Intelligent loss checking and assessment multi-mode system based on vehicle accident scene online evaluation

Abstract

The invention discloses an intelligent loss check and assessment multi-modal system based on vehicle accident scene online evaluation, and the system comprises a user interaction interface module which is used for receiving client input and correcting input deviation through a feedback cycle regulation and control mechanism, and a multi-modal data fusion module which is used for collecting vehicle accident damage data uploaded by a client, and outputting the vehicle accident damage data to a server. The system comprises an accident data acquisition module used for acquiring accident damage data, a three-dimensional reconstruction model module used for constructing a vehicle accident three-dimensional model based on the acquired accident damage data and accurately evaluating the vehicle damage condition, an intelligent loss assessment module used for generating a loss assessment report according to an assessment result and implementing intelligent loss assessment operation, and a data storage module used for storing the accident data, the assessment result and the loss assessment report; according to the on-line evaluation intelligent loss checking and assessment multi-mode system based on the vehicle accident scene, feedback circulation of a user interaction interface is regulated and controlled, so that the problem of inaccurate loss assessment caused by client input deviation is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios. Background Art

[0002] In the field of vehicle accident damage assessment, with the rapid development of artificial intelligence technology, online assessment and intelligent damage assessment methods based on vehicle accident scenarios have gradually been applied. These methods usually use multimodal data, such as images, text, and sensor data, to quickly assess and calculate accident losses through algorithms. The relevant system usually consists of a user interaction interface, a data acquisition module, a model evaluation module, and a feedback loop mechanism.

[0003] The workflow of the existing technology is usually as follows: the user uploads accident-related information through an interactive interface, including pictures of the accident scene, descriptive text and other necessary data. The system generates damage assessment results by analyzing these data and feeds the results back to the user. However, the current technology has significant defects: when the accident information entered by the user is inaccurate or incomplete, such as uploading low-quality pictures, missing descriptive information or misleading input, the damage assessment results generated by the system will often be affected. In this case, the existing system lacks an effective feedback loop control mechanism and cannot automatically adjust or correct user input deviations, resulting in inaccurate damage assessment results. In addition, although some systems have attempted to introduce manual correction and input guidance functions, these methods usually rely on a large amount of manual intervention, which can neither meet real-time requirements nor easily increase the complexity of user operations. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent multi-modal damage assessment system based on online assessment of vehicle accident scenarios, and to regulate the feedback loop of the user interaction interface to solve the problem of inaccurate damage assessment caused by customer input deviation.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios, the system comprising: A user interface module for receiving customer input and correcting input deviations through a feedback loop control mechanism; A multimodal data fusion module connected to the user interface module is used to collect vehicle accident damage data uploaded by customers; The 3D reconstruction model module connected to the multimodal data fusion module is used to construct a 3D model of the vehicle accident based on the collected accident damage data, accurately assess the damage to the vehicle, including simulating the deformation trajectory of the damaged area of ​​the vehicle. The specific formula is: =D; Where a represents the deformation of the damaged area of ​​the vehicle in the accident, t represents the time, b represents the structural parameters of the vehicle material, and D represents the dynamic disturbance force caused by the external force. Indicates the change of shape over time; The intelligent damage assessment module connected to the 3D reconstruction model module is used to generate a damage assessment report based on the assessment results and implement intelligent damage assessment operations, including extracting the volume parameters of the damaged area from the 3D model and calculating the total repair cost. The specific formula is: A = C + V × B; Among them, A represents the total repair cost after the vehicle is damaged, C represents the basic repair cost incurred in all repair plans, V represents the volume parameter of the damaged area of ​​the vehicle extracted from the 3D reconstruction model, and B represents the repair cost per unit volume; The data storage module connected with the user interaction interface module, the multimodal data fusion module, the three-dimensional reconstruction model module and the intelligent damage assessment module is used to store accident data, assessment results and damage assessment reports.

[0006] Preferably, the user interaction interface module receives customer input and corrects the input deviation through a feedback loop control mechanism, including: Record user operations through the user interaction interface and dynamically adjust the interface design. The specific formula is: ; Among them, E represents the utility value of the user interaction interface, F represents the feedback value obtained by the user, and G represents the effort cost paid by the user during the data input process.

[0007] Preferably, the multimodal data fusion module collects the vehicle accident damage data uploaded by the customer including: Extract features from multimodal data such as images, sensors and texts, compare the overall feature gain generated after multimodal data fusion with the single modality features, and calculate the data fusion yield. The specific formula is: H = (L-I) / I; Among them, H represents the yield of multimodal data fusion, L represents the overall feature gain of the fused data, and I represents the feature contribution of single modality data.

[0008] Preferably, the data storage module stores accident data, assessment results and damage assessment reports including: The accident data is stored in blocks according to the characteristics, the total data size and the number of blocks are recorded, and the storage density of the unit data is calculated. The specific formula is: N = n / p; Wherein, N represents the storage density of unit data, n represents the total data size of accident data, and p represents the number of data blocks.

[0009] Preferably, the calculation formula of the volume parameter V of the damaged area of ​​the vehicle extracted from the three-dimensional reconstruction model in the intelligent damage evaluation module is: V=S×h; Among them, V represents the volume parameter of the damaged area of ​​the vehicle extracted from the 3D reconstruction model, S represents the area of ​​the damaged area, and h represents the damaged depth.

[0010] Preferably, the multimodal data fusion module collects the vehicle accident damage data uploaded by the customer, including the vehicle accident picture data uploaded by the customer, the collision sensor data, and the accident text information described by the customer.

[0011] Preferably, the intelligent damage assessment module generates a damage assessment report based on the evaluation results, and implements the intelligent damage assessment operation. The repair cost B per unit volume includes the cost of repair materials, the complexity of the repair process, and the time and labor input of the repair operation.

[0012] Preferably, the user interaction interface module receives the customer input and corrects the feedback value F obtained by the user in the input deviation through the feedback loop control mechanism. The calculation formula is: F = k × W; Among them, F represents the feedback value obtained by the user, k represents the quality factor of the user input data, and W represents the responsiveness of the system to the feedback.

[0013] Preferably, the user interaction interface module receives the customer input and corrects the input deviation through the feedback loop control mechanism. The calculation formula of the effort cost G paid by the user in the data input process is: G = t × c; Among them, G represents the effort cost paid by the user in the process of data input, t represents time, and c represents the complexity coefficient of the user's operation.

[0014] Preferably, the calculation formula of the characteristic contribution I of the single modality data in the vehicle accident damage data uploaded by the customer collected by the multimodal data fusion module is: I=xi×yi; Among them, I represents the feature contribution of single modal data, x i Represents the eigenvalue of a single modal data, y i represents the weight of a single modal data, and i represents the type number of the modal data.

[0015] It can be seen from the above technical solution that the present invention has the following beneficial effects: This multimodal system for intelligent damage assessment based on online assessment of vehicle accident scenarios receives customer input through the user interaction interface module and corrects input deviations through the feedback loop control mechanism. The multimodal data fusion module collects vehicle accident damage data uploaded by customers. The 3D reconstruction model module constructs a 3D model of the vehicle accident based on the collected accident damage data to accurately assess the damage to the vehicle. The intelligent damage assessment module generates a damage assessment report based on the assessment results and implements intelligent damage assessment operations. It can automatically detect and adjust the accident information input by the user, correct user input deviations in real time, avoid inaccurate damage assessment results due to data quality problems, and do not require a lot of manual intervention, which significantly reduces the complexity of operation and improves the system. The ease of use and intelligence level can prompt users to optimize input in real time, greatly reducing the repeated operations caused by inaccurate input, improving user satisfaction, and can quickly complete the collection, processing, evaluation and feedback of accident data to meet the real-time requirements for damage assessment results in motor insurance claims. It improves the comprehensive utilization efficiency of different types of data, gives full play to the characteristics and advantages of various types of data, and generates more comprehensive and accurate damage assessment results. The user's input burden is reduced through intelligent design, and the user can obtain accurate damage assessment results without complex processing of the input data, which greatly reduces the usage threshold, and regulates the feedback loop of the user interaction interface to solve the problem of inaccurate damage assessment caused by customer input deviations. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a schematic diagram of the connection of the system modules of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] like Figure 1 As shown, the present invention provides a technical solution: a multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios, the system comprising: A user interface module for receiving customer input and correcting input deviations through a feedback loop control mechanism; A multimodal data fusion module connected to the user interface module is used to collect vehicle accident damage data uploaded by customers; The 3D reconstruction model module connected to the multimodal data fusion module is used to construct a 3D model of the vehicle accident based on the collected accident damage data, accurately assess the damage to the vehicle, including simulating the deformation trajectory of the damaged area of ​​the vehicle. The specific formula is: =D; Where a represents the deformation of the damaged area of ​​the vehicle in the accident, t represents the time, b represents the structural parameters of the vehicle material, and D represents the dynamic disturbance force caused by the external force. Indicates the change of shape over time; The intelligent damage assessment module connected to the 3D reconstruction model module is used to generate a damage assessment report based on the assessment results and implement intelligent damage assessment operations, including extracting the volume parameters of the damaged area from the 3D model and calculating the total repair cost. The specific formula is: A = C + V × B; Among them, A represents the total repair cost after the vehicle is damaged, C represents the basic repair cost incurred in all repair plans, V represents the volume parameter of the damaged area of ​​the vehicle extracted from the 3D reconstruction model, and B represents the repair cost per unit volume; The data storage module connected with the user interaction interface module, the multimodal data fusion module, the three-dimensional reconstruction model module and the intelligent damage assessment module is used to store accident data, assessment results and damage assessment reports.

[0019] The above system obtains the vehicle accident information input by the customer through the user interaction interface module, and uses the multimodal data fusion module to collect accident-related data. In the three-dimensional reconstruction model module, through mathematical modeling, based on the deformation trajectory and material properties of the damaged area of ​​the accident, a three-dimensional model of the vehicle is constructed, the influence of dynamic disturbance force on the damaged area is analyzed, and an accurate damaged deformation trajectory is formed. Subsequently, the intelligent damage assessment module uses the volume parameters and unit repair cost extracted from the three-dimensional model, combined with the basic repair cost, to generate a damage assessment report and complete the intelligent damage assessment operation. All data are finally archived by the data storage module for subsequent analysis and recording. By integrating multimodal data and three-dimensional reconstruction technology, this system improves the accuracy of vehicle accident damage assessment, significantly reduces the error of human operation, and greatly shortens the time cost of assessment and damage assessment. The introduction of dynamic deformation simulation formula enhances the adaptability of the model to the complex deformation of the damaged area of ​​the accident. In addition, the intelligent damage assessment module improves the transparency and scientificity of the damage assessment plan by accurately calculating the repair cost, further improving user satisfaction.

[0020] The user interface module receives customer input and corrects input deviations through a feedback loop control mechanism including: Record user operations through the user interaction interface and dynamically adjust the interface design. The specific formula is: ; Among them, E represents the utility value of the user interaction interface, F represents the feedback value obtained by the user, and G represents the effort cost paid by the user during the data input process.

[0021] This implementation uses the user interaction interface module to dynamically record user operations, capture deviations in user input data in real time, and make feedback adjustments based on user operating habits and input accuracy to optimize user experience. , the system evaluates the user's interactive utility: F reflects the quality of feedback information obtained after the user inputs, such as the accuracy of system prompts or automatic corrections, G quantifies the operating costs paid by users in the data input process, including time, operation complexity, etc., and E serves as the basis for system design optimization to guide the dynamic adjustment of interface design, such as reducing unnecessary steps, simplifying complex operations, or improving visual prompts, thereby improving interaction efficiency. Through the feedback loop control mechanism, the system can accurately correct the deviation of user input, ensure the integrity and accuracy of data, and enhance the user's interactive experience. By dynamically adjusting the interface design and optimizing the user operation process, the time and operation complexity of users completing input tasks are reduced. The system corrects the user input deviation through the feedback loop mechanism to ensure that the quality of the collected data meets the requirements of subsequent analysis. The application of the utility evaluation formula makes the interface design more intelligent and personalized, the user operation is more intuitive, and the feedback is more timely. By reducing invalid operations and optimizing the interaction path, the user's effort cost is significantly reduced and satisfaction is improved.

[0022] The multimodal data fusion module collects vehicle accident damage data uploaded by customers, including: Extract features from multimodal data such as images, sensors and texts, compare the overall feature gain generated after multimodal data fusion with the single modality features, and calculate the data fusion yield. The specific formula is: H = (L-I) / I; Among them, H represents the yield of multimodal data fusion, L represents the overall feature gain of the fused data, and I represents the feature contribution of single modality data.

[0023] This implementation method uses a multimodal data fusion module to extract and fuse features from multiple source data such as images, sensor data, and text descriptions uploaded by customers: Feature extraction: extract shape, color, and texture features from images, obtain damage locations and mechanical parameters from sensor data, and extract accident descriptions and keywords from text. Feature fusion: Use data fusion algorithms, such as weighted average method or deep learning model, to integrate the features of each modality to generate a unified overall feature representation. Benefit calculation: Use the formula H = (L-I) / I to compare the overall feature gain of the fused data with the contribution of a single modality feature to evaluate the actual benefit of multimodal fusion: L represents the comprehensive feature expression effect generated after fusion, and I represents a single modality, such as the feature expression effect of using only image or sensor data. This data fusion method can effectively improve the model's ability to recognize complex damage features. By fusing multiple modal data, a more comprehensive feature description is generated, and the ability to capture the complexity and details of vehicle damage is improved. Multimodal fusion eliminates the limitations of single modal data, effectively reduces misjudgments due to modal missing or noise, and improves the accuracy of accident assessment. Multimodal fusion achieves synergistic gain of data, fully utilizes the characteristics of image, sensor and text data, improves the overall efficiency of the system, and can dynamically evaluate the effectiveness of different modal combinations, providing a quantitative basis for optimizing system design.

[0024] The data storage module stores accident data, assessment results and damage assessment reports including: The accident data is stored in blocks according to the characteristics, the total data size and the number of blocks are recorded, and the storage density of the unit data is calculated. The specific formula is: N = n / p; Wherein, N represents the storage density of unit data, n represents the total data size of accident data, and p represents the number of data blocks.

[0025] This implementation method organizes and manages accident data through a data storage module. The main process is as follows: According to the characteristics of the accident data, such as vehicle type, damaged area, accident time, etc., the data is divided into several blocks. Each block of data has an independent feature label, which is convenient for fast indexing and calling. The data size and related feature descriptions are recorded for each block. At the same time, the total size n and the number of blocks p of the statistical data are calculated. The storage density of the unit data is calculated by the formula N=n / p to evaluate the efficiency and capacity utilization of the storage system. Storage density is an important indicator for measuring the uniformity of data storage distribution and system load balance. All stored data can be used for subsequent intelligent damage assessment operations, damage assessment report generation and historical data analysis to ensure the comprehensiveness of the system's functions and the efficiency of data management. Through block storage and density calculation, the organization of data storage is optimized, which is convenient for fast indexing and efficient query. Reasonable block division reduces the redundancy and duplication of data storage, improves the utilization of storage resources, and supports more efficient system operation. Through storage density calculation, the system can identify the uniformity of data distribution, providing a basis for subsequent data optimization and expansion. The modular block storage design facilitates system expansion and maintenance, and adapts to data storage needs of different scales.

[0026] The calculation formula of the volume parameter V of the damaged area of ​​the vehicle extracted from the three-dimensional reconstruction model in the intelligent damage evaluation module is: V = S × h; Among them, V represents the volume parameter of the damaged area of ​​the vehicle extracted from the 3D reconstruction model, S represents the area of ​​the damaged area, and h represents the damaged depth.

[0027] In the intelligent damage assessment module, the volume of the damaged area of ​​the vehicle is calculated using a three-dimensional reconstruction model. The specific process is as follows: Damaged area area extraction: From the three-dimensional model, the boundary of the damaged area is determined by edge detection, shape fitting and other algorithms, and its projected area S is calculated. Damaged depth measurement: The depth value h of the damaged area is obtained by the spatial coordinate difference of the three-dimensional model or other depth sensing technology. Volume calculation: Based on the formula V=S×h, the extracted area and depth are multiplied to obtain the volume parameter V of the damaged area. Result output: The volume parameter V is used for subsequent repair cost calculation and damage assessment report generation to ensure the accuracy and reliability of the assessment results. By combining three-dimensional reconstruction technology and formula calculation, the volume parameters of the damaged area can be extracted quickly and accurately, which is suitable for damaged areas with complex shapes. The volume calculation of the damaged area is decomposed into a simple product of area and depth, which reduces the complexity and facilitates algorithm implementation and calculation optimization. The extraction of volume parameters provides a reliable basis for repair cost assessment, and supports the assessment needs of various vehicle types and damaged forms. Automated calculation replaces the traditional manual measurement method, greatly shortens the time of damage assessment, and improves the overall efficiency of the system.

[0028] The multimodal data fusion module collects the vehicle accident damage data uploaded by the customer, including the vehicle accident image data uploaded by the customer, the collision sensor data, and the accident text information described by the customer. The multimodal data fusion module builds a comprehensive accident information input process by integrating the accident pictures, collision sensor data and text descriptions. During the processing, the image data uses computer vision technology to extract the shape, texture and color of the damage, the sensor data uses physical models to analyze the collision intensity, direction and impact area, and the text information uses natural language processing algorithms to analyze the background information such as the time, location and cause of the accident. After fusing these multi-source data, the system uses a unified feature representation model to integrate each modal data into a complete damage assessment input, providing high-quality initial data for three-dimensional reconstruction and nuclear damage calculation. Through this design, the system effectively avoids the assessment bias caused by acquisition errors or insufficient information of a single data modality. The introduction of the multimodal data fusion module greatly improves the comprehensiveness and accuracy of accident assessment. Through the collaborative work of image data, sensor data and text descriptions, the system can cover multi-dimensional data from visual information to physical characteristics to event background, forming a complete understanding of the damage caused by the accident. This data fusion method not only improves the accuracy of the assessment, but also optimizes the efficiency of data utilization, making the accident damage assessment process more scientific and reliable. In addition, the modular design enables the system to flexibly respond to different accident scenarios, with higher applicability and expansion capabilities. By deeply integrating the diverse data provided by users, the user experience is significantly enhanced and the damage assessment errors caused by insufficient information are reduced.

[0029] The intelligent damage assessment module generates a damage assessment report based on the assessment results and implements the intelligent damage assessment operation. The repair cost per unit volume B includes the cost of repair materials, the complexity of the repair process, and the time and labor input of the repair operation. The intelligent damage assessment module dynamically calculates the repair cost per unit volume B based on the volume parameters of the damaged area of ​​the vehicle and the repair cost model. Specifically, the cost of repair materials is determined by the type and quantity of required materials, such as the cost of sheet metal, coating and sealant; the complexity of the repair process is quantified by the difficulty of the repair steps, technical requirements and equipment usage; the time and labor input of the repair operation are estimated based on the actual working hours, the skill level of the operator and the local labor cost standards. By comprehensively analyzing these three factors, the system generates the repair cost per unit volume and calculates the total repair cost based on this to provide data support for the damage assessment report. Finally, the system integrates the repair cost, detailed information of the damaged area and the damage assessment process records into a damage assessment report for reference by users and maintenance organizations. Through the detailed segmentation and quantitative analysis of repair materials, process complexity and labor input, the intelligent damage assessment module significantly improves the scientificity and accuracy of the damage assessment results. The detailed calculation of repair costs not only takes into account the diversity of material selection, but also covers the technical requirements of complex processes and the specific input of manual operations, ensuring that the evaluation results are comprehensive and accurate. In addition, the damage assessment report generated by the system has a clear structure and detailed data, providing users with a transparent reference to repair costs, enhancing the practical value of the report. At the same time, the dynamic adjustment of the repair cost model can adapt to different vehicle types, repair processes and regional standards, further improving the scope of application and flexibility of the system.

[0030] The user interaction interface module receives customer input and corrects the input deviation through the feedback loop control mechanism. The calculation formula of the feedback value F obtained by the user is: F = k × W; Among them, F represents the feedback value obtained by the user, k represents the quality factor of the user input data, and W represents the responsiveness of the system to the feedback.

[0031] The user interaction interface module realizes continuous optimization of user input quality by dynamically recording and analyzing user input data and combining it with the feedback loop control mechanism. The formula F=k×W is used to quantify the feedback value: k reflects the accuracy, completeness and consistency of user input data. The higher the quality of user input, the higher the feedback value; W represents the system's responsiveness to input data, including interface adjustment speed, accuracy of prompt content and improvement effect. Through this mechanism, the system dynamically adjusts the design and feedback strategy of the interactive interface, such as providing real-time correction prompts for low-quality data and shortening feedback time for high-quality data, thereby enhancing the accuracy of user input and interaction efficiency. By introducing the feedback value calculation formula, the system can realize dynamic evaluation of user input quality and feedback effect, and then optimize the adjustment strategy of the interactive interface. High-quality user input data obtains more timely and accurate feedback, which significantly improves the user experience; low-quality data is corrected in real time through system response, ensuring that the final data quality meets the evaluation requirements. This module effectively reduces the subsequent processing errors caused by data input deviations, while optimizing the transparency of the feedback mechanism, enabling users to clearly perceive the response logic of the system, and improving user satisfaction and system efficiency. In addition, quantitative analysis based on feedback value can also provide data support for system development and optimization, and enhance the adaptability and scalability of the system.

[0032] The user interface module receives customer input and corrects the input deviation through the feedback loop control mechanism. The calculation formula of the effort cost G paid by the user in the data input process is: G = t × c; Among them, G represents the effort cost paid by the user in the process of data input, t represents time, and c represents the complexity coefficient of the user's operation.

[0033] The user interaction interface module calculates the user's effort cost G by recording the user's interaction time t during the data input process and analyzing the complexity c of the operation steps. The interaction time is automatically counted by the system from the start of the user's operation to the completion of data input; the operation complexity coefficient is dynamically evaluated by the system based on the number of input steps, interface operation difficulty, and data format requirements. Through the formula G=t×c, the system quantifies the effort cost and provides real-time feedback to the user. For example, when the user's operation time is extended due to an unfriendly interface design, the system will reduce the complexity coefficient c by adjusting the layout of the interactive interface and optimizing the data verification steps, thereby reducing the user's effort cost and improving input efficiency. This mechanism continuously optimizes the user experience through closed-loop feedback and ensures the efficiency and accuracy of data input. By quantifying the user's effort cost during the data input process, the system can identify and optimize the key factors that affect user efficiency, significantly improving the design rationality and operation fluency of the interactive interface. The real-time feedback of the effort cost enables users to perceive the potential efficiency bottlenecks in the input process, while the dynamic regulation of the complexity coefficient effectively reduces unnecessary operation steps and interface interference, reducing the user's input burden. The optimized interface design not only improves the speed and accuracy of data input, but also enhances user satisfaction and participation. In addition, the system's monitoring and analysis of effort costs can provide data support for subsequent version development and further promote the optimization of interfaces and functions.

[0034] The calculation formula of the characteristic contribution I of the single modality data in the vehicle accident damage data uploaded by the multimodal data fusion module is: I = xi × yi; Among them, I represents the feature contribution of a single modal data, xi represents the feature value of a single modal data, yi represents the weight of a single modal data, and i represents the type number of the modal data.

[0035] The multimodal data fusion module calculates the feature contribution I of a single modal data to the fusion result by extracting eigenvalues ​​xi and assigning weights yi according to the characteristics of different modal data. The eigenvalue xi reflects the specific information content of the modal data, such as texture complexity in the image modality and collision intensity in the sensor data; the weight yi is dynamically adjusted according to the importance and credibility of the modal data, for example, when the image is blurred or the sensor data is missing, the weight of the corresponding modality is reduced. Through the formula I=xi×yi, the system quantifies the contribution value of each modality, provides a weight reference for multimodal data fusion, and ensures the balance and accuracy of the final fusion result. The contribution value I of each modal data is finally summarized to generate a comprehensive feature representation. This module significantly improves the accuracy and adaptability of multimodal data fusion by dynamically calculating the feature contribution of a single modal data. The combination of eigenvalues ​​and weights enables the system to flexibly respond to changes in the quality of different modal data, ensuring that important modalities dominate the fusion results, while the influence of secondary or low-quality modalities is effectively weakened. This design improves the system's ability to analyze accident-damaged data, while reducing the interference of missing or abnormal single modal data on the overall evaluation results. In addition, this calculation method also provides a quantitative basis for ranking the importance of modal data, optimizes the data fusion process, and enables the system to exhibit higher stability and reliability in complex accident scenarios.

[0036] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent multi-modal system for damage assessment based on online assessment of vehicle accident scenarios, characterized by: The system comprises: A user interface module for receiving customer input and correcting input deviations through a feedback loop control mechanism; A multimodal data fusion module connected to the user interface module is used to collect vehicle accident damage data uploaded by customers; The 3D reconstruction model module connected to the multimodal data fusion module is used to construct a 3D model of the vehicle accident based on the collected accident damage data, accurately assess the damage to the vehicle, including simulating the deformation trajectory of the damaged area of ​​the vehicle. The specific formula is: =D; Where a represents the deformation of the damaged area of ​​the vehicle in the accident, t represents the time, b represents the structural parameters of the vehicle material, and D represents the dynamic disturbance force caused by the external force. Indicates the change of shape over time; The intelligent damage assessment module connected to the 3D reconstruction model module is used to generate a damage assessment report based on the assessment results and implement intelligent damage assessment operations, including extracting the volume parameters of the damaged area from the 3D model and calculating the total repair cost. The specific formula is: A = C + V × B; Among them, A represents the total repair cost after the vehicle is damaged, C represents the basic repair cost incurred in all repair plans, V represents the volume parameter of the damaged area of ​​the vehicle extracted from the 3D reconstruction model, and B represents the repair cost per unit volume; The data storage module connected with the user interaction interface module, the multimodal data fusion module, the three-dimensional reconstruction model module and the intelligent damage assessment module is used to store accident data, assessment results and damage assessment reports.

2. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 1 is characterized by: The user interaction interface module receives the customer input and corrects the input deviation through the feedback loop control mechanism, including: Record user operations through the user interaction interface and dynamically adjust the interface design. The specific formula is: ； Among them, E represents the utility value of the user interaction interface, F represents the feedback value obtained by the user, and G represents the effort cost paid by the user during the data input process.

3. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 1 is characterized by: The multimodal data fusion module collects the vehicle accident damage data uploaded by the customer, including: Extract features from multimodal data such as images, sensors and texts, compare the overall feature gain generated after multimodal data fusion with the single modality features, and calculate the data fusion yield. The specific formula is: H = (L-I) / I; Among them, H represents the yield of multimodal data fusion, L represents the overall feature gain of the fused data, and I represents the feature contribution of single modality data.

4. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 1 is characterized by: The data storage module stores accident data, assessment results and damage assessment reports including: The accident data is stored in blocks according to the characteristics, the total data size and the number of blocks are recorded, and the storage density of the unit data is calculated. The specific formula is: N = n / p; Wherein, N represents the storage density of unit data, n represents the total data size of accident data, and p represents the number of data blocks.

5. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 1 is characterized by: The calculation formula of the volume parameter V of the damaged area of ​​the vehicle extracted from the three-dimensional reconstruction model in the intelligent damage evaluation module is: V = S × h; Among them, V represents the volume parameter of the damaged area of ​​the vehicle extracted from the 3D reconstruction model, S represents the area of ​​the damaged area, and h represents the damaged depth.

6. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 1 is characterized by: The multimodal data fusion module collects vehicle accident damage data uploaded by the customer, including vehicle accident picture data uploaded by the customer, collision sensor data, and accident text information described by the customer.

7. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 1 is characterized by: The intelligent damage assessment module generates a damage assessment report based on the assessment results and implements the intelligent damage assessment operation. The repair cost B per unit volume includes the cost of repair materials, the complexity of the repair process, and the time and labor input of the repair operation.

8. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 2 is characterized by: The user interaction interface module receives the customer input and corrects the feedback value F obtained by the user in the input deviation through the feedback loop control mechanism. The calculation formula is: F = k × W; Among them, F represents the feedback value obtained by the user, k represents the quality factor of the user input data, and W represents the responsiveness of the system to the feedback.

9. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 2 is characterized by: The user interaction interface module receives the customer input and corrects the input deviation through the feedback loop control mechanism. The calculation formula of the effort cost G paid by the user in the data input process is: G = t × c; Among them, G represents the effort cost paid by the user in the process of data input, t represents time, and c represents the complexity coefficient of the user's operation.

10. The multi-modal system for intelligent damage assessment based on online assessment of vehicle accident scenarios according to claim 3 is characterized by: The calculation formula of the characteristic contribution I of the single modality data in the vehicle accident damage data uploaded by the multimodal data fusion module is: I=xi×yi; Among them, I represents the feature contribution of a single modal data, xi represents the feature value of a single modal data, yi represents the weight of a single modal data, and i represents the type number of the modal data.

Images