ETC transaction anomaly detection and instant release processing system

Abstract

The application discloses an ETC transaction abnormality detection and instant release processing system, relates to the technical field of ETC transactions, and has an abnormality detection and instant release processing system with dynamic environment adaptability, which can effectively solve the problems of declining license plate recognition efficiency and insufficient real-time performance caused by microwave signal reflection, lane interaction interference and external environment influence in a multi-lane environment of a toll plaza, vehicle information is acquired by the system through a license plate recognition module and an ETC information reading module in cooperation, and matching verification is performed by a comparison module, meanwhile, the validity of completed transactions is checked in combination with a transaction checking module, an abnormality detection module specially identifies abnormal vehicles that have not passed the entrance, and in combination with an instant release module, the vehicles that have completed the transactions are quickly released, congestion and mistaken deduction problems caused by mistaken transactions are reduced, and therefore, the transaction accuracy, real-time performance and passing efficiency of the toll system are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of ETC transaction technology, and more specifically to an ETC transaction anomaly detection and immediate release processing system. Background Technology

[0002] With the implementation of a nationwide highway network, ETC (Electronic Toll Collection) has played a crucial role in improving traffic efficiency. However, at toll plazas, due to microwave signal reflection and interference, ETC devices may make erroneous transactions with antennas in lanes other than the intended destination. Such erroneous transactions not only obstruct traffic but can also lead to subsequent management chaos. Therefore, an effective technical solution is needed that integrates license plate recognition and transaction verification mechanisms to achieve rapid detection and immediate processing of ETC transaction anomalies, ensuring the stability and efficiency of the toll collection system.

[0003] The existing technology has the following shortcomings:

[0004] In multi-lane toll plaza environments, complex microwave signal reflections and lane interference mean that even if license plate recognition and ETC transaction information can be compared to some extent, external factors such as light, rain, or vehicle obstruction can still reduce license plate recognition efficiency, failing to meet real-time requirements. Furthermore, when abnormal transactions occur in non-target lanes, the system lacks a rapid verification mechanism to identify and release vehicles, potentially causing persistent congestion or even incorrect toll charges. These issues necessitate the design of an anomaly detection and immediate release system with dynamic environmental adaptability to minimize the impact of erroneous transactions. Summary of the Invention

[0005] The purpose of this invention is to provide a system for detecting and instantly releasing abnormal transactions in ETC transactions, in order to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an ETC transaction anomaly detection and instant release processing system, including a license plate recognition module, an ETC information reading module, a comparison module, a transaction verification module, an anomaly detection module, and an instant release module;

[0007] The license plate recognition module is used to obtain vehicle license plate information;

[0008] The ETC information reading module is used to obtain license plate information from the vehicle's ETC device.

[0009] The comparison module is used to compare the license plate information obtained by the license plate recognition module with the license plate information obtained by the ETC information reading module. If they match, the transaction is allowed.

[0010] The transaction verification module is used to verify the vehicles involved in the transaction after the transaction is completed, in order to verify the validity of the transaction.

[0011] The anomaly detection module is used to detect vehicles that have not passed through the entrance, obtain their ETC information, and determine whether they have made a transaction in the current toll plaza within a preset time.

[0012] The instant release module is used to immediately release a vehicle if it is detected that the transaction has been completed within a preset time.

[0013] Preferably, the license plate recognition module includes: a high-resolution camera for capturing images of vehicle license plates; an optical character recognition unit for analyzing the license plate images and extracting character information; and a data processing unit for sending the extracted license plate information to the comparison module.

[0014] Preferably, the ETC information reading module includes: a 5.8GHz microwave antenna for acquiring license plate and related transaction information from the vehicle's ETC device; a data decoder for decoding the received ETC device information; and a cache unit for temporarily storing the decoded ETC information for subsequent processing.

[0015] Preferably, the comparison module is used to compare the license plate information obtained by the license plate recognition module with the license plate information obtained by the ETC information reading module. If they match, the transaction is allowed. Two strings S1 and S2 are set, with lengths n and m respectively. A two-dimensional matrix D of size (n+1)×(m+1) is constructed and initialized as follows: The initial state represents the editing cost from an empty string to the target string.

[0016] Fill matrix D using the recursive formula: ;in: Finally, the value of matrix D[n][m] is the edit distance between S1 and S2;

[0017] Character similarity is calculated using a formula normalization: ;in, It is the value of matrix D[n][m], and HAK is the character deviation value, which is in the range of [0,1].

[0018] Preferably, the license plate images are converted to grayscale images, all license plate images are scaled to the same resolution, the color histogram of the license plate images is calculated, and the license plate images are divided into RGB or HSV channels; a histogram is calculated for each channel. Normalize each histogram: ; i represents the segmented interval of pixel values, n is the number of intervals in the histogram, Canny edge detection is used to extract the outer border and character edges of the license plate; the contour feature vector Fs is calculated for the detected contours, character segmentation is performed on the image, and the coordinates of each character are extracted; character position feature vectors are constructed. , representing the center coordinates of each character, calculate the outlier similarity between the license plate images of two images using the following expression: In the formula, GHA represents the outlier value of the license plate image similarity.

[0019] Preferably, the character deviation value and the license plate image similarity anomaly value are converted into feature vectors. The feature vectors are used as input to the machine learning model. The machine learning model uses the prediction of the matching degree value label of the license plate information for each set of feature vectors as the prediction objective and minimizes the sum of prediction errors of the matching degree value labels of all license plate information as the training objective. The machine learning model is trained until the sum of prediction errors converges and the model training stops. The matching degree value of the license plate information is determined according to the model output. The machine learning model is a multinomial regression model.

[0020] Preferably, the matching degree value of the obtained license plate information is compared with the gradient matching threshold, which includes a first matching threshold and a second matching threshold, and the first matching threshold is less than the second matching threshold. The matching degree value of the license plate information is compared with the first matching threshold and the second matching threshold respectively.

[0021] If the matching degree value of the license plate information is greater than the second matching threshold, it means that the matching degree of the license plate information is high, and the license plate information is marked as a complete match, allowing the transaction;

[0022] If the matching degree value of the license plate information is greater than or equal to the first matching threshold and less than or equal to the second matching threshold, it indicates that the matching degree of the license plate information is moderate. The license plate information is marked as an incomplete match and further verification is required.

[0023] If the matching degree value of the license plate information is less than the first matching threshold, it means that the matching degree of the license plate information is low. The license plate information will be marked as mismatched and the transaction will be rejected.

[0024] Preferably, in the anomaly detection module, the license plate recognition module or ETC information reading module at the entrance is used to capture the vehicle entry timestamp (Tentry); for exit transactions, the vehicle departure timestamp (Texit) is recorded.

[0025] Calculate the time a vehicle spends within the toll plaza: Compare the length of stay Whether it is within a reasonable range, the maximum allowable stay time is set to Tmax; if the vehicle stay time exceeds Tmax or is less than the reasonable time threshold Tmin, it is marked as an abnormal vehicle.

[0026] Define a path set Pvalid, which includes the matching relationship between the entry station Sentry and the exit station Sexit: (Sentry, Sexit) ∈ Pvalid; if a vehicle path is not in Pvalid, it is marked as a path anomaly.

[0027] Set a reasonable stay time range: Tmin≤Tstay≤Tmax; if it exceeds this range, mark it as abnormal.

[0028] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0029] 1. This invention solves the problem of decreased license plate recognition efficiency caused by microwave signal reflection, lane interaction interference, and external environmental factors in existing technologies by designing a dynamic environment-adaptable anomaly detection and instant release system. It also compensates for the lack of rapid verification and anomaly handling mechanisms in multi-lane environments. The system integrates a license plate recognition module, an ETC information reading module, a comparison module, a transaction verification module, anomaly detection module, and an instant release module. It can efficiently complete vehicle information collection, comparison verification, anomaly detection, and release operations, significantly improving the transaction accuracy and traffic efficiency of the toll collection system, and reducing congestion and incorrect charges caused by erroneous transactions.

[0030] 2. This invention further improves the system's reliability and intelligence through optimized design details. A machine learning model is trained using a multinomial regression model combined with character deviation values ​​and license plate image similarity outliers to achieve efficient prediction of license plate information matching accuracy. Gradient matching thresholds are used to classify matching levels, ensuring transaction accuracy. Abnormal vehicles are identified based on reasonable time intervals and path rules, enhancing the accuracy and real-time performance of anomaly detection. This invention is not only applicable to existing ETC toll collection systems but also has good scalability, allowing for wide application in other scenarios within the intelligent transportation field. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0032] Figure 1 This is a system module diagram of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0034] For examples, please refer to Figure 1 As shown, the ETC transaction anomaly detection and instant release processing system described in this embodiment includes a license plate recognition module, an ETC information reading module, a comparison module, a transaction verification module, an anomaly detection module, and an instant release module.

[0035] The license plate recognition module is used to obtain vehicle license plate information;

[0036] The ETC information reading module is used to obtain license plate information from the vehicle's ETC device.

[0037] The comparison module is used to compare the license plate information obtained by the license plate recognition module with the license plate information obtained by the ETC information reading module. If they match, the transaction is allowed.

[0038] The transaction verification module is used to verify the vehicles involved in the transaction after the transaction is completed, in order to verify the validity of the transaction.

[0039] The anomaly detection module is used to detect vehicles that have not passed through the entrance, obtain their ETC information, and determine whether they have made a transaction in the current toll plaza within a preset time.

[0040] The instant release module is used to immediately release a vehicle if it is detected that the transaction has been completed within a preset time.

[0041] The license plate recognition module includes: a high-resolution camera for capturing images of vehicle license plates; an optical character recognition (OCR) unit for analyzing the license plate images and extracting character information; and a data processing unit for sending the extracted license plate information to the comparison module.

[0042] High-resolution camera: Used to capture vehicle license plate images. This camera features wide dynamic range (WDR) and infrared enhancement to ensure clear license plate information capture under various lighting conditions, such as daytime, nighttime, strong light, shadow, or backlight. The camera is installed above the tollbooth entrance or ETC lane, with the angle adjusted to ensure the target vehicle is covered in the field of view. The resolution is no less than 1080P, supporting high-definition video capture; the frame rate is ≥30fps to ensure fast license plate image capture; and it is equipped with autofocus to adapt to vehicle shooting needs at different distances.

[0043] Optical Character Recognition (OCR) Unit: Analyzes license plate images captured by a high-resolution camera and extracts character information from the image into processable data. Supports recognition of multiple license plate formats, including standards from different countries or regions; incorporates anti-interference algorithms to handle character loss caused by dirt, reflection, occlusion, etc.; provides real-time recognition capabilities with a processing latency of less than 100 milliseconds. Optimized Features: The OCR unit combines image preprocessing techniques (such as edge enhancement, noise filtering, and contrast enhancement) to improve recognition accuracy, especially under harsh environmental conditions.

[0044] The data processing unit formats the license plate character information extracted by the OCR unit and sends it to the comparison module. It binds the character information with additional information such as timestamps and camera numbers to ensure data traceability; it also verifies and encodes the extracted information to guarantee data integrity and transmission security. The data processing unit interacts with other modules in the system (such as the ETC information reading module and the comparison module) through a unified communication protocol (such as TCP / IP or CAN bus protocol).

[0045] A high-resolution camera captures vehicle license plate images in real time and sends them to the OCR unit. The OCR unit extracts characters from the image and completes and optimizes incomplete or blurry characters. The data processing unit formats the extracted license plate character information and adds metadata such as timestamps, then sends the processed information to the comparison module for subsequent matching operations.

[0046] The ETC information reading module includes: a 5.8GHz microwave antenna for acquiring license plate and related transaction information from the vehicle's ETC device; a data decoder for decoding the received ETC device information; and a cache unit for temporarily storing the decoded ETC information for subsequent processing.

[0047] 5.8GHz Microwave Antenna: Communicates bidirectionally with the vehicle's ETC device via the 5.8GHz microwave communication protocol to obtain license plate information and related transaction data (such as vehicle type, user account information, device serial number, etc.). It supports short-range high-speed communication, typically 1-15 meters; features a highly directional antenna design to reduce signal interference from adjacent lanes; and has automatic frequency adjustment to adapt to frequency interference in multi-channel environments. The antenna is equipped with a signal booster to ensure stable ETC device signal reception even in harsh environments (such as rainy, foggy, or toll plazas with severe signal reflection).

[0048] Data Decoder: Decodes the encrypted data packets received by the 5.8GHz microwave antenna to extract the vehicle's ETC information. The raw data packets acquired through the microwave antenna are typically sent in encrypted format. The data decoder decrypts the data packets using a built-in algorithm, extracting key information such as license plate number, transaction records, and user account status. It performs integrity and validity checks on the decoded information, filtering out erroneous or incomplete data packets. It supports multiple ETC protocol standards (such as GB / T20851 China ETC standard); it has a high-performance processor to ensure decoding latency is less than 10 milliseconds, meeting real-time requirements; and it is equipped with an anomaly detection mechanism that triggers an alarm when decoding fails or the signal is interrupted.

[0049] Cache Unit: Used for temporary storage of decoded ETC information to support subsequent comparison and processing. Data Writing: Decoded ETC information is temporarily stored in the cache unit before being transmitted to other modules; Data Management: Cache data is managed using a First-In-First-Out (FIFO) algorithm to ensure the accuracy of the transmission order; Data Cleanup: The cache is automatically cleaned up and storage space is released after successful data transmission or processing. High-speed storage chips are used, supporting high-frequency read and write operations; power-loss protection ensures no data loss in the event of a sudden failure; dynamic cache expansion is supported, adjusting the cache capacity according to traffic load.

[0050] When a vehicle enters the ETC lane, the 5.8GHz microwave antenna detects the vehicle's ETC device and obtains its data packets through the microwave communication protocol.

[0051] The data decoder receives data packets from the antenna, decrypts them, and extracts key data such as license plate number, transaction information, and device serial number. The decoded data is temporarily stored in a cache unit, along with metadata such as timestamp and lane number, for further processing by the comparison and transaction verification modules. If duplicate transactions or abnormal signals are detected, the cache unit can also retain some data for system log recording or manual verification.

[0052] The comparison module is used to compare the license plate information obtained by the license plate recognition module with the license plate information obtained by the ETC information reading module. If they match, the transaction is allowed. The matching module includes: performing error correction processing on some character deviations in the license plate information caused by dirt, obstruction or recognition error based on edit distance or similarity index; and determining whether the license plates belong to the same vehicle by analyzing the overall features of the license plate image (such as color, font and arrangement).

[0053] Given two strings S1 (license plate recognition result) and S2 (license plate information from the ETC device), with lengths n and m respectively. Construct a two-dimensional matrix D of size (n+1)×(m+1), initialized as follows: The initial state represents the editing cost from an empty string to the target string.

[0054] Fill matrix D using the recursive formula: ;in: Finally, the value of matrix D[n][m] is the edit distance between S1 and S2.

[0055] Character similarity is calculated using a formula normalization: ;in, It is the value of matrix D[n][m], and HAK is the character deviation value, which ranges from [0,1]. The closer the value is to 1, the more similar the strings are.

[0056] Convert license plate images to grayscale to remove color interference. Scale all license plate images to the same resolution (e.g., 100×30 pixels) to ensure consistent feature extraction. Reduce image noise using Gaussian blur or median filtering.

[0057] Calculate the color histogram of the license plate image, dividing the image into RGB or HSV channels; calculate the histogram for each channel. Normalize each histogram: i represents the segmented interval of pixel values, and n is the number of intervals in the histogram. Canny edge detection is used to extract the outer border and character edges of the license plate. The contour feature vector Fs is calculated for the detected contours, such as contour area, perimeter, aspect ratio, etc.

[0058] Perform character segmentation on the image and extract the coordinates of each character; construct character position feature vectors. , representing the center coordinates of each character, calculate the outlier similarity between the license plate images of two images using the following expression: In the formula, GHA represents the outlier value of the license plate image similarity.

[0059] Based on the outlier of the license plate image similarity, the outlier is defined as: E = 1 − GHA; when E > Threshold (Threshold is the outlier threshold, such as 0.3), the license plate image is determined to be abnormal.

[0060] Character deviation values ​​and license plate image similarity anomalies are converted into feature vectors. These feature vectors are then used as input to a machine learning model. The machine learning model uses the prediction of the matching degree label of the license plate information for each set of feature vectors as its prediction objective and minimizes the sum of prediction errors for the matching degree labels of all license plate information as its training objective. The machine learning model is trained until the sum of prediction errors converges, at which point the model training stops. The matching degree value of the license plate information is determined based on the model output. The machine learning model is a multinomial regression model.

[0061] The method for obtaining the matching degree value of license plate information is as follows: Obtain the corresponding function expression from the comprehensive feature vector training data of the trained machine learning model. In the formula, This is the model's output function, where HAK is the character-specific deviation value and GHA is the license plate image similarity outlier value. This represents the degree of matching of the license plate information.

[0062] The matching degree value of the obtained license plate information is compared with the gradient matching threshold, which includes a first matching threshold and a second matching threshold, and the first matching threshold is less than the second matching threshold. The matching degree value of the license plate information is compared with the first matching threshold and the second matching threshold respectively.

[0063] If the matching degree value of the license plate information is greater than the second matching threshold, it means that the matching degree of the license plate information is high, and the license plate information is marked as a complete match, allowing the transaction;

[0064] If the matching degree value of the license plate information is greater than or equal to the first matching threshold and less than or equal to the second matching threshold, it indicates that the matching degree of the license plate information is moderate. The license plate information is marked as an incomplete match and further verification is required.

[0065] If the matching degree value of the license plate information is less than the first matching threshold, it means that the matching degree of the license plate information is low. The license plate information will be marked as mismatched and the transaction will be rejected.

[0066] The transaction verification module is used to verify the vehicles involved in the transaction after the transaction is completed, in order to verify the validity of the transaction.

[0067] The transaction verification module consists of: storing historical vehicle transaction information, including license plate number, transaction time, entrance / exit station information, and transaction amount. Technical features: supporting fast queries based on multiple criteria such as license plate number and ETC device number; data retention period can be flexibly set according to system requirements (e.g., 30 days, 90 days).

[0068] Real-time transaction data interface: Obtains real-time transaction data from the current toll station for processing by the verification module. Data sources: Includes license plate recognition module, ETC information reading module, and comparison module. Verification logic unit: Verifies and compares transaction data according to preset rules and logical algorithms to determine the validity of the transaction. Verification dimensions: Time verification: Checks whether the transaction was completed within a reasonable time frame (e.g., whether the exit transaction time is later than the entry time). Route verification: Verifies whether the vehicle has followed a legitimate toll route; checks the transaction's legitimacy in conjunction with the route generation algorithm. Amount verification: Checks whether the transaction amount is consistent with the toll standard corresponding to the route. Duplicate verification: Determines whether the vehicle has made duplicate transactions within a short period (preventing duplicate charges).

[0069] Anomaly Marking Module: When verification fails or anomalies are detected, anomaly markers are generated and the information is recorded in the system log for subsequent analysis and manual intervention. Anomaly Types: Missing Entry Point: No entry transaction record was detected; Duplicate Transaction: The time interval between transaction records is too short; Path Anomaly: The vehicle's entry and exit do not conform to the legal path rules.

[0070] Feedback Unit: This unit sends the verification results back to the system's immediate release module or other relevant modules to determine whether the vehicle is allowed to pass or requires further processing. It receives vehicle transaction information (such as license plate number, transaction time, and entry / exit information) from the real-time transaction data interface. It also accesses the historical transaction database to query the vehicle's historical transaction records by license plate number or ETC device number.

[0071] Transaction validity is verified through the following steps: Time verification: confirming the transaction time is compliant; Route verification: checking if the entry and exit points are within a legal route; Amount verification: comparing the transaction amount with the route's toll standard; Duplicate verification: checking if the vehicle has repeatedly passed the same toll point within a short period. If the verification passes, a "Transaction Valid" flag is generated, allowing the vehicle to continue passing; if the verification fails, an anomaly flag is generated and sent to the anomaly flag module for recording, while simultaneously feeding back to the immediate release module, prompting manual intervention or refusal of passage. The verification result is returned to the system, and the feedback may include: verification status (valid / invalid); anomaly type (if any).

[0072] The anomaly detection module further includes: a time monitoring unit for recording the time after a vehicle enters the toll plaza; an anomaly rule base for storing rules related to vehicles not following the prescribed path or not completing transactions at the entrance; and a detection logic unit for comparing vehicle behavior with the rules to determine whether any anomalies exist.

[0073] The time monitoring unit is mainly used to record and track the time vehicles spend in the toll plaza, thus providing a time basis for anomaly detection. When a vehicle enters the toll plaza, the system automatically captures a timestamp and monitors its dwell time in different areas.

[0074] Use the license plate recognition module or ETC information reading module at the entrance to capture the vehicle's entry timestamp (Tentry); for exit transactions, record the vehicle's departure timestamp (Texit).

[0075] Calculate the time a vehicle spends within the toll plaza: Compare the length of stay Is it within a reasonable range, for example, the maximum allowed stay time is Tmax?

[0076] If a vehicle's dwell time exceeds Tmax or is less than the reasonable time threshold Tmin, it is marked as an abnormal vehicle.

[0077] The exception rule library stores a series of predefined rules used to detect whether vehicles violate traffic regulations. The rules cover various scenarios such as abnormal routes, abnormal transactions, and abnormal dwell times.

[0078] Define a valid path set Pvalid, such as the matching relationship between the entry station Sentry and the exit station Sexit: (Sentry, Sexit) ∈ Pvalid; if a vehicle path is not in Pvalid, it is marked as a path anomaly.

[0079] The trading rules include:

[0080] Rule 1: Vehicles must have an entry transaction record. If an exit transaction is conducted directly without an entry record, it will be marked as abnormal.

[0081] Rule 2: If a vehicle makes repeated transactions within a short period of time (e.g., the entry time interval is less than 5 minutes), it will be marked as abnormal.

[0082] Set a reasonable dwell time range: Tmin ≤ Tstay ≤ Tmax; if it exceeds this range, mark it as abnormal. Develop special rules for special vehicles (such as emergency vehicles and VIP vehicles), such as allowing free passage or detours via specific routes. The abnormal rule base supports dynamic updates, allowing rules to be added, modified, or deleted according to actual needs; rules are regularly synchronized with the toll system's superior management platform to ensure real-time performance and effectiveness.

[0083] The detection logic unit compares vehicle behavior data with rules in the anomaly rule base to determine if the vehicle exhibits any abnormal behavior. It then combines this with time information provided by the time monitoring unit and real-time data from other modules to comprehensively analyze vehicle behavior.

[0084] Receive vehicle timestamps, route information (entry and exit stations), transaction records, and other relevant information.

[0085] Check each rule in the exception rule base to see if it is satisfied: Path exception detection: Check if the vehicle path meets Pvalid;

[0086] Transaction anomaly detection: Verify whether there are any missing or duplicate transactions for a vehicle;

[0087] Time anomaly detection: Compare whether the vehicle dwell time is within the range of Tmin and Tmax.

[0088] If a vehicle violates any rules, an anomaly flag (Flagabnormal) is generated, and the anomaly type is recorded. The anomaly flag is then fed back to the system log and the immediate release module. For vehicles that repeatedly trigger abnormal rules, their abnormal behavior is recorded and reported to the management system; the system also supports generating anomaly statistical reports for manual review and decision-making reference.

[0089] The instant release module is used to immediately release a vehicle if it is detected that the transaction has been completed within a preset time.

[0090] Release Control Unit: Receives transaction verification results from the detection module and controls the release equipment in the toll plaza. Implementation: Control signal interface: Communicates with the gate or traffic light control system via standard protocols (such as RS-485 or CAN bus); Release operation: After confirming the transaction is complete and there are no abnormalities, sends a release signal to open the gate or switch to green light.

[0091] Multi-channel processing unit: In multi-lane environments, it identifies and assigns release signals to ensure vehicles accurately pass through designated lanes. Lane binding: It associates vehicle transaction information with the corresponding lane number; Multi-threaded processing: It processes vehicle release operations in different lanes in parallel to avoid interference and delays.

[0092] Anomaly Handling Unit: Handles incomplete transactions due to anomalies or errors, providing support for subsequent processing. Anomaly Notification: Informs drivers of transaction anomalies via LED display or voice prompts; Manual Operation: Prompts toll station administrators to manually verify and release the abnormal vehicle. Instant Release Module Workflow

[0093] Receive the detection results from the anomaly detection module or transaction verification module, including the following information: vehicle transaction status (valid / invalid); license plate number or ETC device number; detection timestamp and lane number.

[0094] Verification rules include: whether a valid transaction record exists; whether the transaction time is within a reasonable range; and whether the license plate information or ETC device information matches.

[0095] Based on the transaction verification results, the following decisions are made: if the transaction is completed and there are no abnormalities, a release signal is triggered; if the transaction is abnormal (such as duplicate transactions, abnormal paths, etc.), an exception handling process is triggered.

[0096] Control turnstiles or traffic lights: Normal passage: Open the turnstile or switch to green light to allow vehicles to pass; Abnormal handling: Keep the turnstile closed or switch to red light and prompt for manual verification. Store the passage results and related records in the system log for subsequent analysis and management.

[0097] In this embodiment, multiple modules are integrated to achieve efficient management and anomaly handling of Electronic Toll Collection (ETC) transactions. The license plate recognition module and the ETC information reading module work together to acquire vehicle license plate information and ETC device information, respectively. A comparison module matches the two to confirm the legality of the transaction. The transaction verification module further verifies completed transactions to ensure their validity. The anomaly detection module specifically handles vehicles that attempt to complete transactions at the exit without having passed through the entrance, using preset time and transaction rules to determine anomalies. The immediate release module, based on the transaction verification and anomaly detection results, quickly releases the vehicle after confirming the transaction, ensuring the security and efficiency of the toll collection system.

[0098] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0099] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0100] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An ETC transaction anomaly detection and immediate release processing system, characterized in that: It includes a license plate recognition module, an ETC information reading module, a comparison module, a transaction verification module, an anomaly detection module, and an instant release module; The license plate recognition module is used to obtain vehicle license plate information; The ETC information reading module is used to obtain license plate information from the vehicle's ETC device. The comparison module is used to compare the license plate information obtained by the license plate recognition module with the license plate information obtained by the ETC information reading module. If they match, the transaction is allowed. The transaction verification module is used to verify the vehicles involved in the transaction after the transaction is completed, in order to verify the validity of the transaction. The anomaly detection module is used to detect vehicles that have not passed through the entrance, obtain their ETC information, and determine whether they have made a transaction in the current toll plaza within a preset time. The instant release module is used to immediately release a vehicle if it is detected that the transaction has been completed within a preset time. The comparison module compares the license plate information obtained by the license plate recognition module with the license plate information obtained by the ETC information reading module. If a match is found, the transaction is allowed. Two strings, S1 and S2, with lengths n and m respectively, are used to construct a two-dimensional matrix D of size (n+1)×(m+1). Initialization: The initial state represents the editing cost from an empty string to the target string. Fill matrix D using the recursive formula: Among them: when , ,otherwise, Finally, the value of matrix D[n][m] is the edit distance between S1 and S2; Character similarity is calculated using a formula normalization: ;in, It is the value of matrix D[n][m], where HAK is the character deviation value, ranging from [0,1]. Convert the license plate images to grayscale, scale all license plate images to the same resolution, calculate the color histogram of the license plate images, and divide the license plate images into RGB or HSV channels; calculate the histogram for each channel. Normalize each histogram: ; i represents the segmented interval of pixel values, n is the number of intervals in the histogram, Canny edge detection is used to extract the outer border and character edges of the license plate; the contour feature vector Fs is calculated for the detected contours, character segmentation is performed on the image, and the coordinates of each character are extracted; character position feature vectors are constructed. , representing the center coordinates of each character, calculate the outlier similarity between the license plate images of two images using the following expression: In the formula, GHA represents the outlier value of the license plate image similarity. The character deviation values ​​and license plate image similarity anomalies are converted into feature vectors. These feature vectors are then used as input to a machine learning model. The machine learning model uses the prediction of the matching degree label of the license plate information for each set of feature vectors as the prediction objective and minimizes the sum of prediction errors for the matching degree labels of all license plate information as the training objective. The machine learning model is trained until the sum of prediction errors converges, at which point the model training stops. The matching degree value of the license plate information is determined based on the model output. The machine learning model is a multinomial regression model. The matching degree value of the obtained license plate information is compared with the gradient matching threshold, which includes a first matching threshold and a second matching threshold, and the first matching threshold is less than the second matching threshold. The matching degree value of the license plate information is compared with the first matching threshold and the second matching threshold respectively. If the matching degree value of the license plate information is greater than the second matching threshold, it means that the matching degree of the license plate information is high, and the license plate information is marked as a complete match, allowing the transaction; If the matching degree value of the license plate information is greater than or equal to the first matching threshold and less than or equal to the second matching threshold, it indicates that the matching degree of the license plate information is moderate. The license plate information is marked as an incomplete match and further verification is required. If the matching degree value of the license plate information is less than the first matching threshold, it means that the matching degree of the license plate information is low. The license plate information will be marked as mismatched and the transaction will be rejected.

2. The system for detecting and instantly releasing abnormal ETC transactions according to claim 1, characterized in that: The license plate recognition module includes: a high-resolution camera for capturing images of vehicle license plates; an optical character recognition unit for analyzing the license plate images and extracting character information; and a data processing unit for sending the extracted license plate information to the comparison module.

3. The system for detecting and instantly releasing abnormal ETC transactions according to claim 1, characterized in that: The ETC information reading module includes: a 5.8GHz microwave antenna for acquiring license plate and related transaction information from the vehicle's ETC device; a data decoder for decoding the received ETC device information; and a cache unit for temporarily storing the decoded ETC information for subsequent processing.

4. The system for detecting and instantly releasing abnormal ETC transactions according to claim 1, characterized in that: In the anomaly detection module, the license plate recognition module or ETC information reading module at the entrance is used to capture the vehicle entry timestamp (Tentry); for exit transactions, the vehicle departure timestamp (Texit) is recorded. Calculate the time a vehicle spends within the toll plaza: Compare the length of stay Is it within a reasonable range? Set the maximum allowable stay time as Tmax. If the vehicle stays for more than Tmax or less than the reasonable time threshold Tmin, it is marked as an abnormal vehicle. Define a path set Pvalid, which includes the matching relationship between the entry station Sentry and the exit station Sexit: (Sentry, Sexit) ∈ Pvalid; if a vehicle path is not in Pvalid, it is marked as a path anomaly. Set a reasonable stay time range: Tmin≤Tstay≤Tmax; if it exceeds this range, mark it as abnormal.

Images