A virtual reality-based high-voltage equipment live-line work early warning method for motor train units

Abstract

The application discloses a high-voltage equipment live working early warning method for motor train unit based on virtual reality, which comprises the following steps: firstly, collecting three-dimensional data of high-voltage equipment on the roof of a motor train unit, and constructing a virtual simulation model with a one-to-one scale; generating a dynamic safety distance envelope surface according to real-time voltage level, environmental parameters and the length of a working tool, expanding the surface of a live body of the high-voltage equipment to form a three-dimensional closed envelope surface as a prohibited working area; acquiring position data of personnel and tools in real time by using ultra-wideband, inertial measurement and visual positioning technology, mapping the data to the virtual model to generate a virtual avatar; performing collision detection in the virtual environment, and making a risk prediction based on speed and trend; finally, outputting two levels of visual warning signals, i.e., an approaching early warning signal and a danger alarm signal, according to the detection and prediction results, and forming a multi-modal warning output by cooperating with bone conduction earphones and wearable devices. The application realizes intuitive perception of a dangerous area, early risk prediction and physiological adaptive warning, and effectively improves the safety of live working.

Description

Technical Field

[0001] This invention relates to the field of electrified railway operation and maintenance safety technology, specifically to a method for early warning of live-line operation of high-voltage equipment on high-speed trains based on virtual reality. Background Technology

[0002] High-voltage equipment on the roof of high-speed trains is exposed to harsh environments for extended periods and requires regular inspection and maintenance. To improve operational efficiency, some maintenance work must be carried out in a "live" state, that is, while the overhead contact line is not energized, the inspection of roof equipment or simple fault handling is conducted. The core risk of live work lies in the control of safe distance. If the distance between the worker or the hand-held metal tool and the high-voltage live conductor is less than the safe distance, it may cause arc discharge, resulting in personal injury and equipment damage.

[0003] Currently, safety measures for live-line work mainly employ three modes: physical isolation, manual monitoring, and experience-based estimation. Insulating barriers are used to block the contact path between workers and live conductors, but their coverage is limited and affects the worker's field of vision. Dedicated monitors provide verbal reminders, but these are susceptible to errors due to factors such as viewing angle, light intensity, and worker distraction, leading to missed or incorrect judgments. Workers estimate safe distances visually, lacking objective quantitative indicators, resulting in large errors and high risks. While some scenarios attempt to apply virtual reality technology to live-line work, high-precision alignment between the virtual scene and the actual train roof environment has not been achieved, and it cannot adapt to dynamically changing tool lengths and environmental factors. Regarding positioning, existing solutions mostly rely on UWB and IMU fusion positioning; the dense metal on the train roof and strong signal reflection interference easily lead to positioning drift, packet loss, and offset, causing warning failures. In terms of safety warnings, the physiological states of workers, such as heart rate, sweating, fatigue, and tension, are not monitored, and the warning sensitivity cannot be adaptively adjusted based on the worker's condition. Warnings are still issued based on conventional thresholds even under high-risk physiological states, easily leading to safety accidents. Furthermore, existing methods can only trigger warnings after an intrusion into a dangerous area, and cannot predict risks in advance, leaving workers with insufficient reaction time.

[0004] Therefore, there is an urgent need for a live-line work early warning method that can achieve intuitive early warning of dangerous areas, reliable verification of multiple positioning, adaptive early warning of physiological state, and early prediction of risks, in order to solve the shortcomings of the existing technologies mentioned above. Summary of the Invention

[0005] To address the problem mentioned in the background art regarding the inability to predict risks in advance for live-line work in hazardous areas, this invention provides a virtual reality-based early warning method for live-line work on high-voltage equipment of high-speed trains.

[0006] The above-mentioned objective of this application is achieved through the following technical solution:

[0007] A method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality includes the following steps:

[0008] Collect three-dimensional data of the high-voltage equipment on the roof of the high-speed train and construct a one-to-one scale virtual simulation model;

[0009] A dynamic safety distance envelope is generated based on real-time voltage levels, environmental parameters, and tool length.

[0010] The spatial location data of workers and tools are collected in real time using multiple positioning technology and mapped onto the virtual simulation model in real time.

[0011] In a virtual environment, the distance between the workers and tools and the safety distance envelope is calculated, collision detection is performed, and risk prediction is made based on velocity vector and motion trend analysis.

[0012] Based on the collision detection results and risk prediction results, a graded and visualized early warning signal is output.

[0013] By adopting the above technical solution, firstly, three-dimensional data of the high-voltage equipment on the roof of the EMU is collected to construct a one-to-one scale virtual simulation model, providing a precise spatial reference for subsequent positioning mapping and collision detection. Then, a dynamic safety distance envelope is generated based on real-time voltage levels, environmental parameters, and tool lengths, solving the problem that traditional fixed safety distances cannot adapt to environmental changes and tool differences. Next, multi-positioning technology is used to collect the spatial position data of the operators and tools in real time and map it into the virtual model, achieving high-precision alignment between the virtual scene and the real roof environment. Then, collision detection is performed in the virtual environment, and risk prediction is made based on velocity vector and motion trend analysis, issuing early warnings before the operators actually touch the dangerous area. Finally, based on the detection and prediction results, graded and visualized early warning signals are output, enabling operators to intuitively perceive the danger boundary and react in a timely manner, realizing the transformation from passive protection to active early warning.

[0014] In a preferred embodiment, this application can be further configured as follows: the acquisition of three-dimensional data of the high-voltage equipment on the roof of the EMU and the construction of a one-to-one scale virtual simulation model specifically includes:

[0015] Laser 3D scanning technology was used to scan the roof of the EMU to obtain point cloud data;

[0016] Construct a 1:1 scale virtual simulation model that includes the vehicle body, pantograph, high-voltage disconnector, and insulators, and ensure that the size error between the virtual simulation model and the physical equipment is controlled within the preset accuracy range.

[0017] The virtual simulation model is preloaded into the backend processing unit of the AR glasses or VR system worn by the operator.

[0018] By adopting the above technical solution, laser 3D scanning technology is used to scan the roof of the EMU to obtain point cloud data, and a 1:1 scale virtual simulation model including the car body, pantograph, high-voltage disconnect switch and insulators is constructed. This ensures that the size error between the virtual simulation model and the physical equipment is controlled within a preset accuracy range. The virtual simulation model is then preloaded into the back-end processing unit of the AR glasses or VR system worn by the operators. Laser 3D scanning technology can obtain high-precision point cloud data, ensuring that the virtual model and the physical equipment are highly consistent in geometric dimensions and spatial position. This provides an accurate reference coordinate system for the subsequent spatial position mapping of operators and tools. At the same time, preloading the model into the back-end processing unit of the AR glasses or VR system ensures the response speed of real-time positioning mapping and collision detection.

[0019] In a preferred embodiment, this application can be further configured such that: the generation of a dynamic safety distance envelope based on real-time voltage levels, environmental parameters, and tool length specifically includes:

[0020] Get the current relative air humidity and air pressure, and look up the corresponding environmental insulation withstand voltage correction factor;

[0021] Obtain the length of the tool held by the worker;

[0022] Calculate the corrected minimum safe distance, which is equal to the base safe distance multiplied by the environmental correction factor, plus the tool length;

[0023] Using the surface of the live part of the high-voltage equipment as a reference, the modified minimum safe distance is extended outward to form a three-dimensional closed envelope surface, which serves as a dangerous area where work is prohibited.

[0024] By adopting the above technical solution, the current relative air humidity and air pressure are obtained, the corresponding environmental insulation withstand voltage correction coefficient is queried, the length of the tool held by the operator is obtained, and the corrected minimum safe distance is calculated. The corrected minimum safe distance is equal to the basic safe distance multiplied by the environmental correction coefficient plus the tool length. Finally, the corrected minimum safe distance is extended outward from the surface of the live part of the high-voltage equipment to form a three-dimensional closed envelope as the danger zone where work is prohibited. The influence of environmental factors (humidity, air pressure) and the length of the working tool on the safe distance is comprehensively considered. Compared with the fixed safe distance, it has stronger environmental adaptability and scenario specificity, and can ensure sufficient safety margin under different working conditions.

[0025] In a preferred embodiment, this application can be further configured as follows: the real-time acquisition of spatial location data of workers and tools through multi-positioning technology, and the real-time mapping of this data to the virtual simulation model, specifically includes:

[0026] The absolute spatial location of workers and tools is obtained through ultra-wideband positioning technology;

[0027] The positioning deviation caused by high-speed motion is compensated by the inertial measurement unit;

[0028] The AR glasses' front-facing camera collects feature points on the roof of the vehicle, and visual synchronous positioning and mapping technology is used to calculate the relative pose.

[0029] The results of ultra-wideband positioning, inertial measurement unit and visual positioning are fused and verified, outliers are removed and drift is corrected, and if any source is abnormal, the other two sources will fill the gap.

[0030] The real-time position is mapped to the coordinate system of the virtual simulation model using a coordinate transformation matrix to generate a virtual avatar.

[0031] By adopting the above technical solution, the absolute spatial position of the operators and tools is obtained through ultra-wideband (UWB) positioning technology. The positioning deviation caused by high-speed movement is compensated by inertial measurement unit (IMU). The front-facing camera of AR glasses collects feature points on the roof and calculates the relative pose using visual simultaneous localization and mapping (SLAM) technology. Then, the three results of UWB, IMU and visual positioning are fused and verified, outliers are eliminated and drift is corrected. If any source is abnormal, the other two sources will fill the gap. Finally, the real-time position is mapped to the coordinate system of the virtual simulation model using a coordinate transformation matrix to generate a virtual avatar. The three positioning technologies complement each other and cross-validate each other, effectively suppressing signal reflection interference and multipath effects in the dense metal environment of the EMU roof, ensuring continuous and usable positioning, and providing a reliable position data foundation for collision detection.

[0032] In a preferred embodiment, this application can be further configured such that: calculating the distance between the operator and tools and the safety distance envelope in a virtual environment, and performing collision detection and risk prediction, specifically includes:

[0033] Real-time calculation of the shortest distance between the endpoints of workers and tools and the safe distance envelope;

[0034] Collision detection algorithms are used to determine whether there is an intersection or if the distance is less than a safety threshold.

[0035] Risk prediction is performed, which includes calculating the velocity vector and motion trend of the operator and tool endpoints, and predicting whether they will collide with the safe distance envelope within a preset time.

[0036] Based on the predicted movement trends, early warning signals are triggered in advance.

[0037] By adopting the above technical solution, the shortest distance between the endpoints of the operator and tools and the safe distance envelope is calculated in real time in the virtual environment. The collision detection algorithm is used to determine whether there is an intersection or the distance is less than the safe threshold. At the same time, the velocity vector and motion trend of the endpoints of the operator and tools are calculated to predict whether they will collide with the safe distance envelope within a preset time and to trigger an early warning signal in advance based on the motion trend prediction results. On the basis of traditional collision detection, a risk prediction mechanism based on velocity vector and motion trend is introduced, which can identify the intrusion behavior that is about to occur in advance, give the operator sufficient reaction time, and realize the leap from post-event alarm to pre-event prediction.

[0038] In a preferred embodiment, this application can be further configured such that the risk prediction mechanism specifically includes:

[0039] Real-time continuous position data of the operator's hands and tool ends are collected to generate motion trajectories;

[0040] Calculate the velocity and acceleration vectors based on the motion trajectory, and analyze the motion trend;

[0041] The probability of intrusion risk is calculated based on the velocity vector, acceleration vector, and distance between the current position and the safe distance envelope.

[0042] When the probability of intrusion exceeds a preset threshold, a pre-judgment warning is triggered before the operator actually touches the safe distance envelope.

[0043] By adopting the above technical solution, continuous position data of the operator's hands and tool ends are collected in real time to generate a motion trajectory. Based on the motion trajectory, velocity vectors and acceleration vectors are calculated to analyze the motion trend. Then, based on the velocity vector, acceleration vector, and the distance between the current position and the safe distance envelope, the intrusion risk probability is calculated. When the intrusion risk probability exceeds a preset threshold, a pre-judgment warning is triggered before the operator actually touches the safe distance envelope. Through the joint analysis of motion trajectory, velocity vector, and acceleration vector, it is possible to accurately determine whether the operator has a continuous tendency to approach the danger zone, avoiding false alarms caused by instantaneous position fluctuations, and ensuring that the actual intrusion behavior can be identified in time before it occurs.

[0044] In a preferred embodiment, this application can be further configured such that: the step of outputting a graded visual early warning signal based on the collision detection result and the risk prediction result specifically includes:

[0045] The tiered visual early warning signals include proximity warning status and danger alarm status;

[0046] When the distance to the safety envelope is greater than the first threshold but less than the second threshold, it is determined to be a proximity warning state. A semi-transparent halo is rendered around the charged body, and the distance value is displayed at the tip of the tool.

[0047] When the distance to the safety envelope is less than the first threshold or when intrusion has occurred, it is determined to be a dangerous alarm state, and a bright red virtual wall is displayed in the field of vision, accompanied by an audible alarm and vibration feedback.

[0048] By adopting the above technical solution, a two-level warning state is set according to the distance between the operator and tools and the safe distance envelope: when the distance is greater than the first threshold but less than the second threshold, it is determined to be a proximity warning state, a semi-transparent halo is rendered around the live part, and the distance value is displayed at the tip of the tool; when the distance is less than the first threshold or has been intruded, it is determined to be a danger alarm state, a bright red virtual wall is displayed in the field of vision, accompanied by sound alarm and vibration feedback; the graded warning mechanism enables the operator to judge the degree of danger according to different warning signals, and can actively adjust the operation in the proximity warning stage to avoid entering the danger alarm state.

[0049] In a preferred embodiment, this application can be further configured such that the hierarchical visual early warning signal includes:

[0050] In the near-warning state, a semi-transparent yellow halo is rendered around the live parts of the high-voltage equipment. The transparency of the yellow halo decreases as the distance decreases, and the distance value between the tool tip and the safe distance envelope is displayed in real time in the operator's field of vision.

[0051] In the state of danger alarm, a bright red virtual wall is generated in the field of vision. The surface of the virtual wall has a texture flow animation. The texture flow direction is in the direction of intrusion. The area of ​​the virtual wall in the field of vision expands as the degree of danger increases.

[0052] The red virtual wall, together with the graded alarm sounds emitted by the bone conduction headphones and the graded vibration feedback from the wearable device, constitutes a multimodal early warning output.

[0053] By adopting the above technical solution, in the approach warning state, a semi-transparent yellow halo is rendered around the live parts of the high-voltage equipment. The transparency of the yellow halo decreases as the distance decreases, and the distance value between the tool tip and the safe distance envelope is displayed in real time in the operator's field of vision. In the danger alarm state, a bright red virtual wall is generated in the field of vision, with a textured flowing animation on the surface and the flow direction pointing in the direction of intrusion. The area of ​​the virtual wall in the field of vision expands as the degree of danger increases. At the same time, it is combined with the graded alarm sound emitted by the bone conduction headphones and the graded vibration feedback of the wearable device to form a multimodal warning output. The gradual change in the transparency of the yellow halo provides intuitive distance perception, the textured flowing animation of the red virtual wall indicates the direction of intrusion, and the multimodal output of bone conduction headphones and vibration feedback ensures that the operator can perceive danger signals through multiple channels of vision, hearing and touch at the same time, and will not miss the warning information even in noisy environments.

[0054] In a preferred embodiment, this application can be further configured to include a physiological state perception step, specifically including:

[0055] The heart rate, skin conductivity, and posture characteristics of workers are collected in real time using wearable devices.

[0056] Based on the collected data, the stress and fatigue levels of the workers are assessed, and a physiological risk coefficient is output.

[0057] The dynamic safety distance envelope is corrected using the physiological risk coefficient, and the safety distance envelope expands as the physiological risk coefficient increases;

[0058] The physiological risk coefficient is used to adjust the prediction time and early warning sensitivity of collision detection. When the operator is under tension or fatigue, the prediction time is shortened and the early warning advance is increased.

[0059] The physiological risk coefficient is also used to dynamically adjust the intensity of the graded visual early warning signal.

[0060] By adopting the above technical solution, wearable devices are used to collect the heart rate, skin conductivity, and posture characteristics of workers in real time. Based on the collected data, the stress and fatigue levels of workers are judged and a physiological risk coefficient is output. Then, the physiological risk coefficient is used to correct the dynamic safety distance envelope (the safety distance envelope expands as the physiological risk coefficient increases), adjust the prediction time and warning sensitivity of collision detection (shorten the prediction time and increase the warning lead when workers are stressed or fatigued), and dynamically adjust the intensity of the graded visual warning signal. This mechanism incorporates the physiological state of workers into the safety warning system. When workers are stressed or fatigued, the safety boundary is automatically expanded and the warning sensitivity is increased, providing stronger safety protection in high-risk physiological states.

[0061] In a preferred embodiment, this application can be further configured such that the dynamic adjustment of the physiological risk coefficient to the early warning signal specifically includes:

[0062] When the physiological risk coefficient exceeds the first threshold, the trigger distance for the near warning state is increased to a preset multiple of the default value, the initial transparency of the yellow halo is reduced, and the texture flow speed of the virtual wall is accelerated.

[0063] When the physiological risk coefficient exceeds the second threshold, the trigger distance of the danger alarm state is increased, the red virtual wall flashes at a preset frequency, the alarm sound of the bone conduction headphones is accelerated, and the vibration intensity of the wearable device is enhanced, forming a multimodal strong warning output that forces the operator to stop moving.

[0064] By adopting the above technical solution, a graded enhanced early warning system is implemented based on the severity of the physiological risk coefficient: when the physiological risk coefficient exceeds the first threshold, the trigger distance for approaching the warning state is increased to a preset multiple of the default value, the initial transparency of the yellow halo decreases, and the texture flow speed of the virtual wall increases; when the physiological risk coefficient exceeds the second threshold, the trigger distance for the danger alarm state is increased, the red virtual wall flashes at a preset frequency, the alarm sound of the bone conduction headphones is accelerated, and the vibration intensity of the wearable device is enhanced, forming a multimodal strong warning output that forces the worker to stop moving; this mechanism dynamically adjusts the warning trigger conditions and warning intensity according to the worker's physiological risk level, providing earlier and stronger warnings when the worker is in a state of high tension or high fatigue, forcibly intervening in dangerous operations, and fully embodying the people-oriented safety protection concept.

[0065] In summary, this application includes at least one of the following beneficial technical effects:

[0066] 1. Hazard visualization and intuitive boundary perception: Virtual reality technology is used to visualize the invisible safety distance as a visible virtual warning wall, allowing workers to intuitively perceive the danger boundary, completely solving the problem of invisible safety distance and reducing the risk of accidental contact;

[0067] 2. Dynamic Adaptation for Enhanced Safety: The size of the safety envelope is dynamically adjusted based on real-time voltage levels, ambient humidity, and tool length to adapt to different work scenarios and environmental conditions, significantly improving safety compared to a fixed safety distance.

[0068] 3. Triple positioning is accurate and reliable: It adopts UWB+IMU+vision triple positioning cross-check to suppress multipath interference in the metal environment of the vehicle roof. If any source is abnormal, the other two sources will fill in to ensure that the positioning is continuous and usable without loss. The overall positioning accuracy is better than 2cm.

[0069] 4. Proactive prediction and reserve reaction time: By analyzing the hand movement trajectory, velocity vector and acceleration vector of the operator, the risk of intrusion can be predicted in advance, and an early warning can be triggered before the action occurs, so as to reserve sufficient reaction time for the operator and achieve proactive protection;

[0070] 5. Tiered early warning and multimodal feedback: Combining a two-level visual early warning system with yellow halo proximity warning and red virtual wall danger alarm, along with bone conduction headphone sound alarm and wearable device vibration feedback, ensures that the early warning signal can be perceived by the workers in a timely manner;

[0071] 6. Physiological Adaptation, Enhanced Protection in High-Risk Conditions: Real-time monitoring of physiological states such as heart rate, sweating, fatigue / stress, etc., automatically corrects the safety distance envelope, adjusts the warning lead time, and enhances the virtual warning effect, providing earlier and stronger safety protection in high-risk physiological states;

[0072] 7. Adaptable to work without affecting efficiency: In AR glasses assisted mode, the virtual warning wall is superimposed in a semi-transparent form, which does not obstruct the working field of vision, ensuring safety without affecting the efficiency of live-line work, and adapting to the actual needs of live-line maintenance of EMU trains. Attached Figure Description

[0073] Figure 1 This is a flowchart of an embodiment of a live-line working early warning method for high-voltage equipment on a high-speed train based on virtual reality, as described in this application.

[0074] Figure 2 This is a flowchart illustrating the implementation of step S10 in an embodiment of a live-line working early warning method for high-voltage equipment on a high-speed train based on virtual reality in this application.

[0075] Figure 3 This is a flowchart illustrating the implementation of step S20 in an embodiment of a live-line working early warning method for high-voltage equipment on a high-speed train based on virtual reality in this application.

[0076] Figure 4 This is a schematic diagram illustrating the dynamic safety distance envelope generation principle.

[0077] Figure 5 This is a flowchart illustrating the implementation of step S30 in an embodiment of a live-line working early warning method for high-voltage equipment on a high-speed train based on virtual reality in this application.

[0078] Figure 6 This is a logic block diagram of real-time mapping and collision detection;

[0079] Figure 7 This is a flowchart illustrating the implementation of step S40 in an embodiment of a live-line working early warning method for high-voltage equipment on a high-speed train based on virtual reality in this application.

[0080] Figure 8 This is a flowchart illustrating the implementation of step S50 in an embodiment of a live-line working early warning method for high-voltage equipment on a high-speed train based on virtual reality in this application.

[0081] Figure 9 This is a flowchart illustrating the implementation of step S60 in an embodiment of a live-line working early warning method for high-voltage equipment on a high-speed train based on virtual reality, as described in this application. Detailed Implementation

[0082] The following is in conjunction with the appendix Figure 1-9 This application will be described in further detail.

[0083] In one embodiment, such as Figure 1 As shown, this application discloses a method for early warning of live-line work on high-voltage equipment of EMU trains based on virtual reality, which specifically includes the following steps:

[0084] S10: Collect three-dimensional data of the high-voltage equipment on the roof of the EMU and construct a one-to-one virtual simulation model.

[0085] In this embodiment, the 1:1 scale virtual simulation model refers to a digital three-dimensional model that is completely consistent with the actual high-voltage equipment on the roof of a high-speed train in terms of geometric dimensions, spatial location, and structural relationships. The virtual simulation model is used to provide accurate spatial references for workers in augmented reality or virtual reality environments and is the basis for subsequent positioning mapping and collision detection. The high-voltage equipment on the roof of the high-speed train mainly includes key components such as pantographs, high-voltage disconnect switches, and insulators. These components are the main sources of risk in live-line work, so they need to be accurately modeled.

[0086] Specifically, laser 3D scanning technology is used to scan the roof of the EMU to obtain high-precision point cloud data. Laser 3D scanning can collect the geometric information of the roof equipment with millimeter-level precision, ensuring that the size error between the virtual model and the physical equipment is controlled within a preset accuracy range (such as ±2mm). After obtaining the point cloud data, a 1:1 scale virtual simulation model including the car body, pantograph, high-voltage disconnect switch and insulators is constructed using point cloud reconstruction algorithms. Finally, the completed virtual simulation model is preloaded into the back-end processing unit of the AR glasses or VR system worn by the operators to ensure the response speed of subsequent real-time positioning mapping and collision detection.

[0087] S20: Generate a dynamic safety distance envelope based on real-time voltage level, environmental parameters, and tool length.

[0088] In this embodiment, the dynamic safety distance envelope refers to a three-dimensional closed virtual surface formed by extending a certain distance outward from the surface of the live part of the high-voltage equipment. This surface defines the boundary of the dangerous area where work is prohibited. The value of the dynamic safety distance is not fixed, but is calculated based on the real-time voltage level, environmental parameters, and the length of the tools held by the operator. The environmental parameters include the relative humidity and air pressure of the current working environment, because these factors affect the insulation withstand voltage characteristics of the air. The higher the humidity and the lower the air pressure, the worse the air insulation performance, and the greater the required safety distance.

[0089] Specifically, first, the current relative humidity and air pressure are obtained, and the corresponding environmental correction coefficient Kenv is obtained by consulting the preset environmental insulation withstand voltage correction coefficient table. Then, the length Ltool of the operator's handheld tool is obtained; the corrected minimum safe distance is calculated using the formula: Dsafe = Dbase × Kenv + Ltool, where Dbase is the basic safe distance under standard working conditions; finally, using the surface of the live part of the high-voltage equipment as a reference, the corrected minimum safe distance Dsafe is extended outward to form a three-dimensional closed envelope surface, which serves as the danger zone where work is prohibited; this envelope surface is presented in a visual form in the virtual scene, allowing operators to intuitively perceive the danger boundary.

[0090] S30: Real-time spatial location data of workers and tools are collected using multi-positioning technology and mapped onto the virtual simulation model in real time.

[0091] In this embodiment, multi-positioning technology refers to the fusion of ultra-wideband positioning, inertial measurement unit (IMU) and visual simultaneous positioning and modeling. Figure 3 The system employs a multi-source positioning scheme with three positioning methods. Ultra-wideband positioning is an indoor positioning technology based on ultra-wideband pulse signals, capable of providing spatial absolute position with centimeter-level accuracy. An inertial measurement unit (IMU) measures the acceleration and angular velocity of an object using accelerometers and gyroscopes to compensate for positioning deviations caused by high-speed motion. Visual synchronous positioning and mapping technology collects feature points on the roof of the vehicle using the front-facing camera of AR glasses to calculate relative pose. The three positioning methods complement and cross-validate each other, effectively suppressing signal reflection interference and multipath effects in the dense metal environment of the train roof. If any one source fails, the other two sources will compensate, resulting in an overall positioning accuracy better than 2cm.

[0092] Specifically, ultra-wideband positioning technology is used to obtain the absolute spatial position of the worker's head, key joints of the hands, and tool endpoints. Inertial measurement units are used to compensate for positioning deviations caused by high-speed movement in real time. The front-facing camera of AR glasses collects feature points on the vehicle roof (such as insulators, corners of high-voltage equipment, positioning marks, etc.). Visual synchronous positioning and mapping technology is used to calculate the relative pose. Kalman filtering and consistency verification are performed on the three positioning results to fuse them, remove outliers and correct drift. The real-time position is mapped to the coordinate system of the virtual simulation model using a coordinate transformation matrix to generate a virtual avatar that is synchronized with the worker's movements in real time.

[0093] S40: Calculate the distance between the operator and tools and the safety distance envelope in the virtual environment, perform collision detection, and make risk prediction based on velocity vector and motion trend analysis.

[0094] In this embodiment, collision detection refers to the process of calculating the shortest distance between the endpoints of the worker and tools and the safe distance envelope in real time, and using a collision detection algorithm to determine whether there is an intersection or the distance is less than the safe threshold; risk prediction refers to, based on the current collision detection, analyzing the velocity vectors and motion trends of the worker and tool endpoints to predict whether they will collide with the safe distance envelope within a preset time, issuing an early warning in advance, and allowing the worker reaction time, thus upgrading from passive alarm to proactive prediction.

[0095] Specifically, the system calculates the shortest distance between the worker and tool tip (such as the tip of a handheld tool) and the safety distance envelope in real time. Collision detection algorithms (such as the AABB bounding box algorithm or ray method) are used to determine whether there is an intersection or the distance is less than the safety threshold. Continuous position data of the worker's hand and tool tip are collected to generate a motion trajectory. Velocity and acceleration vectors are calculated based on the motion trajectory, and the motion trend is analyzed. Based on the velocity vector, acceleration vector, and the distance between the current position and the safety distance envelope, the probability of intrusion risk is calculated. When the probability of intrusion risk exceeds a preset threshold, a pre-judgment warning is triggered before the worker actually touches the safety distance envelope.

[0096] S50: Based on collision detection results and risk prediction results, output graded and visualized early warning signals.

[0097] In this embodiment, the graded visual early warning refers to presenting different levels and colors of virtual warning effects in the AR / VR field of view based on the distance between the operator and tools and the safe distance envelope and the degree of intrusion risk. The graded early warning is divided into two levels: proximity warning and danger alarm, which enables the operator to judge the degree of danger according to the warning level and proactively adjust the operation in the early stage.

[0098] Specifically, when the distance to the safe envelope is greater than the first threshold but less than the second threshold (e.g., greater than 30cm but less than 100cm), it is determined to be a proximity warning state. A semi-transparent yellow halo is rendered around the live parts of the high-voltage equipment. The transparency of the yellow halo decreases as the distance decreases, allowing the operator to intuitively perceive the trend of distance change. The distance value between the tool tip and the safe distance envelope is displayed in real time in the operator's field of vision. When the distance to the safe envelope is less than the first threshold (e.g., less than 30cm) or has been intruded, it is determined to be a danger alarm state. A bright red virtual wall is generated in the field of vision. The surface of the virtual wall has a textured flowing animation. The texture flowing in the direction of intrusion, and the area of ​​the virtual wall in the field of vision expands as the degree of danger increases. At the same time, in conjunction with the graded alarm sound emitted by the bone conduction headphones and the graded vibration feedback of the wearable device, a multimodal warning output of visual, auditory, and tactile senses is formed to ensure that the operator will not miss the warning information even in noisy environments.

[0099] In one embodiment, such as Figure 2 As shown, in step S10, which involves collecting the three-dimensional data of the high-voltage equipment on the roof of the EMU and constructing a one-to-one scale virtual simulation model, the steps include:

[0100] S11: Laser 3D scanning technology is used to scan the roof of the EMU to obtain point cloud data.

[0101] In this embodiment, laser 3D scanning technology refers to the technology of obtaining a massive set of 3D coordinate points on the target surface by emitting laser pulses to the target object and receiving the reflected signals, and calculating the distance between the target point and the scanner based on the laser flight time or phase difference; point cloud data refers to a data set consisting of millions of 3D spatial coordinate points, each point containing three coordinate components of X, Y, and Z, which can accurately describe the surface geometry of the target object.

[0102] Specifically, the operators used laser 3D scanners to conduct a comprehensive scan of the train's roof. During the scan, the laser scanners collected point cloud data from multiple angles and positions on the roof and the surface of high-voltage equipment to ensure coverage of all key parts, including the top outline of the train body, the pantograph and its base, high-voltage disconnect switches, insulators, and other high-voltage equipment. After the scan was completed, the acquired raw point cloud data was preprocessed, including filtering and denoising (removing outliers and noise points), registration and fusion (unifying the scan data from multiple stations to the same coordinate system), and data simplification (reducing data redundancy while ensuring accuracy). Finally, a high-precision complete point cloud model of the roof equipment was generated.

[0103] S12: Construct a 1:1 scale virtual simulation model including the vehicle body, pantograph, high-voltage disconnector, and insulators, ensuring that the dimensional error between the virtual simulation model and the physical equipment is controlled within the preset accuracy range.

[0104] In this embodiment, the virtual simulation model refers to a three-dimensional digital model reconstructed based on point cloud data; the one-to-one scale means that the geometric dimensions of the virtual model are completely consistent with those of the physical device; the preset accuracy range refers to the size error control target between the virtual model and the physical device, which is usually required to be controlled at the millimeter level (e.g., ±2mm) to ensure the accuracy of subsequent positioning mapping and collision detection.

[0105] Specifically, based on the point cloud data obtained in step S11, a 3D model is reconstructed. First, the preprocessed point cloud is reconstructed into a surface to generate a triangular mesh model. Then, the model is refined and optimized according to the structural characteristics of each component of the high-voltage equipment. During the reconstruction process, the following key equipment models are constructed: vehicle body (as a spatial reference and coordinate system), pantograph (the core component of the high-voltage equipment on the vehicle roof, which is closest to the contact network when the pantograph is raised), high-voltage disconnect switch (a key operating component), and insulator (a high-voltage insulating component whose feature points can be used as visual positioning markers). After the model is constructed, the model is dimensionally verified using standard size reference objects (such as calibration blocks of known size or equipment design drawings) to ensure that the dimensional error between the model and the physical equipment is controlled within the preset accuracy range.

[0106] S13: Preload the virtual simulation model into the backend processing unit of the AR glasses or VR system worn by the operator.

[0107] In this embodiment, preloading refers to loading the virtual simulation model into the background processing unit of the AR glasses or VR system before the operator starts live-line work, so as to avoid lag or delay caused by model loading delay during the operation. The background processing unit of the AR glasses or VR system refers to the processor or computing module responsible for core operations such as image rendering, coordinate calculation, and collision detection. It is usually an embedded graphics processor or edge computing device.

[0108] Specifically, the completed virtual simulation model is exported as a standard 3D model format and transmitted to the backend processing unit of the AR glasses or VR system via a data interface. The backend processing unit stores and manages the model and automatically loads it into video memory or system memory after system startup. When workers wearing AR glasses or VR systems enter the work area, the backend processing unit renders the virtual simulation model in real time based on the current equipment posture and viewpoint information and overlays it onto the real environment. Since the model is pre-loaded, the entire rendering and display process can be completed in milliseconds, ensuring the smoothness and real-time performance of the virtual scene display. In addition, the pre-loaded model also contains semantic information of each high-voltage device (such as equipment type, rated voltage, etc.), providing a data foundation for subsequent dynamic safety distance calculation and collision detection.

[0109] In one embodiment, such as Figure 3 As shown, in step S20, namely, generating a dynamic safety distance envelope based on real-time voltage levels, environmental parameters, and tool length, the steps include:

[0110] S21: Obtain the current relative air humidity and air pressure, and query the corresponding environmental insulation withstand voltage correction coefficient.

[0111] In this embodiment, the environmental insulation withstand voltage correction factor refers to the coefficient value used to correct the influence of environmental factors on air insulation performance, which is usually obtained by looking up a standard environmental parameter reference table; relative air humidity refers to the percentage of actual water vapor content in the air to saturated water vapor content at the same temperature. The higher the humidity, the more significant the decrease in air insulation performance; air pressure refers to the atmospheric pressure value at the work site. The lower the air pressure, the thinner the air, and the lower the insulation withstand voltage.

[0112] Specifically, the smart terminals worn by the workers (such as AR glasses or handheld terminals) collect real-time relative humidity and air pressure data at the current work location through built-in environmental sensors, or obtain environmental parameters by connecting to the on-site meteorological monitoring station via wireless network. After obtaining the relative humidity and air pressure, the system looks up the data in a preset environmental correction coefficient lookup table. This lookup table is based on insulation coordination standards in the field of electrical engineering, and different humidity and air pressure combinations correspond to different correction coefficients. For example, when the relative humidity is 80%, the humidity correction coefficient is 1.1; when the air pressure is 90 kPa, the air pressure correction coefficient is 1.05. The comprehensive environmental correction coefficient, Kenv, is the product of the two values ​​or calculated according to preset rules.

[0113] S22: Obtain the length of the tool held by the operator.

[0114] In this embodiment, the length of the tool held by the operator refers to the effective length of the tool currently being used by the operator (such as a voltage detector, grounding rod, wrench, screwdriver, etc.) in the working direction, that is, the distance from the hand holding position of the tool to the tip of the tool (the end closest to the live conductor).

[0115] Specifically, the system obtains tool length in the following ways: First, an active or passive identification tag (such as an RFID tag or QR code) is pre-installed on the tool to store the tool's model and standard length parameters. When the operator uses the tool, the AR glasses identify the tag on the tool through a camera, automatically reading the tool's type and length information. A posture sensor can also be installed on the tool to detect its extension status and angle in real time. For telescopic tools, the system dynamically calculates the current effective length based on sensor data. If the tool is not equipped with an identification tag, the operator can also manually input the tool type or length value through the AR glasses' gesture or voice interaction interface. The system uses the obtained tool length Ltool as an important parameter for subsequent safe distance calculations.

[0116] S23: Calculate the corrected minimum safe distance, which is equal to the basic safe distance multiplied by the environmental correction factor, plus the tool length.

[0117] In this embodiment, the basic safety distance refers to the minimum safety distance determined according to the voltage level under standard environmental conditions (such as temperature 20°C, relative humidity 50%, and standard atmospheric pressure 101.325 kPa). This value is usually specified by national or industry standards (such as the "Electrical Safety Work Regulations"). The modified minimum safety distance is the actual safety distance after comprehensively considering environmental factors and tool length based on the basic safety distance.

[0118] Specifically, the system first looks up the corresponding basic safety distance Dbase from a preset safety distance standard table based on the real-time voltage level of the current high-voltage equipment. For example, for a 27.5kV EMU traction power supply system, the basic safety distance might be 0.8m. Then, the environmental correction factor Kenv obtained in S21 is multiplied by the basic safety distance Dbase to obtain the environmentally corrected safety distance. Finally, the tool length Ltool obtained in S22 is added to obtain the corrected minimum safety distance Dsafe = Dbase × Kenv + Ltool. For example, if Dbase is 0.8m, Kenv is 1.15, and Ltool is 0.2m, then Dsafe = 0.8 × 1.15 + 0.2 = 1.12m.

[0119] The generation logic of the dynamic safe distance envelope is as follows: Figure 4 As shown, the base safety distance Dbase, together with the environment correction factor Kenv and the tool length Ltool, determines the size of the final safety distance Dsafe.

[0120] S24: Using the surface of the live part of the high-voltage equipment as a reference, extend the modified minimum safety distance outward to form a three-dimensional closed envelope surface, which serves as a dangerous area where work is prohibited.

[0121] In this embodiment, the surface of a live part refers to the outer surface of a conductive part in a high-voltage equipment that carries operating voltage, such as the pantograph's sliding plate, the contacts of a high-voltage disconnector, and the high-voltage end fittings of an insulator; the three-dimensional closed envelope surface refers to a continuous closed three-dimensional curved surface formed by extending outward a certain distance from the surface of the live part, which separates the dangerous area from the safe area; the dangerous area refers to the space in which workers or tools are not allowed to enter, and there is a risk of electric shock once the worker enters the area.

[0122] Specifically, the system uses the three-dimensional surface of each high-voltage equipment's live part in the virtual simulation model as a reference, and uniformly extends outward along its normal direction to generate a continuous three-dimensional closed envelope surface, calculated by S23 and corrected by the minimum safe distance Dsafe. For equipment with complex shapes (such as pantographs), the envelope surface is generated using a distance field-based method, that is, the distance from each point to the equipment surface is calculated in the space around the equipment, and the isosurface with a distance equal to Dsafe is extracted as the envelope surface. The generated three-dimensional envelope surface is displayed in the AR / VR field of view as a semi-transparent colored curved surface, allowing operators to intuitively perceive the boundary of the danger zone. The area inside the envelope surface is a danger zone where work is prohibited, and no part of the operator's body or handheld tools may enter this area. The area outside the envelope surface is a safe zone where operators can move freely. This envelope surface is dynamically updated in real time according to changes in voltage level, environmental parameters, and tool length, ensuring accurate safety boundary prompts under different operating conditions.

[0123] In one embodiment, such as Figure 5 As shown, in step S30, which involves real-time acquisition of spatial location data of workers and tools using multi-positioning technology and real-time mapping of this data to the virtual simulation model, the steps include:

[0124] S31: Obtain the absolute spatial position of personnel and tools through ultra-wideband positioning technology.

[0125] In this embodiment, ultra-wideband positioning technology is a wireless positioning technology based on nanosecond-level narrow pulse signals. It calculates the distance by measuring the flight time of the signal between the positioning tag and the base station, and achieves precise positioning by using the intersection of the distances of multiple base stations. The absolute spatial position refers to the three-dimensional spatial coordinates (X, Y, Z) in the global coordinate system. Ultra-wideband technology has the characteristics of strong anti-multipath interference capability, strong penetration and high positioning accuracy, and can achieve centimeter-level position positioning in the metal-dense environment of the roof of a high-speed train.

[0126] Specifically, at least four ultra-wideband (UWB) positioning base stations are deployed around the work area. These base stations are connected to the back-end processing unit via wired or wireless means. Workers wear UWB positioning tags (which can be integrated into safety helmets, AR glasses, or wearable devices), and miniature UWB positioning tags are also installed on their work tools. The positioning tags transmit UWB pulse signals to the base stations at a preset frequency (e.g., 20Hz). After receiving the signal, the base station calculates the signal flight time. The back-end processing unit performs triangulation based on the ranging data from multiple base stations to obtain the absolute spatial position coordinates of the tag, with a positioning accuracy better than 10cm. Because UWB signals have good anti-multipath interference capabilities, they can maintain stable positioning performance in environments with metal reflections on vehicle roofs.

[0127] S32: Compensates for positioning deviations caused by high-speed motion through an inertial measurement unit.

[0128] In this embodiment, the inertial measurement unit is an attitude measurement device that integrates an accelerometer and a gyroscope. It can measure the acceleration and angular velocity of an object in three-dimensional space. The displacement change can be obtained by double integration of the acceleration. The positioning deviation caused by high-speed motion refers to the lag or deviation in position estimation caused by the signal sampling rate limitation or obstruction when the operator moves quickly.

[0129] Specifically, the inertial measurement unit (IMU) is integrated into the AR glasses worn by the operator and the positioning tag on the tool end, continuously acquiring three-axis acceleration and three-axis angular velocity data at a high sampling frequency (e.g., 200Hz). The background processing unit preprocesses the IMU data, including filtering, noise reduction, and zero-bias correction. When the ultra-wideband positioning signal is briefly interrupted due to obstruction or interference, the system uses the relative displacement calculated by the IMU as a supplement to the position estimate. After the signal is restored, the IMU data and the ultra-wideband data are fused using Kalman filtering to eliminate positioning lag and bias caused by high-speed movement, thereby improving the continuity and smoothness of positioning.

[0130] S33: Collect feature points on the roof of the vehicle using the front-facing camera of the AR glasses, and calculate the relative pose using visual synchronous positioning and mapping technology.

[0131] In this embodiment, visual synchronous positioning and mapping technology refers to the technology of acquiring environmental images through a camera, extracting feature points from the images and tracking their movement between consecutive frames, while estimating the camera's own motion trajectory and constructing an environmental map; roof feature points refer to markers in the roof environment that have stable and unique visual characteristics, such as the edge of an insulator's skirt, the corner of high-voltage equipment, bolt heads, pre-pasted QR code markings, etc.

[0132] Specifically, the front-facing camera of the AR glasses worn by the operators captures images of the rooftop environment in real time at a preset frame rate (e.g., 30fps). The back-end processing unit extracts features from the images, using ORB or SIFT feature extraction algorithms to identify key points in the images and match them with a pre-built visual feature map. Through feature point tracking and triangulation between consecutive frames, the relative pose of the camera is calculated. Visual positioning provides effective supplementary positioning information when the ultra-wideband signal is blocked by metal or interfered with by reflections, and can correct the cumulative errors of the ultra-wideband and IMU, further improving the robustness of positioning.

[0133] S34: Perform fusion verification of the three results of ultra-wideband positioning, inertial measurement unit and visual positioning, remove outliers and correct drift, and replace the other two sources when any one source is abnormal.

[0134] In this embodiment, fusion verification refers to the comprehensive processing of the output data from the three positioning sources. By comparing the consistency between the data from each positioning source, it is determined whether there are outliers or drift. Data fusion is then performed using complementary filtering or Kalman filtering algorithms. Outliers refer to positioning results that deviate significantly from the true position, usually caused by signal blockage, multipath effects, or sensor malfunctions. Drift refers to the trend of positioning results gradually deviating from the true position over time, usually caused by the cumulative error of the inertial measurement unit. When any source is abnormal, the other two sources will fill in the gap. This means that when the fusion verification mechanism determines that the data from a specific positioning source is abnormal, the system automatically switches to the other two positioning sources for fusion calculation to ensure continuous and uninterrupted positioning.

[0135] Specifically, the backend processing unit receives three types of data at a preset period (e.g., 50ms): ultra-wideband positioning coordinates, inertial measurement unit pose increments, and visual positioning pose, and inputs them into the Kalman filter fusion framework. The state vector of the Kalman filter includes three-dimensional spatial position, velocity, attitude angle, and sensor bias, while the observation vector is provided by the three positioning sources respectively. The filter determines the validity of each observation source by calculating the innovation (the difference between the observed value and the predicted value): if the innovation of a source exceeds a preset threshold for a duration exceeding a set tolerance time (e.g., 3 consecutive frames), the source is determined to be abnormal, and the filter automatically resets the observation weight of that source to zero, with the remaining two sources maintaining the positioning solution. When the source signal returns to normal and the innovation falls back to within the threshold, the filter gradually restores its weight. This fusion verification mechanism ensures the continuous availability of the positioning system in complex metal environments such as the roof of a high-speed train, avoiding positioning interruptions caused by single-source loss of lock.

[0136] S35: Use a coordinate transformation matrix to map the real-time position to the coordinate system of the virtual simulation model to generate a virtual avatar.

[0137] In this embodiment, the coordinate transformation matrix is ​​a 4×4 homogeneous transformation matrix, which includes a rotation matrix and a translation vector, and is used to transform the position coordinates in the positioning sensor coordinate system to the virtual simulation model coordinate system; the virtual avatar refers to the digital representation in the virtual simulation model that corresponds to the operator and tools, and its position and posture are kept synchronized with the real operator and tools in real time.

[0138] Specifically, during the system initialization phase, a rigid transformation relationship between the ultra-wideband positioning coordinate system and the virtual simulation model coordinate system is established through hand-eye calibration or standard reference objects (such as preset calibration points), and the coordinate transformation matrix T is solved. During the real-time operation phase, the background processing unit continuously receives the spatial absolute position coordinates of the operator's head, hands, and tool endpoints, multiplies them with the coordinate transformation matrix T, and obtains the corresponding positions in the virtual simulation model coordinate system. Then, based on the mapped position coordinates, a virtual avatar model that is synchronized with the operator's movements in real time is generated at the corresponding position in the virtual simulation model, including a 3D model representation of the head, hands, and tool endpoints. This virtual avatar is used for subsequent collision detection and distance calculation to ensure that the position mapping in the virtual world is consistent with the actual physical world, and the positioning data refresh frequency is not lower than the update frequency required for system collision detection. The real-time rendering of the virtual avatar allows the operator to see the positional relationship between their digital representation and the virtual safety envelope in the AR / VR field of view, and intuitively judge the current safety status.

[0139] The complete logic of the above real-time mapping and collision detection is as follows: Figure 6 As shown, in a virtual environment, collision detection and risk prediction are performed based on the spatial relationship between the real-time position of the virtual avatar and the safe distance envelope.

[0140] In one embodiment, such as Figure 7 As shown, in step S40, which involves calculating the distance between the workers and tools and the safety distance envelope in the virtual environment, and performing collision detection and risk prediction, the steps include:

[0141] S41: Calculate the shortest distance between the endpoints of workers and tools and the safe distance envelope in real time.

[0142] In this embodiment, the shortest distance refers to the closest distance from each key part (such as the head and hands) and tool tip (such as the tool tip) of the virtual avatar of the operator to the safety distance envelope. This distance value is used to determine whether the current operator and tool are close to or have entered the danger zone. The safety distance envelope is a three-dimensional closed surface. The calculation of the shortest distance is essentially the minimum Euclidean distance from a point in the computation space to the three-dimensional surface.

[0143] Specifically, the backend processing unit acquires the real-time coordinates of key parts of the virtual avatar (head center, left and right hands, tool endpoints) in the virtual simulation model coordinate system at a preset frequency (e.g., 30Hz). For each key point, the system calculates the shortest distance to all triangular faces on the safety distance envelope. The specific calculation method is as follows: traverse each triangular face on the envelope and calculate the projection point of the key point onto the plane of that face. If the projection point falls inside the face, the distance from the key point to the face is the perpendicular distance from the point to the plane; if the projection point falls outside the face, the distance is the distance from the key point to the nearest edge or vertex of the face. The minimum value among all faces is taken as the shortest distance from the key point to the safety distance envelope. The calculated distance value is updated in real time with centimeter-level accuracy and used as input data for subsequent collision detection and graded warning.

[0144] S42: Use a collision detection algorithm to determine whether there is an intersection or the distance is less than the safety threshold.

[0145] In this embodiment, the collision detection algorithm is a calculation method used to determine whether two objects in three-dimensional space are in contact or intersecting; the safety threshold refers to the preset minimum allowable safety margin. When the distance between the operator or tool and the safety distance envelope is less than the threshold, the system determines that there is a collision risk; the collision detection adopts a bounding box or distance field-based method, which can complete the spatial relationship judgment between the virtual avatar and the safety envelope in milliseconds.

[0146] Specifically, the system employs a hierarchical bounding box (BVH) collision detection algorithm or a collision detection method based on the signed distance field (SDF). For the BVH-based method, a bounding box tree is first constructed for each part of the virtual avatar, and another set of bounding box trees is constructed for the safe distance envelope. By quickly eliminating non-intersecting bounding box pairs, the detection range is narrowed, and finally, precise intersection testing is performed at the triangle facet level. If any part of the virtual avatar is detected to intersect with the safe distance envelope, or if the shortest distance between the virtual avatar and the envelope is less than a preset safety threshold (e.g., 2cm), a collision is determined. For the SDF-based method, the signed distance field of the safe distance envelope is pre-calculated. During runtime, the corresponding values ​​of each key point in the distance field are queried to quickly determine whether the value is less than zero (intrusion) or less than the safety threshold. The collision detection algorithm outputs the detection result as a collision state flag.

[0147] S43: Perform risk prediction, which includes calculating the velocity vector and motion trend of the operator and tool endpoints, and predicting whether they will collide with the safe distance envelope within a preset time.

[0148] In this embodiment, the velocity vector refers to the motion direction vector of the worker and tool endpoint in three-dimensional space, including magnitude and direction information; the motion trend refers to the future motion direction and velocity change inferred based on the current velocity vector and acceleration vector; the risk prediction refers to, based on the current collision detection, analyzing the velocity vector and motion trend of the worker and tool endpoint to predict whether they will collide with the safe distance envelope within a preset time, issuing an early warning in advance, reserving reaction time for the worker, and realizing the upgrade from passive alarm to active prediction.

[0149] Specifically, the system caches historical position data of each key point within a preset time window (e.g., 0.5 seconds) to form a motion trajectory. It calculates the velocity vector v = (Δx / Δt, Δy / Δt, Δz / Δt) based on the position differences between consecutive frames, and the acceleration vector a = (Δv_x / Δt, Δv_y / Δt, Δv_z / Δt) based on the velocity differences. Based on the current key point position P0, velocity v, and acceleration a, the predicted position P(t) at time t is extrapolated as P(t) = P0 + v·t + 0.5·a·t². If the distance between the predicted position and the safety distance envelope is less than zero (i.e., an intrusion is predicted), the system calculates the time interval T_intrude from the current moment to the predicted intrusion. If T_intrude is less than a preset prediction time threshold (e.g., 1 second), the system determines that there is an impending intrusion risk.

[0150] S44: Trigger early warning signals in advance based on the predicted movement trends.

[0151] In this embodiment, the predictive warning refers to the early warning signal issued based on motion trend analysis before the operator or tool actually touches the safe distance envelope. This signal aims to remind the operator that their current direction of movement may be dangerous, giving the operator sufficient time to adjust their operation. Unlike collision detection based on the current distance, the predictive warning focuses on the risk that is "about to happen" rather than the collision that has "already happened".

[0152] Specifically, when the calculated T_intrude in step S43 is less than the preset prediction time threshold, the system immediately triggers a prediction warning signal. The prediction warning is triggered independently of the collision detection warning, providing an early warning before workers actually enter the hazardous area. The strength of the warning signal is graded according to the magnitude of T_intrude: when T_intrude is less than 0.3 seconds, it is determined to be a high-risk prediction, triggering an emergency warning; when T_intrude is between 0.3 seconds and 1 second, it is determined to be a medium-risk prediction, triggering a normal prediction warning. The prediction warning works in conjunction with the collision warning based on actual distance, providing an early warning before workers approach the hazardous area and triggering a stronger alarm when workers actually contact the hazardous area, forming a dual protection mechanism. In practical applications, when a worker moves a tool towards a live conductor at a relatively high speed, even if the tool tip is still some distance from the safety envelope, the system can predict that the worker is about to enter a dangerous area based on the velocity vector and issue a warning 0.5 seconds in advance. Upon receiving the warning, the worker can stop or slow down their movements in time to avoid actual intrusion. The prediction warning signal is presented in the AR / VR field of view as an independent icon or a flashing prompt, which is different from the visual style of the collision warning, allowing the worker to clearly distinguish between the two different levels of danger: "imminent risk" and "already facing risk".

[0153] In one embodiment, step S43, i.e., the risk prediction mechanism, includes the following steps:

[0154] S431: Real-time acquisition of continuous position data of the operator's hands and tool ends to generate motion trajectories.

[0155] In this embodiment, continuous position data refers to the sequence of position coordinates of the operator's hand and the tool's endpoint in three-dimensional space; the motion trajectory is a path formed by connecting these position coordinates in chronological order, reflecting the spatial position change pattern of the hand and tool during the operation, and is the basic data for calculating velocity vector and acceleration vector.

[0156] Specifically, the background processing unit uses the S30's multi-positioning technology to continuously collect the spatial coordinates of the operator's hands and tool ends at a sampling frequency of 30Hz, forming time-series data. The system maintains a cache queue for each key point, storing the position data within the most recent second. The cache queue adopts the first-in-first-out principle, updating the queue whenever new position data is obtained. The position data in the cache queue are concatenated in chronological order to generate the motion trajectory of the key point, which is stored in the memory of the background processing unit.

[0157] S432: Calculate the velocity vector and acceleration vector based on the motion trajectory, and analyze the motion trend.

[0158] In this embodiment, the velocity vector refers to the first derivative of position with respect to time, representing the rate and direction of position change; the acceleration vector refers to the first derivative of velocity with respect to time, representing the rate and direction of velocity change; and the motion trend refers to the future motion state inferred based on the current velocity vector and acceleration vector.

[0159] Specifically, the system reads the position data sequence from the cache queue and calculates the velocity and acceleration using the difference method: velocity v(t) = (P(t) - P(t-Δt)) / Δt, acceleration a(t) = (v(t) - v(t-Δt)) / Δt. The calculated velocity and acceleration are filtered to eliminate high-frequency noise. Then, the system analyzes whether the velocity direction points to the safe distance envelope, whether the velocity magnitude is increasing, and whether the distance between the current position and the envelope is continuously decreasing.

[0160] S433: Calculate the intrusion risk probability based on the velocity vector, acceleration vector, and distance between the current position and the safe distance envelope.

[0161] In this embodiment, the intrusion risk probability refers to the likelihood that a worker or tool, in its current state of motion, will intrude into the safe distance envelope within a preset time window. The value ranges from 0 to 1, and this probability comprehensively considers three factors: current distance, velocity vector, and acceleration vector.

[0162] Specifically, the system uses a weighted summation method to calculate the intrusion risk probability P_risk = w_d × P_d + w_v × P_v + w_a × P_a, where P_d is the distance-based risk component (the closer the distance, the higher the risk), P_v is the velocity-based risk component (the velocity direction points to the envelope and the greater the velocity, the higher the risk), and P_a is the acceleration-based risk component (the acceleration direction points to the envelope and the greater the acceleration, the higher the risk). w_d, w_v, and w_a are the corresponding weight coefficients. After calculating the weighted sum, the Sigmoid function is used to map the result to a probability value between 0 and 1.

[0163] S434: When the probability of intrusion exceeds a preset threshold, a pre-judgment warning is triggered before the operator actually touches the safe distance envelope.

[0164] In this embodiment, the preset threshold refers to the minimum intrusion risk probability value that triggers the pre-judgment warning, which is set to 0.7 by default; the pre-judgment warning refers to the warning signal issued in advance based on the risk prediction result before the actual collision occurs, which aims to give the operators sufficient reaction time.

[0165] Specifically, the system compares the intrusion risk probability calculated in step S433 with a preset threshold. If the intrusion risk probability exceeds the threshold, the system immediately triggers a pre-judgment warning signal; if it does not exceed the threshold, the system continues to monitor. After the pre-judgment warning is triggered, if the operator's movement direction changes and the intrusion risk probability decreases and falls below the threshold, the system automatically cancels the pre-judgment warning. In practical applications, when an operator moves a tool towards high-voltage equipment at a relatively fast speed, even if the tool tip is still a certain distance from the safety envelope, the system can still quickly calculate a high intrusion risk probability and issue a warning in advance. After receiving the warning, the operator can stop or slow down their movements in time to avoid actual intrusion.

[0166] In one embodiment, such as Figure 8 As shown, in step S50, namely, outputting a graded visual early warning signal based on the collision detection results and risk prediction results, the steps include:

[0167] S51: The hierarchical visual early warning signal includes proximity warning status and danger alarm status.

[0168] In this embodiment, the graded visualization early warning refers to presenting virtual warning effects of different colors and intensities in the AR / VR field of view based on the distance between the operator and tools and the safe distance envelope and the risk level; the proximity warning state refers to the warning level when the operator or tools have approached the safe distance envelope but have not yet entered the danger zone, which is used to remind the operator to pay attention to the current operation and belongs to the early warning stage; the danger alarm state refers to the warning level when the operator or tools have entered or are about to enter the danger zone, which is used to warn the operator to stop the operation immediately and evacuate, which belongs to the emergency alarm stage.

[0169] Specifically, based on the collision detection and risk prediction results in step S40, the system outputs two different levels of warning signals. The two warning states are clearly distinguished in terms of visual presentation, warning intensity, and triggering conditions, enabling operators to accurately judge the degree of danger based on the current warning state and proactively adjust their operating behavior when approaching the warning stage to avoid entering the dangerous alarm state and achieve the safety management goal of graded protection.

[0170] S52: When the distance to the safety envelope is greater than the first threshold but less than the second threshold, it is determined to be a proximity warning state. A semi-transparent halo is rendered around the charged body, and the distance value is displayed at the tip of the tool.

[0171] In this embodiment, the first threshold is the critical distance value for triggering a hazard alarm, and the second threshold is the critical distance value for triggering a proximity warning. Typically, the first threshold is less than the second threshold. The semi-transparent halo refers to a layer of colored semi-transparent luminous effect superimposed around the live parts of high-voltage equipment, used to visually indicate the location of the boundary of the danger zone. The distance value refers to the real-time distance value from the tool tip to the safety distance envelope, displayed in digital form near the tool tip.

[0172] Specifically, when the background processing unit detects that the shortest distance D between the operator or tool tip and the safe distance envelope meets the first threshold < D ≤ second threshold (e.g., 30cm < D ≤ 100cm), the system determines it to be in a proximity warning state. In the virtual simulation model, the system renders a semi-transparent yellow halo around the live parts of the high-voltage equipment. The transparency of the yellow halo decreases as the distance decreases, and the closer the distance, the less transparent the halo becomes, resulting in a stronger warning effect. This allows the operator to intuitively perceive the trend of distance change through the change in the halo's transparency. At the same time, a floating digital label is generated near the tool tip in the operator's field of vision, displaying the current distance value (accurate to centimeters) from the tool tip to the safe distance envelope in real time, providing the operator with precise quantitative distance information.

[0173] S53: When the distance to the safety envelope is less than the first threshold or has been intruded, it is determined to be a dangerous alarm state, a bright red virtual wall is displayed in the field of vision, and sound alarm and vibration feedback are provided.

[0174] In this embodiment, the bright red virtual wall refers to a bright red semi-transparent virtual wall generated in the worker's field of vision to block the intrusion path. It is used to strongly indicate danger and prohibit passage, and to make the worker instinctively stop moving forward through visual pressure. The sound alarm refers to the graded alarm sound emitted through bone conduction headphones, and the vibration feedback refers to the tactile vibration signal emitted through wearable devices. The three together constitute a multimodal early warning output.

[0175] Specifically, when the backend processing unit detects that the shortest distance D between the operator or tool endpoint and the safe distance envelope is ≤ the first threshold (e.g., ≤ 30cm) or has entered the envelope, the system determines it to be in a danger alarm state. In the operator's AR / VR field of view, a bright red virtual wall is generated in front of the intrusion direction. This wall extends from the ground to the top of the equipment, completely blocking the operator's operation path. The surface of the red virtual wall has a textured flowing animation, with the texture flowing in the direction of intrusion, prompting the operator to immediately evacuate in the opposite direction. At the same time, a rapid intermittent alarm sound is emitted through bone conduction headphones, and high-frequency vibration feedback is triggered through wearable devices (such as smart bracelets or vibrating vests). The bright red virtual wall, the sound alarm, and the vibration feedback work together to form a strong multimodal warning signal, forcing the operator to stop the current action and evacuate the danger zone. The area occupied by the red virtual wall in the field of vision increases with the degree of danger. The deeper the intrusion, the larger the proportion of the virtual wall in the field of vision, which brings a stronger sense of visual oppression to the workers and prompts them to instinctively retreat. When the workers retreat to the safe area, the red virtual wall disappears automatically, the alarm sound and vibration feedback stop simultaneously, and the system returns to normal monitoring status.

[0176] In one embodiment, step S51, namely, outputting a graded visual early warning signal based on the collision detection result and risk prediction result, includes the following steps:

[0177] S511: In the approach warning state, a semi-transparent yellow halo is rendered around the live parts of the high-voltage equipment. The transparency of the yellow halo decreases as the distance decreases and the distance value between the tool tip and the safe distance envelope is displayed in real time in the operator's field of vision.

[0178] In this embodiment, the semi-transparent yellow halo refers to the semi-transparent yellow luminous effect superimposed around the live parts of the high-voltage equipment, used to mark the boundary of the danger zone; the transparency decreases with distance because the opacity of the halo is inversely proportional to the distance to the safety distance envelope surface, the closer the distance, the less transparent the halo and the stronger the warning. The distance value refers to the real-time distance value from the tool tip to the envelope surface.

[0179] Specifically, when the system is determined to be approaching a warning state, the background processing unit renders a semi-transparent yellow halo at the safe distance envelope. The transparency decreases linearly with distance: the transparency is higher (about 80%) when the distance is far, only showing a faint outline; the transparency drops to the lowest level (about 20%) when the distance is close to the first threshold. At the same time, a floating digital label is generated near the tool tip in the operator's field of vision, displaying the distance value to the envelope in real time. This allows the operator to both qualitatively perceive the distance change trend and obtain accurate quantitative information to accurately judge the remaining safety margin.

[0180] S512: In the state of danger alarm, a bright red virtual wall is generated in the field of vision. The surface of the virtual wall has a texture flow animation. The flow direction of the texture is in the direction of intrusion. The area of ​​the virtual wall in the field of vision expands as the degree of danger increases.

[0181] In this embodiment, the bright red virtual wall refers to a bright red, semi-transparent virtual wall generated in the field of vision of the workers to block the intrusion path; the texture flow animation refers to the dynamic texture of the wall surface moving along a specific direction to indicate the direction of intrusion and guide the evacuation direction. The area ratio refers to the proportion of the virtual wall in the field of vision, which expands as the degree of danger increases, creating a sense of visual oppression.

[0182] Specifically, when a dangerous alarm is detected, the back-end processing unit generates a bright red virtual wall in front of the intrusion direction, extending from the vehicle roof to the top of the equipment, completely blocking the intrusion path; the wall surface is covered with dynamic texture patterns that flow continuously in the opposite direction of the intrusion, indicating both the intrusion direction and guiding the evacuation direction; the area of ​​the virtual wall dynamically changes with the degree of danger: when the distance is just below the first threshold, it occupies about 30% of the field of vision; as the intrusion deepens, it gradually expands to more than 60%, creating a strong sense of visual oppression, forcing the workers to instinctively stop moving forward and retreat.

[0183] S513: The red virtual wall, together with the graded alarm sound emitted by the bone conduction headphones and the graded vibration feedback of the wearable device, constitutes a multimodal early warning output.

[0184] In this embodiment, bone conduction headphones refer to headphones that transmit sound through vibrations of the skull, do not obstruct the ear canal, and are suitable for noisy environments; graded alarm tones refer to alarm sounds whose frequency and intensity increase with the level of danger. Wearable devices refer to wearable devices such as smart bracelets or vibrating vests that provide tactile vibration feedback.

[0185] Specifically, when entering a danger alarm state, the back-end processing unit simultaneously triggers auditory and tactile warnings; the bone conduction headphones emit graded alarm sounds according to the degree of danger: a low-frequency intermittent sound when the distance approaches the first threshold; the frequency gradually increases as the intrusion progresses; and a high-frequency continuous rapid sound when the envelope has been invaded; the wearable device simultaneously generates graded vibration feedback, with the vibration intensity increasing to the maximum during the danger alarm state. The red virtual wall, alarm sound, and vibration feedback are triggered simultaneously, transmitting danger signals from three dimensions: visual, auditory, and tactile, forming a multimodal strong warning output. Even if the operator ignores the warning in one channel, the synchronous warnings in the other channels can still effectively transmit danger information, ensuring that danger can be perceived in a timely manner and operations can be stopped in any working environment. After evacuating to a safe area, the three warning signals are simultaneously deactivated.

[0186] In one embodiment, such as Figure 9 As shown, it also includes the physiological state perception step:

[0187] S61: Real-time collection of heart rate, skin conductivity and posture characteristics of workers through wearable devices.

[0188] In this embodiment, wearable devices refer to smart bracelets worn by workers or physiological sensors integrated into AR glasses; heart rate reflects the state of cardiovascular activity, skin conductivity reflects the degree of tension (increased sweating and increased conductivity when tense), and posture characteristics reflect the degree of physical fatigue.

[0189] Specifically, workers wear smart bracelets, and the system collects heart rate, skin conductivity, and triaxial acceleration data in real time at a sampling frequency of no less than 10Hz and performs filtering and noise reduction processing on the raw signals.

[0190] S62: Based on the collected data, determine the tension and fatigue levels of the workers and output the physiological risk coefficient.

[0191] In this embodiment, the physiological risk coefficient is a quantitative indicator output after comprehensively assessing the level of tension and fatigue. The value ranges from 1 to 2, with a default value of 1 (normal state). The larger the value, the higher the physiological risk.

[0192] Specifically, the system analyzes heart rate, skin conductivity, and posture characteristics, and uses a multi-index fusion assessment model to output tension and fatigue levels through fuzzy logic or preset rules. The physiological risk coefficient K_life is then calculated using weighted averages.

[0193] S63: The dynamic safety distance envelope is corrected using the physiological risk coefficient, and the safety distance envelope expands as the physiological risk coefficient increases.

[0194] In this embodiment, the corrected final safety distance Dsafe_final = Dsafe × K_life. Since K_life≥1, the safety envelope expands as the physiological risk coefficient increases, and the safety boundary automatically expands under high-risk conditions.

[0195] Specifically, when K_life=1.3, if the original safe distance is 1.0m, the final safe distance will be increased to 1.3m, so that stressed or fatigued workers can receive warnings from a greater distance and have more time to react.

[0196] S64: Adjust the prediction time and early warning sensitivity of collision detection using the physiological risk coefficient. When the operator is under tension or fatigue, shorten the prediction time and increase the early warning lead time.

[0197] In this embodiment, the prediction time refers to the length of the time window used to predict the impact in risk prediction; increasing the warning lead time means triggering the warning earlier; when workers are tense or fatigued, their reaction speed decreases, and they need to receive the warning earlier.

[0198] Specifically, when K_life exceeds 1.3, the prediction time is shortened from 0.5 seconds to 0.3 seconds, the warning lead time is extended from 0.5 seconds to 1.0 seconds, and the intrusion risk probability trigger threshold is reduced accordingly, so that fatigued workers receive the warning signal earlier.

[0199] S65: The physiological risk coefficient is also used to dynamically adjust the intensity of the graded visual early warning signal.

[0200] In this embodiment, the warning signal strength refers to the prominence of visual, auditory, and tactile warnings. The higher the physiological risk, the stronger the warning signal, ensuring that workers can clearly perceive the warning under high-risk conditions.

[0201] Specifically, when K_life increases, the transparency of the yellow halo decreases (making it more noticeable), the brightness and flashing frequency of the red virtual wall increase, the volume of the bone conduction headphones increases, the vibration of the wearable device increases, and the warning intensity returns to its default value after the physiological state returns to normal.

[0202] In one embodiment, step S65, namely the dynamic adjustment of the physiological risk coefficient to the warning signal, includes the following steps:

[0203] S651: When the physiological risk coefficient exceeds the first threshold, the trigger distance of the proximity warning state is increased to a preset multiple of the default value, the initial transparency of the yellow halo is reduced, and the texture flow speed of the virtual wall is accelerated.

[0204] In this embodiment, the first threshold is set to 1.3 by default to distinguish between normal / mild physiological risk and moderate physiological risk; the trigger distance expansion means that the initial trigger distance (second threshold) of the approach warning is multiplied by a preset multiple, so that the warning is triggered earlier at a greater distance; the initial transparency of the yellow halo is reduced to make the warning more eye-catching from the beginning; the virtual wall texture flows faster to enhance the dynamic effect of the visual warning.

[0205] Specifically, when the physiological risk coefficient exceeds the first threshold (K_life > 1.3), the system will increase the proximity warning trigger distance to 1.5 times the default value (e.g., 150cm), reduce the initial transparency of the yellow halo from 80% to 60%, and increase the flow speed of the red virtual wall texture to 1.5 times the default value, so that workers in poor physiological condition can perceive the warning signal earlier and more clearly.

[0206] S652: When the physiological risk coefficient exceeds the second threshold, the trigger distance of the danger alarm state is increased, the red virtual wall flashes at a preset frequency, the alarm sound of the bone conduction headphones is accelerated, and the vibration intensity of the wearable device is enhanced, forming a multimodal strong warning output that forces the operator to stop moving.

[0207] In this embodiment, the second threshold is set to 1.6 by default to distinguish between moderate and high physiological risks; the danger alarm trigger distance expansion means that the first threshold is multiplied by a preset multiple, so that the danger alarm is triggered earlier at a greater distance; the red virtual wall flashes at high frequency (e.g., 3-5Hz), the alarm sound is accelerated, and the vibration is enhanced, all three are output at maximum intensity at the same time to form a mandatory multimodal warning, forcing workers in a state of high tension or high fatigue to stop their actions immediately.

[0208] Specifically, when the physiological risk coefficient exceeds the second threshold (K_life > 1.6), the system expands the danger alarm trigger distance to 1.5 times the default value (e.g., 45cm), the red virtual wall flashes at a high frequency of 3Hz, the bone conduction headphone alarm sound is sped up to twice the default, the wearable device vibration is increased to the maximum intensity, and the three warning channels output at the maximum intensity simultaneously, forming a multimodal strong warning that forces the worker to stop moving; when the physiological risk coefficient falls back below the threshold, each warning parameter automatically returns to the corresponding level.

[0209] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0210] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention. The actual structure is not limited to this. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. A method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality, characterized in that, Includes the following steps: Collect three-dimensional data of the high-voltage equipment on the roof of the high-speed train and construct a one-to-one scale virtual simulation model; A dynamic safety distance envelope is generated based on real-time voltage levels, environmental parameters, and tool length. The spatial location data of workers and tools are collected in real time using multiple positioning technology and mapped onto the virtual simulation model in real time. In a virtual environment, the distance between the workers and tools and the safety distance envelope is calculated, collision detection is performed, and risk prediction is made based on velocity vector and motion trend analysis. Based on the collision detection results and risk prediction results, a graded and visualized early warning signal is output.

2. The method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality according to claim 1, characterized in that, The process of collecting three-dimensional data of the high-voltage equipment on the roof of the high-speed train and constructing a 1:1 scale virtual simulation model specifically includes: Laser 3D scanning technology was used to scan the roof of the EMU to obtain point cloud data; Construct a 1:1 scale virtual simulation model that includes the vehicle body, pantograph, high-voltage disconnector, and insulators, and ensure that the size error between the virtual simulation model and the physical equipment is controlled within the preset accuracy range. The virtual simulation model is preloaded into the backend processing unit of the AR glasses or VR system worn by the operator.

3. The method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality according to claim 1, characterized in that, The process of generating a dynamic safety distance envelope based on real-time voltage levels, environmental parameters, and tool length specifically includes: Get the current relative air humidity and air pressure, and look up the corresponding environmental insulation withstand voltage correction factor; Obtain the length of the tool held by the worker; Calculate the corrected minimum safe distance, which is equal to the base safe distance multiplied by the environmental correction factor, plus the tool length; Using the surface of the live part of the high-voltage equipment as a reference, the modified minimum safe distance is extended outward to form a three-dimensional closed envelope surface, which serves as a dangerous area where work is prohibited.

4. The method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality according to claim 1, characterized in that, The process of collecting spatial location data of workers and tools in real time using multiple positioning technology and mapping it to the virtual simulation model in real time specifically includes: The absolute spatial location of workers and tools is obtained through ultra-wideband positioning technology; The positioning deviation caused by high-speed motion is compensated by the inertial measurement unit; The AR glasses' front-facing camera collects feature points on the roof of the vehicle, and visual synchronous positioning and mapping technology is used to calculate the relative pose. The results of ultra-wideband positioning, inertial measurement unit and visual positioning are fused and verified, outliers are removed and drift is corrected, and if any source is abnormal, the other two sources will fill the gap. The real-time position is mapped to the coordinate system of the virtual simulation model using a coordinate transformation matrix to generate a virtual avatar.

5. The method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality according to claim 1, characterized in that, The process of calculating the distance between the workers and tools and the safety distance envelope in a virtual environment, and performing collision detection and risk prediction, specifically includes: Real-time calculation of the shortest distance between the endpoints of workers and tools and the safe distance envelope; Collision detection algorithms are used to determine whether there is an intersection or if the distance is less than a safety threshold. Risk prediction is performed, which includes calculating the velocity vector and motion trend of the operator and tool endpoints, and predicting whether they will collide with the safe distance envelope within a preset time. Based on the predicted movement trends, early warning signals are triggered in advance.

6. The method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality according to claim 5, characterized in that, The risk prediction mechanism specifically includes: Real-time continuous position data of the operator's hands and tool ends are collected to generate motion trajectories; Calculate the velocity and acceleration vectors based on the motion trajectory, and analyze the motion trend; The probability of intrusion risk is calculated based on the velocity vector, acceleration vector, and distance between the current position and the safe distance envelope. When the probability of intrusion exceeds a preset threshold, a pre-judgment warning is triggered before the operator actually touches the safe distance envelope.

7. The method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality according to claim 1, characterized in that, The step of outputting a tiered, visualized early warning signal based on collision detection results and risk prediction results specifically includes: The tiered visual early warning signals include proximity warning status and danger alarm status; When the distance to the safety envelope is greater than the first threshold but less than the second threshold, it is determined to be a proximity warning state. A semi-transparent halo is rendered around the charged body, and the distance value is displayed at the tip of the tool. When the distance to the safety envelope is less than the first threshold or when intrusion has occurred, it is determined to be a dangerous alarm state, and a bright red virtual wall is displayed in the field of vision, accompanied by an audible alarm and vibration feedback.

8. A method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality, as described in claim 7, is characterized in that... The hierarchical visual early warning signals include: In the near-warning state, a semi-transparent yellow halo is rendered around the live parts of the high-voltage equipment. The transparency of the yellow halo decreases as the distance decreases, and the distance value between the tool tip and the safe distance envelope is displayed in real time in the operator's field of vision. In the danger alarm state, a bright red virtual wall is generated in the field of vision. The surface of the virtual wall has a texture flow animation. The texture flow direction is in the direction of intrusion. The area of ​​the virtual wall in the field of vision expands as the danger level increases. The red virtual wall, together with the graded alarm sounds emitted by the bone conduction headphones and the graded vibration feedback from the wearable device, constitutes a multimodal early warning output.

9. A method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality, as described in claim 1, is characterized in that... It also includes the physiological state perception step, specifically including: The heart rate, skin conductivity, and posture characteristics of workers are collected in real time using wearable devices. Based on the collected data, the stress and fatigue levels of the workers are assessed, and a physiological risk coefficient is output. The dynamic safety distance envelope is corrected using the physiological risk coefficient, and the safety distance envelope expands as the physiological risk coefficient increases; The physiological risk coefficient is used to adjust the prediction time and early warning sensitivity of collision detection. When the operator is under tension or fatigue, the prediction time is shortened and the early warning advance is increased. The physiological risk coefficient is also dynamically used to adjust the intensity of the graded visual early warning signal.

10. A method for early warning of live-line work on high-voltage equipment of high-speed trains based on virtual reality, as described in claim 9, characterized in that, The dynamic adjustment of the physiological risk coefficient to the early warning signal specifically includes: When the physiological risk coefficient exceeds the first threshold, the trigger distance for the near warning state is increased to a preset multiple of the default value, the initial transparency of the yellow halo is reduced, and the texture flow speed of the virtual wall is accelerated. When the physiological risk coefficient exceeds the second threshold, the trigger distance of the danger alarm state is increased, the red virtual wall flashes at a preset frequency, the alarm sound of the bone conduction headphones is accelerated, and the vibration intensity of the wearable device is enhanced, forming a multimodal strong warning output that forces the operator to stop moving.

Images