison plug-in, a maintenance plug-in, and a safety and comfort plug-in;

[0113] The blocking plug-in service submodule 142 includes a plug-in related to politics, a plug-in related to pornography, a plug-in related to explosions, a plug-in related to uncivilized language, a plug-in related to gambling, and a plug-in related to drugs;

[0114] The 3D voice digital human plug-in service submodule 143 includes a 3D car model effect plug-in, a lip shape driving plug-in based on audio features, and a face repair plug-in.

[0115] The business knowledge module 15 includes a vertical knowledge submodule 151, a risk control knowledge submodule 152 and a model basic knowledge submodule 153;

[0116] The basic model module 16 includes a commercial general large model submodule 161 and an open source self-built general large model submodule 162 .

[0117] The digital human service submodule 133 collects and shoots the image, sound, and body movements of real people, extracts voiceprints, and then generates and calibrates digital human videos based on audio feature-based lip shape drive plug-in (Wav2lip), face repair plug-in (GFPGAN), AI voice generation technology (TTS) and other capabilities.

[0118] By constructing a knowledge graph in the automotive field and storing and managing relevant vehicle knowledge in a structured manner, AI robots can quickly retrieve the information users need based on the knowledge graph and provide targeted explanations. At the same time, the knowledge graph can also be continuously updated and expanded to adapt to the development of the automotive industry and the launch of new models.

[0119] The construction of the industry-related information vocabulary is based on the vocabulary and word frequency of vehicle models, using a two-way maximum matching probability algorithm with a high recognition rate (accuracy over 99%), which can automatically identify / correct industry professional vocabulary such as vehicle model names, car model names, and configurations.

[0120] Various model evaluation indicators are used, such as accuracy, recall, and F1-score for classification tasks, and mean square error (MSE) and mean absolute error (MAE) for regression tasks. Evaluation strategies such as cross-validation are designed to comprehensively and accurately evaluate the generalization ability of the model and avoid overfitting.

[0121] The implementation of the AI ​​model in the AI ​​model unit 2 includes the construction of a knowledge base, and the construction of the knowledge base includes the following steps:

[0122] S1. Data collection: collect vehicle model data, news information, external business data and external vehicle model information;

[0123] S2, data preprocessing, cleaning, text segmentation and annotation of the data collected in step S1;

[0124] S3, knowledge extraction and representation, extracting entity relationships and learning knowledge representation;

[0125] S4. Knowledge base storage and management, database storage, and updating of the knowledge base.

[0126] In this technical solution, the construction of the knowledge base is a key link in an AI model. The knowledge graph of the automotive industry is constructed using automobile vertical knowledge, and a knowledge update mechanism is established to regularly collect new automobile data and update the knowledge base. At the same time, the content and structure of the knowledge base are continuously optimized based on user feedback and questions.

[0127] like Figure 6 As shown, the explanation model of the AI ​​explanation robot 11 creates a prompt pool for vehicle model explanation according to the energy type and highlights of the vehicle model. During model training, the information of the prompt pool is optimized according to the subjective feeling of generating the copy. After the target vehicle model is input, the appropriate prompt is selected according to the characteristics of the target vehicle model. The vehicle model explanation information is explained from six aspects: appearance, interior, power, smart cockpit, smart driving, and safety performance. The output copy needs to be corrected by the "parameter performance" module. The corrected content copy is displayed in the intelligent explanation visual content editing background. The operator manually edits the copy and finally outputs it to the 3D car model demonstration unit 1 for display.

[0128] It is worth noting that the standards for manual editing of copywriting by operators are based on the purpose of text optimization and business cooperation (key information promoted by manufacturers and distributors).

[0129] In this technical solution, the output copy needs to be corrected by the "parameter performance" module to prevent the large model from outputting wrong information, such as horsepower and power values ​​may be confused, and the greater the power, the better is common sense.

[0130] like Figure 5 As shown, during the model training process, the dialogue model of the AI ​​dialogue robot 12 utilizes large-scale computing resources and follows the scaling law to train a large model to improve the training speed and model performance;

[0131] Through model self-testing and manual evaluation, the accuracy, recall, and F1 value indicators of the large model are used to measure the quality of the model's answers, and manual labeling is used to improve the model's supervised learning capabilities.

[0132] In this technical solution, during the dialogue stage, the dialogue control module is responsible for coordinating the work of each subsystem and managing the flow and status of the dialogue; the agent distribution technology can assign different tasks to different subsystems or modules for processing; during the dialogue process, the relevant knowledge and information are searched from the knowledge base through hybrid retrieval technology.

[0133] Build a prompt pool containing various prompts and guiding sentences to stimulate the creativity and diversity of the model, such as different question openings, scene descriptions, etc. The reading comprehension model selects the appropriate answer strategy, calls the corresponding model and data resources according to the user's questions and intentions, and finally uses the content creation model to generate copywriting content with different personalities and styles.

[0134] The present invention has formulated a comprehensive testing plan, including multiple stages such as unit testing, integration testing, system testing and user acceptance testing. The 3D car model has been subjected to multi-angle visual inspection, interactive function testing and performance evaluation to ensure the stability and reliability of the car model under various operations. A large number of case tests, boundary value tests and stress tests have also been carried out on the AI ​​model to verify the accuracy and response time of the model under different input conditions.

[0135] The application delivered by the present invention can be hot-deployed, and most common maintenance operations such as software package version upgrades and simple middleware maintenance do not require downtime. It has a detailed hot deployment plan and an error rollback plan, which forms an operation document or script and is delivered to the operation and maintenance team along with the installation and deployment manual for preservation and updating.

[0136] According to the technical development standard specifications, all content generation applications of the present invention need to be connected to the company's risk control model to ensure that the generated content complies with national laws and regulations and ethical behavioral norms; semantic constraints and rules are introduced into the AI ​​model to avoid selecting sensitive words during the generation process; the generated text fragments are monitored in real time during the content generation process, and once a trend of sensitive words may appear, the generation strategy is adjusted immediately; the content generated by the AI ​​model and the content input by the user must be monitored for content compliance through the risk control model. If sensitive words are found, the entire piece of information is filtered.

[0137] The present invention attaches great importance to the security and privacy protection of user data. Encryption technology is used to transmit and store user data to ensure that user information is not leaked. At the same time, we strictly abide by relevant laws and regulations and data protection standards, establish a complete security management system and authority control mechanism, and limit access to user data and operation permissions.

[0138] For automobile manufacturers and dealers, the present invention provides them with a brand new marketing tool; through this invention, they can more vividly and intuitively display the features and advantages of vehicles and attract the attention of potential customers;

[0139] Compared with traditional print advertisements and promotional materials, the interesting interaction of 3D smart car viewing can leave a deeper impression on users and improve the brand's popularity and reputation. At the same time, the present invention can also be combined with social media, online car forums and other platforms for extensive promotion and dissemination to expand the brand's influence.

[0140] In addition, the project can be applied to building online VR / AR showrooms and digital human sales. Through the display of AI+3D car models and digital human Q&A / explanation, sales efficiency can be improved, thereby reducing marketing costs.

[0141] In general, the invention is a new car viewing and selection tool based on "3D car model + AI technology", which upgrades the way users view and select cars online and improves the user experience;

[0142] In the digital age, the automotive industry is also constantly exploring the path of digital transformation. This invention integrates 3D technology and AI technology and applies them to the field of automobile sales. It is an important practice of the digital transformation of the automotive industry. It will promote automobile companies to strengthen the application of digital technology in sales and service links, improve the overall digital level of the industry, enhance the industry's competitiveness and innovation capabilities, and inject new impetus into the sustainable development of the automotive industry.

[0143] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that this is only for illustration and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A car viewing and selection platform based on 3D modeling and AI technology, featuring: It comprises a 3D car model demonstration unit (1), an AI model unit (2) and a real-time communication unit (3), wherein the 3D car model demonstration unit (1) and the AI ​​model unit (2) are connected via the real-time communication unit (3); The interaction instruction of the AI ​​model unit (2) is sent to the 3D car model demonstration unit (1) via the real-time communication unit (3), and the 3D car model demonstration unit (1) executes the interaction instruction.

2. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 1, characterized in that: The 3D car model demonstration unit (1) and the AI ​​model unit (2) are respectively connected to the central control management unit (4).

3. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 1, characterized in that: The 3D car model demonstration unit (1) comprises a data acquisition module (5), a 3D modeling module (6), a rendering module (7), a car model simulation plug-in module (8), a static display module (9) and a dynamic demonstration module (10).

4. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 3, characterized in that: The data acquisition module (5) acquires data by laser scanning or photogrammetry to accurately acquire the appearance data of the vehicle.

5. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 3, characterized in that: The static display module (9) displays contents including the exterior design, interior layout and technical details of the vehicle. The interior layout is combined with a virtual dummy and a luggage compartment set to help users evaluate the comfort of the passenger space and the size requirements of the storage space. The technical details include the layout of the engine and the structure of the suspension system to help users who have a certain understanding of automobile technology to evaluate the performance and reliability of the vehicle. The dynamic demonstration module (10) demonstrates contents including simulating the vehicle's form and state, the performance of the acceleration and braking systems, and demonstrating the function of intelligent driving.

6. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 1, characterized in that: The AI ​​model unit (2) includes an AI explanation robot (11), an AI dialogue robot (12), a tool module (13), a plug-in module (14), a business knowledge module (15) and a basic model module (16).

7. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 6, characterized in that: The AI ​​dialogue robot (12) includes a text dialogue robot (121) and a voice dialogue robot (122); The tool module (13) includes an AIGC service submodule (131), an intelligent voice service submodule (132), a digital human service submodule (133) and a robot service submodule (134); The plug-in module (14) comprises a vertical plug-in service submodule (141), a sealing control plug-in service submodule (142) and a 3D voice digital human plug-in service submodule (143); The business knowledge module (15) includes a vertical knowledge submodule (151), a risk control knowledge submodule (152) and a model basic knowledge submodule (153); The basic model module (16) includes a commercial general large model submodule (161) and an open source self-built general large model submodule (162).

8. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 7, characterized in that: The implementation of the AI ​​model in the AI ​​model unit (2) includes the construction of a knowledge base, and the construction of the knowledge base includes the following steps: S1. Data collection: collect vehicle model data, news information, external business data and external vehicle model information; S2, data preprocessing, cleaning, text segmentation and annotation of the data collected in step S1; S3, knowledge extraction and representation, extracting entity relationships and learning knowledge representation; S4. Knowledge base storage and management, database storage, and updating of the knowledge base.

9. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 1, characterized in that: The explanation model of the AI ​​explanation robot (11) creates a prompt pool for vehicle model explanation according to the energy type and highlights of the vehicle model. During model training, the information of the prompt pool is optimized according to the subjective feeling of the generated copy. After the target vehicle model is input, the appropriate prompt is selected according to the characteristics of the target vehicle model. The vehicle model explanation information is explained from six aspects: appearance, interior, power, smart cockpit, smart driving, and safety performance. The output copy needs to be corrected by the "parameter performance" module. The corrected content copy is displayed in the intelligent explanation visual content editing background. The operator manually edits the copy and finally outputs it to the 3D car model demonstration unit (1) for display.

10. The car viewing and selection platform based on 3D modeling and AI technology as claimed in claim 1, characterized in that: During the model training process, the dialogue model of the AI ​​dialogue robot (12) utilizes large-scale computing resources and follows the scaling law to train a large model, thereby improving the training speed and model performance; Through model self-testing and manual evaluation, the accuracy, recall, and F1 value indicators of the large model are used to measure the quality of the model's answers, and manual labeling is used to improve the model's supervised learning capabilities.

Images