Tunnel emergency parking belt safety protection method and system

Abstract

The embodiment of the application provides a tunnel emergency parking belt safety protection method and system, and belongs to the technical field of tunnel traffic safety protection. The guide device for performing graded early warning and information guidance after abnormal level judgment and the anti-collision device for performing graded early warning and information guidance after abnormal level judgment are arranged on the existing tunnel emergency parking belt structure. The feedback state data after collision is collected by the sensor arranged on the anti-collision device, so that the whole scheme can form a complete protection chain of abnormal identification, level judgment, guidance intervention, independent anti-collision, state evaluation and maintenance output on the basis of the emergency parking belt boundary calibration data and the video target sequence data. Through the front and rear time window backtracking correction processing, the single video data chain can still obtain a relatively stable abnormal identification result under the shielding, reflection or short-time contour abnormal scene, so that the implementability and reliability in the in-service tunnel scene are improved.

Description

Technical Field

[0001] This invention relates to the field of tunnel traffic safety protection technology, specifically to a method and system for safety protection of emergency stopping lanes in tunnels. Background Technology

[0002] Emergency stopping lanes in in-service highway tunnels serve as crucial areas for vehicle breakdowns and driver safety, and their safety design directly impacts tunnel traffic safety. In existing technologies, emergency stopping lanes often connect to the main lanes at right angles, with only a short transition section. This abrupt change in cross-section can easily cause visual and psychological illusions for drivers, especially under conditions of fatigue, distraction, or unstable route judgment. This can lead to vehicles veering off the edge, obliquely intruding, or even directly colliding with the side wall of the emergency stopping lane entrance or the maintenance lane, resulting in serious consequences.

[0003] Existing protection solutions mainly suffer from the following problems: One type of solution uses a rigid protective structure, which, although simple in form, lacks effective buffering capacity during a collision; another type uses a rotational energy absorption method, which, while having a certain energy absorption effect, can easily cause vehicles to bounce back into the main lane in some scenarios; and a third type, the hydraulic anti-rebound structure, while possessing a certain anti-rebound capability, is complex in structure, has high maintenance costs, and suffers from oil leakage and insufficient environmental adaptability. Meanwhile, existing solutions typically lack an effective chain-like protection system encompassing early warning, protection, response, and maintenance. Visual guidance often relies on single static signs, making it difficult to achieve deep integration with physical protection devices, and also making it difficult to balance safety, economy, and feasibility in the context of existing tunnels.

[0004] Therefore, how to create a tunnel emergency stopping lane safety system that can provide graded warnings to drivers through guidance devices, provide buffer and anti-rebound protection through independently installed anti-collision devices, and monitor and evaluate the status of the anti-collision devices through sensors after an impact, without large-scale demolition and alteration of the original tunnel main structure, remains a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a safety protection method and system for emergency stopping lanes in tunnels to solve at least one of the above-mentioned technical problems.

[0006] To achieve the above objectives, the first aspect of the present invention provides a safety protection method for emergency stopping lanes in tunnels, the method comprising: Acquire emergency parking lane boundary calibration data and video target sequence data, and generate a time-series raw dataset based on the boundary calibration data and the video target sequence data; Coordinate mapping, trajectory reconstruction, and boundary association processing are performed on the original time-series dataset to obtain a scene feature sequence that characterizes the positional relationship between the target and the emergency parking lane; Perform time window backtracking correction processing on abnormal moments in the scene feature sequence to obtain the corrected scene feature sequence; An anomaly level is determined based on the corrected scene feature sequence to obtain an anomaly level result, and a set of linkage control instructions for controlling the operation of the guidance device is generated based on the anomaly level result. The system acquires feedback status data collected by sensors installed on the anti-collision device, and performs a consistency determination based on the feedback status data and the anomaly level result to output safety protection result data and maintenance assessment data.

[0007] Optionally, acquire emergency parking lane boundary calibration data and video target sequence data, and generate a time-series raw dataset based on the boundary calibration data and the video target sequence data, including: Acquire the coordinate data of the emergency parking lane entrance, the parking lane boundary line data, the side wall contour data, the buffer zone range data, and the guide device placement data as boundary calibration data; Obtain target detection bounding box data, target center point data, target category data, and frame timestamp data from video frames to serve as video target sequence data; Based on the relationship between the changes in the center point of the same target in adjacent video frames, calculate the target position data, target velocity data, target acceleration data, and target heading angle data; The boundary calibration data and the video target sequence data are organized according to a unified timestamp and a unified coordinate numbering rule to generate the time-series raw dataset.

[0008] Optionally, coordinate mapping, trajectory reconstruction, and boundary association processing are performed on the original time-series dataset to obtain a scene feature sequence characterizing the positional relationship between the target and the emergency parking lane, including: The video target sequence data is mapped to a local coordinate system with the emergency parking lane entrance as the origin to form the target's trajectory vector at time t; Based on the trajectory vector and the parking strip boundary line data, the boundary approximation coefficient of the target is calculated; Based on the trajectory vector and the boundary approximation coefficient, target anomaly offset features and parking lane occupancy features are extracted, and the scene feature sequence is generated.

[0009] Optionally, perform time-window backtracking correction processing on abnormal moments in the scene feature sequence to obtain a corrected scene feature sequence, including: Extract trajectory segments of the same target within a continuous time window and identify abnormal moments when the boundary approximation coefficient or heading angle change exceeds a preset threshold; For the abnormal moment, retrieve the trajectory segments of the adjacent moments before and after, and calculate the continuity feature value, edge sharpness feature value and motion stability feature value corresponding to each moment; The correction weights for each adjacent time step are calculated based on the continuity feature value, the edge sharpness feature value, and the motion stability feature value. Based on the corrected weights, the trajectory vectors and boundary approximation coefficients at abnormal moments are backtracked and corrected to obtain the corrected scene feature sequence.

[0010] Optionally, an anomaly level determination is performed based on the corrected scene feature sequence to obtain an anomaly level result, including: Based on the corrected scene feature sequence, extract the average boundary approximation feature value, abnormal behavior feature value, and parking lane occupancy feature value; The risk state value is calculated based on the extracted average boundary approximation feature value, abnormal behavior feature value, and parking lane occupancy feature value. Calculate the target's arrival time based on the remaining distance from the target's current location to the emergency parking lane entrance and the target's speed; The risk status value and the target arrival time are compared with a preset anomaly level threshold to determine the anomaly level result of Level 1, Level 2, or Level 3 anomaly.

[0011] Optionally, a set of linkage control instructions for controlling the operation of the guidance device is generated based on the anomaly level result, including: Based on the anomaly level results, the target arrival time, and the occupancy characteristics of the target's lane and parking lane, a target guidance object correspondence is constructed. A sequence of warning guidance parameters is generated based on the correspondence between the target guidance objects; If the anomaly level is Level 1, a Level 1 linkage control command is generated to control the guidance device to execute a low-brightness strobe warning; if the anomaly level is Level 2, a Level 2 linkage control command is generated to control the guidance device to execute a low-brightness strobe warning, dynamic sign display, and enhanced entrance lighting; if the anomaly level is Level 3, a Level 3 linkage control command is generated to control the guidance device to execute full-intensity entrance guidance and trigger the upstream emergency sign through the bypass access interface. Write the first-level linkage control command, the second-level linkage control command, or the third-level linkage control command into the linkage control command set.

[0012] Optionally, after obtaining the anomaly level result, the following may also be included: The equivalent mass is determined based on the target category data, and the predicted impact energy is calculated based on the equivalent mass, target velocity, and target heading angle. The collision avoidance assessment index is calculated based on the predicted impact energy, the parking lane occupancy characteristic value, and the risk state value. The expected impact zone code and expected response level code are determined based on the collision avoidance assessment index, and the expected impact zone code and expected response level code are used as reference data for consistency determination.

[0013] Optionally, acquire feedback status data collected by sensors installed on the anti-collision device, and output maintenance assessment data, including: The feedback status data output by the displacement sensor, damping status sensor and online monitoring sensor on the anti-collision device are obtained, wherein the feedback status data includes at least the deformation of the energy absorption unit, the damping retention status, the response segment code and the device online rate; A health score is calculated based on the feedback status data; The maintenance assessment data is output based on the maintenance priority and maintenance work order parameters generated by the health score generation module.

[0014] Optionally, a consistency determination is performed based on the feedback status data and the anomaly level result, and security protection result data is output, specifically including: Based on the feedback status data, the actual response level code, the actual response segment code, and the actual impact response feature value are extracted, and the actual impact response index is calculated based on the actual impact response feature value. The actual impact response index and the actual response segment code are compared with the predicted impact energy, the expected impact segment code, and the anomaly level result, respectively, and a consistency score is calculated. If the consistency score is not lower than a preset threshold, output the linked and effective security protection result data; If the consistency score is lower than a preset threshold, output deviation alarm data and update the maintenance assessment data.

[0015] A second aspect of the present invention provides a safety protection system for emergency stopping lanes in tunnels, the system comprising: Video sensing device, used to acquire video target sequence data; A collision avoidance device is installed on the side wall of the emergency parking lane entrance and the high-frequency impact area of ​​the side wall; wherein, the collision avoidance device includes: a front energy absorption unit, a middle adjustable damping anti-rebound unit, and an end rubber anti-collision strip and an arc-shaped liner. A guiding device is installed at the upstream warning position, the buffer zone and the entrance area of ​​the emergency parking lane. The guiding device includes at least a low-brightness strobe light, dynamic signs, gradient guide lines, colored markings, dense rumble strips and external LED light strips. A status monitoring device is used to collect feedback status data output by sensors on the anti-collision device; A control device, configured to perform the tunnel emergency stopping lane safety protection method as described in any of the preceding claims, is configured to read pre-stored boundary calibration data, perform anomaly level judgment on the video target sequence data, generate a set of linkage control instructions for controlling the operation of the guidance device, and perform consistency judgment and maintenance assessment based on the feedback status data.

[0016] Through the above technical solution, this invention proposes a safety protection method and system for tunnel emergency stopping lanes. By installing guidance devices and anti-collision devices on the existing tunnel emergency stopping lane structure to perform graded early warning and information guidance after anomaly level assessment, and by using sensors installed on the anti-collision devices to collect feedback status data after a collision, the entire solution can form a complete protection chain based on emergency stopping lane boundary calibration data and video target sequence data. This chain includes anomaly identification, level assessment, guidance intervention, independent anti-collision, status assessment, and maintenance output. Through backtracking correction processing of preceding and following time windows, a single video data chain can still obtain relatively stable anomaly identification results even in scenarios with occlusion, reflection, or short-term contour anomalies, thereby improving the feasibility and reliability in in-service tunnel scenarios.

[0017] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall structural relationship of the modular anti-rebound buffer anti-collision architecture in this invention.

[0019] Figure 2 This is a schematic diagram of the internal cross-sectional structure of the adjustable damping anti-rebound unit in this invention.

[0020] Figure 3 This is a flowchart of the intelligent linkage early warning process for safety protection of tunnel emergency stopping lanes in this invention.

[0021] Figure 4 This is a layout diagram of the visual and tactile dual-dimensional guidance system for safety protection of tunnel emergency parking lanes in this invention.

[0022] Figure 5 This is a flowchart of the steps of the tunnel emergency stopping lane safety protection method in this invention. Detailed Implementation

[0023] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0024] like Figures 1 to 5 As shown, this invention provides a safety protection method and system for emergency stopping lanes in tunnels. The tunnel emergency stopping lane safety protection system generally includes a video sensing device, a collision avoidance device, a guidance device, a status monitoring device, and a control device.

[0025] Figure 1 The modular anti-rebound buffer anti-collision architecture is used to show the overall structural relationship. The front energy absorption unit, the middle adjustable damping anti-rebound unit, and the end rubber anti-collision strip and arc liner are arranged sequentially along the side wall of the emergency parking lane entrance and the high-frequency impact area. Figure 2 This diagram illustrates the internal cross-sectional structure of the adjustable damping anti-rebound unit in the middle of the anti-collision device, specifically the fit between the sleeve, damping rod, disc spring assembly, and double O-ring sealing assembly. The front-end energy-absorbing unit utilizes a honeycomb aluminum-polyurethane composite energy-absorbing block, secured to a steel shear pin by M24 chemical expansion bolts, with quick-release slots connecting the modules. The middle section is an adjustable damping anti-rebound mechanism, composed of a sleeve, damping rod, disc spring assembly, and double O-ring sealing assembly. The disc spring preload (adjustable from 10 to 50 kN) is adjusted via a knob, replacing the hydraulic system. The end section consists of a rubber anti-collision strip and an arc-shaped stainless steel liner, fully covering the high-frequency impact zone on the side walls.

[0026] Figure 3 This is used to demonstrate the intelligent linkage early warning process, which performs a three-level anomaly level judgment based on the perception results, and triggers different guidance actions according to the anomaly level. At the same time, it forms feedback coordination by monitoring the facility status through sensors after the impact. Figure 4 The spatial layout of the visual and tactile dual-dimensional guidance system is shown, with low-brightness strobe lights and dynamic signs installed at 30m, gradient guide lines and colored markings installed in the transition section, and encrypted vibration markings and external LED light strips installed at the entrance.

[0027] Regarding safety protection methods for emergency stopping lanes in tunnels, such as Figure 5 As shown, an embodiment of the present invention provides a safety protection method for emergency stopping lanes in tunnels, the method comprising: S1: Obtain emergency parking lane boundary calibration data and video target sequence data, and generate a time-series raw dataset based on the boundary calibration data and the video target sequence data.

[0028] Specifically, the following steps are taken: acquiring the coordinates of the emergency parking lane entrance, the parking lane boundary line, the side wall contour, the buffer zone range, and the placement location of the guidance device as boundary calibration data; acquiring the target detection box data, target center point data, target category data, and frame timestamp data from video frames as video target sequence data; calculating the target position data, target velocity data, target acceleration data, and target heading angle data based on the center point change relationship of the same target in adjacent video frames; and organizing the boundary calibration data and the video target sequence data according to a unified timestamp and unified coordinate numbering rule to generate the time-series raw dataset.

[0029] In this embodiment of the invention, by acquiring emergency stopping lane boundary calibration data and video target sequence data, subsequent anomaly identification, anomaly level judgment, guidance linkage, and post-collision state assessment are all based on a unified data foundation. The boundary calibration data primarily originates from the pre-measurement and calibration results of the emergency stopping lane entrance, boundary line, sidewall contour, buffer zone range, and guidance device placement locations during the system installation or debugging phase. The video target sequence data originates from continuous video frames output by the video sensing device, which, after target detection and tracking, yield information such as the target detection box, center point, category, and frame timestamp.

[0030] In this embodiment, Figure 4 The spatial layout of the guidance system shown is directly written into the boundary calibration data. For example, the location of the low-brightness strobe light and dynamic marker at 30m is used as the upstream warning location calibration point; the start and end ranges of the gradient guide lines and colored markings in the transition section are used as the transition section guidance zone calibration range; and the area where the encrypted oscillation markings and external LED light strips are located at the entrance is used as the entrance enhanced guidance zone. In this way, when the subsequent control device generates guidance control commands, it does not send indiscriminate control results to the guidance device in a general manner, but can map different anomaly levels to specific parameters. Figure 4 This allows for zoned guidance along the upstream warning location, buffer zone, and entrance area, based on the different spatial locations shown. Meanwhile, Figure 1 The anti-collision device shown is only used as a static spatial object in this step for the calibration of the side wall contour and high-frequency impact area.

[0031] Furthermore, in one executable implementation, the control device calculates target position data, target velocity data, target acceleration data, and target heading angle data based on the relationship between the center point changes of the same target in adjacent video frames. This data, along with emergency stopping lane boundary calibration data, is then organized into a time-series raw dataset according to a unified timestamp and coordinate numbering rule. It should be noted that the velocity and heading angle here do not rely on direct output from additional sensing devices, but are calculated from the temporal position changes within the video sequence.

[0032] S2: Perform coordinate mapping, trajectory reconstruction and boundary association processing on the original time series dataset to obtain a scene feature sequence that characterizes the positional relationship between the target and the emergency parking lane.

[0033] Specifically, the video target sequence data is mapped to a local coordinate system with the emergency parking lane entrance as the origin to form the target's trajectory vector at time t; based on the trajectory vector and the parking lane boundary line data, the target's boundary approximation coefficient is calculated; based on the trajectory vector and the boundary approximation coefficient, the target's abnormal offset features and parking lane occupancy features are extracted, and the scene feature sequence is generated.

[0034] In this embodiment of the invention, step S2 is used to transform the discrete target information obtained from video detection into a scene feature sequence that can directly reflect the relationship between the target and the boundary of the emergency parking lane. It should be noted that the scene feature sequence formed in step S2 is not a simple stacking of detection results, but a temporal result obtained through local coordinate mapping, continuous trajectory reconstruction, and boundary approximation analysis. The purpose of this is to ensure that subsequent anomaly level judgment is no longer based on local image phenomena in a single frame, but rather on the continuous motion relationship between the target and the parking lane entrance and boundary.

[0035] In this embodiment of the invention, the video target sequence data is first mapped to a local coordinate system with the emergency parking lane entrance as the origin, forming the target's trajectory vector at time t: ; in, Let represent the vertical position of the i-th target at time t. Indicates horizontal position. Indicates speed, Indicates acceleration. The heading angle is represented by the following formula: The longitudinal and lateral positions are obtained through spatial transformations between the target's center point and the coordinates of the entrance and boundary lines; the velocity and acceleration are calculated by the time difference between the target positions in adjacent frames; and the heading angle is derived from the tangential direction of the trajectory. Thus, the target is transformed from a detected box object in the video into a trajectory object moving continuously within the local space of the emergency stopping lane.

[0036] After obtaining the trajectory vector, the boundary approximation coefficient of the target is calculated based on the trajectory vector and the parking strip boundary line data: ; in, This represents the lateral distance between the i-th target and the boundary line of the parking lane at time t. This data can be calculated from the shortest distance from the target's current position to the boundary line. This indicates the effective width of the parking lane, which is derived from boundary calibration data. This indicates a reference speed, which can be preset according to the tunnel's operating speed limit. Indicates the direction angle of the parking lane boundary. Indicates the reference angle. , and This represents the weighting coefficient. This boundary approach coefficient maps the target's proximity to the edge, approach speed, and intrusion angle to a single index. As the target gets closer to the parking lane boundary, reaches a higher speed, and its course deviates towards the sidewall, the boundary approach coefficient continuously increases, indicating a gradual escalation of the abnormal risk.

[0037] Furthermore, the control device extracts target anomaly offset features and parking lane occupancy features based on the trajectory vector and boundary approximation coefficients. Target anomaly offset features are primarily used to indicate whether the vehicle exhibits abnormal behaviors such as continuous edge-hugging, excessively rapid lateral deviation, or sudden changes in heading angle (characterizing abnormal behaviors such as fatigue, distraction, and driving against the flow of traffic). Parking lane occupancy features are primarily used to indicate whether there are already parked vehicles or obstacles occupying the current parking lane. It should be noted that the parking lane occupancy features described here are crucial for subsequent... Figure 3 The corresponding anomaly level determination is particularly important because, under the same trajectory offset intensity, if the parking lane is already occupied, a level 3 anomaly is more likely to be triggered. Through the extraction and organization of the above features, a scene feature sequence is finally obtained, which serves as the input for step S3.

[0038] S3: Perform backtracking correction processing on the abnormal moments in the scene feature sequence to obtain the corrected scene feature sequence.

[0039] Specifically, trajectory segments of the same target are extracted within a continuous time window to identify abnormal moments where the boundary approximation coefficient or heading angle change exceeds a preset threshold. For abnormal moments, trajectory segments of adjacent moments are retrieved, and the continuity feature value, edge sharpness feature value, and motion stability feature value corresponding to each moment are calculated. The correction weights for each adjacent moment are calculated based on the continuity feature value, edge sharpness feature value, and motion stability feature value. The trajectory vector and boundary approximation coefficient of the abnormal moment are backtracked and corrected based on the correction weights to obtain the corrected scene feature sequence.

[0040] In this embodiment of the invention, step S3 is used to address the issue of abnormal frames that may occur in single video data under the special environment of a tunnel. It should be noted that, to avoid misjudgments caused by glare, reflection, occlusion, and local contamination of a single video source, this embodiment adds a backtracking correction process for preceding and following time windows in the data processing link. In other words, the improvement of this invention does not lie in infinitely adding new sensing devices, but in introducing more complex timing correction actions adapted to the specific scenario within the existing video data, thereby improving the stability of anomaly level judgment.

[0041] In this embodiment of the invention, the control device extracts trajectory segments of the same target within a continuous time window and identifies abnormal moments where the boundary approximation coefficient or heading angle change exceeds a preset threshold. These abnormal moments can be understood as moments in a frame where the target contour is disturbed, resulting in positional jumps, abrupt changes in heading angle, or abnormally shortened boundary distances. If these abnormal moments are not processed, subsequent... Figure 3 The anomaly level judgment might be mistakenly escalated from Level 1 to Level 3 due to single-frame noise, causing the guidance device to unnecessarily enter a high-intensity warning state. Therefore, after identifying an anomaly, the control device retrieves trajectory segments from adjacent time points, calculates continuity feature values, edge sharpness feature values, and motion stability feature values ​​respectively, and uses these to calculate the correction weights for each adjacent time point, thus obtaining the correction contribution of each adjacent time point to the current anomaly time point: ; ; in, This represents the contribution weight of the i-th target at time r to the correction at the anomalous time. This indicates the overall score. The continuous characteristic value is derived from the degree of continuity between position and velocity at different time points. The edge sharpness feature value is derived from the target outline sharpness calculation. This represents a characteristic value of motion stability, derived from the smoothness of changes in velocity, acceleration, and heading. , and represents the weighting coefficient, and L represents the backtracking window length.

[0042] In practical applications, the control device performs backtracking correction on the trajectory vector and boundary approximation coefficient at abnormal moments based on the corrected weights to obtain the corrected scene feature sequence. It should be noted that the corrected scene feature sequence is not entirely new data detached from the original data, but rather the result of temporal compensation of the original abnormal frames; therefore, it maintains a one-to-one correspondence with the output data of step S2. This step effectively reduces abnormal fluctuations caused by partial video occlusion, tunnel wall reflections, and light glare, paving the way for the process in step S4. Figure 3 The three-level anomaly level judgment provides a more stable data foundation.

[0043] S4: Based on the corrected scene feature sequence, perform an anomaly level judgment to obtain an anomaly level result, and generate a set of linkage control instructions for controlling the operation of the guidance device based on the anomaly level result.

[0044] Specifically, the average boundary approximation feature value, abnormal behavior feature value, and parking lane occupancy feature value are extracted based on the corrected scene feature sequence; a risk status value is calculated based on the extracted average boundary approximation feature value, abnormal behavior feature value, and parking lane occupancy feature value; the target arrival time is calculated based on the remaining distance from the target's current position to the emergency parking lane entrance and the target speed; the risk status value and the target arrival time are compared with a preset anomaly level threshold to determine the anomaly level result of Level 1, Level 2, or Level 3 anomaly.

[0045] Following this, based on the anomaly level result, the target arrival time, and the occupancy characteristics of the target's lane and parking lane, a target guidance object correspondence is constructed; a warning guidance parameter sequence is generated based on the target guidance object correspondence; if the anomaly level result is a Level 1 anomaly, a Level 1 linkage control command is generated to control the guidance device to execute a low-brightness strobe warning; if the anomaly level result is a Level 2 anomaly, a Level 2 linkage control command is generated to control the guidance device to execute a low-brightness strobe warning, dynamic sign display, and enhanced entrance lighting; if the anomaly level result is a Level 3 anomaly, a Level 3 linkage control command is generated to control the guidance device to execute full-intensity entrance guidance and trigger upstream emergency signs through a bypass access interface; the Level 1 linkage control command, the Level 2 linkage control command, or the Level 3 linkage control command is written into the linkage control command set.

[0046] It should be noted that after obtaining the anomaly level result, the equivalent mass is determined based on the target category data, and the predicted impact energy is calculated based on the equivalent mass, target speed, and target heading angle; the collision avoidance assessment index is calculated based on the predicted impact energy, the parking lane occupancy characteristic value, and the risk state value; the expected impact segment code and the expected response level code are determined based on the collision avoidance assessment index, and the expected impact segment code and the expected response level code are used as consistency judgment reference data.

[0047] In this embodiment of the invention, the anomaly level judgment is not a general judgment of the vehicle's state as high or low, but rather a corresponding... Figure 3 The three-tiered early warning process generates three distinct outcomes: Level 1 anomaly, Level 2 anomaly, and Level 3 anomaly, which are then used to further control [the situation]. Figure 4 The guiding device shown performs guiding actions of varying intensities at different spatial locations.

[0048] In this embodiment of the invention, the average boundary approximation feature value, abnormal behavior feature value, and parking lane occupancy feature value are first extracted based on the corrected scene feature sequence, and the risk state value is calculated: ; in, This represents the risk state value at time t. The average boundary approximation characteristic value at time t is derived from the statistical results of the target boundary approximation coefficient within the current time window. These represent abnormal behavior feature values, derived from a combination of features such as target lateral offset, edge-hugging driving, and abnormal changes in heading angle. The occupancy feature value represents the parking lane status, derived from the identification results of the target's stationary state within the parking lane. , and This represents the weighting coefficient.

[0049] Furthermore, the control device calculates the target arrival time based on the remaining distance from the target's current position to the emergency stopping lane entrance and the target's speed, expressed as: ; in, This represents the arrival time of the target at time t. Let represent the remaining distance of the i-th target from the entrance of the emergency parking lane at time t. This represents the velocity of the i-th target at time t. This represents the minimum correction rate used to prevent the denominator from being zero. This is achieved by simultaneously considering the risk state value. and target arrival time The control device can follow Figure 3Process completion anomaly level judgment: When the risk status value is low and the target arrival time is long, it is judged as a level 1 anomaly; when the risk status value is medium or the target arrival time is significantly shortened, it is judged as a level 2 anomaly; when the risk status value reaches a high threshold, the target is about to invade the entrance area or the parking lane is occupied, resulting in a significant increase in danger, it is judged as a level 3 anomaly.

[0050] After the anomaly level result is determined, the control device generates a set of linkage control commands for controlling the operation of the guidance device based on the anomaly level result. Specifically, when an anomaly is determined to be Level 1, the control device follows... Figure 4 The upstream early warning location deployment method generates a first-level linkage control command to control the low-brightness strobe lights to perform warnings, thereby realizing the processing logic of lighting the warning lights for the first-level warning; when it is determined to be a second-level anomaly, in addition to maintaining the strobe warning, the control device also generates a second-level linkage control command to control the dynamic sign display and entrance lighting enhancement, so that the guidance intensity of the buffer zone and the entrance area is increased synchronously; when it is determined to be a third-level anomaly, the control device generates a full-intensity entrance guidance command and triggers the upstream emergency sign through the bypass access interface.

[0051] It should be noted that the warning guidance parameter sequence is expressed as follows: ; in, This indicates the flicker frequency parameter. Indicates brightness parameter, This indicates a dynamic identifier display code. This indicates the linkage length parameter of the LED light strip.

[0052] Furthermore, in this step, the control device can also calculate and predict the impact energy based on the target type, speed, and heading angle: ; in, This represents the predicted impact energy at time t. Indicates the number of targets involved in the calculation. This represents the risk contribution weight of the i-th objective. Let represent the equivalent mass of the i-th objective. This represents the velocity of the i-th target at time t. This represents the heading angle of the i-th target at time t. Indicates the direction angle of the parking lane boundary; The collision avoidance assessment index is calculated based on predicted impact energy, parking lane occupancy characteristic value, and risk state value. ; in, Indicates the collision avoidance assessment index. This indicates the occupancy characteristic value of the parking lane. Indicates the risk status value. , and This represents the weighting coefficient. It should be emphasized that the predicted impact energy and collision avoidance assessment index mentioned above are not used for active control of the collision avoidance device in this embodiment, but rather serve as reference data for consistency determination in the subsequent step S5, used to compare the pre-collision prediction results with the post-collision sensor feedback results of the collision avoidance device.

[0053] S5: Obtain feedback status data collected by sensors installed on the anti-collision device, and perform consistency determination based on the feedback status data and the anomaly level result to output safety protection result data and maintenance assessment data.

[0054] The process of acquiring feedback status data collected by sensors installed on the anti-collision device and outputting maintenance assessment data includes: acquiring feedback status data output by displacement sensors, damping status sensors, and online monitoring sensors on the anti-collision device, wherein the feedback status data includes at least the deformation of the energy-absorbing unit, damping retention status, response segment encoding, and device online rate; calculating a health score based on the feedback status data; and generating maintenance priority and maintenance work order parameters based on the health score to output the maintenance assessment data.

[0055] Specifically, the process of performing a consistency determination based on the feedback status data and the anomaly level result, and outputting safety protection result data, includes: extracting the actual response level code, actual response segment code, and actual impact response feature value based on the feedback status data, and calculating the actual impact response index based on the actual impact response feature value; comparing the actual impact response index and the actual response segment code with the predicted impact energy, the expected impact segment code, and the anomaly level result, respectively, and calculating a consistency score; outputting effective safety protection result data if the consistency score is not lower than a preset threshold; and outputting deviation alarm data and updating the maintenance assessment data if the consistency score is lower than the preset threshold.

[0056] In this embodiment of the invention, step S5 is used to collect feedback status data using sensors installed on the anti-collision device after a collision has occurred or the anti-collision device has been subjected to force, and to compare the feedback status data with the anomaly level result and predicted impact parameters obtained in step S4 for consistency assessment and maintenance evaluation. It should be noted that in this embodiment, the guiding device is only responsible for receiving the linkage control command set to execute strobe warnings, dynamic signage display, and entrance lighting enhancement; the feedback link is handled by the displacement sensor, damping status sensor, and online monitoring sensor installed on the anti-collision device.

[0057] In this embodiment of the invention, the feedback state data includes at least the deformation of the energy-absorbing unit, the damping retention state, the response segment code, and the device online rate. The deformation of the energy-absorbing unit originates from the data set in the device. Figure 1 Displacement sensors on the front-end energy-absorbing unit are used to characterize the actual deformation of the energy-absorbing module after impact; the damping holding state is determined by the sensors set on the front-end energy-absorbing unit. Figure 2 The damping status sensor on the adjustable damping anti-rebound mechanism shown is used to characterize the retention status or preload loss of the damping mechanism after an impact; the response segment code comes from the trigger detection units set along different segments of the anti-collision device, used to indicate the actual location of the anti-collision segment in action; the device online rate comes from the statistical results of the status monitoring device on the working status of each sensor. Based on these feedback status data, the control device first calculates a health score based on the feedback status data: ; in, This represents the health score of the m-th protection module. This represents the deformation of the m-th protective module. Indicates the allowable deformation threshold. Indicates the remaining preload value. Indicates the reference preload value. Indicates the online rate of the device. , and This represents the weighting coefficient. The lower the health score, the more priority the anti-collision device in the corresponding section needs to be maintained or replaced, thus forming the module maintenance priority and maintenance work order parameters.

[0058] Furthermore, the control device extracts the actual response level code, the actual response segment code, and the actual impact response feature value based on the feedback status data, and calculates the actual impact response index based on the actual impact response feature value: ; in, The actual impact response index at time t is represented. This indicates the deformation of the energy-absorbing unit. This indicates the damping retention loss. Indicates the number of sensor triggers in response. , and This represents the weighting coefficient.

[0059] Subsequently, the actual impact response index and the actual response segment code are compared with the predicted impact energy, the expected impact segment code, and the anomaly level result, respectively, to calculate a consistency score: in, The consistency score represents the time t. Indicates the predicted impact energy. Indicates the actual shock response index. Indicates the code of the expected impact segment. Indicates the actual response segment code. Indicates the reference value of the section code. Indicates the anomaly level code. Indicates the actual response level code. Indicates the reference value for the grade code. , and Indicates the weighting coefficient. This represents a correction constant. In this way, the control device verifies whether the pre-collision anomaly level assessment and collision risk prediction are consistent with the actual response of the post-collision anti-collision device.

[0060] When the consistency score is not lower than the preset threshold, it indicates that Figure 3 Anomaly identification, anomaly level judgment and Figure 1 The actual force response of the anti-collision device shown has a high degree of consistency. At this time, the control device outputs effective safety protection result data. When the consistency score is lower than the threshold, it indicates that there is a deviation between the front-end abnormality level judgment, the expected impact section or the expected impact intensity and the actual sensor response. The control device further outputs deviation alarm data and updates the maintenance assessment data.

[0061] It should be noted that the present invention also provides a tunnel emergency stopping lane safety protection system, the system comprising: Video sensing device, used to acquire video target sequence data; A collision avoidance device is installed on the side wall of the emergency parking lane entrance and the high-frequency impact area of ​​the side wall; wherein, the collision avoidance device includes: a front energy absorption unit, a middle adjustable damping anti-rebound unit, and an end rubber anti-collision strip and an arc-shaped liner. A guiding device is installed at the upstream warning position, the buffer zone and the entrance area of ​​the emergency parking lane. The guiding device includes at least a low-brightness strobe light, dynamic signs, gradient guide lines, colored markings, dense rumble strips and external LED light strips. A status monitoring device is used to collect feedback status data output by sensors on the anti-collision device; A control device, configured to perform the tunnel emergency stopping lane safety protection method as described in any of the preceding claims, is configured to read pre-stored boundary calibration data, perform anomaly level judgment on the video target sequence data, generate a set of linkage control instructions for controlling the operation of the guidance device, and perform consistency judgment and maintenance assessment based on the feedback status data.

[0062] In this embodiment of the invention, the system includes a video sensing device, a collision avoidance device, a guidance device, a status monitoring device, and a control device. The system includes a video sensing device (which may be a radar sensing device in some embodiments) located at a visible monitoring position upstream of the emergency stopping lane entrance, used to continuously collect video target sequence data covering the main lane, transition section, and entrance area; a collision avoidance device located on the side wall of the emergency stopping lane entrance and the high-frequency impact area of ​​the side wall, comprising at least a front-end energy-absorbing unit, a middle adjustable damping anti-rebound unit, and an end rubber anti-collision strip and an arc-shaped liner, arranged in a modular external configuration; a guidance device located at the warning position upstream of the emergency stopping lane entrance, the transition section, and the entrance area, comprising at least a low-brightness strobe light, dynamic markings, gradient guide lines, colored markings, dense rumble strips, and an external LED light strip; a status monitoring device connected to the displacement sensor, damping status sensor, and online monitoring sensor on the collision avoidance device, used to collect and feedback status data; and a control device used to read pre-stored boundary calibration data, perform anomaly level judgment on the video target sequence data, and generate a set of linkage control instructions for controlling the operation of the guidance device, and also used to perform consistency judgment and maintenance assessment based on sensor feedback from the collision avoidance device.

[0063] In one specific embodiment, the system of the present invention is applied to an in-service two-way four-lane tunnel in a temperate humid climate, with an emergency stopping lane length of 25m, suitable for a speed of 80km / h. The front-end energy-absorbing unit is arranged 4m before the entrance, the mid-section damping mechanism is spaced approximately 1.5m apart, and the damping force is adjustable to 30kN. A strobe light is installed at 30m along the guiding system, yellow and white dual-color warning tape is laid in the transition section, and rumble strips are installed in the entrance area. The system is also bypassed and connected to the existing platform via the GB / T 28181 protocol.

[0064] In this embodiment, the control device first establishes a local coordinate system for the parking lane entrance based on boundary calibration data, and then... Figure 4 The upstream warning location, mitigation zone, and entrance area are all written into the reference coordinates; subsequently, the video sensing device continuously acquires target video frames, and the control device reconstructs the target trajectory vector and boundary approximation coefficient in step S2. When the vehicle shows a slight tendency to brush against the edge while still a certain distance from the entrance, and the risk status value is low, the control device proceeds according to... Figure 3 The logical judgment is a level 1 exception, and only sends... Figure 4 The system outputs a command to illuminate the low-brightness strobe light at 30m; when the vehicle continues to deviate towards the entrance boundary and the target arrival time is rapidly shortened, it is upgraded to a level two anomaly, further controlling the dynamic signage and entrance LED light strip enhancement; when the vehicle has approached and intruded into the entrance area or the parking lane is occupied, the control device determines it to be a level three anomaly, and while maintaining full-intensity entrance guidance, it triggers the upstream emergency sign through the bypass access interface.

[0065] If the vehicle ultimately experiences a side impact during the above process, then Figure 1 The collision avoidance device shown functions independently at the physical level. The front-end energy-absorbing unit first deforms to dissipate energy, the middle adjustable damping anti-rebound unit suppresses rebound based on its own structural characteristics, and the end rubber anti-collision strip and arc-shaped liner cover and protect the high-frequency impact area. After the impact, the displacement sensor, damping state sensor, and online monitoring sensor installed on the collision avoidance device output feedback status data, and the control device then completes the health score and consistency score calculation according to step S5. Thus, in this embodiment, the guidance device undertakes the controllable early warning and guidance functions, while the collision avoidance device undertakes the independent physical protection and monitorable feedback functions. The two form feedback at the data level through the control device and the status monitoring device.

[0066] In another specific embodiment, the system of the present invention is applied to an in-service single-lane dual-lane tunnel in a cold region, with an emergency stopping lane length of 20m and a minimum temperature of -20℃. The mid-range damping mechanism can use a low-temperature adapted damping medium, the energy-absorbing block can be replaced with a cold-resistant polyurethane material, and the sensor protection level can be upgraded to IP68 to meet the application requirements of cold regions.

[0067] In this embodiment, the complete process still consists of steps S1 to S5. The backtracking correction process in step S3 is particularly important in cold-weather scenarios because low temperatures make vehicle headlight reflection, local condensation, and changes in image contrast more likely to cause single-frame contour distortion. The control device reduces the direct impact of abnormal frames on the anomaly level judgment by comprehensively comparing continuity feature values, edge sharpness feature values, and motion stability feature values, thereby maintaining... Figure 3 Smooth transition between Level 1, Level 2, and Level 3 exceptions in the process.

[0068] When the vehicle exhibits only a slight deviation, the system outputs a Level 1 anomaly and takes control. Figure 4 The upstream strobe lights operate; as the vehicle continues to approach the entrance boundary and the arrival time shortens, a level two anomaly is output, and the dynamic signage and entrance LED light strip are enhanced; if the target has a high probability of collision, a level three anomaly is entered, and the upstream emergency sign is triggered. Even if a collision eventually occurs, Figure 1 and Figure 2 The collision avoidance device shown, with its predetermined physical structure, performs the functions of buffering energy absorption and preventing rebound. After a collision, sensors provide feedback on the deformation, damping retention status, and response segment code. The control device then uses this feedback data to make a consistency judgment on the accuracy of the front-end anomaly level assessment. After the trial operation, if the health score of a certain mid-range damping segment remains low, the control device automatically generates the corresponding maintenance priority and maintenance work order parameters, thereby realizing a safety protection process of early warning, protection, and maintenance.

[0069] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0070] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details described above. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not further describe the various possible combinations.

[0071] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the embodiments of the present invention, they should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. A safety protection method for emergency stopping lanes in tunnels, characterized in that, The method includes: Acquire emergency parking lane boundary calibration data and video target sequence data, and generate a time-series raw dataset based on the boundary calibration data and the video target sequence data; Coordinate mapping, trajectory reconstruction, and boundary association processing are performed on the original time-series dataset to obtain a scene feature sequence that characterizes the positional relationship between the target and the emergency parking lane; Perform time window backtracking correction processing on abnormal moments in the scene feature sequence to obtain the corrected scene feature sequence; An anomaly level is determined based on the corrected scene feature sequence to obtain an anomaly level result, and a set of linkage control instructions for controlling the operation of the guidance device is generated based on the anomaly level result. The system acquires feedback status data collected by sensors installed on the anti-collision device, and performs a consistency determination based on the feedback status data and the anomaly level result to output safety protection result data and maintenance assessment data.

2. The safety protection method for tunnel emergency stopping lanes according to claim 1, characterized in that, Acquire emergency parking lane boundary calibration data and video target sequence data, and generate a time-series raw dataset based on the boundary calibration data and the video target sequence data, including: Acquire the coordinate data of the emergency parking lane entrance, the parking lane boundary line data, the side wall contour data, the buffer zone range data, and the guide device placement data as boundary calibration data; Obtain target detection bounding box data, target center point data, target category data, and frame timestamp data from video frames to serve as video target sequence data; Based on the relationship between the changes in the center point of the same target in adjacent video frames, calculate the target position data, target velocity data, target acceleration data, and target heading angle data; The boundary calibration data and the video target sequence data are organized according to a unified timestamp and a unified coordinate numbering rule to generate the time-series raw dataset.

3. The safety protection method for tunnel emergency stopping lanes according to claim 1, characterized in that, The original time-series dataset is subjected to coordinate mapping, trajectory reconstruction, and boundary association processing to obtain a scene feature sequence characterizing the positional relationship between the target and the emergency parking lane, including: The video target sequence data is mapped to a local coordinate system with the emergency parking lane entrance as the origin to form the target's trajectory vector at time t; Based on the trajectory vector and the parking strip boundary line data, the boundary approximation coefficient of the target is calculated; Based on the trajectory vector and the boundary approximation coefficient, target anomaly offset features and parking lane occupancy features are extracted, and the scene feature sequence is generated.

4. The safety protection method for tunnel emergency stopping lanes according to claim 1, characterized in that, Perform time-window backtracking correction processing on abnormal moments in the scene feature sequence to obtain a corrected scene feature sequence, including: Extract trajectory segments of the same target within a continuous time window and identify abnormal moments when the boundary approximation coefficient or heading angle change exceeds a preset threshold; For the abnormal moment, retrieve the trajectory segments of the adjacent moments before and after, and calculate the continuity feature value, edge sharpness feature value and motion stability feature value corresponding to each moment; The correction weights for each adjacent time step are calculated based on the continuity feature value, the edge sharpness feature value, and the motion stability feature value. Based on the corrected weights, the trajectory vectors and boundary approximation coefficients at abnormal moments are backtracked and corrected to obtain the corrected scene feature sequence.

5. The safety protection method for tunnel emergency stopping lanes according to claim 1, characterized in that, An anomaly level determination is performed based on the corrected scene feature sequence to obtain an anomaly level result, including: Based on the corrected scene feature sequence, extract the average boundary approximation feature value, abnormal behavior feature value, and parking lane occupancy feature value; The risk state value is calculated based on the extracted average boundary approximation feature value, abnormal behavior feature value, and parking lane occupancy feature value. Calculate the target's arrival time based on the remaining distance from the target's current location to the emergency parking lane entrance and the target's speed; The risk status value and the target arrival time are compared with a preset anomaly level threshold to determine the anomaly level result of Level 1, Level 2, or Level 3 anomaly.

6. The safety protection method for tunnel emergency stopping lanes according to claim 5, characterized in that, Based on the anomaly level result, a set of linkage control instructions for controlling the operation of the guidance device is generated, including: Based on the anomaly level results, the target arrival time, and the occupancy characteristics of the target's lane and parking lane, a target guidance object correspondence is constructed. A sequence of warning guidance parameters is generated based on the correspondence between the target guidance objects; If the anomaly level is Level 1, a Level 1 linkage control command is generated to control the guidance device to execute a low-brightness strobe warning; if the anomaly level is Level 2, a Level 2 linkage control command is generated to control the guidance device to execute a low-brightness strobe warning, dynamic sign display, and enhanced entrance lighting; if the anomaly level is Level 3, a Level 3 linkage control command is generated to control the guidance device to execute full-intensity entrance guidance and trigger the upstream emergency sign through the bypass access interface. Write the first-level linkage control command, the second-level linkage control command, or the third-level linkage control command into the linkage control command set.

7. The safety protection method for tunnel emergency stopping lanes according to claim 5, characterized in that, After obtaining the anomaly level result, it also includes: The equivalent mass is determined based on the target category data, and the predicted impact energy is calculated based on the equivalent mass, target velocity, and target heading angle. The collision avoidance assessment index is calculated based on the predicted impact energy, the parking lane occupancy characteristic value, and the risk state value. The expected impact zone code and expected response level code are determined based on the collision avoidance assessment index, and the expected impact zone code and expected response level code are used as reference data for consistency determination.

8. The safety protection method for tunnel emergency stopping lanes according to claim 1, characterized in that, Acquire feedback status data collected by sensors installed on the anti-collision device, and output maintenance assessment data, including: The feedback status data output by the displacement sensor, damping status sensor and online monitoring sensor on the anti-collision device are obtained, wherein the feedback status data includes at least the deformation of the energy absorption unit, the damping retention status, the response segment code and the device online rate; A health score is calculated based on the feedback status data; The maintenance assessment data is output based on the maintenance priority and maintenance work order parameters generated by the health score generation module.

9. The safety protection method for tunnel emergency stopping lanes according to claim 8, characterized in that, Based on the feedback status data and the anomaly level result, a consistency determination is performed, and security protection result data is output, specifically including: Based on the feedback status data, the actual response level code, the actual response segment code, and the actual impact response feature value are extracted, and the actual impact response index is calculated based on the actual impact response feature value. The actual impact response index and the actual response segment code are compared with the predicted impact energy, the expected impact segment code, and the anomaly level result, respectively, and a consistency score is calculated. If the consistency score is not lower than a preset threshold, output the linked and effective security protection result data; If the consistency score is lower than a preset threshold, output deviation alarm data and update the maintenance assessment data.

10. A safety protection system for an emergency stopping lane in a tunnel, characterized in that, The system includes: Video sensing device, used to acquire video target sequence data; A collision avoidance device is installed on the side wall of the emergency parking lane entrance and the high-frequency impact area of ​​the side wall; wherein, the collision avoidance device includes: a front energy absorption unit, a middle adjustable damping anti-rebound unit, and an end rubber anti-collision strip and an arc-shaped liner. A guiding device is installed at the upstream warning position, the buffer zone and the entrance area of ​​the emergency parking lane. The guiding device includes at least a low-brightness strobe light, dynamic signs, gradient guide lines, colored markings, dense rumble strips and external LED light strips. A status monitoring device is used to collect feedback status data output by sensors on the anti-collision device; A control device, configured to execute the tunnel emergency stopping lane safety protection method according to any one of claims 1-9, is configured to read pre-stored boundary calibration data, perform anomaly level judgment on the video target sequence data, generate a set of linkage control instructions for controlling the operation of the guidance device, and perform consistency judgment and maintenance assessment based on the feedback status data.

Images