Traffic key vehicle judgment method fusing visual attributes and OBU registration information

Abstract

The invention discloses a traffic key vehicle determination method fusing visual attributes and OBU registration information, and relates to the technical field of traffic supervision, and the method comprises the steps: collecting vehicle visual features through a high-definition checkpoint camera at a traffic node entrance, and generating a first vehicle attribute set; vehicle official registration data is obtained using wireless communication between a roadside unit (RSU) and an on-board unit (OBU) to generate a second set of vehicle attributes. After license plate consistency verification is completed and data alignment is realized, differential credibility empowerment is performed on two types of attribute sets, the high authority of OBU registration information and the characteristic that a visual identification result is influenced by the environment are fully considered, and finally, multi-source data fusion judgment is completed through a preset rule engine, and a vehicle type judgment result is accurately output. And linkage of local storage and platform data is realized.

Description

Technical Field

[0001] This invention relates to the field of traffic monitoring technology, specifically a method for identifying key traffic vehicles that integrates visual attributes and OBU registration information. Background Technology

[0002] In the field of highway traffic supervision, ensuring traffic safety and efficient operation is the core objective. Among them, passenger vehicles and dangerous goods vehicles are key targets of highway traffic safety supervision. Their accurate identification and effective supervision are of vital importance to preventing major traffic accidents and protecting people's lives and property. With the increasing traffic flow and the diversification of vehicle types, traditional manual inspections and single video recognition methods are no longer sufficient to meet the needs of modern traffic supervision. Therefore, how to use advanced technological means to achieve efficient and accurate identification and supervision of passenger vehicles and dangerous goods vehicles has become a key issue that urgently needs to be addressed in the field of highway traffic supervision.

[0003] In existing technologies, license plate and vehicle type recognition using checkpoint cameras is easily affected by natural environmental factors (such as weather and lighting), leading to deviations in the recognition results for license plate color and vehicle type. This results in unstable vehicle attribute recognition results. Furthermore, relying solely on visual features of the vehicle's exterior is insufficient to effectively distinguish similar vehicle types among passenger and hazardous materials transport vehicles, such as tourist charter buses versus regular buses, or hazardous materials transport vehicles versus regular tank trucks. This makes it impossible to implement the identification rules for passenger and hazardous materials transport vehicles using a single visual recognition method. In addition, single video recognition can only collect information related to the vehicle's exterior and cannot assess the vehicle's overall condition. There is a lack of reliable vehicle identity verification mechanisms to effectively verify operating qualifications and dangerous goods transportation qualifications. The ETC on-board unit (OBU) stores official structured registration information such as vehicle type, number of axles, and operating status. This type of highly reliable data is not fully utilized in the existing regulatory system and is not integrated and corrected with visual recognition results, resulting in a waste of data resources. Therefore, the existing key traffic vehicle identification technology relies on a single sensing source, which has the problems of single-dimensional vehicle attribute information, lack of verification of recognition results, and underutilization of data resources, making it difficult to meet the needs of accurate identification and efficient supervision of passenger and dangerous goods vehicles. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method for determining key traffic vehicles by integrating visual attributes and OBU registration information. This method generates a first vehicle attribute set by collecting vehicle visual features from high-definition checkpoint cameras at traffic node entrances, and generates a second vehicle attribute set by acquiring official vehicle registration data through wireless communication between the roadside unit (RSU) and the on-board unit (OBU). After completing license plate consistency verification to achieve data alignment, the two attribute sets are assigned differentiated credibility weights, fully considering the high authority of OBU registration information and the environmental influence on visual recognition results. Finally, a preset rule engine completes the multi-source data fusion determination, accurately outputting the vehicle category determination result, and achieving local storage and platform data linkage.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for determining key traffic vehicles by integrating visual attributes and OBU registration information, the method comprising the following specific steps: S1: Vehicle images are captured by high-definition checkpoint cameras at traffic node entrances, and the visual features of the vehicles are extracted through AI model analysis to generate a first vehicle attribute set containing the vehicle license plate number; S2: The roadside unit (RSU) deployed at the entrance of the traffic node establishes wireless communication with the on-board unit (OBU) of the entering vehicle, reads the standardized official registration data in the OBU, and generates a second vehicle attribute set containing the vehicle's license plate number. S3: Compare the license plate numbers in the first vehicle attribute set and the second vehicle attribute set. If they are inconsistent, mark the vehicle as suspicious and trigger an alarm. If they are consistent, proceed to the credibility weighting step. Assign a fixed high credibility weight to the OBU registration information and assign a dynamic weight to the visual recognition result based on the real-time environmental conditions. S4: Based on the preset rule engine, it integrates the first vehicle attribute set and the second vehicle attribute set after confidence weighting to complete the category determination of passenger and dangerous goods vehicles and generate a determination result with confidence. S5: Store the judgment result in the local database and push it to the traffic node management platform. The relevant information of high-risk vehicles is also reported to the superior traffic supervision platform.

[0006] Furthermore, in S1, the first vehicle attribute set includes vehicle visual attribute information such as license plate color, vehicle type, toll vehicle type, and dangerous goods related markings.

[0007] Furthermore, in step S1, the specific steps for extracting vehicle visual features are as follows: after transmitting the acquired full-domain vehicle image to the edge computing node, the original vehicle image undergoes preprocessing operations of denoising, enhancement, and geometric correction in sequence, using image preprocessing formulas. After optimizing the image pixel features, the region is located using the feature localization formula. Precise localization of vehicle feature regions is performed on the preprocessed image, specifically identifying the license plate region, vehicle outline region, and vehicle marking region. Subsequently, feature extraction and feature matching are performed on each feature region, using a feature matching degree formula. After completing feature matching and determination, the system sequentially performs character recognition of license plate numbers, color gamut feature determination of license plate colors, contour feature matching of vehicle types, appearance feature identification of toll-paying vehicles, and feature detection of hazardous materials-related markings on the vehicle body. The coordinates in the original vehicle image are Pixel value at that location, This represents the coordinates in the preprocessed vehicle image. Pixel value at that location, Represents the pixel gain coefficient. Represents pixel offset; The coordinates in the image are The similarity value between the region and the preset feature template. The coordinates in the image are The actual characteristic value of the region Represents the preset vehicle feature region template feature value; This represents the degree of matching between the extracted vehicle features and the standard features. This represents the actual feature value extracted from the vehicle feature region. These represent the standard feature values ​​corresponding to various visual attributes of a vehicle. The number of feature dimensions representing each visual attribute of a vehicle.

[0008] Furthermore, in S2, the second vehicle attribute set includes vehicle registration attribute information such as vehicle type code, operating attribute identifier, number of vehicle axles, and dangerous goods transportation qualification identifier.

[0009] Furthermore, in S3, after the edge computing node receives the first vehicle attribute set and the second vehicle attribute set, it retrieves the license plate number field from the two attribute sets respectively and performs precise verification by matching characters one by one. If the verification result shows that there is a character difference between the two license plate numbers, the vehicle is immediately marked as a suspicious vehicle and a local alarm mechanism is triggered. If the comparison result is consistent, the vehicle data alignment is completed and the attribute credibility assessment stage is entered. A fixed high credibility weight is assigned to the OBU registration information. Then, combined with the real-time weather, lighting, and actual environmental conditions of the traffic node, such as whether the vehicle is occluded, the corresponding dynamic weight is matched to the visual recognition result, and finally, a vehicle attribute fusion basic dataset with credibility quantification is generated.

[0010] Furthermore, in S3, the specific steps for matching the corresponding dynamic weights to the visual recognition results, based on the real-time environmental conditions of the traffic node, such as weather, lighting, and whether vehicles are obstructing traffic, are as follows: Weather conditions, lighting intensity, and whether vehicles are obstructing traffic are defined as environmental impact dimensions. Multiple environmental impact levels and corresponding single-dimensional weight impact coefficients are preset for each dimension. Real-time environmental monitoring data for each dimension are accurately matched with the preset levels to determine the single-dimensional weight impact coefficients for each dimension. Simultaneously, based on the actual needs of traffic supervision, dimension weight percentages are set for the three core dimensions. A comprehensive environmental impact coefficient is obtained by weighted summation of the dimension weight percentages and the corresponding single-dimensional weight impact coefficients. Finally, the real-time dynamic weight value of the visual recognition result is obtained by multiplying the basic weight value by the comprehensive environmental impact coefficient.

[0011] Furthermore, in S4, the edge computing node retrieves the first and second vehicle attribute sets after credibility weighting, and maps the same-dimensional vehicle attribute information in the two attribute sets to form a unified vehicle attribute fusion analysis dimension. Then, a comprehensive judgment is carried out according to the built-in configurable rule engine. This rule engine is pre-configured with basic judgment rules for passenger vehicles and dangerous goods vehicles, including similar vehicle model distinction rules between tourist charter buses and ordinary buses, and between dangerous goods transport vehicles and ordinary tank trucks. According to the preset weight fusion logic, the OBU registration attribute with a fixed high credibility weight and the visual recognition attribute with a dynamic weight are subjected to single-dimensional weighted fusion calculation. The compliance and attribute matching verification of the core judgment dimensions such as the vehicle's operating qualification, dangerous goods transport qualification, vehicle type, and toll vehicle type are verified one by one. Then, combined with the fusion verification results of each core dimension, the overall confidence of the vehicle category judgment result is calculated to clarify whether the vehicle belongs to the category of passenger vehicles and dangerous goods vehicles and the specific vehicle sub-category.

[0012] Furthermore, in step S4, the OBU registration attribute, which is assigned a fixed high-confidence weight, and the visual recognition attribute, which is assigned a dynamic weight, are subjected to a single-dimensional weighted fusion calculation. The calculation formula is as follows: ,in, It is the first The attribute fusion value of each vehicle's core judgment dimension. It is a fixed high-credibility weight value for OBU registration information; It is the first The OBU registration attribute feature values ​​corresponding to each core vehicle determination dimension. It is the real-time dynamic weight value of the visual recognition result. It is the first The visual recognition attribute feature values ​​corresponding to the core judgment dimensions of each vehicle.

[0013] Furthermore, in S4, the overall confidence level of the vehicle category determination result is calculated by combining the fusion verification results of each core dimension. The calculation formula is as follows: ,in, It is the overall confidence level of the classification results for passenger vehicles and dangerous goods vehicles. It is the first The attribute fusion value of each vehicle's core judgment dimension; It is the first The matching degree of attribute features of each vehicle's core judgment dimension. It is the ordinal identifier of the core judgment dimension of the vehicle; It is the total number of core judgment dimensions for vehicles.

[0014] Compared with existing technologies, this method for determining traffic priority vehicles by integrating visual attributes and OBU registration information has the following advantages: This invention effectively solves the problem of relying on a single sensing source for key traffic vehicle identification by constructing a dual-channel data fusion architecture of checkpoint visual recognition and RSU-OBU communication. It overcomes the shortcomings of single video recognition, such as susceptibility to natural environmental interference, difficulty in distinguishing similar vehicle models, lack of reliable identity verification methods, and ineffective utilization of official OBU registration data. It generates a first vehicle attribute set by collecting vehicle visual features through high-definition checkpoint cameras, and generates a second vehicle attribute set by obtaining official vehicle registration data through communication between the roadside unit and the vehicle unit. It then performs license plate consistency verification and differentiated credibility weighting, and completes multi-source data fusion judgment through a preset rule engine. This greatly improves the dimensionality of vehicle attribute information and enhances the reliability of recognition results. All data processing is completed within the traffic node local area network, realizing local aggregation and real-time fusion of vehicle data, and ensuring the efficiency and security of data processing.

[0015] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0017] Figure 1 A flowchart of a method for determining key traffic vehicles that integrates visual attributes and OBU registration information; Figure 2 A flowchart of step S3 in a method for determining key traffic vehicles that integrates visual attributes and OBU registration information; Figure 3 This is a flowchart of step S4 of a traffic priority vehicle determination method that integrates visual attributes and OBU registration information. Detailed Implementation

[0018] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0019] This invention provides a method for determining key traffic vehicles by integrating visual attributes and OBU registration information. It generates a first vehicle attribute set by collecting vehicle visual features from high-definition checkpoint cameras at traffic node entrances, and generates a second vehicle attribute set by acquiring official vehicle registration data through wireless communication between the roadside unit (RSU) and the on-board unit (OBU). After completing license plate consistency verification to achieve data alignment, the two attribute sets are assigned differentiated credibility weights, fully considering the high authority of OBU registration information and the environmental influence on visual recognition results. Finally, a preset rule engine completes the multi-source data fusion determination, accurately outputting the vehicle category determination result, and achieving local storage and platform data linkage.

[0020] S1: Vehicle images are captured by high-definition checkpoint cameras at traffic node entrances, and the visual features of the vehicles are extracted through AI model analysis to generate a first vehicle attribute set containing the vehicle license plate number; High-definition checkpoint cameras are deployed at the entrance of traffic nodes. When a vehicle enters the camera's acquisition area, the camera is triggered to acquire a full-area image of the vehicle. The acquired image is a high-definition color image that covers the front of the vehicle, the side of the vehicle, and the license plate area, ensuring the integrity of the vehicle's visual features. The acquired full-area image of the vehicle is transmitted in real time to the edge computing node that is equipped with the traffic node. The edge computing node performs preprocessing operations such as denoising, enhancement, and geometric correction on the original vehicle image to eliminate noise, blur, and distortion caused by weather, lighting, and shooting angle.

[0021] Image preprocessing is performed using formulas Complete pixel feature optimization, where: original pixel values Coordinates in the image Pixel grayscale value or RGB color gamut value, pixel gain coefficient The pixel offset is set to 0.8-1.5 based on the actual blur level of the image. The overall image brightness should be set between 0 and 50. If the image has backlighting or low light conditions, increase the brightness appropriately. Value and increase Value; if the image has strong overexposure, reduce it appropriately. Value and decrease This value optimizes image brightness and contrast.

[0022] After image preprocessing, the region feature localization formula is used. Accurately locate vehicle feature regions in the preprocessed image, including: Coordinates in the image The similarity value between the region and the preset vehicle feature template; the smaller the value, the higher the matching degree between the region and the template. Coordinates in the image The actual feature values ​​of the region include the region's contour features, texture features, and color gamut features; The preset vehicle feature region template features the edge calculation nodes pre-entered based on national standard vehicle models, standard license plate sizes, and hazardous materials label styles. This formula is used to pinpoint the license plate region, vehicle exterior outline region, and vehicle marking region in the image, providing precise region ranges for subsequent feature extraction.

[0023] For each locked feature region, feature extraction and feature matching are performed sequentially. Feature matching is performed using the feature matching degree formula. The judgment is complete, including: The value ranges from 0 to 1, representing the matching degree between the extracted vehicle features and the standard features. The closer the value is to 1, the higher the matching degree. These are the actual feature values ​​extracted from the vehicle feature region; Standard feature values ​​corresponding to each visual attribute of the vehicle are pre-entered based on national traffic standards. The feature dimension number of each visual attribute of the vehicle is set according to the actual recognition requirements. For example, the feature dimension number of the license plate number is the number of characters, and the feature dimension number of the license plate color is the RGB three-color dimension.

[0024] Based on the above feature extraction and matching process, the following steps are implemented sequentially: License plate number character recognition: The license plate area is segmented into characters, and the license plate number is accurately recognized by matching the character outline features with standard character templates, including the complete recognition of Chinese characters, letters and numbers; Determining the color gamut characteristics of license plate color: Extract the RGB color gamut values ​​of the license plate area and compare them with the color gamut ranges of standard license plates such as blue, yellow, green, and white plates to determine the license plate color; Vehicle type contour feature matching: Extract the size and shape features of the vehicle's exterior contour area and match them with national standard vehicle model templates such as small passenger cars, large passenger cars, trucks, and tank trucks to determine the vehicle type; Identification of toll-collecting vehicle appearance features: Determine toll-collecting vehicle types by combining appearance features such as vehicle wheelbase, body length, and number of tires, in accordance with the traffic toll vehicle classification standards; Feature detection of vehicle body hazardous materials related markings: Extract texture and pattern features from the marking areas on the vehicle body and match them with standard templates for hazardous materials markings such as explosives, flammable materials, and corrosive substances to detect whether hazardous materials related markings exist on the vehicle body.

[0025] The vehicle visual attribute information obtained above, such as license plate number, license plate color, vehicle type, toll vehicle type, and dangerous goods related markings, are integrated to generate a structured first vehicle attribute set. This attribute set uses the license plate number as a unique identifier to facilitate subsequent matching with the second vehicle attribute set.

[0026] S2: The roadside unit (RSU) deployed at the entrance of the traffic node establishes wireless communication with the on-board unit (OBU) of the entering vehicle, reads the standardized official registration data in the OBU, and generates a second vehicle attribute set containing the vehicle's license plate number. Roadside Units (RSUs) are deployed at traffic node entrances and matched with high-definition checkpoint cameras. When a vehicle enters the RSU wireless communication coverage area, the RSU establishes a dedicated short-range wireless communication with the vehicle's On-Board Unit (OBU). The communication protocol is compatible with mainstream wireless communication standards in the transportation field, enabling the RSU to legally read data from the OBU.

[0027] The RSU reads the standardized official registration data stored in the OBU. This data is the official data entered when the vehicle is registered with the traffic management and transportation management departments, and is authoritative and accurate. After reading, the RSU parses and structures the data to eliminate data format differences and generates a second set of vehicle attributes with the license plate number as the unique identifier.

[0028] The second vehicle attribute set includes vehicle registration attribute information such as license plate number, vehicle type code, operating attribute identifier, number of axles, and dangerous goods transportation qualification identifier. Among them, the vehicle type code corresponds one-to-one with the national standard vehicle model classification; the operating attribute identifier includes types such as tourist passenger transport, road passenger transport, general freight transport, and non-operating transport; the dangerous goods transportation qualification identifier is a binary identifier that directly reflects whether the vehicle has the legal qualification for transporting dangerous chemicals.

[0029] like Figure 2 As shown, S3: Compare the license plate numbers in the first vehicle attribute set and the second vehicle attribute set. If they are inconsistent, the vehicle is marked as suspicious and an alarm is triggered. If they are consistent, the vehicle enters the credibility weighting stage, which assigns a fixed high credibility weight to the OBU registration information and assigns a dynamic weight to the visual recognition result based on the real-time environmental conditions. After receiving the first vehicle attribute set and the second vehicle attribute set, the edge computing node first performs a precise license plate number verification operation: it retrieves the license plate number field from the two attribute sets respectively, performs format unification processing on the license plate number to eliminate differences in character case, full-width and half-width characters, and then performs verification by matching characters one by one to determine whether the two license plate numbers are completely consistent.

[0030] If the verification result shows that there is any difference in any character between the two license plate numbers, the vehicle will be immediately marked as a suspicious vehicle. At the same time, the local alarm mechanism of the traffic node will be triggered. The local alarm will use sound and light alarm to remind the on-site supervisors. At the same time, the license plate number, visual recognition information and OBU registration information of the suspicious vehicle will be pushed to the traffic node management platform to realize the real-time warning of suspicious vehicles.

[0031] If the comparison results show that the two license plate numbers are completely identical, the vehicle data alignment is completed, and the process moves to the attribute credibility assessment and weighting stage. Different credibility weights are assigned to the two types of attribute sets, and finally, a vehicle attribute fusion basic dataset with credibility quantification is generated. The specific weighting process is as follows: Fixed high-reliability weight assignment for OBU registration information: Since the registration data within the OBU is official standardized data with high authority and accuracy, a fixed high-reliability weight is assigned to the OBU registration information (second vehicle attribute set). The weight value is pre-configured according to the actual needs of traffic supervision, with a range of 0.7-0.9. This weight value remains fixed during the judgment process and does not change with the external environment.

[0032] Dynamic weight assignment of visual recognition results: Visual recognition results are easily affected by environmental conditions such as weather, lighting, and vehicle occlusion. Therefore, dynamic weights are assigned to the visual recognition information (first vehicle attribute set) and dynamically calculated based on the real-time environmental conditions of traffic nodes.

[0033] The dynamic weight calculation steps for visual recognition results are as follows: Weather conditions, light intensity, and whether vehicles obstruct the view are identified as the three core environmental impact dimensions affecting the accuracy of visual recognition results. For each dimension, multiple levels of environmental impact are preset, along with corresponding single-dimensional weight impact coefficients. The coefficient values ​​range from 0.5 to 1.0. The closer the coefficient is to 1, the smaller the impact of the environmental conditions on visual recognition, and the higher the credibility of the recognition results; the closer the coefficient is to 0.5, the greater the impact, and the lower the credibility.

[0034] Environmental monitoring equipment is deployed at traffic nodes to collect real-time weather conditions and light intensity data. At the same time, edge computing nodes use image analysis technology to identify whether vehicles are obstructed and the level of obstruction. The real-time monitoring / identification data of each dimension are accurately matched with the preset environmental impact level to determine the single-dimensional weight influence coefficient corresponding to each dimension.

[0035] Based on the actual impact of each environmental dimension on visual recognition, a weight percentage is assigned to each of the three core dimensions, denoted as _____. (weather), (illumination), (Obscured), and satisfies In this embodiment, the following settings are preferred: , , The feature matching is based on the fact that illumination interference is the main influencing factor for visual recognition in actual traffic scenarios.

[0036] The comprehensive environmental impact coefficient is calculated by weighting the dimensional weight percentages and the corresponding single-dimensional weight influence coefficients. The calculation formula is: , The value range is 0.5-1.0.

[0037] Pre-set basic weights for visual recognition results The value ranges from 0.3 to 0.5, combined with the comprehensive environmental impact coefficient. The real-time dynamic weights of the visual recognition results are obtained through product operations. The calculation formula is: And satisfy If the constraint is not satisfied after calculation, for Make fine adjustments to ensure that the sum of the weights is 1.

[0038] After the above weighting is completed, the weighted first vehicle attribute set and the second vehicle attribute set are integrated to generate a basic dataset for vehicle attribute fusion, providing a data foundation for subsequent vehicle category determination.

[0039] like Figure 3 As shown, S4: Based on the preset rule engine, the first vehicle attribute set and the second vehicle attribute set after confidence weighting are integrated to complete the category determination of passenger and dangerous goods vehicles and generate a determination result with confidence. Edge computing nodes retrieve a fusion dataset of vehicle attributes after trust weighting, and rely on a built-in configurable rule engine to determine the categories of passenger vehicles and hazardous material transport vehicles. The rule engine is a modular configuration system that can flexibly modify the determination rules according to adjustments in traffic supervision policies without reconstructing the overall algorithm. It adapts to the key vehicle supervision needs of different regions and stages. The specific determination process is as follows: The visual recognition attributes of the first vehicle attribute set are mapped to the registration attributes of the second vehicle attribute set in the same dimension to form a unified vehicle attribute fusion analysis dimension, eliminating the expression differences between the two types of attribute sets. For example, the visually recognized vehicle type and the vehicle type code registered by the OBU are mapped to the same analysis dimension; the visually recognized dangerous goods related marks and the dangerous goods transportation qualification marks registered by the OBU are mapped to the same analysis dimension, ensuring that attributes in the same dimension can be fused and calculated.

[0040] The rule engine pre-configures basic judgment rules and similar vehicle differentiation rules for passenger buses, highway buses, and hazardous chemical transport vehicles. These passenger buses and hazardous chemical transport vehicles include tourist buses, highway buses, and hazardous chemical transport vehicles. The similar vehicle differentiation rules are used to avoid misclassifying ordinary vehicles as key vehicles. The specific rules are as follows: Tourist buses: The OBU is registered with the operation attribute of tourist passenger transport and the visual recognition vehicle type is large passenger bus and the toll vehicle type is large passenger bus; Highway passenger vehicles: The OBU registration and operation attribute is identified as road passenger transport, and the visual recognition vehicle type is large passenger vehicle, and the toll vehicle type is large passenger vehicle; Hazardous materials transport vehicles: The OBU registration for dangerous goods transport qualification marks is present, the visual identification vehicle type is tank vehicle / box vehicle, and the number of axles matches the registered number of axles; Distinguishing similar vehicle types: Ordinary buses (non-commercial) must be marked as non-commercial and are excluded from the category of passenger vehicles; ordinary tank trucks must be marked as having no dangerous goods transport qualification and are excluded from the category of hazardous chemical transport vehicles.

[0041] According to the preset weighted fusion logic, a one-dimensional weighted fusion calculation is performed on the OBU registration attributes, which are assigned a fixed high-confidence weight, and the visual recognition attributes, which are assigned a dynamic weight. The calculation is performed using the formula... Implementation, including: It is the first The attribute fusion value of each vehicle's core judgment dimension. It is a fixed high-credibility weight value for OBU registration information; It is the first The OBU registration attribute feature values ​​corresponding to each core vehicle determination dimension. It is the real-time dynamic weight value of the visual recognition result. It is the first The visual recognition attribute feature values ​​corresponding to the core judgment dimensions of each vehicle.

[0042] The core dimensions for vehicle assessment include operating qualifications, dangerous goods transportation qualifications, vehicle type, toll-charging vehicle type, and number of axles. A weighted fusion calculation is performed on each core dimension sequentially to obtain the attribute fusion value for each dimension. - .

[0043] Combining the fusion verification results of each core dimension, through the formula Calculate the overall confidence level of the vehicle category determination results, where: This represents the overall confidence level of the classification results for passenger vehicles and hazardous material transport vehicles. The value ranges from 0 to 1, with a value closer to 1 indicating a more reliable classification result. It is the first The attribute fusion value of each vehicle's core judgment dimension; It is the first The matching degree of attribute features for each core vehicle judgment dimension, with a value ranging from 0 to 1, is set according to the importance of that dimension in the judgment of key vehicles, such as dangerous goods transportation qualifications. We assign a weight of 1.0 to the highest-weighted vehicle type. Take 0.8, It is the ordinal identifier of the core judgment dimension of the vehicle; It is the total number of core judgment dimensions for vehicles.

[0044] A confidence threshold of 0.8 is pre-set in the rule engine (adjustable according to regulatory accuracy requirements). If the calculated overall confidence C ≥ 0.8, the judgment result is valid. Based on the fusion analysis results, it is determined whether the vehicle belongs to the category of passenger vehicles and dangerous goods vehicles. If it does, the specific vehicle sub-category is further determined. If C < 0.8, the judgment result is a suspected key vehicle, which requires subsequent manual review. Finally, a vehicle category judgment result with confidence is generated, which includes information such as license plate number, vehicle category, judgment confidence, fusion value of each core dimension, and real-time environmental conditions.

[0045] S5: Store the judgment result in the local database and push it to the traffic node management platform. The relevant information of high-risk vehicles is also reported to the superior traffic supervision platform. The generated judgment results with confidence levels are stored in the local database of traffic nodes in a structured data format. The local database adopts a distributed storage method to ensure the security and traceability of the data. The stored information is retained for at least 90 days to facilitate subsequent verification and tracing by traffic regulatory authorities.

[0046] Meanwhile, the judgment results are pushed to the traffic node management platform in real time via encrypted network transmission. The platform displays the judgment results visually, including real-time traffic information, confidence level, and vehicle attributes of key vehicles, which facilitates real-time monitoring and dispatching by on-site supervisors.

[0047] For high-risk vehicles, their relevant information is simultaneously reported to the superior traffic supervision platform. High-risk vehicles include: key vehicles identified as passenger vehicles or dangerous goods vehicles, suspicious vehicles with inconsistent license plate numbers, vehicles with insufficient confidence in the assessment but suspected to be key vehicles, and vehicles without dangerous goods transportation qualifications but with dangerous goods markings on the vehicle body. The reported information includes the complete set of vehicle attributes, assessment results, alarm type, collection time, traffic node location, etc., to achieve multi-level supervision of key traffic vehicles. The superior platform can coordinate and manage the traffic information of key vehicles at all traffic nodes within its jurisdiction and provide risk warnings.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for determining key traffic vehicles by integrating visual attributes and OBU registration information, characterized in that, The method includes the following specific steps: S1: Vehicle images are captured by high-definition checkpoint cameras at traffic node entrances, and the visual features of the vehicles are extracted through AI model analysis to generate a first vehicle attribute set containing the vehicle license plate number; S2: The roadside unit (RSU) deployed at the entrance of the traffic node establishes wireless communication with the on-board unit (OBU) of the entering vehicle, reads the standardized official registration data in the OBU, and generates a second vehicle attribute set containing the vehicle's license plate number. S3: Compare the license plate numbers in the first vehicle attribute set and the second vehicle attribute set. If they are inconsistent, mark the vehicle as suspicious and trigger an alarm. If they are consistent, proceed to the credibility weighting step. Assign a fixed high credibility weight to the OBU registration information and assign a dynamic weight to the visual recognition result based on the real-time environmental conditions. S4: Based on the preset rule engine, it integrates the first vehicle attribute set and the second vehicle attribute set after confidence weighting to complete the category determination of passenger and dangerous goods vehicles and generate a determination result with confidence. S5: Store the judgment result in the local database and push it to the traffic node management platform. The relevant information of high-risk vehicles is also reported to the superior traffic supervision platform.

2. The method for determining key traffic vehicles by integrating visual attributes and OBU registration information according to claim 1, characterized in that, In S1, the first vehicle attribute set includes vehicle visual attribute information such as license plate color, vehicle type, toll vehicle type, and dangerous goods related markings.

3. The method for determining key traffic vehicles by integrating visual attributes and OBU registration information according to claim 1, characterized in that, In step S1, the specific steps for extracting vehicle visual features are as follows: after transmitting the acquired full-domain vehicle image to the edge computing node, the original vehicle image is sequentially subjected to preprocessing operations of denoising, enhancement, and geometric correction, using image preprocessing formulas. After optimizing the image pixel features, the region is located using the feature localization formula. Precise localization of vehicle feature regions is performed on the preprocessed image, specifically identifying the license plate region, vehicle outline region, and vehicle marking region. Subsequently, feature extraction and feature matching are performed on each feature region, using a feature matching degree formula. After completing feature matching and determination, the system sequentially performs character recognition of license plate numbers, color gamut feature determination of license plate colors, contour feature matching of vehicle types, appearance feature identification of toll-paying vehicles, and feature detection of hazardous materials-related markings on the vehicle body. The coordinates in the original vehicle image are Pixel value at that location, This represents the coordinates in the preprocessed vehicle image. Pixel value at that location, Represents the pixel gain coefficient. Represents pixel offset; The coordinates in the image are The similarity value between the region and the preset feature template. The coordinates in the image are The actual characteristic value of the region Represents the preset vehicle feature region template feature value; This represents the degree of matching between the extracted vehicle features and the standard features. This represents the actual feature value extracted from the vehicle feature region. These represent the standard feature values ​​corresponding to various visual attributes of a vehicle. The number of feature dimensions representing each visual attribute of a vehicle.

4. The method for determining key traffic vehicles by integrating visual attributes and OBU registration information according to claim 1, characterized in that, In S2, the second vehicle attribute set includes vehicle registration attribute information such as vehicle type code, operating attribute identifier, number of axles, and dangerous goods transportation qualification identifier.

5. The method for determining key traffic vehicles by integrating visual attributes and OBU registration information according to claim 1, characterized in that, In step S3, after receiving the first vehicle attribute set and the second vehicle attribute set, the edge computing node retrieves the license plate number field from the two attribute sets respectively and performs precise verification by matching characters one by one. If the verification result shows that there is a character difference between the two license plate numbers, the vehicle is immediately marked as a suspicious vehicle and a local alarm mechanism is triggered. If the comparison result is consistent, the vehicle data alignment is completed and the attribute credibility assessment stage is entered. A fixed high credibility weight is assigned to the OBU registration information. Then, combined with the actual environmental conditions of the traffic node, such as the weather, lighting, and whether the vehicle is occluded, the corresponding dynamic weight is matched to the visual recognition result, and finally, a vehicle attribute fusion basic dataset with credibility quantification is generated.

6. The method for determining key traffic vehicles by integrating visual attributes and OBU registration information according to claim 1, characterized in that, In step S3, the specific steps for matching dynamic weights to the visual recognition results based on real-time environmental conditions such as weather, lighting, and whether vehicles are obstructing traffic nodes are as follows: Weather conditions, lighting intensity, and whether vehicles are obstructing traffic are defined as environmental impact dimensions. Multiple environmental impact levels and corresponding single-dimensional weight impact coefficients are preset for each dimension. Real-time environmental monitoring data for each dimension are accurately matched with the preset levels to determine the single-dimensional weight impact coefficients for each dimension. Simultaneously, based on actual traffic supervision needs, dimension weight percentages are set for the three core dimensions. A comprehensive environmental impact coefficient is obtained by weighted summation of the dimension weight percentages and the corresponding single-dimensional weight impact coefficients. Finally, the real-time dynamic weight value of the visual recognition result is obtained by multiplying the basic weight value by the comprehensive environmental impact coefficient.

7. The method for determining key traffic vehicles by integrating visual attributes and OBU registration information according to claim 1, characterized in that, In step S4, the edge computing node retrieves the first and second vehicle attribute sets after credibility weighting, and maps the same-dimensional vehicle attribute information in the two attribute sets to form a unified vehicle attribute fusion analysis dimension. Then, it performs a comprehensive judgment based on the built-in configurable rule engine. This rule engine is pre-configured with basic judgment rules for passenger vehicles and dangerous goods vehicles, including similar vehicle model distinction rules between tourist charter buses and ordinary buses, and between dangerous goods transport vehicles and ordinary tank trucks. According to the preset weight fusion logic, the OBU registration attribute with a fixed high credibility weight and the visual recognition attribute with a dynamic weight are weighted and fused in a single dimension. The compliance and attribute matching verification of the core judgment dimensions of vehicle operation qualification, dangerous goods transport qualification, vehicle type, and toll vehicle type are verified one by one. Then, combined with the fusion verification results of each core dimension, the overall confidence of the vehicle category judgment result is calculated to clarify whether the vehicle belongs to the category of passenger vehicles and dangerous goods vehicles and the specific vehicle sub-category.

8. The method for determining key traffic vehicles by integrating visual attributes and OBU registration information according to claim 7, characterized in that, In step S4, the OBU registration attribute, which is assigned a fixed high-confidence weight, and the visual recognition attribute, which is assigned a dynamic weight, are subjected to a single-dimensional weighted fusion calculation. The calculation formula is as follows: ,in, It is the first The attribute fusion value of each vehicle's core judgment dimension. It is a fixed high-credibility weight value for OBU registration information; It is the first The OBU registration attribute feature values ​​corresponding to each core vehicle determination dimension. It is the real-time dynamic weight value of the visual recognition result. It is the first The visual recognition attribute feature values ​​corresponding to the core judgment dimensions of each vehicle.

9. The method for determining key traffic vehicles by integrating visual attributes and OBU registration information according to claim 7, characterized in that, In step S4, the overall confidence level of the vehicle category determination result is calculated by combining the fusion verification results of each core dimension. The calculation formula is as follows: ,in, It is the overall confidence level of the classification results for passenger vehicles and dangerous goods vehicles. It is the first The attribute fusion value of each vehicle's core judgment dimension; It is the first The matching degree of attribute features of each vehicle's core judgment dimension. It is the ordinal identifier of the core judgment dimension of the vehicle; It is the total number of core judgment dimensions for vehicles.

Images