Smuggled vehicle identification method based on highway ETC gantry data

Abstract

The present invention discloses a method for identifying smuggled vehicles based on highway ETC gantry data. The method comprises the following steps: within a preset high-frequency highway section, screening vehicle trips that are captured by both the starting and ending gantry license plate recognition systems and have gantry transaction records or gantry license plate recognition records at the source of goods within a set time period before entering the high-frequency highway section; calculating the number of gantry transactions and average speed for each vehicle trip; identifying vehicle trips that meet a first condition, characterized by the following characteristics: the number of gantry transactions is zero and the average speed is greater than a speed threshold; and identifying vehicle trips that meet a second condition, characterized by at least one of the following characteristics: rapid entry and exit at the source of goods, U-turns on the main line, or abnormal stops within the gantry section at the source of goods. The present invention can improve the timeliness and accuracy of smuggled vehicle identification.

Description

Technical Field

[0001] The present invention relates to the technical field related to big data, and more specifically, to a method for identifying smuggled vehicles based on highway ETC gantry data. Background Art

[0002] Traditional smuggling monitoring relies on manual inspections and simple equipment, making it difficult to analyze the massive amounts of highway traffic data. The daily volume of ETC gantry transaction records and license plate recognition data generated by highways is enormous, containing multi-dimensional information such as vehicle travel time, location, and route. However, traditional methods lack automated data processing capabilities, making it difficult to effectively screen out vehicle trips suspected of smuggling from this massive data. For example, for vehicles that frequently travel between cargo sources such as ports and border checkpoints and specific highway sections, traditional methods can only perform manual comparisons or simple rule-based screening, making it difficult to identify abnormal traffic patterns through data correlation, resulting in inefficient monitoring of vehicles with long-term, high-frequency traffic. Furthermore, some vehicles evade regulation by obscuring license plates or blocking traffic media, resulting in missing ETC transaction records or incomplete gantry license plate recognition information. Traditional systems are also severely inadequate in processing these data anomalies.

[0003] Therefore, it is necessary to design a technical solution that can overcome the above-mentioned defects to a certain extent. Summary of the Invention

[0004] An object of the present invention is to provide a method for identifying smuggled vehicles based on highway ETC gantry data, which can improve the timeliness and accuracy of smuggled vehicle identification.

[0005] To achieve these objectives and other advantages of the present invention, according to one aspect of the present invention, a method for identifying smuggled vehicles based on highway ETC gantry data is provided, comprising: S1: within a preset high-frequency passage section of the highway, screening vehicle trips that are simultaneously captured by the license plate recognition systems of the starting gantry and the ending gantry, and have a cargo source gantry transaction record or a gantry license plate recognition record within a set time period before entering the high-frequency passage section of the highway; S2: calculating the number of gantry transactions and the average speed of each vehicle trip within the high-frequency passage section; S3: identifying vehicle trips that meet the first condition, the first condition having the following characteristics : The number of gantry transactions without duplicates is zero, and the average speed is greater than the speed threshold; S4:: Extract the vehicle license plate number corresponding to the vehicle trip that does not meet the first condition but meets any of the conditions of the number of gantry transactions without duplicates being zero or the average speed being greater than the speed threshold, obtain the gantry license plate recognition record of the cargo source within the set time period before entering the high-frequency traffic section, and identify the vehicle trip that meets the second condition. The second condition has the following characteristics: at least one of the behaviors of quickly getting on and off the highway at the cargo source, turning around on the main line, or abnormal staying in the gantry section of the cargo source; S5: Add the license plate number corresponding to the vehicle trip that meets the first condition or the second condition to the suspected license plate database.

[0006] Furthermore, the speed threshold is 80% of the design speed of the high-frequency traffic section; the set time range before entering the high-frequency traffic section is 1 hour to 4 hours.

[0007] Furthermore, the gantry transaction record includes: license plate number, ETC gantry number, passing time, passing number; the gantry license plate recognition record extraction fields include: license plate number, ETC gantry number, passing time.

[0008] Furthermore, rapid entry and exit of highways include the time from the vehicle entering the highway to the vehicle exiting the highway within the gantry section of the cargo source being shorter than the preset time threshold; mainline U-turn behavior includes the vehicle generating continuous passage records in opposite directions within the same gantry section of the main line of the highway; abnormal stay behavior between gantry sections at the cargo source includes the vehicle staying between adjacent gantries at the cargo source for a time exceeding the stay threshold.

[0009] Furthermore, the S3 also includes: S31: dividing the high-frequency passage interval into at least 3 sub-intervals according to the ETC gantry distribution between the starting gantry and the ending gantry; S32: calculating the segmented speed of each sub-interval based on the vehicle's passage time stamp at adjacent gantries and the geographical spacing between the gantries; S33: if the following conditions are met at the same time, it is determined that the vehicle trip meets the condition that the average speed is greater than the speed threshold: the average speed of the entire journey is greater than the speed threshold; there is at least one sub-interval with a segmented speed greater than 85% of the design speed; the segmented speeds of more than 50% of the sub-intervals are greater than 75% of the design speed.

[0010] Furthermore, the S3 also includes: if there is a gantry license plate recognition record in the vehicle's journey within the high-frequency traffic area, then check whether there is a valid ETC transaction record within 24 hours before the vehicle enters the high-frequency traffic area, and count the transaction success rate of other vehicles passing through the high-frequency traffic area during the same period. If the success rate is ≥95%, the signal is judged to be normal; when the following conditions are met at the same time, it is judged that the vehicle journey meets the condition that the number of gantry transactions deduplication is zero: the vehicle ETC equipment status is normal; the gantry signal reliability is normal; the vehicle's gantry license plate recognition record capture rate in the high-frequency traffic area is ≥80%.

[0011] Furthermore, the S4 specifically includes: step S41: according to the cargo type and the approved load capacity stated in the vehicle driving license, the vehicle is divided into three categories: ordinary cargo vehicles, cold chain transport vehicles, and hazardous chemicals transport vehicles, and the site type of the cargo source gantry is obtained, and the site type includes ports, border checkpoints, and logistics parks; step S42: based on the historical passage records of the vehicle type at the same type of cargo source gantry in the past thirty days, the fast on and off highway benchmark value and the abnormal stay benchmark value are calculated by the historical average time and the historical average stay time; step S43: detect the fast on and off highway behavior, main line U-turn behavior and abnormal stay behavior in the cargo source gantry interval according to the following rules The conditions are as follows: rapid on- and off-highway behavior: the actual time taken by the vehicle from entering the highway from the cargo source gantry to exiting the highway is less than 60% of the rapid on- and off-highway reference value obtained in step S42; mainline U-turn behavior: the vehicle generates continuous passage records in opposite directions within the same gantry section of the main line of the highway, and there is no construction filing record in the highway management system during this period; abnormal stop behavior: the actual stay time of the vehicle between adjacent gantries at the cargo source is greater than 1.5 times the abnormal stop reference value obtained in step S42; step S44: if at least one of the rapid on- and off-highway behavior, mainline U-turn behavior, and abnormal stop behavior between the gantry sections at the cargo source occurs, it is determined that the vehicle trip meets the second condition.

[0012] Furthermore, the S43 further includes: obtaining the real-time traffic index T when the vehicle passes the cargo source gantry real ; Classify historical traffic data into peak hours, flat hours and low hours; dynamically adjust the fast on and off highway benchmark value and abnormal stop benchmark value through the time correction factor and traffic index correction factor. The corrected fast on and off highway benchmark value = K time ×K traffic × historical average time, corrected abnormal stay benchmark value = (K time / K traffic )×historical average length of stay; where: K time K is the time correction factor, which is 1.2 during peak hours, 1.0 during flat hours, and 0.8 during low hours. traffic is the traffic index correction factor, Ktraffic =1+0.5×(T real -0.6), takes effect when Treal>0.6, otherwise takes 1.0.

[0013] Furthermore, the S5 also includes: S51: constructing a dynamic scoring model for suspected vehicles, and generating a comprehensive risk value according to the following formula: comprehensive risk value = Σ (behavior weight coefficient × number of behavior occurrences) × spatial correlation coefficient × time activity factor; wherein, when configuring the behavior weight coefficient, mainline U-turn behavior > rapid high-speed on and off-highway behavior at the cargo source > abnormal stop behavior in the gantry section of the cargo source > average speed greater than the speed threshold > gantry transaction deduplication number is zero; spatial correlation coefficient is determined by the gantry position of the cargo source; time activity factor = 1 + ln (number of abnormal behaviors in the past 7 days + 1); S52: dynamically adjust the monitoring intensity of the license plate number according to the predefined interval in which the comprehensive risk value is located.

[0014] Furthermore, the method for generating the spatial correlation coefficient includes: constructing an electronic map containing smuggling risk areas; dividing the correlation level based on the topological relationship between the gantries at the source of goods and the risk areas; strong correlation: the gantries are located within the geographical fence of key supervision places, medium correlation: the straight-line distance between the gantries and the high-frequency case-involved areas is ≤ a preset threshold; weak correlation: other gantries that do not meet the above conditions; assigning spatial correlation coefficients according to the correlation level: the strong correlation level is assigned the first-tier coefficient, the medium correlation level is assigned the second-tier coefficient, and the weak correlation level is assigned the third-tier coefficient, where the first-tier coefficient > the second-tier coefficient > the third-tier coefficient.

[0015] The present invention has at least the following beneficial effects:

[0016] This invention precisely locates the travel routes of smuggling vehicles by defining high-frequency traffic zones and time periods associated with the source of goods, combining multi-dimensional screening of gantry transaction records and license plate recognition data. This avoids the blind screening of massive amounts of data required by traditional methods, reduces manual intervention, and significantly reduces data processing costs. A multi-layered identification system, encompassing deduplicated gantry transaction counts, average speed, and abnormal behavior at the source of goods, focuses on both abnormal characteristics of travel trajectories (such as no transaction records and speeding) and in-depth analysis of abnormal behavior patterns around the source of goods (such as rapid entry and exit from highways, mainline U-turns, and unusual stops). This creates a three-dimensional logic for identifying suspicious vehicles, effectively distinguishing between normal traffic and suspected smuggling.

[0017] Other advantages, objectives and features of the present invention will be reflected in part from the following description and will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flowchart of an embodiment of the present application. DETAILED DESCRIPTION

[0019] The present invention is described in further detail below so that those skilled in the art can implement the invention with reference to the description.

[0020] It should be understood that terms such as "having," "comprising," and "including" used in the embodiments of this application do not exclude the presence or addition of one or more other elements or combinations thereof. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are intended only to explain the relative positional relationships and movement of components in a specific posture. If the specific posture changes, the directional indications will also change accordingly. When an element is referred to as being "fixed to" or "disposed on" another element, it can be directly on the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element through an intervening element. References to "first," "second," etc. in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features designated as "first" or "second" may explicitly or implicitly include at least one of such features.

[0021] It should be noted that the technical solutions between the various embodiments of the present application can be combined with each other, but it must be based on the fact that ordinary technicians in this field can implement it. When the combination of technical solutions is mutually contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0022] like Figure 1As shown, the embodiment of the present application provides a method for identifying smuggled vehicles based on highway ETC gantry data, including: S1: within a preset high-frequency passage section of the highway, screening vehicle trips that are captured by both the starting gantry and the end gantry license plate recognition systems and have a cargo source gantry transaction record or a gantry license plate recognition record within a set time period before entering the high-frequency passage section of the highway; S2: calculating the number of gantry transactions and the average speed of each vehicle trip within the high-frequency passage section; S3: identifying vehicle trips that meet the first condition, the first condition having the following characteristics: the number of gantry transactions is zero, And the average speed is greater than the speed threshold; S4: Extract the vehicle license plate number corresponding to the vehicle trip that does not meet the first condition but meets any of the conditions that the number of gantry transactions is zero or the average speed is greater than the speed threshold, obtain the gantry license plate recognition record of the cargo source within the set time period before entering the high-frequency traffic area, and identify the vehicle trip that meets the second condition. The second condition has the following characteristics: at least one of the following behaviors: rapid entry and exit of the highway at the cargo source, U-turn on the main line, or abnormal stop in the gantry area of ​​the cargo source; S5: Add the license plate number corresponding to the vehicle trip that meets the first condition or the second condition to the suspected license plate database.

[0023] For example, "pre-set high-traffic highway sections" require statistical analysis of historical data to select continuous sections with more than 50 one-way traffic counts within the past 30 days, such as the 50-kilometer mainline section of a provincial highway between Port A and Logistics Park B. The license plate recognition systems at the starting and ending gantries can utilize the license plate recognition algorithm powered by Huawei's Atlas 500 Smart Station to capture vehicle images and analyze license plate numbers in real time. The equipment is installed in the center of the top crossbar of the gantries to ensure full lane coverage. The cargo source gantries include dedicated channel gantries connecting ports and border checkpoints. Transaction records for these gantries are generated by Jinyi Technology's ETC lane system and include the license plate number, gantry number (e.g., G15-0321), travel time (accurate to the second), and pass number (system-generated UUID). License plate recognition records are captured using a Hikvision DS-2CD864EF-E camera, and the extracted fields are aligned with the license plate number, gantry number, and timestamp in the transaction records.

[0024] The screening logic of S1 is as follows: first, all license plate recognition records of the starting point (such as K100+000 gantry) and the end point (such as K150+000 gantry) of the high-frequency passage section between 0:00 and 24:00 on the same day are retrieved from the provincial transportation data center, and the initial trip set is formed by license plate number association; then, for each trip, check whether there is at least one transaction record or license plate recognition record in the cargo source gantry list (such as including 5 port-specific gantries) within 1 hour, 2 hours or 4 hours (optional according to regional transportation characteristics) before the time of entering the starting gantry, and screening is achieved through a database left join operation.

[0025] In S2, the number of gantry transactions without duplicates is a unique count of ETC transaction records within the travel time range according to the gantry number. For example, if a vehicle passes through three gantries twice in a high-frequency interval, the number after deduplication is 3; the average speed is calculated as the actual mileage of the interval (the route length from the start point to the end point is obtained through the GIS system, such as 50 kilometers) divided by the travel time (the travel time from the end point minus the travel time from the start point, converted into hours).

[0026] The first condition for S3 requires both a zero deduplication transaction count (i.e., no ETC transactions were triggered during the high-frequency range) and an average speed exceeding a threshold (e.g., 80% of the design speed). S4 then reverse-checks the vehicle's license plate identification records for the cargo source within a set timeframe before it entered the high-frequency range. It then identifies abnormal behavior based on pre-defined behavioral rules (e.g., rapid highway entry and exit is defined as entering and exiting a port gantry within one hour).

[0027] In this example, by correlating high-frequency intervals with the temporal and spatial relationships of the cargo source, we first filter out vehicle trips with potential smuggling risks. We then identify initial anomalies through dual verification of transaction and speed data. Finally, we conduct a secondary screening based on behavioral characteristics surrounding the cargo source. Compared to traditional monitoring methods that rely solely on a single data dimension, this layered approach enables a more systematic integration of highway traffic data, gradually narrowing the scope of suspects. This overcomes the difficulty of extracting effective information from massive amounts of data and provides a structured data foundation for subsequent, accurate identification.

[0028] In another embodiment, the speed threshold is 80% of the design speed of the high-frequency traffic section; and the set time range before entering the high-frequency traffic section is 1 hour to 4 hours.

[0029] For example, the "high-frequency traffic section design speed" is determined according to the "Highway Engineering Technical Standards". For example, the design speed of a two-way six-lane expressway is 120km / h, and the corresponding speed threshold is 96km / h; the design speed of a two-way four-lane mountainous section is 80km / h, and the threshold is 64km / h. The time ranges of 1 hour, 2 hours, and 4 hours are set to correspond to different scenarios such as short-distance connection, mid-distance transportation, and long-distance transfer. For example, the high-frequency interval adjacent to the port can be selected as 1 hour, and the high-frequency interval for inter-provincial transportation can be selected as 4 hours. During specific screening, the system uses the timestamp of the vehicle entering the starting gantry to intercept the time window of the corresponding length and query whether there is a valid record of the gantry at the source of goods within the window.

[0030] In this embodiment, the speed threshold is linked to the design speed of the road section, avoiding the adaptability problem of the fixed threshold under different road conditions. For example, differentiated standards are adopted for plain highways and mountain highways, making the speeding judgment more in line with actual driving specifications. The tiered design with a set time range can cover the common transportation cycle of smuggling vehicles from the source of goods to entering the high-frequency section. For example, in a certain border smuggling scenario, the typical transportation time for vehicles from the border checkpoint to the high-frequency section is 2.5 hours. Selecting a range of 1-4 hours can effectively include this period. Compared with the traditional method of setting thresholds based on experience, this parameter setting method is more scientific and flexible, and improves the accuracy and applicability of data screening.

[0031] In another embodiment, the gantry transaction record includes: license plate number, ETC gantry number, passing time, passing number; the gantry license plate recognition record extraction fields include: license plate number, ETC gantry number, passing time.

[0032] For example, gantry transaction records are generated by the interaction between the ETC lane's RSU device (such as Wanjie Technology's CPC-500) and the onboard OBU. The pass number serves as the primary key to ensure record uniqueness. Gantry numbers use a nationally standardized 12-digit encoding scheme, and the transit time is calibrated to the second using the device's built-in Beidou timing module. Gantry license plate recognition records are collected by a Uniview Technology ANPR-3000 camera installed 50 meters upstream of the gantry. License plates are identified using a deep learning algorithm. The extracted fields are strictly aligned with the license plate number, gantry number, and timestamp in the transaction record, facilitating subsequent data fusion.

[0033] When storing data, transaction records and license plate recognition records are stored in the highway toll system database and video surveillance database, respectively, and are correlated and matched based on license plate number and travel time. When processing a vehicle's trip, all transaction records within the high-frequency range are first retrieved, and SQL statements are used to calculate the number of duplicates. License plate recognition records are also retrieved to verify whether the vehicle actually passed through each key gantries, ensuring data integrity.

[0034] In this embodiment, the field structure of the two types of data is clearly defined, providing a standardized input format for subsequent data processing. The pass number in the transaction record can be used to track individual passes, while the gantry number enables precise positioning. The license plate recognition record serves as supplementary verification of transaction data, particularly in the event of a vehicle OBU device failure or signal obstruction. The license plate recognition record confirms the vehicle's actual passage. This dual-source data collection approach addresses the vulnerability of traditional single data sources to equipment failure, improving data reliability and the fault tolerance of the recognition process.

[0035] In another embodiment, the rapid entry and exit behavior of the highway includes the time from the vehicle entering the highway to the vehicle exiting the highway within the gantry section of the cargo source being shorter than a preset time threshold; the main line U-turn behavior includes the vehicle generating continuous passage records in opposite directions within the same gantry section of the main line of the highway; the abnormal stay behavior between the gantry sections of the cargo source includes the vehicle staying between adjacent gantries at the cargo source for a time exceeding the stay threshold.

[0036] For example, the "gantry interval at the cargo source" refers to the highway entrance and exit gantries corresponding to the cargo source and the connecting sections between them, such as the area between the G18-0123 entrance gantry and the G18-0124 exit gantry corresponding to a port. The "preset time threshold" is set according to the normal travel time of different cargo sources, and can be selected as 30 minutes, 45 minutes or 60 minutes. For example, if the traffic flow around the logistics park is smooth, the threshold can be set to 30 minutes. Due to the complex loading and unloading operations in the port area, the threshold can be set to 60 minutes. To determine the behavior of quickly getting on and off the highway, it is necessary to extract the vehicle's entry time (entrance gantry sign recognition time) and exit time (exit gantry sign recognition time) in the interval, calculate the time difference and compare it with the threshold.

[0037] The "same gantry section" in the "Mainline U-turn" section refers to the mainline highway between two adjacent interchanges, such as the section between K20+000 and K30+000. Consecutive passes in opposite directions must occur within a one-hour time difference and have the same gantry number (e.g., first passing through the K25+000 gantry heading toward Beijing, then again 15 minutes later, heading toward Shanghai). The highway management system API is also used to check whether there is a construction closure registered during that period. If no such record is available, the record is considered an anomaly.

[0038] In the "Abnormal Dwell Behavior" category, "adjacent gantries at the cargo source" refers to the two closest gantries within the cargo source, such as entry gantry A and exit gantry B at a border checkpoint. Dwell time is calculated as the difference between the two gantries' transit times. For example, if a vehicle passes through entry gantry A at 10:00 and exit gantry B at 1:30 PM, the dwell time is 3.5 hours. The dwell threshold is set based on the cargo source type; it can be set to 3 hours in a port area or 2 hours in a logistics park. Exceeding the threshold is considered an abnormal dwell.

[0039] In this example, by defining specific rules and thresholds for three types of abnormal behavior, the system can automatically identify unusual traffic patterns around cargo sources based on gantry data. Compared to the fuzzy judgments of traditional manual analysis, this quantitative behavior definition and standardized calculation method significantly improves the operability and accuracy of abnormal behavior identification. This effectively monitors evasive tactics commonly used by smuggling vehicles, such as rapid entry and exit and mid-route turnarounds, addressing the problem of missed detections caused by unclear behavioral characteristics in traditional methods.

[0040] In another embodiment, the S3 also includes: S31: dividing the high-frequency passage interval into at least 3 sub-intervals according to the ETC gantry distribution between the starting gantry and the ending gantry; S32: calculating the segmented speed of each sub-interval based on the vehicle's passage timestamps at adjacent gantries and the geographical spacing between the gantries; S33: if the following conditions are met at the same time, it is determined that the vehicle trip meets the condition that the average speed is greater than the speed threshold: (a) the average speed of the entire journey > the speed threshold; (b) there is at least one sub-interval with a segmented speed > 85% of the design speed; (c) the segmented speeds of more than 50% of the sub-intervals are > 75% of the design speed.

[0041] For example, when dividing the high-frequency traffic section in S31, if the starting gantry is K100+000 and the ending gantry is K160+000, with two ETC gantries located in between, K120+000 and K140+000, then the section is divided into three sub-sections: K100-K120, K120-K140, and K140-K160, each 20 kilometers long. When calculating the segment speed in S32, the vehicle's passing times at adjacent gantries are obtained (for example, the K100 gantry's passing times are 10:00:00 and the K120 gantry's passing times are 10:15:00). Using the geographic information system, the distance between the two gantries is 20 kilometers. The segment speed is calculated as 80 km / h based on the distance divided by the time difference (in hours).

[0042] In condition (a) of S33, the speed threshold is 80% of the design speed for the high-frequency section (e.g., 96 km / h for a design speed of 120 km / h). The average speed for the entire journey is calculated as: total mileage divided by total driving time. Condition (b) requires that the segmented speed of at least one sub-section exceed 85% of the design speed (e.g., 102 km / h for a design speed of 120 km / h). Condition (c) requires that the segmented speeds of more than 50% of the sub-sections exceed 75% of the design speed (e.g., 90 km / h). For example, if the segmented speeds of three sub-sections are 100 km / h, 95 km / h, and 85 km / h, 100 > 102 does not hold, 95 > 90 does, and 85 > 90 does not. Since more than 50% (1.5) of the segments are not satisfied, the condition is not triggered.

[0043] In this embodiment, by subdividing the high-frequency interval into multiple subintervals and calculating the segmented speed, and combining the multi-condition judgment of the full-distance average speed and the segmented speed, the problem that a single average speed may mask local speeding or mid-route stagnation is avoided. For example, if a vehicle's full-distance average speed just exceeds the threshold, but the speeds in most subintervals are below the reasonable range, the system will not mistakenly judge it as an abnormality; on the other hand, vehicles that exceed the speed in some subintervals and whose overall speed meets the characteristics will be accurately identified. Compared with the traditional judgment method that only relies on the full-distance average speed, this multi-dimensional speed analysis method can more carefully capture the driving anomalies of vehicles in different sections of the road, effectively eliminate misjudgments caused by local congestion or speeding in individual sections, and improve the reliability of speed anomaly identification.

[0044] In another embodiment, S3 also includes: if there is a gantry license plate recognition record in the high-frequency traffic section during the vehicle journey, then checking whether there is a valid ETC transaction record within 24 hours before the vehicle enters the high-frequency traffic section, and counting the transaction success rate of other vehicles passing through the high-frequency traffic section during the same period. If the success rate is ≥95%, the signal is determined to be normal; when the following conditions are met at the same time, it is determined that the vehicle journey meets the condition that the number of gantry transactions deduplication is zero: the vehicle ETC device status is normal; the gantry signal reliability is normal; the vehicle's gantry license plate recognition record capture rate in the high-frequency traffic section is ≥80%.

[0045] For example, "valid ETC transaction records" refer to those with a "successful" transaction status and have not been revoked. This is determined by querying the transaction log of the vehicle's OBU device within the 24 hours prior to entering the high-frequency zone. The "transaction success rate of other vehicles during the same period" measures the proportion of successful ETC transactions among all vehicles that passed through the high-frequency zone during the same time period (with an error of ±15 minutes) as the vehicle. If 96 out of 100 vehicles have successful transactions, the success rate is 96%, meeting the ≥95% signal normality requirement.

[0046] "Normal vehicle ETC device status" is defined as the device being activated and the most recent transaction (regardless of whether it occurred during a high-frequency interval) being successful. The device status flag can be checked through the highway network toll collection system. The "Gantry Sign Record Capture Rate" is calculated as the ratio of the number of gantries captured by the sign recognition system (such as Hikvision's sign recognition equipment) during a high-frequency interval to the number of gantries actually passed. For example, if a vehicle passes through five gantries and the sign recognition record shows four, the capture rate is 80%. If the number of deduplicated transactions is zero (i.e., no valid ETC transaction records), and the device and signal conditions are normal, and the capture rate is ≥80%, then the behavior is considered to be an intentional attempt to evade transactions, excluding equipment failure or signal issues.

[0047] In this embodiment, triple verification of the zero-duplicate transaction count ensures the validity of the transaction, avoiding misjudgments caused by ETC equipment failures and gantry signal coverage issues. For example, if a vehicle's OBU battery depletes and a transaction fails, the system will detect the abnormal device status and exclude it from suspicion. If a gantry signal failure causes multiple transactions to fail on the same day, the success rate of other vehicles' transactions (e.g., below 95%) can be used to determine the signal anomaly and avoid batch misjudgments. This multi-level data verification mechanism significantly improves the rigor of abnormal behavior identification compared to the traditional method of simply determining an anomaly based on a zero transaction record, reduces false alarms caused by technical factors, and enables the monitoring system to focus more on smuggled vehicles deliberately evading regulation.

[0048] In another embodiment, the S4 specifically includes: step S41: according to the cargo type and the approved load capacity stated in the vehicle driving license, the vehicle is divided into three categories: ordinary cargo vehicles, cold chain transport vehicles, and hazardous chemicals transport vehicles, and the site type of the cargo source gantry is obtained, and the site type includes ports, border checkpoints, and logistics parks; step S42: based on the historical passage records of the vehicle type at the same type of cargo source gantry in the past thirty days, the fast on and off highway benchmark value and the abnormal stop benchmark value are calculated by the historical average time and the historical average stay time; step S43: according to the following rules, the fast on and off highway behavior, the main line U-turn behavior and the abnormal stop between the cargo source gantries are detected Stay behavior: Rapid on / off highway behavior: The actual time taken by the vehicle to enter the highway from the cargo source gantry to exit the highway is less than 60% of the rapid on / off highway benchmark value obtained in step S42; Mainline U-turn behavior: The vehicle has continuous passage records in the opposite direction within the same gantry section of the main line of the highway, and there is no construction filing record in the highway management system during this period; Abnormal stay behavior: The actual stay time of the vehicle between adjacent gantries at the cargo source is greater than 1.5 times the abnormal stay benchmark value obtained in step S42; Step S44: If at least one of the rapid on / off highway behavior, mainline U-turn behavior and abnormal stay behavior between the gantry sections at the cargo source occurs, it is determined that the vehicle trip meets the second condition.

[0049] For example, vehicle classification in S41 is achieved by reading the "Vehicle Type" field from the electronic driving license file (e.g., the Ministry of Transport's Motor Vehicle Information Database). Vehicles with a certified load capacity ≤ 1.5 tons are classified as general cargo vehicles, those equipped with refrigeration equipment are classified as cold chain transport vehicles, and those marked "Dangerous Goods Transport" are classified as hazardous chemical transport vehicles. The type of gantry site at the cargo source is associated with the gantry's geographic location. For example, a gantry located in a customs-controlled area is marked as a "border checkpoint," while one connected to a large logistics warehouse is marked as a "logistics park."

[0050] In S42, the historical average time is calculated as the average of the time differences between entry and exit of a certain type of vehicle at the same type of cargo source gantry (such as port gantry) in the past 30 days. For example, the historical average time taken by ordinary cargo vehicles at the port gantry is 90 minutes, so the benchmark value for fast entry and exit of the highway is 90 minutes; the historical average stay time is the average of the stay time between adjacent gantries. For example, the average stay time of cold chain transport vehicles at adjacent gantries at the border checkpoint is 120 minutes, and the benchmark value for abnormal stay is 120 minutes.

[0051] In S43, a rapid on / off trip is considered when the actual duration is less than 60% of the baseline (e.g., 90 minutes x 60% = 54 minutes). An abnormal stop is considered when the actual dwell time is greater than 1.5 times the baseline (e.g., 120 minutes x 1.5 = 180 minutes). Mainline U-turns still require a check for construction records, which is achieved by invoking the Provincial Department of Transportation's construction management system API.

[0052] In this embodiment, by modeling the classification of vehicle types and cargo source types, and combining historical data to calculate dynamic benchmark values, abnormal behavior judgments are made to better fit the normal operating characteristics of different vehicles. For example, cold chain transport vehicles normally stay for a long time due to the need for cargo loading and unloading. By setting exclusive benchmark values ​​based on historical data, it is possible to avoid misjudging reasonable stops as abnormalities; the rapid passage of ordinary trucks in logistics parks may meet normal transportation needs. By comparing with the historical average time consumption, the extremely short time consumption that is significantly abnormal can be accurately identified. Compared with the traditional unified threshold judgment method, this differentiated analysis method effectively reduces the misjudgment rate caused by differences in vehicle types and scenarios, and improves the adaptability of the recognition system to complex transportation scenarios.

[0053] In another embodiment, the S43 also includes: obtaining the real-time traffic index Treal when the vehicle passes through the cargo source gantry; classifying the historical traffic data according to peak hours (7:00-9:00, 17:00-19:00), off-peak hours (9:00-17:00) and low-peak hours (19:00-7:00 the next day); dynamically adjusting the fast on and off-highway benchmark value and the abnormal stop benchmark value through the time period correction factor and the traffic index correction factor, the corrected fast on and off-highway benchmark value = Ktime × Ktraffic × historical average time consumption, the corrected abnormal stop benchmark value = (Ktime / Ktraffic) × historical average stop duration; where: Ktime is the time period correction factor, which is 1.2 during peak hours, 1.0 during off-peak hours, and 0.8 during low-peak hours; Ktraffic is the traffic index correction factor, Ktraffic=1+0.5×(Treal-0.6), which takes effect when Treal>0.6, otherwise it takes 1.0.

[0054] For example, the real-time traffic index, Treal, is obtained by accessing the traffic API of AutoNavi or Baidu Maps and reflects the degree of congestion on roads surrounding the cargo source portal (0-1 indicates smooth traffic, 1-2 indicates light congestion, and so on). Historical data is automatically categorized into peak, off-peak, and low-peak periods based on vehicle travel time. For example, a vehicle traveling at 8:30 AM is considered peak traffic, while a vehicle traveling at 2:00 PM is considered off-peak.

[0055] When calculating the correction factor, if Treal = 0.8 (mild congestion), then Ktraffic = 1 + 0.5 × (0.8 - 0.6) = 1.1. During peak hours, Ktime = 1.2, and the corrected baseline for rapid on / off access is 1.2 × 1.1 × the historical average time (e.g., 90 minutes) = 118.8 minutes. The actual time must be less than 118.8 × 60% = 71.28 minutes to be considered abnormal. The baseline for abnormal dwell time is (1.2 / 1.1) × 120 minutes = 130.9 minutes. The actual dwell time exceeds 130.9 × 1.5 = 196.4 minutes to be considered abnormal. If Treal = 0.5 (unimpeded traffic), then Ktraffic = 1.0. During off-peak hours, Ktime = 0.8, and the corrected baseline for rapid on / off access is 0.8 × 1.0 × 90 = 72 minutes, while the baseline for abnormal dwell time is (0.8 / 1.0) × 120 = 96 minutes.

[0056] In this embodiment, by introducing time period and traffic index correction factors, the benchmark value can be dynamically adjusted based on real-time traffic conditions and time period characteristics, solving the problem that traditional fixed benchmark values ​​cannot adapt to differences in congested or unobstructed scenarios. For example, during peak hours, traffic is congested, and the time required for normal vehicle travel increases. The corrected benchmark value for rapid on / off highways is correspondingly increased, avoiding misjudging the normal time required due to congestion as "slow." During off-peak hours, when traffic is smooth, the benchmark value is lowered, allowing for more sensitive capture of abnormally rapid travel. This dynamic correction mechanism enables the system to adapt to complex traffic environments in real time, significantly improving the accuracy of abnormal behavior determination and scenario adaptability, and is particularly effective in identifying the covert transportation activities of smuggled vehicles that exploit peak hours in the morning and evening.

[0057] In another embodiment, the S5 also includes: S51: constructing a dynamic scoring model for suspicious vehicles, and generating a comprehensive risk value according to the following formula: comprehensive risk value = Σ (behavior weight coefficient × number of behavior occurrences) × spatial correlation coefficient × time activity factor; wherein, when configuring the behavior weight coefficient, mainline U-turn behavior > rapid high-speed on and off-highway behavior at the cargo source > abnormal stop behavior in the gantry section of the cargo source > average speed greater than the speed threshold > gantry transaction deduplication number is zero; spatial correlation coefficient is determined by the gantry position of the cargo source; time activity factor = 1 + ln (number of abnormal behaviors in the past 7 days + 1); S52: dynamically adjust the monitoring intensity of the license plate number according to the predefined interval in which the comprehensive risk value is located.

[0058] For example, the behavior weighting coefficients in S51 are set as follows: 3.0 for mainline U-turns, 2.5 for rapid on / off traffic, 2.0 for unusual stops, 1.5 for unusual speeds, and 1.0 for zero transactions. The spatial correlation coefficient is divided into three levels: strong correlation (gantry located within the border checkpoint's geofence) with a coefficient of 1.5, moderate correlation (gantry distance to a high-frequency crime zone ≤ 5 kilometers) with a coefficient of 1.2, and weak correlation (other situations) with a coefficient of 1.0. When calculating the temporal activity factor, if a vehicle has experienced two unusual behaviors in the past seven days, the factor is 1 + ln(2 + 1) ≈ 2.0986.

[0059] Example of a comprehensive risk value calculation: A vehicle made one mainline U-turn (3.0 × 1), one rapid highway entry / exit (2.5 × 1), a strong correlation (1.5) at the source portal, and two total anomalies over the past seven days (time factor ≈ 2.0986). The comprehensive risk value is (3.0 + 2.5) × 1.5 × 2.0986 ≈ 17.31. S52 predefined intervals, such as 0-10 for low risk (routine monitoring), 10-20 for medium risk (increased data verification frequency), and 20+ for high risk (triggering a real-time alert).

[0060] In this embodiment, a dynamic scoring model incorporating behavioral weights, spatial correlations, and temporal activity factors is constructed to quantitatively assess the risk level of suspected vehicles. Compared to traditional methods that rely solely on single-behavior assessments, this model comprehensively considers the severity of abnormal behavior (e.g., mainline U-turns are given a higher weight), the risk level of the cargo source (vehicles in strongly correlated areas have a higher risk factor), and the frequency of recent abnormalities (the activity factor amplifies the risk of frequently abnormal vehicles), thereby allocating monitoring resources to high-risk vehicles. For example, vehicles that frequently make mainline U-turns near border checkpoints will have a significantly increased overall risk score, triggering real-time alerts and facilitating precise control by law enforcement. This differentiated monitoring strategy effectively addresses the resource waste associated with traditional "one-size-fits-all" monitoring methods and improves the overall efficiency of smuggled vehicle control.

[0061] In another embodiment, the method for generating the spatial correlation coefficient includes: constructing an electronic map including smuggling risk areas; dividing the correlation level based on the topological relationship between the gantries at the source of goods and the risk areas; strong correlation: the gantries are located within the geographical fence of key supervision places, medium correlation: the straight-line distance between the gantries and the high-frequency case-involved areas is ≤ a preset threshold; weak correlation: other gantries that do not meet the above conditions; assigning spatial correlation coefficients according to the correlation level: the strong correlation level is assigned the first-tier coefficient, the medium correlation level is assigned the second-tier coefficient, and the weak correlation level is assigned the third-tier coefficient, where the first-tier coefficient > the second-tier coefficient > the third-tier coefficient.

[0062] For example, the "smuggling risk zone" is constructed by integrating historical high-incidence locations (such as border smuggling corridors and port smuggling terminals) provided by customs anti-smuggling departments into an electronic map. Amap's API is then used to draw a geofence (a circular area with a 500-meter radius). The preset threshold is set at 5 kilometers, meaning the straight-line distance between the associated portal and the risk zone must not exceed 5 kilometers. This can be calculated using Baidu Maps' distance measurement tool.

[0063] When categorizing the correlation levels, a gantry located within the border checkpoint's geofence (latitude and longitude deviation ≤ 50 meters) is considered strongly correlated; another gantry located 4.8 kilometers from a smuggling terminal is considered moderately correlated; and a gantry located in a logistics park, far from all risk areas, is considered weakly correlated. Spatial correlation coefficients are assigned to three levels: 1.5, 1.2, and 1.0, and stored in a system configuration table for use by the risk scoring model.

[0064] In this embodiment, a geographic information system (GIS) model is used to associate risk areas with gantries, enabling the system to automatically identify vehicle traffic around high-risk areas. For example, a gantry located within the fence of a border checkpoint will have its associated vehicles assigned a higher risk weighting for abnormal behavior. Even a single instance of rapid entry or exit from the highway will significantly increase the overall risk value, resulting in priority monitoring. This spatially-based risk-addition mechanism, compared to traditional monitoring methods that don't consider geographic factors, can more accurately target vehicles in high-risk areas for smuggling. This resolves the issue of ambiguous risk classification in cross-regional monitoring and provides law enforcement with more targeted early warning information.

[0065] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. A smuggled vehicle identification method based on highway ETC gantry data is characterized by: include: S1: Within a preset high-frequency highway access zone, screen the vehicles that are captured by both the starting and ending gantry license plate recognition systems, and have a cargo source gantry transaction record or gantry license plate recognition record within a set time period before entering the high-frequency highway access zone; S2: Calculate the number of gantry transactions and average speed for each vehicle within the high-frequency travel interval. The number of gantry transactions is a unique count of ETC transaction records within the travel time range by gantry number. S3: Identify a vehicle trip that meets a first condition, where the first condition has the following characteristics: the number of gantry transactions without duplicates is zero, and the average speed is greater than a speed threshold; S4: Extract the vehicle license plate number corresponding to the vehicle trip that does not meet the first condition but meets any of the conditions of zero gantry transaction deduplication number or average speed greater than the speed threshold, obtain the vehicle gantry license plate recognition record at the cargo source within a set time period before entering the high-frequency traffic zone, and identify the vehicle trip that meets the second condition, which has the following characteristics: at least one of the following: rapid entry and exit of the highway at the cargo source, U-turn on the main line, or abnormal stop in the cargo source gantry zone; S5: Add the license plate number corresponding to the vehicle itinerary that meets the first condition or the second condition to the suspected license plate database.

2. The method for identifying smuggled vehicles based on highway ETC gantry data according to claim 1, characterized in that: The speed threshold is 80% of the design speed of the high-frequency traffic section; the set time range before entering the high-frequency traffic section is 1 hour to 4 hours.

3. The method for identifying smuggled vehicles based on highway ETC gantry data according to claim 1, characterized in that: Gantry transaction records include: license plate number, ETC gantry number, passage time, and passage number; The fields extracted from the gantry license plate recognition record include: license plate number, ETC gantry number, and passage time.

4. The method for identifying smuggled vehicles based on highway ETC gantry data according to claim 1, characterized in that: Rapid entry and exit of the highway includes the vehicle entering and exiting the highway within the gantry area of ​​the cargo source, with the duration being shorter than the preset time threshold; Mainline U-turn behavior includes the continuous passage records of vehicles in opposite directions within the same gantry section of the main line of the expressway; Abnormal dwelling behavior between gantries at the cargo source location includes the vehicle's dwelling time between adjacent gantries at the cargo source location exceeding the dwelling threshold.

5. The method for identifying smuggled vehicles based on highway ETC gantry data according to claim 1, characterized in that: Said S3 further comprises: S31: Divide the high-frequency traffic section into at least three subsections based on the distribution of ETC gantries between the starting gantries and the ending gantries; S32: Calculate the segmented speed of each sub-interval based on the vehicle's passing time stamps at adjacent gantries and the gantries' geographical distance; S33: If the following conditions are met at the same time, it is determined that the vehicle travel meets the condition that the average speed is greater than the speed threshold: The average speed of the entire journey is greater than the speed threshold; There is at least one sub-interval with a segment speed greater than 85% of the design speed; The segmented speed of more than 50% of the sub-intervals is greater than 75% of the design speed.

6. The method for identifying smuggled vehicles based on highway ETC gantry data as claimed in claim 4, characterized in that: The S4 specifically includes: Step S41: Based on the cargo type and rated load capacity stated in the vehicle license, the vehicle is classified into three categories: general cargo vehicle, cold chain transport vehicle, and hazardous chemical transport vehicle. The type of site to which the gantry at the cargo source belongs is obtained, including ports, border checkpoints, and logistics parks. Step S42: Based on the historical passage records of the vehicle type at the same type of cargo source gantry in the past thirty days, the rapid access and abnormal stop benchmark values ​​are calculated using the historical average time consumption and the historical average stop duration; Step S43: Detecting behaviors of quickly getting on and off the highway at the cargo source, mainline U-turns, and abnormal stops between gantry sections at the cargo source according to the following rules: Rapid on / off highway behavior: The actual time taken by the vehicle from entering the highway from the cargo source gantry to exiting the highway is less than 60% of the rapid on / off highway benchmark value obtained in step S42; Mainline U-turn behavior: Vehicles have continuous passage records in opposite directions within the same gantry section of the expressway mainline, and there is no construction record in the expressway management system during the corresponding period; Abnormal dwell behavior: The actual dwell time of the vehicle between adjacent gantries at the cargo source is greater than 1.5 times the abnormal dwell reference value obtained in step S42; Step S44: If at least one of the following behaviors occurs: rapid entry and exit of the expressway at the cargo source, U-turn on the main line, and abnormal stop at the gantry section at the cargo source, it is determined that the vehicle journey meets the second condition.

7. The method for identifying smuggled vehicles based on highway ETC gantry data according to claim 1, characterized in that: The S5 further includes: S51: Construct a dynamic scoring model for suspicious vehicles and generate a comprehensive risk value according to the following formula: Comprehensive risk value = Σ(behavior weight coefficient × number of behavior occurrences) × spatial correlation coefficient × time activity factor When configuring the behavior weight coefficient, the following criteria are met: Mainline U-turn > Rapidly getting on and off the highway at the cargo source > Abnormal stop behavior in the cargo source gantry area > Average speed greater than the speed threshold > The number of gantry transactions with zero deduplication; The spatial correlation coefficient is determined by the cargo source gantry location; The temporal activity factor = 1 + ln(Number of abnormal behaviors in the past 7 days + 1); S52: Dynamically adjust the monitoring intensity of the license plate number according to the predefined interval of the comprehensive risk value.

8. The method for identifying smuggled vehicles based on highway ETC gantry data according to claim 7, characterized in that: Methods for generating spatial correlation coefficients include: Constructing electronic maps of smuggling risk areas; The association level is divided based on the topological relationship between the gantry at the source of goods and the risk area; strong association: the gantry is located within the geographical fence of the key supervision site; medium association: the straight-line distance between the gantry and the high-frequency case-involved area is ≤ the preset threshold; weak association: other gantry that does not meet the above conditions; The spatial correlation coefficient is assigned according to the correlation level: the first-tier coefficient is assigned to the strong correlation level, the second-tier coefficient is assigned to the medium correlation level, and the third-tier coefficient is assigned to the weak correlation level, among which the first-tier coefficient > the second-tier coefficient > the third-tier coefficient.

Images