Immersive safety accident reproduction system and method based on digital twinning

Abstract

The invention relates to the technical field of accident reproduction, and particularly discloses an immersive safety accident reproduction system and method based on digital twinning, and the system comprises a scene rendering end which builds a refined original scene rendering model according to the facility texture of an accident scene video, three-dimensional remote sensing topographic data and a preset scene facility original model; the dynamic fusion rendering end is used for carrying out fusion dynamic rendering on multiple attributes, action tracking and accident development representation data of an accident scene person and a refined original scene rendering model to generate a safety accident reproduction dynamic model; the immersive display end is used for dividing a safety accident reproduction dynamic model based on accident process key point moments to obtain corresponding local models, and displaying one or more local models to a user through virtual reality or augmented reality equipment; the accuracy and authenticity of safety accident reproduction and the user experience are improved, the accident safety management level can be improved, and the immersive interaction experience and participation sense of the user are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of accident reconstruction, and particularly relates to an immersive safety accident reconstruction system and method based on digital twin. Background Art

[0002] In today's highly complex and diverse social and industrial environments, the frequency and severity of safety accidents are increasing continuously, posing a huge threat to people's lives, property, and social stability. To effectively prevent and respond to safety accidents, it is particularly important to deeply study the occurrence mechanism, process, and consequences of accidents. As an emerging digital means, digital twin technology provides new ideas and methods for the reconstruction and analysis of safety accidents. Currently, in the field of safety accident reconstruction, existing technical means have many limitations. Common methods mostly rely on traditional two-dimensional images, text descriptions, or relatively rough three-dimensional models to present accident situations. These methods have obvious deficiencies in showing the realism and details of the accident scene.

[0003] However, for the model construction of scene facilities in existing systems, relatively simple and general models are mostly used, and it is impossible to carry out refined and personalized design and construction for different types of scenes and specific facilities. Only relying on a single modeling data source, lacking the multi-modal data fusion of accident scene video texture features and three-dimensional remote sensing terrain, resulting in the distortion rate of the spatial topological relationship of scene facilities. This leads to the inability to accurately show the unique structures, functions, and operating states of facilities in different scenes during accident reconstruction, and it is difficult to meet the needs of in-depth analysis and research of accidents. Moreover, the traditional method adopts the splicing mode of static scenes + preset animations, and it is impossible to realize the real-time dynamic coupling of personnel movement tracking data, accident development parameters, and scene models, and the data synchronization error is generally high. Existing VR systems only support fixed-time-axis playback, and do not establish a spatio-temporal slicing mechanism based on key moment points of the accident process (such as collision critical points, energy release peak points), resulting in a relatively high missing rate of immersive analysis of key causative factors.

[0004] Therefore, the present invention proposes an immersive safety accident reconstruction system and method based on digital twin. Summary of the Invention

[0005] The present invention provides an immersive safety accident reproduction system and method based on digital twin, including: The scene rendering end can build a refined original scene rendering model by combining the texture features of the original accident scene facilities, the three-dimensional remote sensing terrain data, and the original scene facility models under each preset scene type, providing a high-quality basic scene for subsequent dynamic rendering. The dynamic fusion rendering end dynamically renders the multi-attribute characterization features, action tracking data, and multi-dimensional accident development characterization data of the people with the original scene model, making the reproduction model more realistically and comprehensively reflect the accident situation. The immersive display end divides the dynamic model according to the key moments of the accident process and displays it to the user through virtual reality or augmented reality devices, enabling the user to experience the accident reproduction process more targeted and immersively. It can reproduce the occurrence process of safety accidents more accurately, providing intuitive and effective means for accident analysis, prevention, and training. It improves the accuracy, authenticity, and user experience of safety accident reproduction, helps to improve the safety management level, and enhances the user's immersive interaction experience and sense of participation. Through the multi-source data fusion architecture, dynamic rendering engine, and spatio-temporal slicing display mechanism of the present invention, the defects described in the background technology are solved, and the reduction accuracy of accident details is improved.

[0006] The present invention provides an immersive safety accident reproduction system based on digital twin, including:

[0007] A scene rendering end, configured to build a refined original scene rendering model based on the texture features of the original accident scene facilities in the accident scene video, the three-dimensional remote sensing terrain data of the target accident location, and the original scene facility models under each preset scene type;

[0008] A dynamic fusion rendering end, configured to perform fusion dynamic rendering on the multi-attribute characterization features, action tracking data, and multi-dimensional accident development characterization data of all people in the accident scene with the refined original scene rendering model to obtain a safety accident reproduction dynamic model;

[0009] An immersive display end, configured to divide the safety accident reproduction dynamic model based on all the key moments of the accident process at the accident scene, obtain a safety accident partial reproduction dynamic model corresponding to each key moment of the accident process, and display the safety accident partial reproduction dynamic model corresponding to a single or multiple key moments of the accident process to the user through a virtual reality device or an augmented reality device.

[0010] Preferably, the scene rendering end includes:

[0011] A scene rough construction module, configured to build a rough original scene model based on the selection instruction input by the user and the original scene facility models under each preset scene type;

[0012] The scene refinement rendering module is used to perform fine rendering on the rough original scene model based on the texture features of the original accident scene facilities in the accident scene video and the 3D remote sensing terrain data of the target accident location, and obtain the refined original scene rendering model.

[0013] Preferably, the dynamic fusion rendering terminal includes:

[0014] The human interaction analysis module is used to analyze the human interaction response data of all human scene interaction spatio-temporal nodes based on the multi-attribute characterization features and action tracking data of all people in the accident scene, and generate the human interaction spatio-temporal record thread of the accident scene;

[0015] The accident development analysis module is used to analyze the multi-dimensional accident development characterization data based on the accident scene video, and generate the accident development spatio-temporal record thread of the accident scene based on the multi-dimensional accident development characterization data;

[0016] The dynamic interaction analysis module is used to perform fusion dynamic rendering on the refined original scene rendering model by combining the action tracking data of all people, the human interaction response data of all human scene interaction spatio-temporal nodes, and the multi-dimensional accident development characterization data with the human interaction spatio-temporal record thread and the accident development spatio-temporal record thread of the accident scene, and obtain the dynamic model of safety accident reconstruction.

[0017] Preferably, the human interaction analysis module includes:

[0018] The character characterization and action tracking sub-module is used to perform multi-attribute feature characterization and action tracking on all people in the accident scene video, and obtain the multi-attribute characterization features and action tracking data of all people in the accident scene;

[0019] The human dynamic model generation sub-module is used to generate the 3D action tracking models of all people in the accident scene based on the multi-attribute characterization features and action tracking data of all people in the accident scene;

[0020] The interaction node recognition sub-module is used to identify all human scene interaction spatio-temporal nodes based on the action tracking data of all people in the accident scene;

[0021] The interaction response analysis sub-module is used to analyze the human interaction response data of all human scene interaction spatio-temporal nodes based on the accident scene video;

[0022] The interaction spatio-temporal record generation sub-module is used to generate the human interaction spatio-temporal record thread of the accident scene based on the human interaction response data of all human scene interaction spatio-temporal nodes.

[0023] Preferably, the interaction node recognition sub-module includes:

[0024] A speed and acceleration derivation unit, which is used to determine the three-dimensional coordinates of each person at each moment based on the action tracking data of all persons at the accident scene, and determine the speed and acceleration of each person in the three coordinate dimensions at each moment based on the three-dimensional coordinates of each person at each moment;

[0025] A rate of change vector construction unit, which is used to calculate the velocity rate of change vector and acceleration rate of change vector of each person at each moment based on the speed and acceleration of each person in the three coordinate dimensions at each moment;

[0026] An action rate of change calculation unit, which is used to calculate the comprehensive action rate of change of each person at each moment based on the magnitude of the velocity rate of change vector and the magnitude of the acceleration rate of change vector of each person at each moment;

[0027] An interaction spatio-temporal node analysis unit, which is used to determine all the artificial scene interaction spatio-temporal nodes of all persons based on the central coordinates of each facility in the refined original scene rendering model, the three-dimensional coordinates of each person at each moment, and the comprehensive action rate of change of each person at each moment.

[0028] Preferably, the interaction spatio-temporal node analysis unit includes:

[0029] A spatial correlation degree calculation subunit, which is used to calculate the spatial correlation degree between each person and all facilities in the refined original scene rendering model at each moment based on the distance between the central coordinates of each facility in the refined original scene rendering model and the three-dimensional coordinates of each person at each moment;

[0030] A spatio-temporal correlation factor calculation subunit, which is used to calculate the spatio-temporal correlation factor between each person and all facilities in the refined original scene rendering model at each moment based on the comprehensive action rate of change of each person at each moment and the spatial correlation degree between each person and all facilities in the refined original scene rendering model at each moment;

[0031] An interaction spatio-temporal node analysis subunit, which is used to determine all the artificial scene interaction spatio-temporal nodes of all persons based on the spatio-temporal correlation factor between each person and all facilities in the refined original scene rendering model at each moment.

[0032] Preferably, the method for the interaction spatio-temporal node analysis subunit to determine all the artificial scene interaction spatio-temporal nodes of all persons based on the spatio-temporal correlation factor between each person and all facilities in the refined original scene rendering model at each moment includes:

[0033] Filter out all moments when the spatio-temporal correlation factor between a single person and all facilities in the refined original scene rendering model at the corresponding moment is not less than the spatio-temporal correlation threshold among all moments, and regard these moments as all artificial scene interaction time nodes of the corresponding person. Then summarize all artificial scene interaction time nodes of the corresponding person and the three-dimensional coordinates of the corresponding person at all artificial scene interaction time nodes, and regard them as all artificial scene interaction spatio-temporal nodes of the corresponding person until all artificial scene interaction spatio-temporal nodes of all persons are determined.

[0034] Preferably, the immersive display terminal includes:

[0035] An accident process key point recognition module for recognizing all accident process key point moments in the safety accident reproduction dynamic model;

[0036] An accident process division module for dividing the safety accident reproduction dynamic model based on all accident process key point moments to obtain a local safety accident reproduction dynamic model corresponding to each accident process key point moment;

[0037] An immersive display module for displaying the local safety accident reproduction dynamic model corresponding to the corresponding accident process key point moment to the user based on the immersive interaction instruction input by the user and through a virtual reality device or an augmented reality device.

[0038] Preferably, the accident process key point recognition module includes:

[0039] A dynamic feature change rate analysis sub-module for generating a multi-dimensional state vector of each moment at the accident scene based on the safety accident reproduction dynamic model, and generating a dynamic feature change rate vector of each moment at the accident scene based on the multi-dimensional state vector of each moment at the accident scene;

[0040] An accident key point comprehensive index analysis sub-module for analyzing the correlation between different features at each moment at the accident scene based on the dynamic feature change rate vector of each moment at the accident scene, and determining the accident key point comprehensive index of each moment at the accident scene based on the dynamic feature change rate vector of each moment at the accident scene and the correlation between different features;

[0041] A key point moment screening sub-module for screening out all moments when the accident key point comprehensive index is not less than the accident key point comprehensive index threshold among all moments as all accident process key point moments in the safety accident reproduction dynamic model.

[0042] The present invention provides an immersive safety accident reproduction method based on digital twin, which is applied to any of the above immersive safety accident reproduction systems based on digital twin, and includes:

[0043] S1: Based on the original accident scene facility texture features in the accident scene video, the three-dimensional remote sensing terrain data of the target accident location, and the original scene facility models under each preset scene type, build a refined original scene rendering model;

[0044] S2: Integrate and dynamically render the multi-attribute characterization features, action tracking data of all characters in the accident scene, and multi-dimensional accident development characterization data with the refined original scene rendering model to obtain a dynamic model for safety accident reconstruction;

[0045] S3: Based on all the key points of the accident process at the accident scene, divide the dynamic model for safety accident reconstruction to obtain a local dynamic model for safety accident reconstruction corresponding to each key point of the accident process, and display the local dynamic model for safety accident reconstruction corresponding to a single or multiple key points of the accident process to the user through virtual reality devices or augmented reality devices.

[0046] The beneficial effects of the present invention compared with the prior art are as follows: The scene rendering end can combine the original accident scene facility texture features, three-dimensional remote sensing terrain data, and the original scene facility model to build a refined original scene rendering model, providing a high-quality basic scene for subsequent dynamic rendering. The dynamic fusion rendering end fuses and dynamically renders the multi-attribute characterization features, action tracking data of characters, and multi-dimensional accident development characterization data with the original scene model, making the reconstruction model more realistically and comprehensively reflect the accident situation. The immersive display end divides the dynamic model according to the key points of the accident process at all times and displays it to the user through virtual reality or augmented reality devices, enabling the user to more specifically and immersively experience the accident reconstruction process. It can more accurately reproduce the occurrence process of safety accidents, providing an intuitive and effective means for accident analysis, prevention, and training. It improves the accuracy, authenticity, and user experience of safety accident reconstruction, helps to improve the safety management level, enhances the user's immersive interaction experience and sense of participation. Through the multi-source data fusion architecture, dynamic rendering engine, and spatio-temporal slicing display mechanism of the present invention, the defects described in the background art are solved, and the reduction accuracy of accident details is improved.

[0047] Other features and advantages of the present invention will be described in the subsequent specification, and part of them will become obvious from the specification or be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in this application document.

[0048] The technical solutions of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0049] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the accompanying drawings:

[0050] Figure 1 It is a schematic diagram of an immersive safety accident reproduction system based on digital twin in an embodiment of the present invention;

[0051] Figure 2 It is a schematic diagram of the execution logic of an immersive safety accident reproduction system based on digital twin in an embodiment of the present invention;

[0052] Figure 3 It is a flowchart of a method for immersive safety accident reproduction based on digital twin in an embodiment of the present invention. Detailed Embodiments

[0053] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0054] Embodiment 1:

[0055] The present invention provides an immersive safety accident reproduction system based on digital twin. Referring to Figure 1 and Figure 2 , it includes:

[0056] A scene rendering terminal, which is used to build a refined original scene rendering model based on the original accident scene facility texture features in the accident scene video, the three-dimensional remote sensing terrain data of the target accident location, and the original scene facility models under each preset scene type;

[0057] A dynamic fusion rendering terminal, which is used to perform fusion dynamic rendering on the multi-attribute characterization features and action tracking data of all the people in the accident scene and the multi-dimensional accident development characterization data with the refined original scene rendering model to obtain a safety accident reproduction dynamic model;

[0058] An immersive display terminal, which is used to divide the safety accident reproduction dynamic model based on all the key moments of the accident process at the accident scene to obtain a safety accident partial reproduction dynamic model corresponding to each key moment of the accident process, and display the safety accident partial reproduction dynamic model corresponding to a single or multiple key moments of the accident process to the user through virtual reality devices or augmented reality devices.

[0059] In this embodiment, the accident scene video refers to the video material that records the actual situation of the safety accident scene, which contains various information such as the environmental conditions at the accident scene, the status of facilities, and the activities of people.

[0060] In this embodiment, the texture features of the original accident scene facilities are the significant feature manifestations such as the surface texture, patterns, colors, etc. presented by various facilities in the accident scene.

[0061] In this embodiment, the location of the target accident is the specific geographical location where the safety accident studied actually occurred.

[0062] In this embodiment, the 3D remote sensing terrain data is the 3D related data about the terrain of the location of the target accident obtained through remote sensing technology, which can clearly reflect the actual situation such as the undulation of the terrain.

[0063] In this embodiment, each preset scene type is a different category of accident scene types preset in advance, such as factory workshop scenes, road traffic accident scene types, etc.

[0064] In this embodiment, the original model of scene facilities under each preset scene type is the initial model of scene facilities (such as buildings, structures, equipment, road greening, vehicles, tools, etc. inside and outside the factory area) that has not been finely processed yet, constructed for each preset scene type.

[0065] In this embodiment, the refined original scene rendering model is the scene model obtained after comprehensive fine processing and rendering of elements such as the texture features of on-site facilities and terrain data based on the preliminarily established model.

[0066] In this embodiment, the multi-attribute characterization features of people are the features that describe in detail the attributes of people in the accident scene from multiple aspects, such as age features, gender features, clothing features, etc.

[0067] In this embodiment, the action tracking data is the relevant data used to record the action conditions of people in the accident scene.

[0068] In this embodiment, the multi-dimensional accident development characterization data is the relevant data that describes the accident development process and status from multiple different aspects, such as the scale, influence range, severity, etc. of the accident.

[0069] In this embodiment, the dynamic model for reproducing safety accidents is the model generated after fusing various relevant data of the accident scene, which can dynamically display the entire process of the accident occurrence.

[0070] In this embodiment, all key moments of the accident process at the accident scene are specific time points with key significance and important value during the entire process of the accident occurrence, such as the moment when the accident starts to occur, the moment when key events appear, etc.

[0071] In this embodiment, the local reproduction dynamic model of the safety accident corresponding to each key moment of the accident process is a dynamic model generated for each key moment in the accident process, which can separately display the detailed local conditions of the accident at that specific moment.

[0072] The beneficial effects of the above technology are as follows: The scene rendering end can build a refined original scene rendering model by combining the texture features of the original accident scene facilities, the three-dimensional remote sensing terrain data, and the original model of the scene facilities, providing a high-quality basic scene for subsequent dynamic rendering. The dynamic fusion rendering end fuses and dynamically renders the multi-attribute characterization features of the characters, the action tracking data, and the multi-dimensional accident development characterization data with the original scene model, making the reproduction model more realistically and comprehensively reflect the accident situation. The immersive display end divides the dynamic model according to the key moments of the accident process and displays it to the user through virtual reality or augmented reality devices, enabling the user to more targeted and immersively experience the accident reproduction process. It can more accurately reproduce the occurrence process of safety accidents, providing intuitive and effective means for accident analysis, prevention, and training. It improves the accuracy, authenticity, and user experience of safety accident reproduction, helps to enhance the safety management level, and enhances the user's immersive interaction experience and sense of participation. Through the multi-source data fusion architecture, dynamic rendering engine, and spatio-temporal slicing display mechanism of the present invention, the defects described in the background technology are solved, and the accuracy of restoring accident details is improved.

[0073] Embodiment 2:

[0074] Based on Embodiment 1, the scene rendering end includes:

[0075] A rough scene construction module, which is used to build a rough original scene model based on the selection instruction input by the user and the original model of the scene facilities under each preset scene type;

[0076] A scene refinement rendering module, which is used to perform fine rendering on the rough original scene model based on the texture features of the original accident scene facilities in the accident scene video and the three-dimensional remote sensing terrain data of the target accident location to obtain a refined original scene rendering model.

[0077] In this embodiment, the selection instruction input by the user is an instruction given by the user according to their own needs to guide the system operation.

[0078] In this embodiment, building a rough original scene model based on the selection instruction input by the user and the original model of the scene facilities under each preset scene type means constructing a general original scene model according to the instruction given by the user and combining the initial facility models under various preset scene types set in advance.

[0079] In this embodiment, the rough original scene model is finely rendered based on the texture features of the original accident scene facilities in the accident scene video and the three-dimensional remote sensing terrain data of the target accident location. Obtaining the refined original scene rendering model means using the texture characteristics of the facilities in the accident scene video and the terrain data of the accident location to further refine and render the initially built scene model, thereby obtaining a more accurate and realistic scene rendering model.

[0080] The beneficial effects of the above technical solutions are as follows: The scene rough construction module quickly constructs a rough original scene model according to the selection instructions and the original model input by the user, improving the efficiency of scene construction. The scene refinement rendering module uses the texture features of the facilities at the accident scene and the three-dimensional remote sensing terrain data to finely render the rough model, making the scene more realistic and accurate. This step-by-step construction method improves the quality and details of the scene while ensuring efficiency. It can provide a more realistic basic model for subsequent dynamic fusion rendering. It improves the efficiency and effect of scene rendering and helps to reproduce the accident scene more realistically.

[0081] Embodiment 3:

[0082] Based on Embodiment 1, the dynamic fusion rendering end includes:

[0083] The human interaction analysis module is used to analyze the human interaction response data of all human-scene interaction spatio-temporal nodes based on the multi-attribute characterization features and action tracking data of all people at the accident scene, and generate a human interaction spatio-temporal record thread for the accident scene;

[0084] The accident development analysis module is used to analyze multi-dimensional accident development characterization data based on the accident scene video, and generate an accident development spatio-temporal record thread for the accident scene based on the multi-dimensional accident development characterization data;

[0085] The dynamic interaction analysis module is used to perform fusion dynamic rendering of the action tracking data of all people, the human interaction response data of all human-scene interaction spatio-temporal nodes, and the multi-dimensional accident development characterization data with the refined original scene rendering model based on the human interaction spatio-temporal record thread and the accident development spatio-temporal record thread of the accident scene, and obtain a dynamic model for reproducing the safety accident.

[0086] In this embodiment, the human-scene interaction spatio-temporal node refers to the combination point of the specific time and space position where a person interacts with the scene facilities at the accident scene.

[0087] In this embodiment, the human interaction response data of the human-scene interaction spatio-temporal node is the relevant data generated by a person's interaction behavior with the scene at the human-scene interaction spatio-temporal node (for example, the data indicating the change in the form of the facility after the task makes an interaction behavior with the scene).

[0088] In this embodiment, the spatiotemporal recording thread of human interaction at the accident scene is sequence data formed by continuously recording the time and space information of the interaction between people and scenes at the accident scene.

[0089] In this embodiment, the multi-dimensional accident development characterization data is analyzed based on the accident scene video, which is to obtain relevant data that can describe the accident development from multiple aspects through in-depth research on the accident scene video.

[0090] In this embodiment, the accident development spatiotemporal record thread of the accident scene is generated based on the multi-dimensional accident development representation data, which constructs a record sequence reflecting the changes in the accident development in time and space based on the acquired multi-dimensional accident development representation data.

[0091] In this embodiment, based on the human interaction spatiotemporal recording thread of the accident scene and the accident development spatiotemporal recording thread, the action tracking data of all characters, the human interaction response data of all human-scene interaction spatiotemporal nodes, and the multi-dimensional accident development characterization data are integrated with the refined original scene rendering model for dynamic rendering. The dynamic model for reproducing a safety accident is obtained by combining the spatiotemporal recording information of human interaction at the accident scene and the accident development, together with the character action tracking, human interaction response and multi-dimensional accident development data, with the refined original scene rendering model, and generating a dynamic model that can show the full picture of the accident through dynamic rendering.

[0092] The beneficial effects of the above technical solutions are as follows: the human interaction analysis module obtains human interaction response data and generates human interaction space-time recording threads by analyzing the multi-attribute characterization features and motion tracking data of the characters, providing detailed information for a comprehensive understanding of the human factors in the accident. The accident development analysis module analyzes the multi-dimensional accident development characterization data based on the accident scene video, and generates an accident development space-time recording thread, which helps to accurately grasp the development process of the accident. The dynamic interaction analysis module combines the human interaction space-time recording thread and the accident development space-time recording thread for fusion dynamic rendering, making the dynamic model of safety accident reproduction more in line with the actual situation. It can more comprehensively and truly reflect the dynamic relationship between the characters and the environment in the accident, and the development of the accident. It improves the accuracy and authenticity of the dynamic model of safety accident reproduction, providing a more valuable reference for accident analysis and prevention.

[0093] Embodiment 4:

[0094] Based on Example 3, the human interaction analysis module includes:

[0095] The character portrayal and action tracking sub-module is used to conduct multi-attribute feature portrayal and action tracking on all the characters in the accident scene video, and obtain the multi-attribute portrayal features and action tracking data of all the characters in the accident scene;

[0096] The character dynamic model generation sub-module is used to generate a three-dimensional action tracking model of all the characters in the accident scene based on the multi-attribute portrayal features and action tracking data of all the characters in the accident scene;

[0097] The interaction node recognition sub-module is used to identify all the human-scene interaction spatio-temporal nodes based on the action tracking data of all the characters in the accident scene;

[0098] The interaction response analysis sub-module is used to analyze the human interaction response data of all the human-scene interaction spatio-temporal nodes based on the accident scene video;

[0099] The interaction spatio-temporal record generation sub-module is used to generate a human interaction spatio-temporal record thread of the accident scene based on the human interaction response data of all the human-scene interaction spatio-temporal nodes.

[0100] In this embodiment, conducting multi-attribute feature portrayal and action tracking on all the characters in the accident scene video, and obtaining the multi-attribute portrayal features and action tracking data of all the characters in the accident scene means carefully describing each character in the accident scene video in multiple attributes and simultaneously tracking their actions, so as to obtain the multi-faceted feature descriptions of these characters and the relevant data of their actions.

[0101] In this embodiment, generating a three-dimensional action tracking model of all the characters in the accident scene based on the multi-attribute portrayal features and action tracking data of all the characters in the accident scene means using the multi-attribute features and action tracking data of the characters obtained at the accident scene to construct a model that can display the three-dimensional action conditions of these characters.

[0102] In this embodiment, the human interaction response data of all human-scene interaction spatio-temporal nodes analyzed based on the accident scene video is the relevant data obtained by studying the accident scene video, which shows the reaction of the scene facilities caused by the interaction actions of people at specific spatio-temporal nodes of the human-scene interaction. To analyze the human interaction response data of all human-scene interaction spatio-temporal nodes based on the accident scene video, it is necessary to first preprocess the video, adjust its resolution to an appropriate state and unify the frame rate for subsequent processing; then use deep learning object detection algorithms such as YOLO and FasterR-CNN to detect people and scene facilities respectively, and then use multi-object tracking algorithms such as DeepSORT to obtain their motion trajectories; then extract key frames, and determine the interaction spatio-temporal nodes based on the position and posture changes of people and scene facilities in the key frames and adjacent frames; finally, at these nodes, use the pose estimation algorithm to analyze the human interaction actions, and at the same time observe and record the reaction data such as the displacement and deformation of the scene facilities under the interaction actions.

[0103] In this embodiment, generating the human interaction spatio-temporal record thread of the accident scene based on the human interaction response data of all human-scene interaction spatio-temporal nodes is to sort out the record sequence that can reflect the changes of human interaction in time and space in the accident scene according to the obtained human interaction response data at the human-scene interaction spatio-temporal nodes.

[0104] The beneficial effects of the above technical solutions are as follows: The character characterization and action tracking sub-module conducts multi-attribute feature characterization and action tracking on people, providing a rich data basis for subsequent analysis. The human dynamic model generation sub-module generates a three-dimensional action tracking model, making the action performance of people more intuitive and clear. The interaction node recognition sub-module can accurately identify the human-scene interaction spatio-temporal nodes and capture the key interaction moments. The interaction response analysis sub-module analyzes the interaction response data to deeply understand the behavior of people at the interaction nodes. The interaction spatio-temporal record generation sub-module generates the human interaction spatio-temporal record thread, providing important clues and basis for the overall dynamic fusion rendering. It improves the accuracy and comprehensiveness of the analysis of human interaction, and helps to more realistically reproduce the human factors in the accident.

[0105] Embodiment 5:

[0106] Based on Embodiment 4, the interaction node recognition sub-module includes:

[0107] A speed and acceleration derivation unit, which is used to determine the three-dimensional coordinates of each person at each moment based on the action tracking data of all people in the accident scene, and determine the speed and acceleration of each person in the three coordinate dimensions at each moment based on the three-dimensional coordinates of each person at each moment;

[0108] A rate-of-change vector construction unit, configured to calculate a velocity rate-of-change vector and an acceleration rate-of-change vector for each person at each moment based on the velocity and acceleration of each person in three coordinate dimensions at each moment;

[0109] An action rate-of-change calculation unit, configured to calculate a comprehensive action rate-of-change for each person at each moment based on the magnitudes of the velocity rate-of-change vectors and the acceleration rate-of-change vectors of each person at each moment;

[0110] An interaction spatio-temporal node analysis unit, configured to determine all artificial scene interaction spatio-temporal nodes of all persons based on the central coordinates of each facility in the refined original scene rendering model, the three-dimensional coordinates of each person at each moment, and the comprehensive action rate-of-change of each person at each moment.

[0111] In this embodiment, determining the velocity and acceleration of each person in three coordinate dimensions at each moment based on the three-dimensional coordinates of each person at each moment means calculating, according to the spatial position coordinates of the person at each specific time point, the moving velocity of the person in the X, Y, and Z coordinate directions at this moment and the rate of change of the velocity (acceleration) using relevant mathematical methods.

[0112] In this embodiment, calculating the velocity rate-of-change vector and the acceleration rate-of-change vector for each person at each moment based on the velocity and acceleration of each person in three coordinate dimensions at each moment means using the velocity and acceleration data of the person in three directions at each moment and further obtaining, through mathematical operations, vectors that can describe the rate of change of velocity and acceleration. Among them, the velocity rate-of-change vector is composed of the velocity rates-of-change in the X, Y, and Z axis directions at a single moment, and the acceleration rate-of-change vector is composed of the acceleration rates-of-change in the X, Y, and Z axis directions at a single moment.

[0113] In this embodiment, calculating the comprehensive action rate-of-change for each person at each moment based on the magnitudes of the velocity rate-of-change vectors and the acceleration rate-of-change vectors of each person at each moment means comprehensively calculating the magnitudes of the velocity rate-of-change vectors and the acceleration rate-of-change vectors of the person at each moment to obtain a value that can reflect the overall intensity of the action change of this person at this moment. For example: Using the weighted summation method, the sum of the product of the magnitude of the velocity rate-of-change vector of each person at each moment and the corresponding weight coefficient and the product of the magnitude of the acceleration rate-of-change vector and the corresponding weight coefficient is regarded as the comprehensive action rate-of-change of each person at the corresponding moment, and the sum of the two weight coefficients is 1.

[0114] The beneficial effects of the above technical solutions are as follows: The speed and acceleration derivation unit calculates the speed and acceleration by determining the three-dimensional coordinates of a person, providing basic motion parameters for subsequent analysis. The rate-of-change vector construction unit calculates the speed rate-of-change vector and the acceleration rate-of-change vector, more comprehensively describing the changes in the person's motion. The action rate-of-change calculation unit obtains the comprehensive action rate-of-change, which can quantitatively measure the intensity and degree of change of a person's actions. The interactive spatio-temporal node analysis unit determines the interactive spatio-temporal nodes by comprehensively considering the coordinates of scene facilities, the three-dimensional coordinates of a person, and the comprehensive action rate-of-change, improving the accuracy and reliability of recognition. It can more accurately identify the human-scene interactive spatio-temporal nodes, providing key time points and location information for subsequent interactive response analysis and dynamic fusion rendering.

[0115] Embodiment 6:

[0116] Based on Embodiment 5, the interactive spatio-temporal node analysis unit includes:

[0117] The spatial correlation degree calculation subunit is used to calculate the spatial correlation degree between each person at each moment and all facilities in the refined original scene rendering model based on the distance between the center coordinates of each facility in the refined original scene rendering model and the three-dimensional coordinates of each person at each moment (which is the sum of the weighted reciprocals of the distances between the person and all facilities):

[0118]

[0119] In the formula, S(t) is the spatial correlation degree between the currently calculated single person and all facilities in the refined original scene rendering model at time t, m is the total number of facilities in the refined original scene rendering model, γ j is the weight coefficient of the jth facility in the refined original scene rendering model, d j (t) is the distance between the center coordinates of the jth facility in the refined original scene rendering model and the three-dimensional coordinates of the currently calculated single person at time t,

[0120] The spatio-temporal correlation factor calculation subunit is used to calculate the spatio-temporal correlation factor between each person at each moment and all facilities in the refined original scene rendering model based on the comprehensive action rate-of-change of each person at each moment and the spatial correlation degree between each person at each moment and all facilities in the refined original scene rendering model (combining the action changes in the time dimension and the correlation situation in the spatial dimension):

[0121] C(t) = R(t) * S(t)

[0122] Wherein, C(t) is the spatio-temporal correlation factor between each character and all facilities in the refined original scene rendering model at time t, and R(t) is the comprehensive action change rate of each character at time t;

[0123] The interactive spatio-temporal node analysis subunit is used to determine all the artificial scene interactive spatio-temporal nodes of all characters based on the spatio-temporal correlation factors between each character and all facilities in the refined original scene rendering model at each moment.

[0124] In this embodiment, the spatial correlation degree between each character and all facilities in the refined original scene rendering model at each moment is a metric value used to measure the tightness of the mutual relationship in spatial position between a single character and all facilities in the refined original scene rendering model at a specific moment.

[0125] In this embodiment, the spatio-temporal correlation factor between each character and all facilities in the refined original scene rendering model at each moment comprehensively considers the action situation of the character at a specific moment and its spatial relationship with all facilities, and is an index used to more comprehensively describe the comprehensive correlation degree between the character and the scene facilities in time and space at this moment.

[0126] The beneficial effects of the above technical solutions are as follows: The spatial correlation degree calculation subunit obtains the spatial correlation degree by calculating the distance between the character and the facility, quantifying the positional relationship between the character and the environmental facilities. The spatio-temporal correlation factor calculation subunit comprehensively considers the comprehensive action change rate and the spatial correlation degree, and more comprehensively evaluates the correlation degree between the character and the facility in time and space. The interactive spatio-temporal node analysis subunit determines the interactive spatio-temporal nodes based on the spatio-temporal correlation factors, making the determination of the nodes more scientific and accurate. It can more accurately capture the moments and positions with important interactive significance in time and space between the character and the scene facilities. By performing multi-dimensional feature extraction on the action tracking data, quantifying the spatial correlation between the character and the scene facilities, calculating the spatio-temporal correlation factors, and making judgments based on the set thresholds, the function of accurately identifying all artificial scene interactive spatio-temporal nodes from the action tracking data is realized. It improves the accuracy and reliability of the analysis of interactive spatio-temporal nodes, providing strong support for more realistically reproducing the human interactions in the accident.

[0127] Embodiment 7:

[0128] Based on Embodiment 6, the method for the interactive spatio-temporal node analysis subunit to determine all the artificial scene interactive spatio-temporal nodes of all characters based on the spatio-temporal correlation factors between each character and all facilities in the refined original scene rendering model at each moment includes:

[0129] Filter out all moments when the spatio-temporal correlation factor between a single person and all facilities in the refined original scene rendering model at the corresponding moment is not less than the spatio-temporal correlation threshold among all moments, and regard these moments as all the artificial scene interaction time nodes of the corresponding person. Then summarize all the artificial scene interaction time nodes of the corresponding person and the three-dimensional coordinates of the corresponding person at all the artificial scene interaction time nodes, and regard them as all the artificial scene interaction spatio-temporal nodes of the corresponding person until all the artificial scene interaction spatio-temporal nodes of all persons are determined.

[0130] In this embodiment, the spatio-temporal correlation threshold is a standard value preset for judging whether the spatio-temporal correlation degree between a person and a scene facility reaches an important level. If the spatio-temporal correlation factor between a person and a facility exceeds this threshold, it is considered to have important interaction significance.

[0131] In this embodiment, the artificial scene interaction time node refers to a specific time point in the accident scene where there is a significant interaction between a person and the scene.

[0132] The beneficial effects of the above technical solutions are as follows: By setting the spatio-temporal correlation threshold, it is possible to accurately screen out the moments with significant interaction significance, avoiding the interference of irrelevant or secondary interaction moments. Summarizing the three-dimensional coordinates of a person at the interaction time node together completely determines the artificial scene interaction spatio-temporal node, providing accurate position information for subsequent analysis and rendering. This determination method has clear criteria and operability, improving the accuracy and consistency of the determination of interaction spatio-temporal nodes. It can effectively highlight the key artificial interaction moments and positions, helping to more accurately reproduce the important interaction scenes in the accident. It enhances the scientificity and effectiveness of the determination of artificial scene interaction spatio-temporal nodes, providing more valuable interaction data for the safety accident reproduction system.

[0133] Embodiment 8:

[0134] Based on Embodiment 1, the immersive display terminal includes:

[0135] An accident process key point recognition module for recognizing all accident process key point moments in the safety accident reproduction dynamic model;

[0136] An accident process division module for dividing the safety accident reproduction dynamic model based on all accident process key point moments to obtain a safety accident partial reproduction dynamic model corresponding to each accident process key point moment;

[0137] An immersive display module for, based on the immersive interaction instruction input by the user and through a virtual reality device or an augmented reality device, displaying the safety accident partial reproduction dynamic model corresponding to the corresponding accident process key point moment to the user.

[0138] In this embodiment, the dynamic model of safety accident reproduction is divided based on all key moments of the accident process, and the local dynamic model of safety accident reproduction corresponding to each key moment of the accident process is obtained. That is, based on all the key time points in the accident occurrence process, the dynamic model of the entire safety accident reproduction is decomposed into multiple parts, and each part is a local dynamic model corresponding to a certain key moment of the accident, which can separately display the specific situation of the accident at that moment.

[0139] In this embodiment, based on the immersive interaction instructions input by the user and through a virtual reality device or an augmented reality device, the local dynamic model of safety accident reproduction corresponding to the corresponding key moment of the accident process is displayed to the user. That is, in accordance with the instructions given by the user to deeply understand and experience, using virtual reality or augmented reality technical means, the local dynamic model of safety accident reproduction corresponding to the key moment of the accident process specified by the user is presented to the user, giving the user a feeling of being on the scene.

[0140] The beneficial effects of the above technical solutions are as follows: The accident process key point identification module can accurately find out the critical moments in the dynamic model of safety accident reproduction, providing a basis for subsequent model division. The accident process division module divides the model according to the key moments, making each local reproduction dynamic model more targeted and independent. The immersive display module displays according to the user's immersive interaction instructions, enhancing the user's sense of participation and experience. It can meet the user's attention and in-depth understanding needs for specific accident process key points, improving the flexibility and pertinence of the display. It improves the accuracy and user-friendliness of the safety accident reproduction display, helping to better play the role of the system in accident analysis and training.

[0141] Embodiment 9:

[0142] Based on Embodiment 8, the accident process key point identification module includes:

[0143] The dynamic feature change rate analysis sub-module is used to generate a multi-dimensional state vector for each moment at the accident scene based on the dynamic model of safety accident reproduction, and generate a dynamic feature change rate vector for each moment at the accident scene based on the multi-dimensional state vector for each moment at the accident scene;

[0144] The accident key point comprehensive index analysis sub-module is used to analyze the correlation between different features at each moment at the accident scene based on the dynamic feature change rate vector for each moment at the accident scene, and determine the accident key point comprehensive index for each moment at the accident scene based on the dynamic feature change rate vector for each moment at the accident scene and the correlation between different features;

[0145] The key moment screening sub-module is used to screen out all moments when the comprehensive index of accident key points is not less than the threshold of the comprehensive index of accident key points from all moments as all accident process key moments in the dynamic model of safety accident reproduction.

[0146] In this embodiment, the multi-dimensional state vector of each moment at the accident scene generated by the dynamic model of safety accident reproduction is to create a vector that can describe its state from multiple aspects for each specific time point at the accident scene through the dynamic model of safety accident reproduction. For example, for moment t, the multi-dimensional feature vector X(t) = [x1(t), x2(t), …, x n (t)], where x i (t) represents the value of the i-th feature dimension at moment t. For example, x1(t) can be the displacement of the key object in the accident, x2(t) is its speed, x3(t) is the acceleration, etc.

[0147] In this embodiment, generating the dynamic feature change rate vector of each moment at the accident scene based on the multi-dimensional state vector of each moment at the accident scene is to calculate a vector that can reflect the degree of change of these states according to the multi-dimensional state vector of each moment:

[0148]

[0149] where Δt is an extremely small time interval used to calculate the change rate of the feature in a short time. This vector reflects the degree of change of each feature dimension near moment t.

[0150] In this embodiment, analyzing the correlation between different features of each moment at the accident scene based on the dynamic feature change rate vector of each moment at the accident scene is to use the dynamic feature change rate vector of each moment to study the mutual correlation and influence relationship between various features of the accident scene at that moment. For example, introducing the feature change correlation matrix C(t), whose element C ij (t) represents the correlation between the i-th feature change rate and the j-th feature change rate at moment t:

[0151]

[0152] where R i (t) and R j (t) are the i-th and j-th elements in R(t) respectively, R i (t k ) and R j (t k ) are the i-th and j-th elements in R(t k ) respectively, and are their average values within a time window N, t kAt the k-th moment within a time window N, the correlation matrix helps to discover the synergistic or abnormal relationships between feature changes.

[0153] In this embodiment, based on the dynamic feature change rate vectors at each moment of the accident scene and the correlations between different features, the comprehensive accident key point index at each moment of the accident scene is determined by integrating the dynamic feature change rate vectors at each moment and the correlations between different features, and calculating a comprehensive index that can comprehensively reflect the importance of that moment in the accident. For example, define the comprehensive accident key point index at time t as:

[0154]

[0155] In the formula, α i is the weight of the i-th feature change rate, reflecting its importance for accident key point judgment, β ij is the weight of the correlation matrix element, reflecting the contribution of the correlation between different feature changes to accident key point judgment,

[0156] In this embodiment, the threshold of the comprehensive accident key point index is a preset standard value used to judge whether the importance of each moment of the accident scene reaches a critical level. If the comprehensive accident key point index of a certain moment exceeds this threshold, it is considered that this moment is the critical key point moment of the accident.

[0157] The beneficial effects of the above technical solutions are as follows: The dynamic feature change rate analysis sub-module generates multi-dimensional state vectors and dynamic feature change rate vectors, providing basic data for subsequent analysis. The comprehensive accident key point index analysis sub-module determines the comprehensive index by analyzing the correlations, and can comprehensively evaluate the importance of each moment. The key point moment screening sub-module screens out the key point moments by setting the threshold, ensuring the accuracy and effectiveness of the recognition. It can accurately identify the critical key point moments in the dynamic model of safety accident reproduction, providing key nodes for subsequent model division and display. It improves the accuracy and reliability of accident process key point recognition, and helps to more specifically display important accident processes.

[0158] Embodiment 10:

[0159] The present invention provides an immersive safety accident reproduction method based on digital twin, which is applied to any one of the immersive safety accident reproduction systems based on digital twin in Embodiments 1 to 9, referring to Figure 3 , including:

[0160] S1: Based on the original accident scene facility texture features in the accident scene video, the three-dimensional remote sensing terrain data of the target accident location, and the original scene facility models under each preset scene type, build a refined original scene rendering model;

[0161] S2: Integrate and dynamically render the multi-attribute characterization features, action tracking data of all characters, and multi-dimensional accident development representation data in the accident scene with the refined original scene rendering model to obtain a dynamic model for safety accident reconstruction.

[0162] S3: Divide the dynamic model for safety accident reconstruction based on all the key points of the accident process in the accident scene to obtain a local dynamic model for safety accident reconstruction corresponding to each key point of the accident process, and display the local dynamic model for safety accident reconstruction corresponding to a single or multiple key points of the accident process to the user through virtual reality devices or augmented reality devices.

[0163] The beneficial effects of the above technology are as follows: Step S1 can build a refined original scene rendering model by combining the texture features of the original accident scene facilities, three-dimensional remote sensing terrain data, and the original model of the scene facilities, providing a high-quality basic scene for subsequent dynamic rendering. Step S2 integrates and dynamically renders the multi-attribute characterization features, action tracking data of characters, and multi-dimensional accident development representation data with the original scene model, making the reconstruction model more realistically and comprehensively reflect the accident situation. Step S3 divides the dynamic model according to the key points of the accident process and displays it to the user through virtual reality or augmented reality devices, enabling the user to experience the accident reconstruction process more targeted and immersively. It can more accurately reproduce the occurrence process of safety accidents, providing intuitive and effective means for accident analysis, prevention, and training. It improves the accuracy, authenticity, and user experience of safety accident reconstruction, helps to enhance the safety management level, and enhances the user's immersive interaction experience and sense of participation. The defects described in the background technology are solved through the multi-source data fusion architecture, dynamic rendering engine, and spatio-temporal slicing display mechanism of the present invention, improving the accuracy of restoring accident details.

[0164] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and its equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. An immersive safety accident reconstruction system based on digital twins, characterized in that: include: The scene rendering end is used to build a refined original scene rendering model based on the texture features of the original accident scene facilities in the accident scene video, the three-dimensional remote sensing terrain data of the target accident location, and the original model of the scene facilities under each preset scene type; The dynamic fusion rendering end is used to fuse the multi-attribute characterization features and motion tracking data of all characters at the accident scene and the multi-dimensional accident development representation data with the refined original scene rendering model for dynamic rendering to obtain a dynamic model for reproducing safety accidents; The immersive display terminal is used to divide the dynamic model of safety accident reconstruction based on all key moments of the accident process at the accident scene, obtain the local dynamic model of safety accident reconstruction corresponding to each key moment of the accident process, and display the local dynamic model of safety accident reconstruction corresponding to a single or multiple key moments of the accident process to the user through a virtual reality device or an augmented reality device.

2. The immersive safety accident reconstruction system based on digital twin according to claim 1 is characterized in that: Scene rendering end, including: A scene rough construction module is used to construct a rough original scene model based on the selection instructions input by the user and the original model of the scene facilities under each preset scene type; The scene refinement rendering module is used to perform refinement rendering on the rough original scene model based on the texture features of the original accident scene facilities in the accident scene video and the three-dimensional remote sensing terrain data of the target accident location to obtain a refined original scene rendering model.

3. The immersive safety accident reconstruction system based on digital twin according to claim 1 is characterized in that: Dynamic fusion rendering end, including: The human interaction analysis module is used to analyze the human interaction response data of all human scene interaction spatiotemporal nodes based on the multi-attribute characterization features and action tracking data of all characters at the accident scene, and generate a spatiotemporal recording thread of human interaction at the accident scene; The accident development analysis module is used to analyze multi-dimensional accident development representation data based on the accident scene video, and generate a spatiotemporal recording thread of the accident development at the accident scene based on the multi-dimensional accident development representation data; The dynamic interaction analysis module is used to dynamically render the human interaction response data of all characters and all human-scene interaction spatiotemporal nodes based on the human interaction spatiotemporal recording thread at the accident scene and the accident development spatiotemporal recording thread, and the multi-dimensional accident development representation data with the refined original scene rendering model to obtain a dynamic model for safety accident reproduction.

4. The immersive safety accident reconstruction system based on digital twin according to claim 3 is characterized in that: Human interaction analysis module, including: The character depiction and motion tracking submodule is used to perform multi-attribute feature depiction and motion tracking on all characters in the accident scene video, and obtain multi-attribute depiction features and motion tracking data of all characters in the accident scene; A character dynamic model generation submodule is used to generate a three-dimensional motion tracking model of all characters at the accident scene based on the multi-attribute characterization features and motion tracking data of all characters at the accident scene; The interaction node identification submodule is used to identify all human scene interaction spatiotemporal nodes based on the motion tracking data of all people at the accident scene; The interactive response analysis submodule is used to analyze the human interactive response data of all human scene interactive spatiotemporal nodes based on the accident scene video; The interaction spatiotemporal record generation submodule is used to generate the human interaction spatiotemporal record thread of the accident scene based on the human interaction response data of all human scene interaction spatiotemporal nodes.

5. The immersive safety accident reconstruction system based on digital twin according to claim 4 is characterized in that: The interactive node identification submodule includes: A speed and acceleration derivation unit, for determining the three-dimensional coordinates of each person at each moment based on the motion tracking data of all persons at the accident scene, and determining the speed and acceleration of each person in three coordinate dimensions at each moment based on the three-dimensional coordinates of each person at each moment; A change rate vector construction unit, used to calculate a velocity change rate vector and an acceleration change rate vector of each character at each moment based on the velocity and acceleration of each character in three coordinate dimensions at each moment; An action change rate calculation unit, used to calculate the comprehensive action change rate of each character at each moment based on the modulus of the velocity change rate vector and the modulus of the acceleration change rate vector of each character at each moment; The interactive space-time node analysis unit is used to determine all artificial scene interactive space-time nodes of all characters based on the central coordinates of each facility in the refined original scene rendering model, the three-dimensional coordinates of each character at each moment, and the comprehensive change rate of each character's action at each moment.

6. The immersive safety accident reconstruction system based on digital twin according to claim 5 is characterized in that: Interactive spatiotemporal node analysis unit, including: A spatial correlation calculation subunit, for calculating the spatial correlation between each character and all facilities in the refined original scene rendering model at each moment based on the distance between the center coordinates of each facility in the refined original scene rendering model and the three-dimensional coordinates of each character at each moment; A spatiotemporal correlation factor calculation subunit, for calculating the spatiotemporal correlation factor between each character and all facilities in the refined original scene rendering model at each moment based on the comprehensive change rate of each character's actions at each moment and the spatial correlation between each character and all facilities in the refined original scene rendering model at each moment; The interactive space-time node analysis subunit is used to determine all artificial scene interactive space-time nodes of all characters based on the space-time correlation factors between each character and all facilities in the refined original scene rendering model at each moment.

7. The immersive safety accident reconstruction system based on digital twin according to claim 6 is characterized in that: The method of the interactive spatiotemporal node analysis subunit determining all artificial scene interactive spatiotemporal nodes of all characters based on the spatiotemporal correlation factors between each character and all facilities in the refined original scene rendering model at each moment includes: Among all the moments, all the moments in which the spatiotemporal correlation factor between a single character and all the facilities in the refined original scene rendering model at the corresponding moment is not less than the spatiotemporal correlation threshold are screened out, and are regarded as all the man-made scene interaction time nodes of the corresponding character, and all the man-made scene interaction time nodes of the corresponding character and the three-dimensional coordinates of the corresponding character at all the man-made scene interaction time nodes are summarized as all the man-made scene interaction spatiotemporal nodes of the corresponding character, until all the man-made scene interaction spatiotemporal nodes of all the characters are determined.

8. The immersive safety accident reconstruction system based on digital twin according to claim 1 is characterized in that: Immersive display terminal, including: The accident process key point identification module is used to identify all the key points of the accident process in the safety accident reconstruction dynamic model; The accident process division module is used to divide the safety accident reconstruction dynamic model based on all the key moments of the accident process, and obtain the safety accident local reconstruction dynamic model corresponding to each key moment of the accident process; The immersive display module is used to display the local reproduction dynamic model of the safety accident corresponding to the key moments of the accident process to the user through a virtual reality device or an augmented reality device based on the immersive interactive instructions input by the user.

9. The immersive safety accident reconstruction system based on digital twin according to claim 8 is characterized in that: The module for identifying the key points of the accident process includes: A dynamic feature change rate analysis submodule is used to generate a multi-dimensional state vector at each moment of the accident scene based on a safety accident reconstruction dynamic model, and to generate a dynamic feature change rate vector at each moment of the accident scene based on the multi-dimensional state vector at each moment of the accident scene; The accident key comprehensive index analysis submodule is used to analyze the correlation between different features at each moment of the accident scene based on the dynamic feature change rate vector at each moment of the accident scene, and determine the accident key comprehensive index at each moment of the accident scene based on the dynamic feature change rate vector at each moment of the accident scene and the correlation between different features; The key moment screening submodule is used to screen out all moments in which the accident key point comprehensive index is not less than the accident key point comprehensive index threshold as all the accident process key moments in the safety accident reconstruction dynamic model.

10. An immersive safety accident reconstruction method based on digital twins, characterized in that: The immersive safety accident reconstruction system based on digital twins applied to any one of claims 1 to 9 comprises: S1: Based on the texture features of the original accident scene facilities in the accident scene video, the 3D remote sensing terrain data of the target accident location, and the original model of the scene facilities under each preset scene type, a refined original scene rendering model is built; S2: Dynamically render the multi-attribute characterization features and motion tracking data of all characters at the accident scene and the multi-dimensional accident development representation data, and obtain a dynamic model for reproducing safety accidents; S3: Divide the dynamic model of safety accident reconstruction based on all accident process key points at the accident scene, obtain the local dynamic model of safety accident reconstruction corresponding to each accident process key point, and display the local dynamic model of safety accident reconstruction corresponding to single or multiple accident process key points to the user through virtual reality equipment or augmented reality equipment.

Images