A large model-based digital twin modeling method

Abstract

The application relates to the technical field of digital twinning, in particular to a digital twinning modeling method based on a large model. The method comprises the following steps: acquiring three-dimensional point cloud data to form a plurality of initial segmentation regions; correcting the edges of the initial segmentation regions to obtain a region to be modeled; modeling and model matching are performed on each entity target in the region to be modeled according to an entity feature set to generate a corresponding three-dimensional model; the driving state of a mobile trolley is monitored and warned in real time based on a large model, and the warning condition is verified according to the simulated driving speed of the mobile trolley; and the size of the initial segmentation region is adjusted based on the verification result. The application improves the accuracy of the model by correcting the edges, ensures that the boundary of the region to be modeled is more in line with the actual situation, converts the actual object into a digital model, imports the model into a virtual environment for in-depth analysis, adjusts the modeling parameters according to the analysis result, and enhances the real-time performance and accuracy of modeling.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a digital twin modeling method based on a large model. Background Technology

[0002] Digital twins refer to virtual models of physical entities. Through real-time data and analysis techniques, they can reflect the state, behavior, and performance of physical entities. They utilize large-scale machine learning models to model and simulate complex systems. Existing technologies can use feature line fitting and region growing algorithms to achieve rapid point cloud segmentation for modeling production workshops, solving the problems of insufficient and over-segmentation and achieving good segmentation integrity. However, the modeling accuracy and efficiency for complex environments are not ideal. Therefore, there is an urgent need for a rapid point cloud segmentation method that optimizes segmentation parameters through real-time and virtual data to improve modeling accuracy.

[0003] Chinese patent document CN117634210A discloses a digital twin modeling method, including a data collection module, a data storage module, a simulation calculation module, a model building module, a data comparison module, and a data modification module. The data collection module is used to collect various data information of the digital twin system, the data storage module is used to store the collected data information, and the simulation calculation module is used to calculate the collected data and cooperate with the model building module to complete the modeling of the virtual project. It can be seen that the existing digital twin modeling methods lack real-time monitoring and early warning of the state of dynamic targets based on large models, verification of the accuracy of early warning situations, and adjustment of the size of the initial segmented region based on the verification results, resulting in low model accuracy. Summary of the Invention

[0004] To address this, the present invention provides a digital twin modeling method based on a large model, which overcomes the problem of low model accuracy in the prior art due to the lack of monitoring and verification of the driving state of the mobile vehicle to optimize the initial segmentation region.

[0005] To achieve the above objectives, this invention provides a digital twin modeling method based on a large model, comprising:

[0006] The production workshop is fully scanned using a calibrated lidar. The scan data is preprocessed to obtain three-dimensional point cloud data. Based on the density distribution of the three-dimensional point cloud data, the production workshop is initially segmented to form several initial segmented areas.

[0007] Edge contours are extracted from any of the initial segmented regions to calculate the real-time contour error. The edges of the initial segmented regions are then corrected based on the real-time contour error to obtain the region to be modeled.

[0008] Based on the entity feature set, modeling and model matching are performed on each entity target in the area to be modeled to generate a corresponding 3D model, which is then stored in the model database. The 3D model includes a static 3D model and a dynamic 3D model.

[0009] The three-dimensional model is imported into the simulated production workshop. The driving status of the mobile vehicle is monitored and warned in real time based on the large model. The warning situation is verified according to the simulated driving speed of the mobile vehicle. The size of the initial segmented area is adjusted based on the verification results.

[0010] Furthermore, the production workshop is initially segmented based on the density distribution of the 3D point cloud data, including:

[0011] The three-dimensional point cloud data is divided into multiple squares, and the density of feature points corresponding to each square is calculated to analyze the density distribution of each square.

[0012] The production workshop was initially divided based on the density analysis results;

[0013] In this process, the points in each square are counted to calculate the feature point density corresponding to each square. The feature point density is the ratio of the number of points in the square to the area of ​​the square.

[0014] Furthermore, modeling and matching the entity targets within the region to be modeled based on the entity feature set includes,

[0015] Extract the feature information of the region to be modeled as an entity feature set;

[0016] The entity feature set is matched in the feature database, and the entity targets are divided into static targets and dynamic targets according to the matching results;

[0017] For static targets, the geometry of the model is defined based on the corresponding entity feature set, and material and texture information is applied to generate a static 3D model.

[0018] For dynamic targets, real-time coordinates and timestamps are obtained to calculate the moving speed of the dynamic target in the production workshop and to generate a dynamic 3D model.

[0019] Furthermore, correcting the edges of the initial segmented region based on the real-time contour error includes:

[0020] The actual contour error is compared with the standard contour error, and the initial segmented region is smoothed based on the comparison result.

[0021] Specifically, when the actual contour error is determined to be greater than the standard contour error, B-spline curves are used to smooth the edges; the real-time contour error is the sum of the squares of the differences between each actual edge point and the corresponding extracted edge point.

[0022] Furthermore, real-time monitoring of the mobile vehicle's driving status based on a large model includes:

[0023] The simulated driving status of the mobile vehicle is analyzed, and when an abnormal risk is detected in the driving status, an early warning is issued. The simulated driving speed and the real-time driving speed of the mobile vehicle are compared to verify the real-time accuracy of the early warning.

[0024] Furthermore, the analysis of the simulated driving state of the mobile vehicle includes,

[0025] The real-time safe distance between the mobile vehicle and the marked target is obtained, and the real-time safe distance is compared with the standard safe distance.

[0026] Based on the comparison results, it is determined that there is no abnormal risk in the driving status, or the driving status may be analyzed for abnormal risks by combining the deviation trend of the moving car.

[0027] Furthermore, determining that there are no abnormal risks in the driving status based on the comparison results includes:

[0028] When the real-time safe distance is less than or equal to the standard safe distance, the driving status is determined to have any abnormal risks based on the deviation trend;

[0029] When the real-time safe distance is greater than the standard safe distance, it is determined that there is no abnormal risk in the driving status.

[0030] Furthermore, determining whether there are any abnormal risks in the driving status based on the deviation trend includes:

[0031] Obtain the simulated offset direction of the moving vehicle, and determine the offset trend of the moving vehicle based on the simulated offset direction.

[0032] If the simulated offset direction does not begin to deviate, it is determined that there is an abnormal risk in the driving state, and an early warning is issued. The simulated driving speed and the real-time driving speed of the mobile vehicle are also compared.

[0033] Furthermore, verifying the real-time accuracy of the early warning prompts includes,

[0034] The real-time speed of the mobile cart in the simulated production workshop is obtained, and the real-time speed is compared with the standard speed.

[0035] If the real-time accuracy is less than the standard accuracy, the verification result is determined to be incorrect.

[0036] If the real-time accuracy is greater than or equal to the standard accuracy, the verification result is considered correct.

[0037] Furthermore, adjusting the size of the initial segmented region based on the verification results includes:

[0038] If the verification result is determined to be incorrect, the size of the initial segmented region is adjusted.

[0039] Compared with existing technologies, the beneficial effects of this invention are as follows: Data is collected using calibrated and adjusted lidar equipment to ensure the accuracy and reliability of measurements; preprocessing includes noise reduction, filtering, and point cloud simplification to ensure data quality and improve the efficiency of subsequent processing; the point cloud is grouped using the K-means clustering algorithm to identify different regions; edge correction improves model accuracy and ensures the boundaries of the area to be modeled are more realistic; real-world objects are transformed into digital models, and these models are imported into a virtual environment for in-depth analysis; real-time data is combined with the model to dynamically update the 3D model, ensuring it always reflects the latest state, enhancing the real-time performance and accuracy of digital twins; real-time monitoring of the moving vehicle's movement compares the data displayed in the virtual workshop with the dynamic time-varying data in the actual workshop, analyzing errors; large errors indicate low model accuracy, so the size of the segmented regions is adjusted to improve model accuracy; dynamic monitoring ensures production safety, provides timely warnings, and improves the accuracy of model warnings, effectively preventing potential safety hazards.

[0040] Furthermore, through communication between the database and the physics engine, the coordinate, velocity, and environmental information of the dynamic target are mapped into the virtual space, thereby presenting the dynamic changes within the workshop in the virtual environment, which simulates the production workshop. In the dynamic environment, the 3D model is updated based on real-time data. By storing the 3D model in the model database, it is easy to perform dynamic updates, ensuring that the model always reflects the latest state. By combining the real-time data collected by sensors with the model, the state or position of the model is updated, enhancing the real-time performance and accuracy of the digital twin. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating the digital twin modeling method based on a large model according to an embodiment of the present invention;

[0042] Figure 2 This is a schematic diagram of the process of initial division of the production workshop according to an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram illustrating the process of correcting the edges of the initial segmented region according to an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram illustrating the process of modeling and matching various entity targets within the area to be modeled according to an embodiment of the present invention. Detailed Implementation

[0045] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0046] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0047] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0048] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0049] Please see Figure 1 The diagram shown is a flowchart illustrating a digital twin modeling method based on a large model, according to an embodiment of the present invention. The present invention provides a digital twin modeling method based on a large model, comprising:

[0050] Step S1: Use a calibrated lidar to perform a full scan of the production workshop, preprocess the scan data to obtain three-dimensional point cloud data, and perform initial segmentation of the production workshop based on the density distribution of the three-dimensional point cloud data to form several initial segmentation areas.

[0051] Step S2: Extract the edge contour of any of the initial segmented regions to calculate the real-time contour error, and correct the edge of the initial segmented region based on the real-time contour error to obtain the region to be modeled.

[0052] Step S3: Model and match the entity targets in the area to be modeled according to the entity feature set, generate the corresponding three-dimensional model, and store it in the model database. The three-dimensional model includes a static three-dimensional model and a dynamic three-dimensional model.

[0053] Step S4: Import the three-dimensional model into the simulated production workshop, monitor and warn the driving status of the mobile car in real time based on the large model, verify the warning situation according to the simulated driving speed of the mobile car, and adjust the size of the initial segmented area based on the verification results.

[0054] In this embodiment, the calibrated lidar refers to a lidar device that has been calibrated and adjusted to ensure the accuracy and reliability of its measurements. Preprocessing includes noise reduction, filtering, and point cloud simplification to ensure data quality and improve the efficiency of subsequent processing. The K-means clustering algorithm is used to group the point cloud and identify different regions. Edge correction is used to improve the accuracy of the model and ensure that the boundaries of the region to be modeled are more in line with reality. By converting real objects into digital models and importing the models into a virtual environment for in-depth analysis, and by combining real-time data with the model, the 3D model can be dynamically updated to ensure that it always reflects the latest state, enhancing the real-time performance and accuracy of the digital twin. By monitoring the movement of the mobile vehicle in real time, the data displayed in the virtual workshop in real time is compared with the dynamic time-varying data in the actual workshop to analyze the error. If the error is large, it indicates that the model accuracy is low, so the size of the segmented region is adjusted to improve the model accuracy. Dynamic monitoring ensures production safety, provides timely warnings, and improves the accuracy of model warnings, effectively preventing potential safety hazards.

[0055] See Figure 2 As shown, it is a schematic diagram of the process of initial division of the production workshop according to an embodiment of the present invention;

[0056] Specifically, the initial segmentation of the production workshop based on the density distribution of the 3D point cloud data includes:

[0057] Step S101: Divide the three-dimensional point cloud data into multiple squares;

[0058] Step S102: Calculate the feature point density corresponding to each square to analyze the density distribution of each square.

[0059] Step S103: The production workshop is initially divided according to the density analysis results;

[0060] In this process, the points in each square are counted to calculate the feature point density corresponding to each square. The feature point density is the ratio of the number of points in the square to the area of ​​the square.

[0061] In this embodiment, the K-means clustering algorithm is used to analyze whether the density of feature points in the grid is similar, so that grids with similar feature point densities are grouped into the same class, forming multiple initial segmentation regions. The initial grid size is set to 2×2×2 to achieve preliminary point cloud segmentation. According to the data distribution, the number of clusters k is selected, and k is 3. First, Z-score normalization is used to process the feature point density. Then, the K-means clustering algorithm is executed to cluster the normalized feature point density, thereby grouping similar grids into the same class and forming multiple initial segmentation regions. The initial segmentation regions represent different functional areas or equipment areas in the production workshop. The process of extracting the edge contour of any of the initial segmentation regions is to use the OpenCV image processing library to apply the Canny edge detection algorithm to each initial segmentation region to obtain the edge contour. The actual edge data is collected by the sensor to obtain the real edge information corresponding to the extracted edge contour. The least squares method is used to calculate the error between the extracted contour and the actual edge as the real contour error. The sum of the squares of the differences between each actual edge point and the corresponding extracted edge point is calculated to obtain the corresponding real contour error.

[0062] See Figure 3 As shown, it is a schematic diagram of the process of correcting the edge of the initial segmented region according to an embodiment of the present invention;

[0063] Specifically, correcting the edges of the initial segmented region based on the real-time contour error includes,

[0064] Step S201: Compare the actual contour error with the standard contour error;

[0065] Step S202: Smooth the initial segmented region based on the comparison results;

[0066] Specifically, when the actual contour error is determined to be greater than the standard contour error, B-spline curves are used to smooth the edges; the real-time contour error is the sum of the squares of the differences between each actual edge point and the corresponding extracted edge point.

[0067] See Figure 4 As shown, it is a schematic diagram of the process of modeling and matching the entity targets in the area to be modeled according to an embodiment of the present invention;

[0068] Specifically, modeling and matching the entity targets within the region to be modeled based on the entity feature set includes:

[0069] Step S301: Extract the feature information of the region to be modeled as an entity feature set;

[0070] Step S302: Match the entity feature set in the feature database, and classify each entity target into static targets and dynamic targets according to the matching results;

[0071] Step S303: For static targets, define the geometry of the model according to the corresponding entity feature set, apply material and texture information, and generate a static 3D model.

[0072] For dynamic targets, real-time coordinates and timestamps are obtained to calculate the moving speed of the dynamic target in the production workshop and to generate a dynamic 3D model.

[0073] In this embodiment, static targets include buildings and processing equipment, while dynamic targets include vehicles and personnel. Identified dynamic targets are matched against a pre-set dynamic 3D model library. Upon successful matching, the model is instantiated, importing the object into virtual space to form a virtual model corresponding to the real-world object. Ultra-wideband (UWB) positioning technology is used to monitor the movement and trajectory of dynamic targets in the production workshop in real time, recording their real-time coordinates and timestamps, and calculating their movement speed within the workshop. Furthermore, light intensity, temperature and humidity values, and sound decibel values ​​within the workshop are detected using photosensors, temperature and humidity sensors, and sound sensors. The type, perceived movement speed, and environmental monitoring data (light, temperature, humidity, sound) are transmitted to the database. Through communication between the database and the physics engine, the coordinates, speed, and environmental information of the dynamic target are mapped into the virtual space, thus presenting the dynamic changes within the workshop in the virtual environment, which simulates the production workshop. In the dynamic environment, the 3D model is updated based on real-time data. By storing the 3D model in the model database, it is easy to perform dynamic updates, ensuring that the model always reflects the latest state. By combining the real-time data collected by the sensors with the model, the state or position of the model is updated, enhancing the real-time performance and accuracy of the digital twin.

[0074] Specifically, real-time monitoring of the mobile vehicle's driving status based on a large model includes:

[0075] The simulated driving status of the mobile vehicle is analyzed, and when an abnormal risk is detected in the driving status, an early warning is issued. The simulated driving speed and the real-time driving speed of the mobile vehicle are compared to verify the real-time accuracy of the early warning.

[0076] Specifically, the analysis of the simulated driving state of the mobile vehicle includes,

[0077] The real-time safe distance between the mobile vehicle and the marked target is obtained, and the real-time safe distance is compared with the standard safe distance.

[0078] Based on the comparison results, it is determined that there is no abnormal risk in the driving status, or the driving status may be analyzed for abnormal risks by combining the deviation trend of the moving car.

[0079] In this embodiment, the marked target represents equipment in the production workshop that has certain dangers or special characteristics and cannot be collided with or approached. The standard safety distance represents the set safety distance threshold from the marked target. The set value is larger than the danger distance, generally set to be greater than 10% of the danger distance. When it is determined that the real-time safety distance is greater than the standard safety distance, the current driving state is determined to be normal. When it is determined that the real-time safety distance is less than or equal to the standard safety distance, there may be a tendency to reach the danger range. Then, by analyzing the turning trend of the moving car, that is, by analyzing the simulated offset direction based on the comparison result of the standard offset angle and the real-time offset angle, it is determined whether there is a risk in the driving state.

[0080] Specifically, determining that there is no abnormal risk in the driving status based on the comparison results includes:

[0081] When the real-time safe distance is less than or equal to the standard safe distance, the driving status is determined to have any abnormal risks based on the deviation trend;

[0082] When the real-time safe distance is greater than the standard safe distance, it is determined that there is no abnormal risk in the driving status.

[0083] Specifically, determining whether there are abnormal risks in the driving status based on the deviation trend includes:

[0084] Obtain the simulated offset direction of the moving vehicle, and determine the offset trend of the moving vehicle based on the simulated offset direction.

[0085] If the simulated offset direction begins to deviate, it is determined that there is no abnormal risk in the driving state;

[0086] If the simulated offset direction does not begin to deviate, it is determined that there is an abnormal risk in the driving state, and an early warning is issued. The simulated driving speed and the real-time driving speed of the mobile vehicle are also compared.

[0087] In this embodiment, the standard offset angle of the moving trolley is calculated and compared with the actual offset angle. Based on the comparison result, it is determined whether the simulated offset direction has started to shift. Two tangent lines are drawn from the closest point of the moving trolley to the edge contour of the current marked target. The straight line containing the real-time movement direction of the moving trolley is taken as the axis. The angle between any tangent line and the axis is obtained as the standard offset angle between the moving target and the current marked target. The angle between the real-time movement direction of the moving trolley and the axis is obtained as the real-time offset angle. The standard offset angle and the real-time offset angle are then compared.

[0088] If the real-time offset angle is less than the standard offset angle, it is determined that the simulated offset direction has not started to offset.

[0089] If the real-time offset angle is greater than or equal to the standard offset angle, the simulated offset direction is determined to start shifting.

[0090] Specifically, verifying the real-time accuracy of early warning prompts includes,

[0091] The real-time speed of the mobile cart in the simulated production workshop is obtained, and the real-time speed is compared with the standard speed.

[0092] If the real-time accuracy is less than the standard accuracy, the verification result is determined to be incorrect.

[0093] If the real-time accuracy is greater than or equal to the standard accuracy, the verification result is considered correct.

[0094] Wherein, real-time accuracy = 1 - I(real-time driving speed - standard driving speed) / real-time driving speed I.

[0095] In this embodiment, the real-time accuracy rate represents the degree to which the real-time driving speed is relatively close to the standard driving speed. The value range is 0-1. The closer the real-time accuracy rate is to 1, the closer the real-time speed is to the standard speed.

[0096] Specifically, adjusting the size of the initial segmented region based on the verification results includes:

[0097] If the verification result is determined to be incorrect, the size of the initial segmented region is adjusted.

[0098] Specifically, the current segmented region area is adjusted to the corrected segmented region area to adjust the size of the initial segmented region; the corrected segmented region area = current segmented region area × [1-j×(standard accuracy - real-time accuracy) / standard accuracy], where j is the adjustment coefficient, j = 0.2.

[0099] By dynamically adjusting the size of the initial segmentation region based on real-time error monitoring, the accuracy and reliability of the model can be improved.

[0100] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0101] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A large model-based digital twin modeling method, characterized in that, include, A calibrated lidar was used to perform a comprehensive scan of the production workshop, and the scan data was preprocessed. To acquire three-dimensional point cloud data, and to perform initial segmentation of the production workshop based on the density distribution of the three-dimensional point cloud data, forming several initial segmented areas; Edge contour extraction is performed on any of the initial segmented regions to calculate the real-time contour error, based on the actual... The contour error is used to correct the edges of the initial segmented region to obtain the region to be modeled. Based on the entity feature set, modeling and model matching are performed on each entity target in the area to be modeled to generate a corresponding 3D model, which is then stored in the model database. The 3D model includes a static 3D model and a dynamic 3D model. The three-dimensional model is imported into the simulated production workshop. The driving status of the mobile car is monitored and warned in real time based on the large model. The warning situation is verified according to the simulated driving speed of the mobile car. The size of the initial segmented area is adjusted based on the verification results. Determining whether there are abnormal risks in the driving status based on the deviation trend includes: Obtain the simulated offset direction of the moving vehicle, and determine the offset trend of the moving vehicle based on the simulated offset direction. If the simulated offset direction does not start to deviate, it is determined that there is an abnormal risk in the driving state, a warning prompt is issued, and the simulated driving speed and real-time driving speed of the mobile vehicle are compared. Real-time monitoring of the driving status of a mobile vehicle based on a large model includes: The simulated driving status of the mobile vehicle is analyzed, and when an abnormal risk is detected in the driving status, an early warning is issued. The simulated driving speed and the real-time driving speed of the mobile vehicle are compared to verify the real-time accuracy of the early warning. Verification of the real-time accuracy of early warning prompts includes, The real-time speed of the mobile cart in the simulated production workshop is obtained, and the real-time speed is compared with the standard speed. If the real-time accuracy is less than the standard accuracy, the verification result is determined to be incorrect. If the real-time accuracy is greater than or equal to the standard accuracy, the verification result is considered correct.

2. The digital twin modeling method based on a large model according to claim 1, characterized in that, The initial segmentation of the production workshop is performed based on the density distribution of the 3D point cloud data, including: The three-dimensional point cloud data is divided into multiple squares, and the density of feature points corresponding to each square is calculated to analyze the density distribution of each square. The production workshop was initially divided based on the density analysis results; In this process, the points in each square are counted to calculate the feature point density corresponding to each square. The feature point density is the ratio of the number of points in the square to the area of ​​the square.

3. The digital twin modeling method based on a large model according to claim 1, characterized in that, Modeling and matching of entity targets within the region to be modeled based on entity feature sets includes, Extract the feature information of the region to be modeled as an entity feature set; The entity feature set is matched in the feature database, and the entity targets are divided into static targets and dynamic targets according to the matching results; For static targets, the geometry of the model is defined based on the corresponding entity feature set, and material and texture information is applied to generate a static 3D model. For dynamic targets, real-time coordinates and timestamps are obtained to calculate the moving speed of the dynamic target in the production workshop and to generate a dynamic 3D model.

4. The digital twin modeling method based on a large model according to claim 1, characterized in that, Correcting the edges of the initial segmented region based on real-time contour errors includes... The actual contour error is compared with the standard contour error, and the initial segmented region is smoothed based on the comparison result. Specifically, when the actual contour error is determined to be greater than the standard contour error, B-spline curves are used to smooth the edges; the real-time contour error is the sum of the squares of the differences between each actual edge point and the corresponding extracted edge point.

5. The digital twin modeling method based on a large model according to claim 1, characterized in that, The analysis of the simulated driving state of the mobile vehicle includes, The real-time safe distance between the mobile vehicle and the marked target is obtained, and the real-time safe distance is compared with the standard safe distance. Based on the comparison results, it is determined that there is no abnormal risk in the driving status, or the driving status may be analyzed for abnormal risks by combining the deviation trend of the moving car.

6. The digital twin modeling method based on a large model according to claim 5, characterized in that, Based on the comparison results, it was determined that there were no abnormal risks in the driving status, including: When the real-time safe distance is less than or equal to the standard safe distance, the driving status is determined to have any abnormal risks based on the deviation trend; When the real-time safe distance is greater than the standard safe distance, it is determined that there is no abnormal risk in the driving status.

7. The digital twin modeling method based on a large model according to claim 1, characterized in that, Adjusting the size of the initial segmented region based on the verification results includes... If the verification result is determined to be incorrect, the size of the initial segmented region is adjusted.

Images