A vehicle overload detection method and apparatus

Abstract

This invention relates to the field of vehicle overload technology and discloses a vehicle overload detection method and device, comprising the following steps: acquiring acoustic emission signals, geomagnetic disturbance signals, infrared thermal radiation signals, distributed vibration signals, and vehicle contour data when a vehicle passes through through a multi-physics field sensing step; performing spatiotemporal alignment and kinematic calculation on the acquired multi-physics field signals through a vehicle behavior feature analysis step; by extracting vehicle lateral sway, instantaneous impact, and tire ground contact features, this invention can effectively identify cheating behaviors such as S-shaped driving, skipping the weighbridge, and rushing the weighbridge. At the same time, the system dynamically adjusts the overload threshold according to real-time road conditions, weather, and historical false alarm rates, significantly reducing false alarms and re-inspection pressure, and achieving intelligent adaptive overload judgment; based on this, this invention uses blockchain evidence storage technology to encapsulate overload judgment logs, cheating tags, and license plate information, and hashes them onto the blockchain to form an immutable and complete law enforcement evidence chain.

Description

Technical Field

[0001] This invention relates to the field of vehicle overload technology, specifically to a method and apparatus for detecting vehicle overload. Background Technology

[0002] Overloaded vehicles seriously endanger the structural safety of highway bridges, cause traffic accidents, and disrupt the transportation market order, making them a key target for traffic management both domestically and internationally. Currently, existing vehicle overload detection technologies are mainly divided into two categories: static weighing and dynamic weighing. While static weighing (such as electronic weighbridges and portable axle load meters) offers high accuracy, it requires vehicles to be stopped for inspection, resulting in low efficiency, potential traffic congestion at toll stations or overload control stations, and significant manual intervention, making it difficult to achieve continuous, non-stop monitoring of vehicles traveling on highways. Dynamic weighing technology uses sensors embedded in the road surface (such as bending plate sensors, piezoelectric quartz sensors, and resistance strain gauge sensors) to measure the axle load of moving vehicles, improving detection efficiency to some extent, but still suffers from the following technical drawbacks: 1. Limited detection dimensions and weak anti-interference capabilities: Existing dynamic weighing systems mainly rely on a single weighing sensor (such as a bending plate or strain gauge), which is easily affected by environmental factors such as temperature, humidity, road surface smoothness, and electromagnetic interference (such as lightning strikes), resulting in large weighing errors (usually ±6% to ±10%). Furthermore, the system cannot correct the errors in real time, making it difficult to effectively identify and respond to cheating behaviors: Some drivers use methods such as S-shaped driving, skipping the scale, rushing the scale, and obscuring license plates to interfere with the weighing results. Traditional systems lack the multimodal joint detection capability for abnormal driving trajectories and instantaneous impact behaviors, allowing overloaded vehicles to evade supervision. 2. Incomplete and easily tampered evidence chain in law enforcement: The weighing records, vehicle information and judgment results of the existing system are mostly stored in local databases, lacking anti-tampering mechanisms. It is difficult to provide complete evidence with judicial validity in law enforcement disputes. Poor adaptability and lagging model updates: The overload threshold of the traditional system is fixed or can only be adjusted manually. It cannot be dynamically optimized according to real-time road conditions, weather changes and historical false alarm rates. Once the detection model is deployed, it is difficult to use actual weighing data for incremental learning and remote upgrades, resulting in a decrease in the accuracy of the system after long-term operation. Therefore, there is an urgent need to provide a vehicle overload detection method and device that can integrate multi-physics field perception, identify cheating behavior, establish an tamper-proof evidence chain, and have self-evolution capabilities to overcome the shortcomings of existing technologies. Summary of the Invention

[0003] The purpose of this invention is to provide a method and apparatus for detecting vehicle overload, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting vehicle overload, comprising the following steps: The system collects acoustic emission signals, geomagnetic disturbance signals, infrared thermal radiation signals, distributed vibration signals, and vehicle contour data as the vehicle passes through via a multi-physics sensing process. The vehicle behavior feature analysis steps are used to perform spatiotemporal alignment and kinematic calculation on the collected multi-physics field signals to extract behavioral parameters including axle shape, velocity, lateral sway amplitude and tire contact geometry. By fusing bending plate sensor data and acoustic vibration indirect load inversion results through a dynamic and reliable weighing process, and using geomagnetic laser information for correction, a preliminary weight estimate with confidence is generated. The overload adaptive judgment step dynamically adjusts the overload threshold based on real-time road conditions, weather levels, and historical false alarm rates, and uses fuzzy logic and evidence-weighted voting to output the overload probability and level. By using anti-cheating and blockchain evidence storage steps, abnormal driving trajectories and instantaneous impact behaviors are detected in real time. The detection results, vehicle identity information and judgment logs are encapsulated and hashed and uploaded to the blockchain. Incremental learning is performed using a manually verified sample library through model self-evolution and remote calibration steps, and the updated model parameters and error compensation table are sent to the roadside detection equipment.

[0005] Preferably, the multi-physics sensing step further includes: collecting tire-road contact acoustic patterns using a low-frequency acoustic emission sensor array; calculating the number and spacing of axles using a triaxial geomagnetic disturbance detection module; identifying braking thermal anomalies using an infrared thermal imaging and temperature compensation module; capturing roadbed vibration waveforms and calculating wheel and axle loads using a distributed fiber optic vibration sensing module; measuring the vehicle body subsidence and tire ground contact length using a lidar contour scanning submodule; and providing microsecond-level time synchronization for each submodule through a synchronization timing and trigger controller.

[0006] Preferably, the vehicle behavior feature analysis step further includes: performing cross-correlation delay estimation and interpolation resampling on multi-source signals to achieve spatiotemporal alignment; fusing the light curtain occlusion sequence and geomagnetic disturbance abrupt change edges, and using a one-dimensional convolutional network to output the number of axes and axis shape; calculating the vehicle speed, acceleration, and lateral sway amplitude using the lidar point cloud and the trigger time difference of the inductive loop; reconstructing the ground imprint of each tire based on the acoustic emission spatial spectrum and the longitudinal peak distribution of fiber vibration; and combining the above parameters into an initial feature pool of load with a confidence label.

[0007] Preferably, the dynamic reliable weighing step further includes: directly measuring the instantaneous pressure curve of the wheel axle using an embedded bending plate sensor; mapping the acoustic emission array and fiber optic vibration waveform to indirect load estimation using a pre-trained deep residual network; correcting the system error of the bending plate sensor based on the geomagnetic disturbance integral area and the vehicle magnetic characteristic model; applying consistency constraints to the axle group using the car body subsidence measured by lidar; calculating the internal consistency of multi-source data and outputting the reliable weighing score; and using adaptive Kalman filtering to fuse the bending plate main measurement and auxiliary inversion results, while simultaneously correcting for slippage based on vehicle speed, temperature, and humidity using a dynamic error compensation table.

[0008] Preferably, the overload adaptive determination step further includes: dynamically calculating a basic threshold based on real-time road conditions and weather levels; using fusion total weight, confidence level, abnormal behavior indicators, and historical false alarm rates as inputs, and outputting the overload probability membership degree using fuzzy logic reasoning; synthesizing evidence from three independently estimated weights (bend plate, acoustic vibration, and geomagnetic laser) based on real-time confidence levels, and outputting a weighted overload confidence interval; and outputting a graded early warning signal based on the overload probability and confidence interval, while simultaneously generating a determination log.

[0009] Preferably, the vehicle overload detection device includes: The multi-physics sensing unit is used to collect acoustic emission, geomagnetic disturbance, infrared thermal radiation, distributed vibration, and lidar contour data when a vehicle passes by. The vehicle behavior feature analysis unit is connected to the multi-physics sensing unit and is used to perform spatiotemporal alignment and kinematic calculation on the sensing signals to extract axle shape, speed, lateral sway amplitude and tire contact geometry parameters. The dynamic reliable weighing unit is connected to the vehicle behavior feature analysis unit and the multi-physics field sensing unit respectively. It is used to fuse the bending plate sensor data and the acoustic vibration inversion results, and to use geomagnetic laser information for correction to generate a weight estimate with confidence. An overload adaptive determination unit, connected to the dynamic reliable weighing unit and the vehicle behavior feature analysis unit, is used to dynamically adjust the overload threshold and output the overload probability and level using fuzzy logic and evidence-weighted voting. The anti-cheating and blockchain evidence storage unit is connected to the vehicle behavior feature analysis unit, dynamic trusted weighing unit and overload adaptive judgment unit. It is used to identify abnormal driving trajectory and instantaneous impact behavior, and encapsulate the judgment log and evidence and hash them onto the blockchain. The model self-evolution and remote calibration unit is connected to the multi-physics sensing unit, dynamic reliable weighing unit and overload adaptive judgment unit, and is used to perform incremental learning using the sample library and issue updated model parameters and error compensation tables.

[0010] Preferably, the multi-physics sensing unit includes: a low-frequency acoustic emission sensing array, a three-axis geomagnetic disturbance detection module, an infrared thermal imaging and temperature compensation module, a distributed fiber optic vibration sensing module, a lidar contour scanning submodule, and a synchronization timing and triggering controller; the synchronization timing and triggering controller is connected to the other submodules respectively, and is used to provide microsecond-level time synchronization and trigger the acquisition window.

[0011] Preferably, the vehicle behavior feature analysis unit includes: a multi-source signal spatiotemporal alignment module, a vehicle separation and axle type identification submodule, a kinematic parameter estimation module, a tire contact geometry inversion module, and a load initial feature pool module; the load initial feature pool module is connected to the kinematic parameter estimation module and the tire contact geometry inversion module respectively, and is used to generate feature vectors with confidence labels and send them to the dynamic confidence weighing unit.

[0012] Preferably, the dynamic reliable weighing unit includes: an embedded bending plate sensor module, an acoustic vibration indirect load inversion module, a geomagnetic laser fusion correction module, a multi-mode confidence assessment module, a Kalman fusion filtering module, and a dynamic error compensation table module; the Kalman fusion filtering module is connected to the embedded bending plate sensor module, the acoustic vibration indirect load inversion module, and the geomagnetic laser fusion correction module respectively, and is used to output the optimal estimated axle load and total weight; the dynamic error compensation table module stores slip compensation coefficients under different vehicle speeds, temperatures, and humidity and is periodically updated by the model self-evolution unit.

[0013] Preferably, the anti-cheating and blockchain evidence storage unit includes: a vehicle trajectory anomaly identification module, an instantaneous impact detection module, a vehicle identity anti-tampering module, an evidence chain encapsulation module, and a hash-on-chain and joint enforcement module; the input end of the vehicle trajectory anomaly identification module is connected to the vehicle behavior feature analysis unit to obtain the lateral sway amplitude time series; the input end of the instantaneous impact detection module is connected to the dynamic trusted weighing unit to obtain the original bending plate signal; the evidence chain encapsulation module is connected to the vehicle trajectory anomaly identification module, the instantaneous impact detection module, the vehicle identity anti-tampering module, and the overload adaptive judgment unit, respectively, and is used to encapsulate the overload judgment log, cheating label, and license plate information into a data packet; the hash-on-chain and joint enforcement module is connected to the evidence chain encapsulation module and is used to upload the hash value of the data packet to the blockchain for access by the enforcement platform.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention integrates multi-physics field sensing data such as acoustic emission, geomagnetism, infrared, fiber optic vibration, and lidar, and combines Kalman filtering and confidence assessment to significantly improve dynamic weighing accuracy and environmental robustness. By extracting vehicle lateral sway, instantaneous impact, and tire ground contact characteristics, it can effectively identify cheating behaviors such as S-shaped driving, skipping the scale, and rushing the scale, ensuring the fairness of detection. At the same time, the system dynamically adjusts the overload threshold based on real-time road conditions, weather, and historical false alarm rates. Combined with fuzzy logic and evidence-weighted voting, it greatly reduces false alarms and re-inspection pressure, achieving intelligent adaptive overload judgment. Building upon this foundation, this invention employs blockchain evidence storage technology to encapsulate overload judgment logs, cheating tags, and license plate information, hashing and uploading them to the blockchain to form an immutable and complete chain of law enforcement evidence, thereby enhancing regulatory credibility. The system supports model self-evolution, periodically updating the inversion network and error compensation table through edge sampling and cloud-based incremental learning, and remotely distributing the updates. This results in low long-term maintenance costs and self-diagnostic capabilities. The fully automated and modular design allows for flexible deployment at key nodes such as highway entrances and provincial borders, comprehensively improving the intelligence level of overload management. Attached Figure Description

[0015] Figure 1 This is a block diagram of the overall structure of the vehicle overload detection method and device of the present invention; Figure 2 This is a structural block diagram of the vehicle behavior detection unit in this invention; Figure 3 This is a structural block diagram of the vehicle weighing unit in this invention; Figure 4 This is a structural block diagram of the vehicle overload determination unit in this invention; Figure 5 This is a structural block diagram of the vehicle anti-cheating unit in this invention; Figure 6 This is a structural block diagram of the vehicle anti-cheating determination unit in this invention; Figure 7 This is a structural block diagram of the vehicle detection control unit in this invention; Figure 8 This is a structural block diagram of the vehicle perception data alignment unit in this invention; Figure 9 This is a structural block diagram of the vehicle data accuracy determination unit in this invention; Figure 10 This is a structural block diagram of the vehicle detection data uplink unit in this invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 As shown in Figure 10, a vehicle overload detection method includes six core functional units: a multi-physics field perception unit, a vehicle behavior feature analysis unit, a dynamic reliable weighing unit, an overload adaptive judgment unit, an anti-cheating and blockchain evidence storage unit, and a model self-evolution and remote calibration unit. The system comprises several components: a multi-physics sensing unit (MPU) for collecting multi-dimensional signals such as acoustic emission, geomagnetic disturbance, infrared thermal radiation, and vibration response from passing vehicles, providing raw data for subsequent analysis; a vehicle behavior feature analysis unit for spatiotemporal alignment and kinematic calculation of the sensing signals, extracting behavioral parameters such as axle shape, speed, and lateral offset; a dynamic and reliable weighing unit for integrating a bending plate sensor with an indirect inversion model to generate a preliminary weight estimate and its confidence level; an overload adaptive judgment unit for dynamically adjusting thresholds based on real-time road conditions and historical false alarm rates, and introducing fuzzy decision-making logic; an anti-cheating and blockchain evidence storage unit for real-time detection of abnormal behaviors such as S-shaped driving, skipping weighbridges, and obscuring license plates, hashing the detection results, vehicle characteristics, timestamps, and other information onto the blockchain; and a model self-evolution and remote calibration unit for using an edge-cloud collaborative mechanism to periodically perform incremental learning on the detection model using actual weighing data and sending updated parameters to roadside equipment. The six units are connected in series with a feedback loop: the perception unit outputs raw features to the behavior analysis unit, and the analysis results are simultaneously sent to the dynamic weighing unit and the anti-cheating unit; the judgment unit integrates the weighing results with confidence, abnormal behavior indicators and adaptive thresholds, and outputs the overload probability and level; the evidence storage unit solidifies the evidence of the entire process; the self-evolution unit periodically pulls the optimized detection and judgment models from the cloud to form a closed loop, which significantly improves the detection robustness and law enforcement credibility in complex scenarios.

[0018] Furthermore, the multi-physics sensing unit includes the following six sub-modules: (1) Low-frequency acoustic emission sensing array—a MEMS microphone array buried on both sides of the road surface to collect 10~500Hz acoustic wave signals generated by tire-road contact, used to invert the tire ground pressure distribution; (2) Triaxial geomagnetic disturbance detection module—using the distortion effect of the geomagnetic field on the vehicle's ferromagnetic components when they pass through, to estimate the number of axles, spacing, and approximate load range; (3) Infrared thermal imaging and temperature compensation module—capturing the temperature field of the brake drum and tire tread, combined with ambient temperature compensation, to identify long-term braking or severe... (4) Distributed fiber optic vibration sensing module - an interferometric fiber optic cable laid longitudinally along the road to capture the submicron level vibration waveform of the roadbed caused by the passing of vehicles, and to calculate the wheel axle load by measuring the wave velocity and amplitude attenuation; (5) LiDAR contour scanning submodule - a low-beam lidar installed on the gantry to measure the amount of the vehicle compartment sinking and the tire ground contact length in real time after the vehicle is fully loaded; (6) Synchronization time and trigger controller - based on the IEEE1588 protocol, it provides microsecond-level time synchronization for all submodules and triggers the acquisition window according to the ground induction coil or millimeter-wave radar signal. The output of each submodule in this unit is the original time sequence signal or image, which is packaged by the synchronization controller and sent to the vehicle behavior feature analysis unit, while retaining a copy of the original data for the self-evolution unit to perform offline analysis.

[0019] The vehicle behavior feature analysis unit includes five sub-modules: (1) Multi-source signal spatiotemporal alignment module - based on cross-correlation time delay estimation and interpolation resampling, the sampling points of acoustic, magnetic, vibration and other signals are aligned to the same time reference; (2) Vehicle separation and axle type recognition sub-module - fused light curtain occlusion sequence and geomagnetic disturbance abrupt change edge, and used a one-dimensional convolutional network to output the number of axles, and loaded axle type and tire width information; (3) Kinematic parameter estimation module - using the continuous frame lidar point cloud and ground induction coil trigger time difference to calculate vehicle speed, acceleration and lateral sway amplitude; (4) Tire grounding geometry inversion module - based on the spatial spectrum of acoustic emission array and the peak distribution of fiber vibration along the longitudinal direction of the road, the grounding imprint length and width of each tire are reconstructed; (5) Load initial feature pool module - the above axle type, speed, sway amplitude, grounding geometry and other parameters are combined into a fixed length feature vector, and each feature is attached with a real-time credibility label (e.g., the credibility of lidar is reduced due to rain and fog). The unit's output is bidirectionally connected to the dynamic trusted weighing unit: on one hand, it sends the initial feature pool to the weighing unit as auxiliary input; on the other hand, it receives residual information from the weighing unit for online correction of the feature extractor's weights. Simultaneously, the "lateral sway amplitude" and "axis shape information" in the feature pool are directly sent to the anti-cheating and blockchain evidence storage unit as the basis for judging S-shaped driving or serpentine movement.

[0020] The dynamic reliable weighing unit consists of six sub-modules: (1) Embedded bending plate sensor module - two rows of staggered resistance strain gauge bending plates (range 0~20t, sampling rate 2kHz) directly measure the instantaneous pressure curve of the wheel axle; (2) Acoustic vibration indirect load inversion module - using a pre-trained deep residual network to map the acoustic emission array and fiber optic vibration waveform to the estimated load of each wheel; (3) Geomagnetic laser fusion correction module - correcting the system error of the bending plate sensor based on the geomagnetic disturbance integral area and the vehicle magnetic feature model, and using the amount of car body sinking measured by the laser radar. Consistency constraints are applied to subsequent axle groups; (4) Multi-mode confidence assessment module - calculates the internal consistency between each sensor source (such as the variance of the ratio of the peak value of the bending plate to the peak value of the acoustic inversion), and outputs a weighing confidence score of 0 to 1; (5) Kalman fusion filtering module - takes the bending plate measurement as the main input, and uses acoustic vibration inversion and geomagnetic correction as the observation update, and uses adaptive Kalman filtering to iteratively output the optimal estimated axle weight, axle group weight and total weight; (6) Dynamic error compensation table module - stores the slip difference compensation coefficients for different vehicle speeds, temperatures and road surface humidity, which are periodically updated by the self-evolution unit and corrected by looking up the table. This unit receives the speed, axle type and confidence label provided by the vehicle behavior feature analysis unit, and its output fused weight and confidence score are sent to the overload adaptive judgment unit; at the same time, the original bending plate signal and the intermediate inversion results are sent to the anti-cheating unit through the internal bus to detect instantaneous impact cheating behaviors such as "jumping the scale" and "rushing the scale".

[0021] The overload adaptive judgment unit consists of four sub-modules working together: (1) Dynamic threshold generator - based on the latest road surface condition (friction coefficient, smoothness index) and current weather level (rain / snow / fog) issued by the self-evolution unit, dynamically calculates the basic threshold = statutory limit × (1 + system error margin), and the error margin is adaptively adjusted with vehicle speed and lateral sway index; (2) Fuzzy logic reasoning module - using the total weight, confidence level, abnormal behavior signs, and historical false alarm rate as input, designs 5 levels of overload probability membership functions, and outputs "definitely overloaded" and "suspected overloaded". Intermediate conclusions such as "determined not to be overloaded"; (3) Evidence weighted voting device - to prevent misjudgment caused by the failure of a single sensor, the weights of the three independent estimated weights of the bending plate, acoustic vibration, and geomagnetic laser are synthesized into DS evidence according to the real-time confidence level, and the weighted overload confidence interval is given; (4) Graded early warning and execution interface - according to the overload probability and confidence interval, different levels of alarm signals are output (Level 1: information board prompts re-inspection; Level 2: linkage to toll station prohibits entry; Level 3: automatic notification to road administration to intercept), and all input parameters of the judgment process are packaged into "judgment log" for the evidence storage unit to solidify. The input sources of this unit include the fused weight of the dynamic reliable weighing unit, the abnormal behavior flag of the behavior feature analysis unit, and the threshold coefficient issued by the self-evolution unit; its output alarm signal directly drives the variable information board and the barrier machine, and at the same time sends the judgment log to the anti-cheating and blockchain evidence storage unit in binary format.

[0022] The anti-cheating and blockchain evidence storage unit has five sub-modules: (1) a vehicle trajectory anomaly identification module—using lidar and geomagnetic disturbance array to reconstruct the lateral deviation trajectory of the vehicle when passing through, and detect typical cheating modes such as S-shaped driving, serpentine movement, and sudden acceleration and deceleration; (2) an instantaneous impact detection module—performing wavelet decomposition on the original signal of the bending plate to identify the instantaneous spike of the first axle and the missing mode of subsequent axles caused by the jumping of the weighbridge; (3) a vehicle identity anti-tampering module—combining high-definition capture of license plate OCR, electronic license plate RFID reading and writing, and body color / decoration features. The lightweight comparison outputs the license plate consistency confidence level; (4) Evidence chain encapsulation module - the overload judgment log, key intermediate data of each sub-module (soundprint spectrum, vibration waveform segment, cheating label), license plate information, time location stamp, and device ID are encapsulated into verifiable data packets according to the national standard GB / T29315; (5) Hash on-chain and joint law enforcement module - the hash value of the data packet is uploaded to the consortium blockchain (traffic supervision chain) through 5GRSU, the original data is stored in the toll station edge server, and a query interface is provided for road administration and traffic police to access in real time. This module is directly connected to the judgment unit and the behavior feature analysis unit: the driving trajectory abnormality recognition module needs the lateral swing amplitude time series provided by the behavior feature unit, and the instantaneous impact detection module needs to obtain the original bending plate signal stream of the dynamic reliable weighing unit. The final generated on-chain credentials can be connected to the public security traffic integrated command platform through the cloud platform to form an undeniable law enforcement credentials.

[0023] The model self-evolution and remote calibration unit includes four sub-modules: (1) Edge data sampling sub-module - randomly saves the original data of the entire vehicle process that has been manually reviewed (re-inspected static weighbridge) at a ratio of 1% to form a sample library with truth labels; (2) Cloud incremental training engine - uses the sample library once a month to incrementally learn the acoustic vibration inversion network, fuzzy rule weights, and dynamic error compensation table to avoid catastrophic forgetting, and at the same time uses the federated learning framework to aggregate the local updates of multiple toll stations; (3) Model compression and secure distribution module - converts the updated model into TensorFlowLite format and distributes it to the edge industrial control computer of each detection point through the TLS encrypted channel; (4) Online self-diagnosis and alarm module - monitors the signal quality of each sensor (such as the self-test pulse response of the acoustic emission array and the attenuation of fiber optic power), and automatically switches to the degradation mode and notifies the maintenance center when the number of effective channels is lower than the threshold. This unit interacts with the sensing unit (acquiring a copy of the original data), the weighing unit (updating the error compensation table), and the judgment unit (issuing dynamic threshold coefficients), forming a complete closed loop from data acquisition and model optimization to parameter updates. This ensures that the system maintains a stable and reliable overload detection capability regardless of changes in the geographical area and season.

[0024] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0025] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting vehicle overload, characterized in that: Includes the following steps: The system collects acoustic emission signals, geomagnetic disturbance signals, infrared thermal radiation signals, distributed vibration signals, and vehicle contour data as the vehicle passes through the multi-physics field sensing process. The vehicle behavior feature analysis steps are used to perform spatiotemporal alignment and kinematic calculation on the collected multi-physics field signals to extract behavioral parameters including axle shape, velocity, lateral sway amplitude and tire contact geometry. By fusing bending plate sensor data and acoustic vibration indirect load inversion results through a dynamic and reliable weighing process, and using geomagnetic laser information for correction, a preliminary weight estimate with confidence is generated. The overload adaptive judgment step dynamically adjusts the overload threshold based on real-time road conditions, weather levels, and historical false alarm rates, and uses fuzzy logic and evidence-weighted voting to output the overload probability and level. By using anti-cheating and blockchain evidence storage steps, abnormal driving trajectories and instantaneous impact behaviors are detected in real time. The detection results, vehicle identity information and judgment logs are encapsulated and hashed and uploaded to the blockchain. Incremental learning is performed using a manually verified sample library through model self-evolution and remote calibration steps, and the updated model parameters and error compensation table are sent to the roadside detection equipment.

2. The vehicle overload detection method according to claim 1, characterized in that: The multi-physics sensing steps further include: collecting tire-road contact acoustic patterns using a low-frequency acoustic emission sensor array; calculating the number and spacing of axles using a triaxial geomagnetic disturbance detection module; identifying braking thermal anomalies using an infrared thermal imaging and temperature compensation module; capturing roadbed vibration waveforms and calculating wheel and axle loads using a distributed fiber optic vibration sensing module; measuring the vehicle body subsidence and tire ground contact length using a lidar contour scanning submodule; and providing microsecond-level time synchronization for each submodule through a synchronization timing and trigger controller.

3. The vehicle overload detection method according to claim 1, characterized in that: The vehicle behavior feature analysis steps further include: performing cross-correlation delay estimation and interpolation resampling on multi-source signals to achieve spatiotemporal alignment; fusing light curtain occlusion sequences and geomagnetic disturbance abrupt edges, and using a one-dimensional convolutional network to output the number of axes and axis shape; calculating vehicle speed, acceleration, and lateral sway amplitude using lidar point clouds and ground loop trigger time differences; reconstructing the ground imprint of each tire based on the acoustic emission spatial spectrum and the longitudinal peak distribution of fiber vibration; and combining the above parameters into an initial feature pool of load with a confidence label.

4. The vehicle overload detection method according to claim 1, characterized in that: The dynamic reliable weighing step further includes: directly measuring the instantaneous pressure curve of the wheel axle using an embedded bending plate sensor; mapping the acoustic emission array and fiber optic vibration waveform to indirect load estimation using a pre-trained deep residual network; correcting the system error of the bending plate sensor based on the integral area of ​​geomagnetic disturbance and the vehicle magnetic characteristic model; constraining the axle group consistency using the amount of carriage subsidence measured by lidar; calculating the internal consistency of multi-source data and outputting the reliable weighing score; and using adaptive Kalman filtering to fuse the main measurement and auxiliary inversion results of the bending plate, while simultaneously correcting for slippage based on vehicle speed, temperature, and humidity using a dynamic error compensation table.

5. The vehicle overload detection method according to claim 1, characterized in that: The overload adaptive judgment step further includes: dynamically calculating the basic threshold based on real-time road conditions and weather levels; using fusion total weight, confidence level, abnormal behavior indicators, and historical false alarm rates as inputs, and outputting the overload probability membership degree using fuzzy logic reasoning; synthesizing evidence from three independently estimated weights (bend plate, acoustic vibration, and geomagnetic laser) based on real-time confidence levels, and outputting the weighted overload confidence interval; and outputting graded early warning signals based on the overload probability and confidence interval, while simultaneously generating a judgment log.

6. The vehicle overload detection device according to claim 1, characterized in that: include: The multi-physics sensing unit is used to collect acoustic emission, geomagnetic disturbance, infrared thermal radiation, distributed vibration, and lidar contour data when a vehicle passes by. The vehicle behavior feature analysis unit is connected to the multi-physics sensing unit and is used to perform spatiotemporal alignment and kinematic calculation on the sensing signals to extract axle shape, speed, lateral sway amplitude and tire contact geometry parameters. The dynamic reliable weighing unit is connected to the vehicle behavior feature analysis unit and the multi-physics field sensing unit respectively. It is used to fuse the bending plate sensor data and the acoustic vibration inversion results, and to use geomagnetic laser information for correction to generate a weight estimate with confidence. An overload adaptive determination unit, connected to the dynamic reliable weighing unit and the vehicle behavior feature analysis unit, is used to dynamically adjust the overload threshold and output the overload probability and level using fuzzy logic and evidence-weighted voting. The anti-cheating and blockchain evidence storage unit is connected to the vehicle behavior feature analysis unit, dynamic trusted weighing unit and overload adaptive judgment unit. It is used to identify abnormal driving trajectory and instantaneous impact behavior, and encapsulate the judgment log and evidence and hash them onto the blockchain. The model self-evolution and remote calibration unit is connected to the multi-physics sensing unit, dynamic reliable weighing unit and overload adaptive judgment unit, and is used to perform incremental learning using the sample library and issue updated model parameters and error compensation tables.

7. The vehicle overload detection device according to claim 1, characterized in that: The multi-physics sensing unit includes: a low-frequency acoustic emission sensing array, a three-axis geomagnetic disturbance detection module, an infrared thermal imaging and temperature compensation module, a distributed fiber optic vibration sensing module, a lidar contour scanning submodule, and a synchronization timing and triggering controller; the synchronization timing and triggering controller is connected to the other submodules respectively, and is used to provide microsecond-level time synchronization and trigger the acquisition window.

8. The vehicle overload detection device according to claim 1, characterized in that: The vehicle behavior feature analysis unit includes: a multi-source signal spatiotemporal alignment module, a vehicle separation and axle type identification submodule, a kinematic parameter estimation module, a tire contact geometry inversion module, and a load initial feature pool module; the load initial feature pool module is connected to the kinematic parameter estimation module and the tire contact geometry inversion module respectively, and is used to generate feature vectors with confidence labels and send them to the dynamic confidence weighing unit.

9. A vehicle overload detection device according to claim 1, characterized in that: The dynamic reliable weighing unit includes: an embedded bending plate sensor module, an acoustic vibration indirect load inversion module, a geomagnetic laser fusion correction module, a multi-mode confidence assessment module, a Kalman fusion filtering module, and a dynamic error compensation table module. The Kalman fusion filtering module is connected to the embedded bending plate sensor module, the acoustic vibration indirect load inversion module, and the geomagnetic laser fusion correction module, respectively, and is used to output the optimal estimated axle load and total weight. The dynamic error compensation table module stores the slip compensation coefficients under different vehicle speeds, temperatures, and humidity levels and is updated periodically by the model self-evolution unit.

10. A vehicle overload detection device according to claim 1, characterized in that: The anti-cheating and blockchain evidence storage unit includes: a vehicle trajectory anomaly identification module, an instantaneous impact detection module, a vehicle identity anti-tampering module, an evidence chain encapsulation module, and a hash-on-chain and joint enforcement module. The input of the vehicle trajectory anomaly identification module is connected to the vehicle behavior feature analysis unit to obtain the time series of lateral sway amplitude. The input of the instantaneous impact detection module is connected to the dynamic trusted weighing unit to obtain the original bending plate signal. The evidence chain encapsulation module is connected to the vehicle trajectory anomaly identification module, the instantaneous impact detection module, the vehicle identity anti-tampering module, and the overload adaptive judgment unit, respectively, and is used to encapsulate the overload judgment log, cheating label, and license plate information into a data packet. The hash-on-chain and joint enforcement module is connected to the evidence chain encapsulation module and is used to upload the hash value of the data packet to the blockchain for access by the enforcement platform.

Images