Collision sensing intelligent guardrail system and method based on Beidou positioning

Abstract

The invention discloses a collision sensing intelligent guardrail system and method based on Beidou positioning, and the method comprises the following steps: S1, collecting sensor data along a guardrail, and carrying out the preprocessing; s2, transmitting the preprocessed sensor data to an edge server, and forwarding the preprocessed sensor data to a cloud server through a cellular network; s3, constructing a gated convolutional network based on two channels, and inputting the state vector sequence into the gated convolutional network; s4, performing multi-layer nested coding and decoding operation on the collision feature vector, constructing a collision judgment function, and judging whether collision abnormity is triggered or not; s5, generating a collision early warning data packet when the collision level label is in an activated state; and S6, transmitting the collision alarm data packet to a road traffic command center for road maintenance personnel to quickly respond and deal with the collision alarm data packet. According to the invention, high-precision real-time perception and intelligent early warning of the collision state of the road guardrail are realized, and the automation level and response efficiency of traffic safety monitoring are obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent perception and embedded sensor control technology, and in particular to a collision perception intelligent guardrail system and method based on Beidou positioning. Background Art

[0002] Against the backdrop of modern urbanization and the rapid development of high-speed transportation, road traffic safety issues are increasingly receiving widespread attention from all sectors of society. Especially on highways, urban elevated roads, mountainous winding roads, and other sections of roads with high speeds and complex terrain, guardrails, as core protective facilities at the physical safety boundary, play an important role in preventing vehicles from running off the road, alleviating the impact of accidents, and guiding driving directions. However, with the increase in road traffic density and the frequent occurrence of extreme driving behaviors, the probability of guardrail collision accidents continues to rise, which in turn places higher demands on smooth traffic, the safety of people's lives and property, and road maintenance. In order to improve the intelligent monitoring capabilities of guardrail safety status, domestic and foreign scholars and engineering practices have proposed a series of guardrail collision detection and response methods.

[0003] Existing guardrail detection technology primarily relies on installing various physical sensors on the guardrail itself, such as strain gauges, vibration sensors, pressure sensors, cameras, and millimeter-wave radar. These sensors sense the physical changes in the guardrail when it is impacted or displaced in real time, transmitting this information via wireless networks to backend servers for analysis. Some systems also utilize manual video monitoring to ensure a rapid response to emergencies. This traditional approach offers certain advantages in real-time performance and integrity during project implementation, enabling monitoring and assessment of the guardrail status in specific areas, and has played a crucial role in handling some traffic accidents.

[0004] However, the guardrail collision detection system based on the above-mentioned traditional methods also faces many limitations. The first is the intensive hardware deployment and high construction cost. In order to ensure the detection accuracy, it is often necessary to deploy multiple sensor nodes and cameras at a high density, especially in highway sections of dozens or even hundreds of kilometers. This solution will lead to a significant increase in the initial investment cost. In addition, these devices have high requirements for power supply, signal transmission and protection level, which further increases the economic burden of operation and maintenance. Secondly, the system communication link is complex and the data transmission is significantly limited. Most existing equipment relies on Wi-Fi or cellular networks to transmit sensor and video data. In complex terrains such as mountainous areas and tunnels where the network signal is unstable or the bandwidth is limited, the data is subject to problems such as packet loss, delay, and bit error, which seriously affect the real-time performance and reliability of the system.

[0005] Thirdly, traditional systems generally use static rule matching or manual threshold recognition models in their data processing strategies, which are unable to dynamically adapt to diverse road conditions and sudden impact characteristics. This method of relying on manual experience to set parameters not only has a high false alarm rate and serious missed alarms, but also cannot provide active early warnings and trend judgments, resulting in a delayed response mechanism and difficulty supporting the traffic control center to quickly and accurately intervene. In particular, in complex situations where it is necessary to accurately identify guardrails that are slightly deformed, undergoing continuous vibration disturbances, or have hidden damage, traditional methods have a serious lack of recognition capabilities. In addition, some systems still rely on manual video judgment and confirmation, which further increases the workload of personnel and is prone to problems such as fatigue judgment and misjudgment.

[0006] Therefore, how to provide a collision-aware intelligent guardrail system and method based on Beidou positioning is an urgent problem that technicians in this field need to solve. Summary of the Invention

[0007] One purpose of the present invention is to propose a collision-aware intelligent guardrail system and method based on Beidou positioning. The present invention makes full use of Beidou high-precision positioning technology, low-power acceleration sensor, extended Kalman filter algorithm and dual-channel gated convolutional network structure, and describes in detail the entire process from data acquisition, state estimation, feature extraction to collision judgment. It has the advantages of low deployment cost, fast response speed, high recognition accuracy and adaptability to complex environments.

[0008] According to an embodiment of the present invention, a collision-aware intelligent guardrail method based on Beidou positioning includes the following steps: S1, collect sensor data along the guardrail and perform preprocessing; S2, transmits the preprocessed sensor data to the edge server, performs nonlinear time series fitting and smoothing filtering operations to obtain the state vector sequence, and forwards it to the cloud server via the cellular network; S3. Construct a dual-channel gated convolutional network, input the state vector sequence into the gated convolutional network, perform feature extraction and fusion, and output a collision feature vector; S4. Perform multi-layer nested encoding and decoding operations on the collision feature vector to obtain a reduced-dimensional feature representation, and construct a collision determination function to determine whether a collision anomaly is triggered, and mark the collision level label; S5. When the collision level tag is activated, a collision warning data packet is generated, wherein the collision warning data packet includes the collision occurrence time, guardrail position coordinates, and collision severity level; S6. Transmit the collision warning data packet to the road traffic command center for road maintenance personnel to quickly respond and handle.

[0009] Optionally, the sensor data includes spatial coordinate data and acceleration data, the spatial coordinate data is provided by the GNSS positioning module, and the acceleration data is provided by the MPU6050 sensor module, and each module is connected in series through a bus to form a group, and each group contains 20 to 50 modules.

[0010] Optionally, the preprocessing includes timestamp alignment, coordinate solution, abnormal drift removal, signal denoising, missing value interpolation and data standardization.

[0011] Optionally, the edge server is equipped with solar panels and batteries to supply electricity to the entire formation, is installed at the edge of the road, has cellular communication capabilities, is connected to the formation at one end, and is connected to the cloud server at the other end.

[0012] Optionally, the gated convolutional network is constructed based on a dual-channel mechanism, the first channel receives acceleration data and extracts time-dependent features, the second channel receives spatial coordinate data and extracts displacement morphological features, the output features of the two channels are fused in a residual connection structure, and a dynamic attention gating mechanism is introduced to perform weighted reconstruction of the time-dependent features and the displacement morphological features to form a collision feature vector.

[0013] Optionally, the collision level labels include vehicle collision with guardrail, man-made damage to guardrail, road subsidence, large displacement of guardrail, geological deformation of road and no collision.

[0014] Optionally, the S2 specifically includes: S21, obtain the spatial coordinate data and acceleration data of each acquisition node in the series structure, where the spatial coordinate data is the three-dimensional position data , acceleration data is three-axis acceleration data , Indicates the collection time index, Indicates the The geographical coordinates of the moment, Indicates the acceleration values in three orthogonal directions at the corresponding moment; S22. Align the timestamps of the three-dimensional coordinates and acceleration data of each acquisition node, interpolate missing samples, and perform linear interpolation to unify the sampling frequency and construct an observation sequence set. ; S23, transmit the preprocessed observation sequence set to the edge server, apply nonlinear time series fitting operation and extended Kalman filter algorithm to perform nonlinear state estimation and noise suppression on the observation sequence, and obtain the state vector sequence after filtering. ,in Represents the current position estimate, represents the estimated value of speed, represents the estimated value of acceleration; S24, forwarding the state vector sequence to the cloud server via the cellular network, and the nonlinear state estimation uses the following Kalman state prediction and update, ; in, Indicates the The estimated state value at time t, Indicates the The predicted state value at time t, Indicates the The Kalman gain at time t, represents the time-aligned observation vector, represents the observation matrix.

[0015] Optionally, the S3 specifically includes: S31, build a dual-channel gated convolutional network to transform the state vector sequence and As the input of the first channel and the second channel respectively; S32, will Input to the first channel, perform one-dimensional convolution extraction operation, and generate time-dependent feature tensors ,The first channel uses a convolution kernel with a gated activation function to locally model the acceleration sequence in the time dimension; S33, will Input to the second channel, perform one-dimensional convolution extraction operation, and generate displacement morphological feature tensor , the second channel models the displacement morphological features in the coordinate sequence through a convolutional structure; S34, the output tensors of the two channels and Fusion in the residual connection structure to construct the fusion feature tensor , the fusion process introduces a dynamic attention gating mechanism to form a collision feature vector through weighted reconstruction: ; in, Indicates the The collision feature vector output at the moment, Indicates the The time-dependent feature tensor at time instant, Indicates the The displacement morphological characteristic tensor at time , represents the attention gating weight matrix, represents the gate bias vector, represents the hyperbolic tangent activation function, Represents the Sigmoid activation function.

[0016] Optionally, the S4 specifically includes: S41, the collision feature vector Input the nested encoding network structure, perform layer-by-layer nonlinear mapping and compression, and obtain a multi-layer potential feature representation sequence ,in represents the feature vector of the first layer encoding output, represents the feature vector of the second layer encoding output, The feature vector representing the output of the third layer code; S42, will Input the decoding structure, perform reverse mapping and dimensionality reduction reconstruction, and generate a reconstructed feature vector. The dimensionality reduction reconstruction process retains the collision core features and compresses redundant information; S43. Based on the reconstructed feature vector, a collision determination function is constructed to determine whether the current moment is an abnormal collision state. The expression of the collision determination function is: ; in, Indicates the The reconstructed eigenvector at time , represents the first linear mapping matrix, represents the second linear mapping matrix, represents the first bias vector, represents the second bias vector, represents a nonlinear activation function, represents the Softmax normalization function, Indicates the The collision judgment label at the moment includes the vehicle collision guardrail, guardrail damage, road subsidence, guardrail displacement, geological deformation and no collision state.

[0017] A collision-aware intelligent guardrail system based on Beidou positioning according to an embodiment of the present invention includes: The data processing module is used to collect spatial coordinate data and acceleration data along the guardrail and perform data preprocessing operations; An edge computing module is used to transmit the pre-processed sensor data to the edge server, perform nonlinear time series fitting and smoothing filtering operations, generate a guardrail state vector sequence, and forward the state vector sequence to the cloud server via the cellular network; A feature extraction module is used to construct a dual-channel gated convolutional network, input the state vector sequence into the gated convolutional network, perform feature extraction and fusion, and output a guardrail collision feature vector; The collision determination module is used to perform multi-layer nested encoding and decoding operations on the guardrail collision feature vector to obtain a reduced-dimensional feature representation, and to construct a collision determination function to determine whether a collision anomaly is triggered and mark the collision level label; A warning generation module, configured to generate a collision warning data packet when the collision level tag is activated, wherein the collision warning data packet includes the collision occurrence time, guardrail position coordinates, and collision severity level; The response and disposal module is used to transmit the collision alarm data packet to the road traffic command center for road maintenance personnel to respond and handle quickly.

[0018] The beneficial effects of the present invention are: First, the present invention integrates the GNSS high-precision positioning chip and the MPU6050 three-axis acceleration sensor to construct a multi-dimensional sensor data acquisition mechanism for the guardrail status. Combined with the bus-type serial deployment structure and the distributed computing architecture of the edge server, high-density, low-power data perception along the guardrail is achieved, which significantly reduces the overall construction and maintenance costs of the system and overcomes the deployment difficulties brought about by traditional solutions that rely on dense camera deployment and high-power sensors.

[0019] Secondly, the present invention introduces the extended Kalman filter algorithm on the edge processing side to perform nonlinear state estimation and noise suppression on the original sensor data, effectively improving the stability and reliability of the data; at the same time, based on the dual-channel gated convolutional network structure, the acceleration features and spatial position features are parallel modeled and deeply integrated respectively, and then the core collision semantic information is extracted through the nested encoding-decoding mechanism, and the collision judgment function is combined to realize the classification and recognition of various collision states, which greatly improves the accuracy and intelligence level of collision detection, and breaks through the response lag and misjudgment problems of traditional static rules and manual thresholding methods.

[0020] Finally, by combining cellular networks with IOT communication structures, the present invention can achieve an alarm response within seconds after a collision occurs, generate a structured collision data packet and push it to the traffic command center in real time, thereby achieving rapid positioning, intelligent classification and remote response to guardrail impact events, significantly improving the processing efficiency and decision-making support capabilities of road safety incidents, and providing high-value data support for subsequent road health monitoring and intelligent traffic big data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 This is a flow chart of a collision-aware intelligent guardrail method based on Beidou positioning proposed by the present invention; Figure 2 This is a schematic diagram of data processing and forwarding for a collision-aware intelligent guardrail method based on Beidou positioning proposed in the present invention; Figure 3 This is a structural diagram of a collision-aware intelligent guardrail method based on Beidou positioning proposed by the present invention; Figure 4 This is a module structure diagram of a collision-aware intelligent guardrail system based on Beidou positioning proposed by the present invention; Figure 5 This is a global workflow diagram of a collision-aware intelligent guardrail system based on Beidou positioning proposed by the present invention; Figure 6 This is a schematic diagram of the functional modules of a collision-aware intelligent guardrail system based on Beidou positioning proposed in the present invention. DETAILED DESCRIPTION

[0022] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0023] refer to Figure 1-3 A collision-aware intelligent guardrail method based on Beidou positioning includes the following steps: S1, collect sensor data along the guardrail and perform preprocessing; S2, transmits the preprocessed sensor data to the edge server, performs nonlinear time series fitting and smoothing filtering operations to obtain the state vector sequence, and forwards it to the cloud server via the cellular network; S3. Construct a dual-channel gated convolutional network, input the state vector sequence into the gated convolutional network, perform feature extraction and fusion, and output a collision feature vector; S4. Perform multi-layer nested encoding and decoding operations on the collision feature vector to obtain a reduced-dimensional feature representation, and construct a collision determination function to determine whether a collision anomaly is triggered, and mark the collision level label; S5. When the collision level tag is activated, a collision warning data packet is generated, wherein the collision warning data packet includes the collision occurrence time, guardrail position coordinates, and collision severity level; S6. Transmit the collision warning data packet to the road traffic command center for road maintenance personnel to quickly respond and handle.

[0024] The present invention provides a collision-aware intelligent guardrail method based on Beidou positioning. Through six consecutive steps, it constructs a complete guardrail collision monitoring, feature recognition, warning generation and response disposal process, with full-process closed-loop monitoring capabilities, significantly improving the automated processing efficiency and real-time response capabilities of road safety incidents.

[0025] In this embodiment, the sensor data includes spatial coordinate data and acceleration data. The spatial coordinate data is provided by the GNSS positioning module, and the acceleration data is provided by the MPU6050 sensor module. Each module is connected in series through a bus to form a group, and each group contains 20 to 50 modules.

[0026] The present invention integrates the GNSS positioning module and the MPU6050 acceleration sensor module on the guardrail node and adopts a bus series structure to form a group. This not only ensures the integrity and multidimensionality of the collected data, but also effectively reduces the deployment cost and equipment energy consumption, and improves the system's deployment adaptability and maintenance convenience.

[0027] In this embodiment, the preprocessing includes timestamp alignment, coordinate solution, abnormal drift removal, signal denoising, missing value interpolation and data standardization.

[0028] The present invention adopts preprocessing operations such as timestamp alignment, coordinate solution, anomaly removal, denoising and standardization to effectively improve the spatiotemporal consistency and stability of the collected data, providing a high-quality input basis for subsequent state estimation and feature modeling.

[0029] In this embodiment, the edge server is equipped with solar panels and batteries to supply power to the entire formation. It is installed on the edge of the road and has cellular communication capabilities. One end is connected to the formation and the other end is connected to the cloud server.

[0030] The present invention utilizes edge servers combined with solar power supply and cellular communication to achieve a low-power, long-endurance, and high-coverage edge computing architecture, effectively improving the system's real-time processing capabilities and remote data upload efficiency, and reducing dependence on on-site manual intervention.

[0031] In this embodiment, the gated convolutional network is constructed based on a dual-channel mechanism. The first channel receives acceleration data and extracts time-dependent features, and the second channel receives spatial coordinate data and extracts displacement morphological features. The output features of the two channels are fused in a residual connection structure, and a dynamic attention gating mechanism is introduced to perform weighted reconstruction of the time-dependent features and the displacement morphological features to form a collision feature vector.

[0032] The present invention designs a dual-channel gated convolutional network structure based on acceleration and coordinate data, and realizes the deep fusion of time-dependent features and displacement morphological features through residual connection and dynamic attention gating mechanism, which effectively improves the system's recognition accuracy and generalization ability of collision types and state changes.

[0033] In this embodiment, the collision level labels include vehicle collision with guardrail, man-made damage to guardrail, road subsidence, large displacement of guardrail, geological deformation of road, and no collision.

[0034] The present invention clearly classifies and outputs six collision level labels, covering different types of structural abnormalities such as vehicle collision, human damage, road subsidence, etc. It has high-resolution, multi-dimensional state recognition capabilities, which facilitates accurate identification and targeted disposal by traffic management departments.

[0035] In this embodiment, S2 specifically includes: S21, obtain the spatial coordinate data and acceleration data of each acquisition node in the series structure, where the spatial coordinate data is the three-dimensional position data , acceleration data is three-axis acceleration data , Indicates the collection time index, Indicates the The geographical coordinates of the moment, Indicates the acceleration values in three orthogonal directions at the corresponding moment; S22. Align the timestamps of the three-dimensional coordinates and acceleration data of each acquisition node, interpolate missing samples, and perform linear interpolation to unify the sampling frequency and construct an observation sequence set. ; S23, transmit the preprocessed observation sequence set to the edge server, apply nonlinear time series fitting operation and extended Kalman filter algorithm to perform nonlinear state estimation and noise suppression on the observation sequence, and obtain the state vector sequence after filtering. ,in Represents the current position estimate, represents the estimated value of speed, represents the estimated value of acceleration; S24, forwarding the state vector sequence to the cloud server via the cellular network, and the nonlinear state estimation uses the following Kalman state prediction and update, ; in, Indicates the The estimated state value at time t, Indicates the The predicted state value at time t, Indicates the The Kalman gain at time t, represents the time-aligned observation vector, represents the observation matrix.

[0036] The present invention constructs a nonlinear state estimation process and uses the extended Kalman filter algorithm to jointly model acceleration and displacement data to generate a state vector sequence, effectively suppressing environmental noise interference and improving the stability of collision data and the accuracy of structure recognition.

[0037] In this embodiment, S3 specifically includes: S31, build a dual-channel gated convolutional network to transform the state vector sequence and As the input of the first channel and the second channel respectively; S32, will Input to the first channel, perform one-dimensional convolution extraction operation, and generate time-dependent feature tensors ,The first channel uses a convolution kernel with a gated activation function to locally model the acceleration sequence in the time dimension; S33, will Input to the second channel, perform one-dimensional convolution extraction operation, and generate displacement morphological feature tensor , the second channel models the displacement morphological features in the coordinate sequence through a convolutional structure; S34, the output tensors of the two channels and Fusion in the residual connection structure to construct the fusion feature tensor , the fusion process introduces a dynamic attention gating mechanism to form a collision feature vector through weighted reconstruction: ; in, Indicates the The collision feature vector output at the moment, Indicates the The time-dependent feature tensor at time instant, Indicates the The displacement morphological characteristic tensor at time , represents the attention gating weight matrix, represents the gate bias vector, represents the hyperbolic tangent activation function, Represents the Sigmoid activation function.

[0038] The present invention designs a convolutional neural network structure under a gating mechanism, models the state vector by channels, and introduces residual fusion and attention weighting mechanisms to extract collision features from multiple scale levels, thereby enhancing the model's ability to express the coupling relationship between multi-source data.

[0039] In this embodiment, the S4 specifically includes: S41, the collision feature vector Input the nested encoding network structure, perform layer-by-layer nonlinear mapping and compression, and obtain a multi-layer potential feature representation sequence ,in represents the feature vector of the first layer encoding output, represents the feature vector of the second layer encoding output, The feature vector representing the output of the third layer code; S42, will Input the decoding structure, perform reverse mapping and dimensionality reduction reconstruction, and generate a reconstructed feature vector. The dimensionality reduction reconstruction process retains the collision core features and compresses redundant information; S43. Based on the reconstructed feature vector, a collision determination function is constructed to determine whether the current moment is an abnormal collision state. The expression of the collision determination function is: ; in, Indicates the The reconstructed eigenvector at time , represents the first linear mapping matrix, represents the second linear mapping matrix, represents the first bias vector, represents the second bias vector, represents a nonlinear activation function, represents the Softmax normalization function, Indicates the The collision judgment label at the moment includes the vehicle collision guardrail, guardrail damage, road subsidence, guardrail displacement, geological deformation and no collision state.

[0040] The present invention constructs a nested encoding and decoding structure to perform multi-layer dimensionality reduction extraction on the collision feature vector, and constructs a collision judgment function through nonlinear activation and Softmax function, thereby achieving high-precision judgment and classification labeling of multiple types of abnormal states and improving the robustness of system anomaly detection.

[0041] refer to Figure 4-6 , a collision-aware intelligent guardrail system based on Beidou positioning, including: The data processing module is used to collect spatial coordinate data and acceleration data along the guardrail and perform data preprocessing operations; An edge computing module is used to transmit the pre-processed sensor data to the edge server, perform nonlinear time series fitting and smoothing filtering operations, generate a guardrail state vector sequence, and forward the state vector sequence to the cloud server via the cellular network; A feature extraction module is used to construct a dual-channel gated convolutional network, input the state vector sequence into the gated convolutional network, perform feature extraction and fusion, and output a guardrail collision feature vector; The collision determination module is used to perform multi-layer nested encoding and decoding operations on the guardrail collision feature vector to obtain a reduced-dimensional feature representation, and to construct a collision determination function to determine whether a collision anomaly is triggered and mark the collision level label; A warning generation module, configured to generate a collision warning data packet when the collision level tag is activated, wherein the collision warning data packet includes the collision occurrence time, guardrail position coordinates, and collision severity level; The response and disposal module is used to transmit the collision alarm data packet to the road traffic command center for road maintenance personnel to respond and handle quickly.

[0042] At the system structure level, the present invention proposes a complete hardware and algorithm combination architecture, covering various functional units such as data acquisition, edge processing, feature extraction, collision judgment, warning generation and response disposal, and constructs a guardrail collision monitoring solution with intelligent perception, rapid response, low power consumption and high reliability.

[0043] Example 1: To verify the feasibility of the present invention, it was applied to a typical high-accident section of a highway trunk line. This section spans a variety of terrain features and features complex road conditions such as high traffic density, limited nighttime visibility, and small curve radiuses. Traditional guardrails often struggle to provide timely feedback on impact information in the event of a collision, delaying maintenance response. Accident identification often relies on manual video playback, resulting in low efficiency and a high rate of misjudgment. To improve the accuracy and speed of guardrail collision perception, the proposed Beidou-based intelligent guardrail collision sensing method was deployed on this section and subjected to continuous cycle testing and comparative verification.

[0044] During the application process, GNSS positioning modules and MPU6050 acceleration sensor modules were installed every 8 meters along the two-way guardrails of the test section. Each group of sensor nodes was connected to an edge server using a bus. The edge server was powered by solar panels and connected to the cloud via a 4G communication network. The guardrail sensor system collected spatial coordinate data and triaxial acceleration data in real time. The extended Kalman filter algorithm was used to perform nonlinear modeling and smoothing of the multidimensional state, generating a state vector sequence. The state vector was then input into a dual-channel gated convolutional network to extract time-dependent features and displacement morphological features. Deep semantic features were then extracted through a nested encoding and decoding module, ultimately outputting a collision level label.

[0045] During the test, various typical impact scenarios were simulated, including contact impacts with the guardrail from various directions, speeds, and vehicle types. These included minor side impacts, human damage, and slow deformation caused by geological subsidence. Compared to traditional guardrail monitoring methods that rely solely on camera recordings and manual inspections, the system developed by the present invention achieved second-level alarms in all tested events, with an average positioning error of less than 1.3 meters, a collision classification accuracy rate of 96.2%, and a 95.7% consistency with manual verification results.

[0046] Despite high-frequency traffic disruption, the system maintained stable data transmission and a low false alarm rate. The system also demonstrated excellent noise suppression, preventing false alarms from vehicles passing close by. For continuous weak disturbances or low-speed displacement-type deformation scenarios, the system, through deep feature reconstruction and weighted recognition using an attention mechanism, successfully identified four road subsidence incidents and two cases of human intervention. All were reported more than four hours before manual inspections, effectively enhancing the system's proactive anomaly warning capabilities.

[0047] To verify the structured data management capabilities, the system can automatically generate an early warning data packet containing fields such as collision time, coordinate points, acceleration peak, speed change, judgment label, etc. after detecting a collision anomaly, for real-time display and archiving analysis by the traffic control background system, greatly improving the data utilization value and providing a basis for road big data analysis and maintenance scheduling.

[0048] In the test scenario, the present invention significantly improved the perception sensitivity, response speed and processing accuracy of guardrail collision events. Compared with the traditional mode, it reduced the manual investigation workload by more than 65% and shortened the response delay time by about 80%, and has good promotion and application value.

[0049] Table 1 Comparison of core performance indicators of the present invention in guardrail collision sensing applications Test Number Impact type Actual trigger time (seconds) System alarm delay (seconds) Positioning error (meters) Decision Label Recognition accuracy (%) Peak acceleration (g) System false alarms T01 Mild frontal impact 3.1 1.2 1.1 Vehicle collision guardrail 97.5 2.8 none T02 Moderate frontal impact 2.6 1.1 1.0 Vehicle collision guardrail 99.1 4.2 none T03 Angle side impact 4.3 1.6 1.4 Vehicle collision guardrail 94.7 1.9 none T04 Guardrail artificial shaking 5.0 1.3 1.2 Guardrail damaged by human factors 92.8 0.8 none T05 Slow geological subsidence - - 1.5 Road subsidence 95.2 0.5 none T06 Warning friction without collision - - - No collision 100 0.2 none T07 Severe high-speed collision 2.2 0.9 1.0 Vehicle collision guardrail 98.3 6.1 none T08 Guardrail prying and destruction 4.8 1.5 1.3 Guardrail damaged by human factors 91.6 0.9 none Table 1 shows the response effect, positioning accuracy, and recognition accuracy of the present invention to various guardrail abnormal conditions in real test scenarios, further verifying its high robustness and practicality in complex traffic environments.

[0050] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A collision-aware intelligent guardrail method based on Beidou positioning, characterized in that: The steps include: S1, collect sensor data along the guardrail and perform preprocessing; S2, transmits the preprocessed sensor data to the edge server, performs nonlinear time series fitting and smoothing filtering operations to obtain the state vector sequence, and forwards it to the cloud server via the cellular network; S3. Construct a dual-channel gated convolutional network, input the state vector sequence into the gated convolutional network, perform feature extraction and fusion, and output a collision feature vector; S4. Perform multi-layer nested encoding and decoding operations on the collision feature vector to obtain a reduced-dimensional feature representation, and construct a collision determination function to determine whether a collision anomaly is triggered, and mark the collision level label; S5. When the collision level tag is activated, a collision warning data packet is generated, wherein the collision warning data packet includes the collision occurrence time, guardrail position coordinates, and collision severity level; S6. Transmit the collision warning data packet to the road traffic command center for road maintenance personnel to quickly respond and handle.

2. The collision-aware intelligent guardrail method based on Beidou positioning according to claim 1 is characterized in that: The sensor data includes spatial coordinate data and acceleration data. The spatial coordinate data is provided by the GNSS positioning module, and the acceleration data is provided by the MPU6050 sensor module. Each module is connected in series through a bus to form a group, and each group contains 20 to 50 modules.

3. The collision-aware intelligent guardrail method based on Beidou positioning according to claim 1 is characterized in that: The preprocessing includes timestamp alignment, coordinate solution, abnormal drift removal, signal denoising, missing value interpolation and data normalization.

4. The collision-aware intelligent guardrail method based on Beidou positioning according to claim 1 is characterized in that: The edge server is equipped with solar panels and batteries to supply electricity to the entire train. It is installed on the edge of the road and has cellular communication capabilities. One end is connected to the train and the other end is connected to the cloud server.

5. The collision-aware intelligent guardrail method based on Beidou positioning according to claim 1 is characterized in that: The gated convolutional network is built based on a dual-channel mechanism. The first channel receives acceleration data and extracts time-dependent features, and the second channel receives spatial coordinate data and extracts displacement morphological features. The output features of the two channels are fused in a residual connection structure, and a dynamic attention gating mechanism is introduced to perform weighted reconstruction of the time-dependent features and displacement morphological features to form a collision feature vector.

6. The collision-aware intelligent guardrail method based on Beidou positioning according to claim 1 is characterized in that: The collision level labels include vehicle collision with guardrail, man-made damage to guardrail, road subsidence, large displacement of guardrail, geological deformation of road and no collision.

7. The collision-aware intelligent guardrail method based on Beidou positioning according to claim 1 is characterized in that: The S2 specifically includes: S21, obtain the spatial coordinate data and acceleration data of each acquisition node in the series structure, where the spatial coordinate data is the three-dimensional position data , acceleration data is three-axis acceleration data , Indicates the collection time index, Indicates the The geographical coordinates of the moment, Indicates the acceleration values in three orthogonal directions at the corresponding moment; S22. Align the timestamps of the three-dimensional coordinates and acceleration data of each acquisition node, interpolate missing samples, and perform linear interpolation to unify the sampling frequency and construct an observation sequence set. ; S23, transmit the preprocessed observation sequence set to the edge server, apply nonlinear time series fitting operation and extended Kalman filter algorithm to perform nonlinear state estimation and noise suppression on the observation sequence, and obtain the state vector sequence after filtering. ,in Represents the current position estimate, represents the estimated value of speed, represents the estimated value of acceleration; S24, forwarding the state vector sequence to the cloud server via the cellular network, and the nonlinear state estimation uses the following Kalman state prediction and update, ； in, Indicates the The estimated state value at time t, Indicates the The predicted state value at time t, Indicates the The Kalman gain at time t, represents the time-aligned observation vector, represents the observation matrix.

8. The collision-aware intelligent guardrail method based on Beidou positioning according to claim 1 is characterized in that: The S3 specifically includes: S31, build a dual-channel gated convolutional network to transform the state vector sequence and As the input of the first channel and the second channel respectively; S32, will Input to the first channel, perform one-dimensional convolution extraction operation, and generate time-dependent feature tensors ,The first channel uses a convolution kernel with a gated activation function to locally model the acceleration sequence in the time dimension; S33, will Input to the second channel, perform one-dimensional convolution extraction operation, and generate displacement morphological feature tensor , the second channel models the displacement morphological features in the coordinate sequence through a convolutional structure; S34, the output tensors of the two channels and Fusion in the residual connection structure to construct the fusion feature tensor , the fusion process introduces a dynamic attention gating mechanism to form a collision feature vector through weighted reconstruction: ； in, Indicates the The collision feature vector output at the moment, Indicates the The time-dependent feature tensor at time instant, Indicates the The displacement morphological characteristic tensor at time , represents the attention gating weight matrix, represents the gate bias vector, represents the hyperbolic tangent activation function, Represents the Sigmoid activation function.

9. The collision-aware intelligent guardrail method based on Beidou positioning according to claim 1 is characterized in that: The S4 specifically includes: S41, the collision feature vector Input the nested encoding network structure, perform layer-by-layer nonlinear mapping and compression, and obtain a multi-layer potential feature representation sequence ,in represents the feature vector of the first layer encoding output, represents the feature vector of the second layer encoding output, The feature vector representing the output of the third layer code; S42, will Input the decoding structure, perform reverse mapping and dimensionality reduction reconstruction, and generate a reconstructed feature vector. The dimensionality reduction reconstruction process retains the collision core features and compresses redundant information; S43. Based on the reconstructed feature vector, a collision determination function is constructed to determine whether the current moment is an abnormal collision state. The expression of the collision determination function is: ； in, Indicates the The reconstructed eigenvector at time , represents the first linear mapping matrix, represents the second linear mapping matrix, represents the first bias vector, represents the second bias vector, represents a nonlinear activation function, represents the Softmax normalization function, Indicates the The collision judgment label at the moment includes the vehicle collision guardrail, guardrail damage, road subsidence, guardrail displacement, geological deformation and no collision state.

10. A collision-aware intelligent guardrail system based on Beidou positioning, which implements the collision-aware intelligent guardrail method based on Beidou positioning according to any one of claims 1 to 9, characterized in that: include: The data processing module is used to collect spatial coordinate data and acceleration data along the guardrail and perform data preprocessing operations; An edge computing module is used to transmit the pre-processed sensor data to the edge server, perform nonlinear time series fitting and smoothing filtering operations, generate a guardrail state vector sequence, and forward the state vector sequence to the cloud server via the cellular network; A feature extraction module is used to construct a dual-channel gated convolutional network, input the state vector sequence into the gated convolutional network, perform feature extraction and fusion, and output a guardrail collision feature vector; The collision determination module is used to perform multi-layer nested encoding and decoding operations on the guardrail collision feature vector to obtain a reduced-dimensional feature representation, and to construct a collision determination function to determine whether a collision anomaly is triggered and mark the collision level label; A warning generation module, configured to generate a collision warning data packet when the collision level tag is activated, wherein the collision warning data packet includes the collision occurrence time, guardrail position coordinates, and collision severity level; The response and disposal module is used to transmit the collision alarm data packet to the road traffic command center for road maintenance personnel to respond and handle quickly.

Images