Highway green channel goods detection system based on license plate recognition and internal structure feature extraction
The multi-sensor collaborative detection system has solved the problem of identifying the internal structure of goods in vehicle inspections during highway green channels, enabling efficient verification and stable supervision without stopping, thus improving traffic efficiency and supervision effectiveness.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- KUNMING UNIV OF SCI & TECH
- Filing Date
- 2026-02-05
- Publication Date
- 2026-06-12
AI Technical Summary
The existing vehicle inspection methods for highway green channels rely on manual experience, which makes it difficult to accurately identify the internal structure of goods. Goods are easily obscured, covered, or mixed, resulting in inconsistent inspection standards and problems such as low traffic efficiency and insufficient supervision.
The system employs a multi-sensor collaborative detection system, including a license plate recognition and reservation management module, a cargo information feature extraction unit, a structural information acquisition array, a lidar module, a dynamic weighing module, an infrared temperature acquisition module, and an intelligent identification and judgment center. This system enables vehicle identification, acquisition and automatic judgment of cargo compartment internal structural information, and comprehensive judgment based on loading density, temperature consistency, and structural consistency.
It enables cargo verification without stopping, identifies obstruction, covering and mixed loading, improves traffic efficiency, ensures the stability and reliability of judgment results, and forms a traceable intelligent supervision system.
Smart Images
Figure CN122200618A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent traffic detection and information recognition technology, specifically a highway green channel cargo detection system based on license plate recognition and internal structural feature extraction. Background Technology
[0002] The highway green channel policy primarily targets vehicles transporting fresh agricultural products, implementing fast passage and toll reduction measures at toll stations. To ensure the authenticity and effectiveness of the policy implementation, accurate verification of the type of goods transported and the loading conditions of vehicles entering the green channel is necessary. Currently, the verification methods for green channel vehicles mainly rely on manual inspection, on-site visual confirmation, and random sampling, with some toll stations supplementing with video surveillance images, manual weighing, or visual comparison. These inspection methods largely depend on human experience and are greatly affected by factors such as environment, personnel, lighting, and obstruction by the cargo container, making it difficult to standardize inspection standards and resulting in fluctuations in accuracy.
[0003] In actual traffic, some vehicles may evade tolls or regulations by using methods such as obscuring, covering, stacking, or placing goods in layers to mix or falsely declare cargo. Since the interior of the cargo compartment cannot be directly observed, relying solely on external images, vehicle weight, or declared information is insufficient to reflect the true internal structure of the goods, making it difficult to effectively identify acts such as mixing, camouflage, and substituting inferior goods. Furthermore, most existing inspection methods require vehicles to briefly stop for verification, easily causing congestion at toll stations and reducing traffic efficiency. Simultaneously, the lack of automatic recording and unified storage mechanisms for inspection results weakens subsequent traceability and cross-regional comparison capabilities, hindering continuous supervision. Summary of the Invention
[0004] To address the aforementioned issues, this invention aims to propose a highway green channel cargo inspection system based on license plate recognition and internal structural feature extraction. This system is installed within highway toll station lanes and automatically verifies green channel vehicles through multi-sensor collaborative detection. It achieves vehicle identity information acquisition, cargo compartment internal structural information collection, external parameters and loading status analysis, and completes intelligent cargo consistency determination and release control.
[0005] To achieve the above objectives, the present invention provides the following technical solution: The system described in this invention includes a license plate recognition and reservation management module (1), a cargo information feature extraction unit (2), a structural information acquisition array (3), a lidar module (4), a dynamic weighing module (5), an infrared temperature acquisition module (6), an intelligent recognition and judgment center (7), an output power adaptive control module (8), a cloud monitoring platform (9), a lifting height limit gantry (10), a barrier gate lifting arm (11), a channel wheel positioning speed bump (12), a toll booth (13), and a high-speed safety guardrail (14).
[0006] The license plate recognition and reservation management module (1) is located at the entrance of the detection channel and consists of a license plate recognition camera, a recognition terminal, and a data call interface. This module recognizes the vehicle license plate in real time when the vehicle enters the channel and uses the license plate as a unique index to access the vehicle reservation records stored in the cloud monitoring platform (9) to obtain information including the type of reserved goods, the vehicle's allowable load capacity, cargo box size parameters, historical passage records, and abnormal passage markers, forming a basic vehicle information set.
[0007] The license plate recognition and reservation management module (1) establishes a data synchronization channel with the intelligent recognition and judgment center (7), and uses the basic vehicle information as a priori constraint for subsequent cargo consistency comparison and anomaly screening. At the same time, the module can automatically mark vehicles that have not made reservations based on the recognition results, triggering manual review or verification processes to ensure the automatic start of the detection process and the pre-management of traffic order.
[0008] This module enables continuous closed-loop control for rapid vehicle identification, synchronized access to cloud archives, and automatic triggering of the inspection process, eliminating the delays and instabilities caused by manual inquiry and registration.
[0009] Preferably, the cargo information feature extraction unit (2) and the structural information acquisition array (3) are respectively arranged on both sides of the detection channel to form a cargo compartment internal structural feature acquisition module. The cargo information feature extraction unit (2) includes an internal detection excitation source, a modulation control circuit and a waveform timing controller, which are used to apply a structural detection excitation signal to the cargo compartment during the process of the vehicle passing through the detection area, so that the cargo stacking area forms a detectable internal structural response field.
[0010] The structural information acquisition array (3) consists of multi-channel receiving units distributed along the height and lateral directions of the cargo compartment, used to synchronously receive the response signals inside the cargo compartment at multiple points during vehicle movement. After signal preprocessing, channel registration and cross-section reconstruction, the acquired response signals are used to obtain the cross-sectional structural feature map of the cargo compartment, which can reflect the internal loading morphology characteristics such as the uniformity of cargo stacking, the existence of cavities and interlayers, and the proportion of mixed foreign objects.
[0011] To offset the temporal misalignment caused by vehicle speed fluctuations, lateral offsets and attitude changes, a synchronous triggering mechanism and a cross-sectional temporal calibration mechanism are set between the cargo information feature extraction unit (2) and the structural information acquisition array (3) to achieve cross-sectional consistency and spatial registration accuracy of the internal structural diagram.
[0012] This module enables contactless acquisition of structural information of non-visible areas inside the cargo compartment, providing direct structural basis for subsequent cargo consistency assessment.
[0013] Preferably, the lidar module (4) and the infrared temperature acquisition module (6) are jointly installed on the lifting height-limiting gantry (10), and can be adaptively positioned according to the external dimensions of the cargo box of the passing vehicle through the height adjustment mechanism. The lifting height-limiting gantry (10) includes a height drive actuator, a beam sensor support structure, and a position feedback and safety limit device. The lidar module performs a three-dimensional contour scan of the outer surface of the vehicle cargo box to obtain discrete point cloud data of the cargo box length, width, and height. The intelligent recognition and judgment center establishes a cargo box space model based on the point cloud data and calculates the effective loading volume. V The infrared temperature acquisition module uses a matrix infrared detector to perform top-down imaging of the temperature field on the exterior of the cargo compartment, obtaining a temperature distribution matrix. T ( x , y This is used to reflect the consistency of cargo origin and internal heat and mass transfer characteristics. The dynamic following adjustment of the gantry ensures that the two sensing modules maintain an effective imaging distance and stable opening angle under different vehicle heights and passing speeds, guaranteeing the comparability and long-term stability of the measurement data.
[0014] Preferably, the dynamic weighing module (5) and the wheel positioning speed bump (12) are installed on the ground of the detection channel and consist of multiple stress sensing units, wheel positioning mechanisms, and signal acquisition devices. When the vehicle passes through the detection area at low speed, the tires act sequentially on the stress sensing units, and the module performs time-series acquisition of the axle load to obtain the axle load sequence. F a The intelligent identification and judgment center (7) filters and determines the path stability of the collected sequence, eliminates dynamic interference caused by vehicle lateral deviation, braking, etc., and inverts the loading mass based on the vehicle axle configuration, vehicle weight and standard load conversion coefficient. m This module enables mass measurement under non-stop conditions and can be combined with the aforementioned cargo compartment volume. V Calculate loading density D This is used to determine whether there is any behavior of using green channel goods for passage under low load or smuggling other goods.
[0015] Preferably, the infrared temperature acquisition module (6) is used to acquire the temperature distribution on the surface of the cargo compartment, and the intelligent identification and judgment center (7) calculates the temperature stability parameter Δ based on the temperature field gradient characteristics, regional temperature difference statistical characteristics, and local temperature anomaly significance characteristics. T If the goods originate from the same source and are stacked tightly, the temperature field will be uniform; if there are mixed goods from different sources, sandwiched covers, or empty fillings, the surface temperature will show a discrete distribution or local abrupt changes. Temperature field analysis can provide auxiliary criteria for determining the homogeneity of goods, and together with loading density and internal structure consistency parameters, it constitutes a multi-source feature set.
[0016] Preferably, the intelligent recognition and judgment center (7) is the core decision-making unit of the system, including a data preprocessing module, a feature extraction module, a fusion judgment model, and a threshold dynamic update module. This center first performs distortion correction, edge enhancement, and texture layering analysis on the structural feature map to extract internal structural consistency parameters. E Then load density D Temperature stability Δ T Consistency with internal structure E Input the fusion judgment model to calculate the confidence index S of green channel goods. The fusion judgment model adopts an updatable weighting strategy, allowing the feature weights to be adaptively adjusted under different terminals, seasons, and cargo types. The intelligent identification and judgment center, based on... S The interval distribution automatically generates release, re-inspection, or interception instructions, and records the judgment process and the basis data completely.
[0017] Preferably, the output power adaptive control module (8) adjusts the power output according to the real-time traffic speed of the vehicle. v Imaging quality evaluation indicators for internal structural feature maps Q The output power of the cargo information feature extraction unit (2) is adjusted in a closed loop. The module includes an imaging quality sampling unit, a power controller, a limiter, and a feedback correction channel. When the vehicle speed is high or the image texture contrast decreases, the excitation power is automatically increased; when the image is oversaturated or the edge noise is enhanced, the power is automatically reduced, thereby achieving dynamic stability of the structural imaging clarity. The minimum available power principle ensures the detection safety boundary and avoids ineffective high-power operation.
[0018] Preferably, the barrier gate lifting arm (11) serves as the release control execution end, realizing vehicle passage management through instruction interaction with the intelligent identification and judgment center (7). The cloud monitoring platform (9) processes the basic data, judgment parameters, and confidence index of each detection. S Values and execution records are encrypted and stored, and can be accessed and compared uniformly across regions and sites. The cloud-based monitoring platform (9) also includes a threshold adaptive update module and a model performance evaluation module, which can automatically adjust the judgment parameters and model weights based on long-term operating data, thereby forming a closed-loop monitoring system and realizing continuous optimization of system performance and strategy linkage.
[0019] Compared with existing technologies, this invention provides a highway green channel cargo inspection system based on license plate recognition and internal structural feature extraction, which has the following beneficial effects: 1) This invention allows vehicles to complete cargo verification by passing through the detection area at a low and constant speed, without stopping or requiring manual intervention. This avoids traffic congestion caused by vehicle queuing and unpacking for inspection, thereby significantly improving the efficiency of toll station traffic.
[0020] 2) This invention introduces a cargo compartment internal structure perception mechanism, which can directly reflect the cargo stacking status and internal loading form. It can identify mixed loading and camouflage behaviors such as covering, bottom layer replacement, and interlayer stuffing, thus solving the technical bottleneck that traditional visual inspection cannot reveal the true situation inside the cargo compartment.
[0021] 3) This invention establishes a joint judgment model that integrates three dimensions: loading density, temperature consistency, and structural consistency. It no longer relies on a single detection index, which is beneficial for maintaining the stability and reliability of the judgment results under different cargo types, seasonal temperature differences, and vehicle conditions.
[0022] 4) The present invention sets an adaptive output power control strategy, which can automatically adjust the detection excitation intensity according to the vehicle speed and the imaging quality of the structural diagram. Under conditions of continuous traffic, vehicle speed fluctuation or vehicle size change, it can still ensure clear imaging and accurate feature extraction, thereby improving the robustness and long-term operational reliability of the system.
[0023] 5) This invention achieves full-process traceability of detection data, judgment results and execution actions through a cloud-based monitoring platform, and supports adaptive updates of model parameters and cross-site policy synchronization, forming a traceable, verifiable and sustainably evolving green channel intelligent monitoring system. Attached Figure Description
[0024] Figure 1 Figure (a) is a top view of the system structure of the present invention, and Figure (b) is a three-dimensional layout view of the system structure. Figure 2 This is a schematic diagram illustrating the acquisition of cargo structural feature information according to the present invention; Figure 3 This is a schematic diagram of the lifting height-limiting gantry structure and sensor arrangement of the present invention; Figure 4 This is a schematic diagram of the dynamic weighing mass inversion process of the present invention; Figure 5 This is a schematic diagram of the multi-source fusion determination process of the present invention; Figure 6 This is a schematic diagram of the adaptive output power control model of the present invention; Explanation of reference numerals in the attached diagram: 1. License plate recognition and reservation management module; 2. Cargo information feature extraction unit; 3. Structural information acquisition array; 4. LiDAR module; 5. Dynamic weighing module; 6. Infrared temperature acquisition module; 7. Intelligent recognition and judgment center; 8. Output power adaptive control module; 9. Cloud monitoring platform; 10. Lifting height limit gantry; 11. Barrier gate lifting arm; 12. Lane wheel positioning speed bump; 13. Toll booth; 14. Highway safety guardrail. Detailed Implementation
[0025] 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.
[0026] like Figure 1 As shown, a highway green channel cargo inspection system based on license plate recognition and internal structural feature extraction is set up in the green channel of the highway toll station, forming a serial inspection channel of entrance recognition area—cargo structural feature acquisition area—intelligent judgment area—on-site release area.
[0027] The entrance recognition area is equipped with a license plate recognition and reservation management module (1). The internal structure acquisition area is equipped with cargo information feature extraction units (2) and structural information acquisition arrays (3) on both sides. The end of the judgment and release area is equipped with a lifting height-limiting gantry (10), and a laser radar module (4) and an infrared temperature acquisition module (6) are installed on its crossbeam. The ground is equipped with a dynamic weighing module (5) and a wheel positioning speed bump (12). The exit side is equipped with a barrier gate lifting arm (11). Each detection unit is connected to the intelligent recognition and judgment center (7) through a data interaction link, and the data and logs are synchronized to the cloud supervision platform (9).
[0028] Specifically, the license plate recognition and reservation management module (1) consists of a license plate recognition camera, a recognition terminal, and a data call interface. When a green channel vehicle to be detected enters the entrance recognition area, the license plate image is collected and the character recognition is completed. The recognized license plate is used as an index to retrieve reservation records, declared cargo type, vehicle load limit, cargo box size parameters, historical passage and abnormal markings from the cloud supervision platform (9), forming a basic vehicle information set, which is pushed to the intelligent recognition and judgment center (7) in real time as a priori constraint for subsequent consistency judgment.
[0029] Optionally, vehicles without appointments or with missing information can be marked to trigger a manual review process, without affecting the continued execution of subsequent automatic detection.
[0030] Furthermore, the cargo information feature extraction unit (2) and the structural information acquisition array (3) are arranged opposite to each other on both sides of the channel of the internal structure acquisition area to form an internal structure feature acquisition module.
[0031] The cargo information feature extraction unit (2) includes an internal detection excitation source, a modulation control circuit, and a waveform timing controller; during the passage of the vehicle, a structural detection excitation signal is applied to the interior of the cargo compartment to form a detectable internal structural response field in the cargo stacking area.
[0032] The structural information acquisition array (3) consists of multi-channel receiving units distributed along the height and lateral directions of the cargo compartment. It performs multi-point synchronous reception of internal responses, and obtains the structural feature map of the internal cross-section of the cargo compartment through channel registration and cross-section reconstruction. The cargo structural feature information acquisition method is as follows: Figure 2 As shown.
[0033] To reduce cross-sectional misalignment caused by vehicle speed fluctuations / lateral offsets, a synchronous triggering mechanism and cross-sectional timing calibration are set between the cargo information feature extraction unit (2) and the structural information acquisition array (3) to ensure the spatial consistency and comparability of adjacent cross-sections.
[0034] Furthermore, the lifting height-limiting gantry (10) includes a height drive actuator, a crossbeam sensor support structure, and a position feedback and safety limit device. Its purpose is to maintain a fixed effective imaging distance and a stable opening angle to adapt to different vehicle heights, such as... Figure 3 As shown.
[0035] The lidar module (4) performs a three-dimensional contour scan of the outer surface of the cargo compartment and outputs lidar point clouds; the intelligent recognition and judgment center (7) establishes a cargo compartment space model based on this and calculates the effective loading volume. V .
[0036] The infrared temperature acquisition module (6) uses a matrix infrared detector to image the temperature field of the cargo compartment surface from a top-down perspective, forming a temperature distribution matrix. T ( x , y This is used to distinguish whether the goods carried by green channel vehicles are refrigerated goods.
[0037] Furthermore, the dynamic weighing module (5) consists of multiple stress sensing units, wheel positioning mechanisms, and signal acquisition units. When the vehicle passes through the wheel positioning speed bump (12) at a low and constant speed, the wheels sequentially act on the dynamic weighing module, and the module's stress sensing units collect the axle load sequence. F a The intelligent identification and judgment center (7) performs dynamic filtering, path stability verification, vehicle weight compensation, and axle type correction to obtain the actual load mass. m Combined with the effective loading volume of the cargo box V Loading density can be calculated D = m / V ,like Figure 4 As shown.
[0038] Furthermore, the intelligent identification and judgment center (7) consists of a data preprocessing module, a multi-source data extraction module, a multi-source fusion model, and a judgment and threshold dynamic update module. It is the core decision-making unit of the system. The multi-source fusion judgment process is as follows: Figure 5 As shown, the specific steps include: S101, Data Preprocessing Module, specifically including: The cross-sectional structural feature map of the cargo compartment output by the structural information acquisition array (3) is subjected to distortion correction, noise filtering, edge enhancement and cross-sectional time alignment in sequence, and a structural map sequence of uniform specifications is output. Calculate the effective loading volume of the lidar point cloud output by lidar module (4). V ; The axle load sequence collected by the dynamic weighing module (5) F a The actual loaded mass is obtained through mass inversion. m .
[0039] S102, Multi-source data extraction module: including structural feature extraction, loading parameter calculation and temperature feature extraction; The structural feature extraction specifically involves inputting a sequence of structural diagrams of uniform specifications, and sequentially performing regional texture uniformity analysis, layer stacking pattern extraction, and local anomaly salience identification to obtain the consistency parameters of the internal structure of the cargo compartment. E； The loading parameter calculation specifically involves inputting the actual loaded mass. m and effective loading volume V， Calculate loading density D = m / V ; The temperature feature extraction specifically involves forming a temperature distribution matrix from the infrared temperature acquisition module (6). T ( x , y The temperature stability parameter Δ is obtained by comprehensively quantifying the regional temperature difference, gradient, and abrupt change characteristics. T ; S103, Multi-source fusion model: based on loading density D Temperature stability parameter Δ T and consistency parameters of the internal structure of the cargo compartment E As input, the loading rationality index is obtained through monotonic mapping. S D Temperature stability index S T Structural consistency index S E Establish a multi-source fusion model and output a confidence index for green channel goods. S ; The multi-source fusion model can employ weighted mapping or a learned classifier, described in a general form in the instruction manual. The expression for calculating the confidence index of green channel goods is as follows: , in, S D To ensure the reasonableness of the loading indicators, S T For temperature stability indicators, S E As a structural consistency indicator, The weighting coefficients for the rationality index of loading, This is the weighting coefficient for the temperature stability index. The weighting coefficients for the structural consistency index; weighting coefficients , , The cloud-based monitoring platform (9) updates the data periodically based on historical re-inspection results.
[0040] S104, Judgment and Threshold Dynamic Update Module: Based on the confidence index of green channel goods. S The interval distribution and two-level threshold settings automatically generate release / re-inspection / interception instructions, and leave a trace of the entire process of input, judgment, and output; The threshold is adaptively adjusted by the cloud monitoring platform (9) according to site / season / category.
[0041] Set two threshold levels θ 1 and θ 2, and θ 1> θ 2: When S ≥ θ 1. Release; when θ 2≤ < θ 1. Re-inspection; when < θ 2. Interception and auditing. The threshold is adaptively adjusted by the cloud-based monitoring platform (9) according to site / season / category.
[0042] The cloud-based monitoring platform (9) periodically updates the weight coefficients and two-level thresholds based on the results of the re-inspection and audit, and sends them to the field to achieve policy adaptation.
[0043] Furthermore, the output power adaptive control module (8) adjusts the output power of the cargo information feature extraction unit (2). P Perform closed-loop regulation, such as Figure 6 As shown. Its basis includes real-time vehicle traffic speed. v With imaging quality evaluation indicators Q (Comprehensive indicators such as texture contrast, edge strength, or entropy taken from the structure map) are used to evaluate the error term of image quality. e Q = Q target QThe power correction can be calculated using a typical control law, which can be expressed as follows: in, P 0 is the reference power. k 1 represents the speed adjustment coefficient. k 2 is the error adjustment coefficient; and it is set as follows: P min ≤ P out ≤ P max The safety limit is maintained. This closed loop maintains the structural imaging sharpness and system safety boundary under dynamic passage conditions.
[0044] The barrier gate lifting arm (11) acts as the execution end, and implements the lifting / holding / lowering of the arm according to the instructions of the judgment center (7) to realize the separation of vehicle release and re-inspection.
[0045] The cloud-based monitoring platform (9) is used for encrypted storage and cross-site access of detection results, model parameters, and execution records; and provides threshold adaptive updates and model performance evaluation capabilities to drive online optimization of weights and thresholds with long-term running data, supporting full lifecycle traceability and regional collaborative supervision.
[0046] Preferably, in order to ensure the cross-sectional correspondence of the structural feature map, a master-slave synchronous trigger is established in the internal structure acquisition area: the channel photoelectric gate or mileage trigger is used as the master trigger signal, and the cargo information feature extraction unit (2) and the structural information acquisition array (3) acquire information synchronously; a short-period phase reference sequence is used to correct the cross-sectional time drift.
[0047] To address the parallax mismatch caused by vehicle yaw and lateral tilt, cross-sectional lateral registration is performed using array lateral calibration points and lidar point clouds, and secondary reconstruction is triggered when necessary to restore sequence consistency.
[0048] Preferably, to ensure field of view coverage, the spacing between the two sides of the internal structure acquisition area can be arranged to be 3.2–4.0 m, and the effective height coverage of the array is 1.2–3.2 m; the height travel of the lifting height-limiting gantry covers 3.0–5.5 m, and mechanical limit and emergency stop are provided.
[0049] Preferably, the vehicle passes through at a low, constant speed (e.g., 5–15 km / h); if the speed exceeds the limit, the entrance will prompt the vehicle to slow down and mark the vehicle as having priority for review.
[0050] Preferably, the structure map section update period is, for example, 10–20 ms; the point cloud frame rate is, for example, 10–20 fps; the temperature frame rate is, for example, 10–30 fps; and the weighing sampling frequency is, for example, 500–1 kHz.
[0051] Generally, the axle load sequence is clustered by axle group and a wheel position window sliding average is performed to remove peak artifacts caused by braking and yaw; the self-weight compensation parameters are obtained from vehicle files and on-site calibration.
[0052] Generally speaking, for T ( x , y Perform regional segmentation, extract the mean and variance of temperature for each region, Δ T It can be defined as a combination of the mean or the maximum and minimum difference of the partition variance.
[0053] Generally, image recognition is performed on the structure diagram and the types in the green channel goods database are identified and distinguished. Then, gray-level co-occurrence matrix texture features, hierarchical direction consistency (e.g., based on structure tensor) and anomaly saliency (e.g., local contrast or sparse reconstruction residual) are fused to obtain a 0–1 normalized consistency score.
[0054] In the aforementioned power regulation process, the clarity evaluation index Q With target clarity Q target Deviation between e Q = Q target Q Used to correct the output power so that the imaging of structural features is always kept within the discriminable range. Q target The resolution range, calculated by the system during the installation and calibration phase using a typical cargo sample set, reflects the clarity range within which structural textures can be reliably identified. Meanwhile, to ensure system operational safety, the output power is limited by… P min and P max It satisfies at any time P min ≤ P out ≤ P max This is to avoid overpowering or imaging failure.
[0055] To ensure the safety and stability of the system operation, the cargo information feature extraction unit adopts the minimum available power principle and sets an upper limit for the duty cycle during operation. Each detection unit is equipped with over-temperature, over-current, and emergency shutdown interlock protection. When an internal structure detection module malfunctions, the system automatically degrades to the basic detection mode consisting of license plate recognition, dynamic weighing, lidar shape measurement, and infrared temperature imaging. At the same time, it issues a manual verification prompt to the site to ensure the continuity of the detection process and the reliability of the results.
[0056] Generally, the workflow of the detection system described in this invention is as follows: Entry recognition: sequentially reads license plate number, retrieves archive, forms basic information set, and triggers detection sequence.
[0057] Internal structure acquisition: The cargo information feature extraction unit (2) sends an excitation to the structure information acquisition array (3) to synchronously collect and identify the cargo features carried by the green channel vehicle.
[0058] Appearance / Temperature / Weighing: Output effective loading volume using lidar point cloud data. V Infrared thermal imaging forms a temperature distribution matrix. T ( x , y The actual loaded mass is obtained through mass inversion. m .
[0059] Features and Fusion: Computation D Δ T , E Input fusion model output S， Generate control commands.
[0060] Access control and record keeping: Sequentially execute gate operation, record the entire process, archive in the cloud, and update thresholds / weights online.
[0061] Furthermore, without altering the core concept of this invention, the aforementioned system can be optionally configured and expanded according to different station conditions, vehicle characteristics, and regulatory requirements.
[0062] Optionally, a "pre-mounted gantry" can be added between the internal structure acquisition area and the gantry to form two acquisition / overlay reconstructions, thereby improving efficiency. E The robustness.
[0063] Optionally, the structural information acquisition array (3) can use replaceable subarrays (dense / sparse) to adapt to different cargo box heights; or a top subarray can be added above the aisle to improve the detection of the upper stacked layers.
[0064] Optionally, stricter thresholds or mandatory inspection strategies can be adopted for certain high-risk categories (such as perishable / high-value agricultural products); fast-track weighting can be applied to reputable fleets to reduce the re-inspection rate.
[0065] Optionally, edge nodes can be deployed on the channel side to preprocess the structure map and point cloud, reduce backhaul bandwidth, and only upload features and indexes in case of anomalies.
[0066] To ensure that the system of this invention remains stable and reliable under different scenarios and long-term continuous operation conditions, this invention also establishes a set of calibration, maintenance and fault tolerance mechanisms to ensure the accuracy consistency of each measurement link and the overall reliability of the system.
[0067] Preferably, the array geometry, time synchronization, and external parameter calibration of lidar-array-infrared thermal imaging are performed based on the standard calibration board or standard frame upon entry; daily calibration is automatically performed based on the "health detection frame".
[0068] Preferably, a blackbody or standard temperature-controlled source is used for multi-point temperature calibration of infrared thermal imaging; Δ values are given for different seasons. T Baseline migration compensation.
[0069] Preferably, the weighing link is checked periodically using a standard weight cart or metering cart; when the drift exceeds the threshold, an alarm is automatically triggered and the system switches to a safety net strategy.
[0070] Preferably, when any subsystem malfunctions, the system maintains its remaining capacity and explicitly marks "partial data missing," prioritizing its entry into the re-inspection queue to ensure passage order and safety.
[0071] The following description uses a green channel vehicle transporting "fresh potatoes" as an example to illustrate the complete process of using the system of this invention. This embodiment is used to demonstrate the system's operation and does not constitute a limitation on the scope of protection of this invention.
[0072] Before passing through the expressway, the vehicle had completed the cargo declaration through the expressway green channel reservation platform. The declaration information was "fresh potatoes, loading weight 8.2 tons, vehicle load capacity 10 tons, cargo box dimensions 4.2 m × 2.1 m × 2.0 m".
[0073] When a vehicle enters the green channel lane of the toll station, the license plate recognition and reservation management module (1) automatically collects the license plate image and identifies the license plate number as "Yun A×××××". The system then retrieves the vehicle's reservation registration, historical passage records, cargo type and load limit parameters from the cloud monitoring platform (9) and sends the information to the intelligent recognition and judgment center (7) to form a basic vehicle information set.
[0074] After the vehicle enters the internal structure detection area, the cargo information feature extraction unit (2) automatically applies a structure detection excitation signal to the inside of the cargo compartment, and the structure information acquisition array (3) synchronously receives the cargo stacking structure response signal.
[0075] After channel registration and cross-section reconstruction, the resulting cross-sectional structural feature map shows that the cargo is potatoes, which are evenly stacked inside the cargo compartment, with no layering, concealment of cavities, or mixing of foreign objects found. The feature analysis module extracts internal structural consistency parameters, obtaining: A higher value indicates stronger consistency.
[0076] The vehicle continues to move forward and passes through the lifting height-limiting gantry (10). The lidar module (4) performs point cloud scanning on the shape of the cargo box and calculates the effective loading volume: The dynamic weighing module (5) collects the vehicle axle load sequence, and obtains the load mass through filtering, compensation and inversion calculation: This leads to the loading density: Infrared temperature acquisition module (6) acquires the surface temperature matrix of the cargo compartment. T ( x , y The temperature distribution stability parameters were calculated as follows: The temperature distribution was relatively uniform, and no significant differences between high and low temperature areas were observed.
[0077] The intelligent recognition and judgment center (7) will use internal structural consistency parameters E Loading density D With temperature stability parameter Δ T Inputting the multi-source fusion model yields the confidence index for green channel goods: because S ≥ θ 1 (for example) θ (1=0.85), the system automatically determines that the vehicle's cargo is genuine, not mixed, and meets the green channel conditions.
[0078] The barrier gate lifting arm (11) is raised, and vehicles can pass directly without stopping. The entire detection process lasts about 8–12 seconds.
[0079] The vehicle identification records, cargo compartment internal structure feature maps, point cloud data, temperature matrix, loading density, and confidence index collected during this inspection process are all included. S Both the release instructions and the release instructions are stored synchronously on the cloud-based regulatory platform (9) for subsequent regional collaborative supervision and anomaly review.
[0080] 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 highway green channel cargo inspection system based on license plate recognition and internal structural feature extraction, characterized in that, The system includes the following components installed on the detection channel: license plate recognition and reservation management module (1), cargo information feature extraction unit (2), structural information acquisition array (3), lidar module (4), dynamic weighing module (5), infrared temperature acquisition module (6), intelligent recognition and judgment center (7), output power adaptive control module (8), cloud monitoring platform (9), lifting height limit gantry (10), barrier gate lifting arm (11), and channel wheel positioning speed bump (12). Among them, the license plate recognition and reservation management module (1), cargo information feature extraction unit (2), structural information acquisition array (3), lidar module (4), dynamic weighing module (5), infrared temperature acquisition module (6), output power adaptive control module (8) are connected to the intelligent recognition and judgment center (7) through a data interaction link; The license plate recognition and reservation management module (1), the intelligent recognition and judgment center (7), the barrier gate lifting arm (11) and the cloud monitoring platform (9) are connected through a two-way communication channel; The lidar module (4) and the infrared temperature acquisition module (6) are installed on the lifting height-limiting gantry (10); The dynamic weighing module (5) and the wheel positioning speed bump (12) are installed on the ground of the detection channel; The cargo information feature extraction unit (2) and the structural information acquisition array (3) are positioned opposite each other on both sides of the detection channel.
2. The system according to claim 1, characterized in that, The license plate recognition and reservation management module (1) is set at the entrance of the detection channel. It is used to recognize the vehicle license plate in real time when the vehicle enters, and retrieve the corresponding reservation record, declared cargo type, load limit, cargo box size parameters and historical passage record from the cloud supervision platform (9) using the recognized vehicle license plate information as an index, forming a basic vehicle information set and sending it to the intelligent recognition and judgment center (7).
3. The system according to claim 1, characterized in that, The cargo information feature extraction unit (2) and the structural information acquisition array (3) are respectively set on opposite sides of the detection channel; The cargo information feature extraction unit (2) is used to apply a structural detection excitation signal to the interior of the cargo compartment during the period when the vehicle passes through the detection area; The structural information acquisition array (3) consists of multi-channel receiving units distributed along the height and lateral directions of the cargo compartment. It is used to receive the internal structural response of the cargo compartment generated under the action of the excitation signal at multiple points synchronously, and to reconstruct the cross-sectional structural feature map of the cargo compartment based on the received response signal sequence. The cargo information feature extraction unit (2) and the structural information acquisition array (3) are provided with a synchronous triggering and cross-sectional timing calibration mechanism.
4. The system according to claim 1, characterized in that, The lifting height limiting gantry (10) includes a height drive actuator, a beam sensor support structure, a position feedback and safety limit device. Its working process is as follows: the height is adaptively adjusted according to the external dimensions of the cargo box of the passing vehicle, so that the laser radar module (4) and the infrared temperature acquisition module (6) maintain an effective imaging distance and a stable opening angle. The working process of the lidar module (4) is as follows: it performs a three-dimensional contour scan on the outer surface of the vehicle cargo box to obtain lidar point cloud data; the intelligent recognition and judgment center (7) establishes a cargo box space model based on the lidar point cloud data and calculates the effective loading volume. V ; The infrared temperature acquisition module (6) works as follows: it uses a matrix infrared detector to image the temperature field of the cargo compartment surface from a top-down perspective, forming a temperature distribution matrix. T ( x , y ).
5. The system according to claim 1, characterized in that, The dynamic weighing module (5) and the wheel positioning speed bump (12) are installed on the ground of the detection channel. They consist of multiple stress sensing units, wheel positioning mechanisms and signal acquisition devices. When the vehicle passes through the detection channel at low speed, the tires act on the stress sensing units in sequence. The module performs time-series acquisition of axle loads to obtain the axle load sequence. F a ; The working process of the intelligent identification and judgment center (7) includes: processing the axle load sequence F a Dynamic filtering, path stability verification, vehicle weight compensation, and axle profile correction are performed to invert and obtain the actual load mass. m Based on the loading mass m With the effective loading volume V Calculate loading density D .
6. The system according to claim 1, characterized in that, The intelligent identification and judgment center (7) is configured to perform the green channel cargo inspection and judgment method, including the following steps: S101. The cross-sectional structural feature map of the cargo compartment output by the structural information acquisition array (3) is subjected to distortion correction, noise filtering, edge enhancement and cross-sectional time alignment in sequence, and a structural map sequence of uniform specifications is output. Calculate the effective loading volume of the lidar point cloud output by lidar module (4). V ; The axle load sequence collected by the dynamic weighing module (5) F a The actual loaded mass is obtained through mass inversion. m； S102. Based on the unified specification of the structural diagram sequence, effective loading volume and actual loading mass, perform structural feature extraction, loading parameter calculation and temperature feature extraction; The structural feature extraction specifically involves inputting a sequence of structural diagrams of uniform specifications, and sequentially performing regional texture uniformity analysis, layer stacking pattern extraction, and local anomaly salience identification to obtain the consistency parameters of the internal structure of the cargo compartment. E； The loading parameter calculation specifically involves inputting the actual loaded mass. m and effective loading volume V， Calculate loading density D = m / V ; The temperature feature extraction specifically involves forming a temperature distribution matrix from the infrared temperature acquisition module (6). T ( x , y The temperature stability parameter Δ is obtained by comprehensively quantifying the regional temperature difference, gradient, and abrupt change characteristics. T ; S103, based on loading density D Temperature stability parameter Δ T and consistency parameters of the internal structure of the cargo compartment E As input, the loading rationality index is obtained through monotonic mapping. S D Temperature stability index S T Structural consistency index S E Establish a multi-source fusion model and output a confidence index for green channel goods. S ; S104, Based on the confidence index of green channel goods S The interval distribution and two-level threshold settings generate release instructions, re-inspection instructions, or interception instructions.
7. The system according to claim 6, characterized in that, The calculation formula for the confidence index of green channel goods is as follows: ， in, S D To ensure the reasonableness of the loading indicators, S T For temperature stability indicators, S E As a structural consistency indicator, The weighting coefficients for the rationality index of loading, This is the weighting coefficient for the temperature stability index. The weighting coefficients for the structural consistency index; weighting coefficients , , The cloud-based monitoring platform (9) updates the data periodically based on historical re-inspection results.
8. The system according to claims 1 and 6, characterized in that: The gate lifting arm (11) is arranged below the lifting height-limiting gantry (10) and serves as the release control execution end; The intelligent identification and judgment center (7) completes the green channel cargo confidence index. S After calculation, the corresponding release command, re-inspection command or interception command is sent to the gate lifting arm (11). The gate lifting arm (11) performs lifting, holding, or lowering actions according to the received instructions.
9. The system according to claim 1, characterized in that: The output power adaptive control module (8) is used to adjust the output power according to the real-time traffic speed of the vehicle. v And imaging quality evaluation indicators of internal structural feature maps Q The output power of the cargo information feature extraction unit (2) P out Dynamic adjustments are made to keep the texture contrast and edge sharpness of the internal structural feature map within a preset discriminable range; The expression for calculating the output power is as follows: P out = P 0+ k 1· v + k 2· e Q in, P 0 is the reference power. k 1 represents the speed adjustment coefficient. k 2 is the error adjustment coefficient. v For speed, e Q This is used to evaluate the image quality error term; and by setting an upper limit for the output power. P max and lower limit of output power P min right P out Implement safety limit control.