Vehicle checking method under traffic checkpoint distribution and electronic equipment

Abstract

The invention discloses a vehicle troubleshooting method under traffic gate distribution and electronic equipment, and relates to the technical field of intelligent traffic management, and the method comprises the steps: constructing a multi-gate collaborative network, collecting vehicle basic data, and constructing a cross-scene global data pool; performing stage processing on the cross-scene global data pool to determine a target vehicle; splicing the vehicle space-time trajectory to generate a cross-scene complete driving map; analyzing and acquiring a cross-scene violation behavior recognition result according to the cross-scene linkage strategy; and triggering a cross-scene early warning instruction and predicting a potential pass gate to carry out cross-scene early warning. The technical problems of single-gate data isolation, insufficient multi-gate collaboration and weak strategy linkage capability of a traditional traffic gate are solved, and the technical effects of data intercommunication and strategy linkage among a plurality of traffic gates, formation of a closed-loop troubleshooting system, and improvement of the accuracy and efficiency of vehicle management and control under a complex traffic network are achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic management technology, and in particular to a vehicle screening method and electronic equipment under traffic checkpoint distribution. Background Technology

[0002] In the field of intelligent traffic management, traffic checkpoints are crucial for vehicle screening and control. Current technologies only allow a single checkpoint to collect vehicle data from a localized area. Data from checkpoints in different scenarios is stored independently, lacking real-time communication and spatiotemporal calibration mechanisms. This results in weak policy linkage capabilities, with each checkpoint's policies executing independently, making it impossible to achieve dynamic early warnings based on cross-scenario data correlation. Consequently, the efficiency of controlling abnormal vehicles in complex traffic networks is low.

[0003] Traditional vehicle screening methods suffer from data isolation and insufficient collaboration, making it impossible to form a global trajectory across scenarios. This makes it difficult to identify vehicles involved in cross-regional violations or cases, and it is also impossible to complete early warning deployment before the target vehicle arrives at the next checkpoint. As a result, screening efficiency is low, early warning is delayed, and it is difficult to meet the needs of precise vehicle control in complex traffic networks. Summary of the Invention

[0004] This application provides a vehicle screening method and electronic device under traffic checkpoint distribution, which is used to solve the technical problems of isolated data from a single traffic checkpoint, insufficient coordination among multiple checkpoints, and weak strategy linkage capabilities in traditional traffic checkpoints.

[0005] The first aspect of this application provides a vehicle screening method under traffic checkpoint distribution. The method includes: constructing a multi-checkpoint collaborative network; traversing any single checkpoint in the multi-checkpoint collaborative network to collect basic vehicle data in real time and constructing a cross-scenario full-domain data pool; performing hierarchical processing on the cross-scenario full-domain data pool; performing local strategy processing on the basic vehicle data of each vehicle in the cross-scenario full-domain data pool to determine the target vehicle; combining the target vehicle with the cross-scenario full-domain data pool to perform spatiotemporal trajectory splicing and feature association to generate a complete cross-scenario driving map; according to a preset cross-scenario linkage strategy, performing multi-dimensional analysis on the cross-scenario full-domain data pool based on the complete cross-scenario driving map to obtain cross-scenario violation behavior identification results; triggering a cross-scenario early warning command based on the cross-scenario violation behavior identification results; predicting the potential passage checkpoints for the target vehicle based on the cross-scenario early warning command and the complete cross-scenario driving map; and issuing a cross-scenario early warning to the target vehicle through a multi-checkpoint strategy linkage module.

[0006] A second aspect of this application provides an electronic device comprising: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement a vehicle screening method under traffic checkpoint distribution.

[0007] This application proposes one or more technical solutions, which have at least the following technical effects:

[0008] This application constructs a multi-checkpoint collaborative network, collects basic vehicle data, and builds a cross-scenario, full-domain data pool. Through hierarchical processing, spatiotemporal trajectory stitching, and feature association, a complete driving map is generated. Combined with a preset cross-scenario linkage strategy, multi-dimensional analysis is performed to trigger cross-scenario early warning commands and predict potential checkpoints for target vehicles. This accurately identifies vehicles involved in cases, with abnormal trajectories, or violating regulations across scenarios, making vehicle screening and control in complex traffic networks more precise and efficient. It achieves data interoperability and strategy linkage among multiple traffic checkpoints, forming a closed-loop screening system and improving the accuracy and efficiency of vehicle control in complex traffic networks. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating a vehicle screening method under a traffic checkpoint distribution system provided in an embodiment of this application.

[0011] Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0012] Explanation of reference numerals in the attached drawings: Input device 301, processor 302, memory 303, output device 304. Detailed Implementation

[0013] This application provides a vehicle screening method and electronic device under traffic checkpoint distribution, which is used to solve the technical problems of isolated data from a single traffic checkpoint, insufficient coordination among multiple checkpoints, and weak strategy linkage capabilities in traditional traffic checkpoints.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0015] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0016] Example 1, as Figure 1 As shown, a vehicle screening method under traffic checkpoint distribution is described, wherein the method includes:

[0017] Step A100: Construct a multi-checkpoint collaborative network, traverse any single checkpoint in the multi-checkpoint collaborative network, collect basic vehicle data in real time, and construct a cross-scenario global data pool.

[0018] In this embodiment, the multi-checkpoint collaborative network is used to achieve real-time data exchange and spatiotemporal calibration between checkpoints in different scenarios.

[0019] Specifically, a multi-checkpoint collaborative network is constructed, video streams from a single checkpoint are collected and parsed to generate basic vehicle data, and after transmission, spatiotemporal calibration and fusion, a cross-scenario full-domain data pool is constructed. The specific steps are explained in detail in A110-A160.

[0020] Step A200: Perform hierarchical processing on the cross-scenario global data pool, execute local strategy processing on the basic data of each vehicle in the cross-scenario global data pool, and determine the target vehicle.

[0021] Optionally, the basic vehicle data in the global data pool can be divided into priority groups and queues can be built. After processing by the local policy engine, the target vehicles can be selected. The specific steps are explained in detail in A210-A230.

[0022] Step A300: Combining the target vehicle, perform spatiotemporal trajectory stitching and feature association based on the cross-scene full-domain data pool to generate a complete cross-scene driving map.

[0023] In this embodiment of the application, the cross-scenario complete driving map is used to comprehensively present the complete driving path and behavioral characteristics of the target vehicle in different scenarios such as urban main roads, highway intersections, and toll stations.

[0024] In one embodiment of this application, target vehicle-related information is extracted from the global data pool, fused with the local policy processing results, and then input into the map generation module to construct a complete driving map across scenarios. The specific steps are described in detail in A310-A330.

[0025] Step A400: Based on the preset cross-scenario linkage strategy, perform multi-dimensional analysis on the cross-scenario full-domain data pool based on the cross-scenario complete driving map to obtain the cross-scenario violation behavior recognition results.

[0026] Specifically, based on the preset linkage strategy, the target vehicle's traffic behavior characteristics are extracted from the full-domain data pool in combination with the driving map. After indicator modeling and risk assessment, cross-scenario violations are identified. The specific steps are explained in detail in A410-A430.

[0027] Step A500: Based on the cross-scenario violation recognition results, trigger a cross-scenario warning command, and predict the potential access checkpoints for the target vehicle based on the cross-scenario complete driving map. Then, issue a cross-scenario warning to the target vehicle through the multi-checkpoint strategy linkage module.

[0028] In this embodiment, the multi-checkpoint strategy linkage module is the core module used to achieve cross-scenario early warning and response coordination.

[0029] Optionally, a warning instruction can be triggered based on the violation identification result, and the potential passage checkpoints of the target vehicle can be predicted by combining the driving map and the warning priority can be determined. A linkage strategy can be issued to complete the cross-checkpoint warning. The specific steps are explained in detail in A510-A540.

[0030] Furthermore, step A100 in the method provided in this application embodiment includes:

[0031] A110: Deploy distributed edge nodes at multiple traffic checkpoints to construct the multi-checkpoint collaborative network.

[0032] A120: Traverse any single checkpoint in the multi-checkpoint collaborative network, call the edge node of the corresponding checkpoint, and obtain the video stream of the checkpoint.

[0033] A130: Perform real-time analysis on the video stream, extract dynamic information of the vehicle as it passes through the checkpoint, and generate the vehicle's basic data.

[0034] A140: The vehicle basic data is transmitted to the central node through the edge node according to a preset synchronous communication protocol.

[0035] A150: The central node performs unified timestamp calibration and spatial location mapping processing on the vehicle basic data to generate standardized multi-source data.

[0036] A160: The standardized multi-source data is fused to construct the cross-scenario global data pool.

[0037] In this embodiment, the distributed edge nodes are front-end data acquisition and transmission units deployed at multiple traffic checkpoints such as urban main roads, highway intersections, and toll stations, serving as the foundational nodes for constructing a multi-checkpoint collaborative network. The central node is the upper-level computing and control node deployed within the traffic checkpoint collaborative network, used to receive basic vehicle data uploaded by multiple distributed edge nodes.

[0038] Specifically, distributed edge nodes are first deployed at multiple traffic checkpoints, such as main urban roads and highway intersections, to build a multi-checkpoint collaborative network and form a physical infrastructure for data interoperability. As front-end data acquisition units, the edge nodes can connect to the monitoring equipment at the checkpoints in real time. When a vehicle passes through any checkpoint, the network traverses each checkpoint and calls the corresponding edge node to obtain the checkpoint's video stream, providing raw information input for subsequent data analysis.

[0039] The acquired video stream is analyzed in real time, and computer vision technology is used to extract dynamic information of vehicles passing through checkpoints, generating basic vehicle data including license plate number, vehicle image, passage time, direction of passage, speed, vehicle size, color information, and spatial location information. For example, license plate numbers are extracted from video frames using license plate recognition algorithms, and vehicle speed is calculated using frame difference or optical flow methods. This multi-dimensional data provides fundamental support for vehicle identification and behavior analysis.

[0040] The generated basic vehicle data is transmitted to the central node via edge nodes according to a pre-defined synchronous communication protocol. The central node, acting as the upper-level computing and control node in the traffic checkpoint collaborative network, is responsible for receiving the basic vehicle data uploaded from multiple distributed edge nodes. This communication protocol ensures the real-time performance and accuracy of data transmission; for example, the use of the MQTT protocol enables low-latency data push, avoiding the data lag problems caused by traditional asynchronous transmission.

[0041] The central node performs unified timestamp calibration and spatial location mapping on the received vehicle basic data. Through a global clock synchronization mechanism, the time deviation between different checkpoints is controlled within seconds. Simultaneously, based on a GIS geographic information system, the spatial locations of each checkpoint are mapped to a unified coordinate system, ensuring spatiotemporal consistency and generating standardized multi-source data. For example, after calibration, the time error between a city checkpoint and a highway checkpoint is no more than 1 second, and the spatial location accuracy reaches the meter level, laying the foundation for cross-scenario data fusion.

[0042] Finally, the standardized multi-source data is merged, and through data cleaning, deduplication, and correlation integration, a cross-scenario, full-domain data pool is constructed. This data pool integrates vehicle traffic data from different scenarios such as urban roads, highways, and toll stations, forming a vehicle behavior database covering the entire domain. For example, it can store in real time the complete passage record and corresponding data of a vehicle from city checkpoint A to highway checkpoint B and then toll station C.

[0043] By deploying distributed edge nodes to build a collaborative network, parsing video streams in real time to extract multi-dimensional data, transmitting data based on synchronization protocols, performing spatiotemporal calibration of central nodes, and fusing multi-source data, real-time interoperability and spatiotemporal unification of data from multiple checkpoints were achieved. A cross-scenario, full-domain data pool was constructed, providing comprehensive and accurate data support for subsequent vehicle trajectory stitching, violation identification, and other tasks, effectively solving the problems of data isolation and insufficient collaboration in traditional technologies.

[0044] Furthermore, step A200 in the method provided in this application embodiment includes:

[0045] A210: Prioritize the vehicle basic data in the cross-scenario global data pool and construct a hierarchical processing queue.

[0046] A220: Input the basic vehicle data in the hierarchical processing queue one by one into the local policy engine, perform rapid deployment processing or passage verification processing, and generate local policy processing results.

[0047] A230: Target vehicles are selected based on the results of the local policy processing.

[0048] In this embodiment, the local policy engine is a core functional module deployed at the local end of the traffic checkpoint for performing fast policy processing of vehicle basic data.

[0049] Optionally, firstly, the basic vehicle data in the cross-scenario global data pool is prioritized and a three-dimensional dynamic weighting mechanism is adopted:

[0050] Traffic frequency dimension: Based on time window, vehicles that pass through ≥3 times in 24 hours (such as logistics delivery vehicles) and vehicles that cross regions ≥2 times in 72 hours are judged as high frequency and high priority; ordinary vehicles that pass through ≤2 times in 1 week are classified as low frequency and low priority.

[0051] Vehicle type dimension: Dangerous goods transport vehicles (including flammable and explosive goods transport vehicles), vehicles under investigation (vehicles on the blacklist of the public security system), and emergency rescue vehicles are automatically classified as high priority; ordinary family cars and commercial buses are classified as medium priority; and non-commercial light vehicles are classified as low priority.

[0052] Traffic time dimension: Trucks traveling in the urban core area during restricted hours (such as 7-9 am and 5-7 pm) and dangerous goods vehicles traveling on highways during nighttime restricted hours (22:00-6:00 am) are classified as high priority; vehicles traveling during regular hours are given secondary priority based on type and frequency.

[0053] This mechanism automatically marks hazardous materials transport vehicles that pass through restricted urban areas three times within a 24-hour period as having the highest priority, while ordinary commuter vehicles that pass through once outside of restricted hours are classified as having a low priority, forming a dynamically adjustable tiered processing queue.

[0054] Based on the above priority classification, a hierarchical processing queue is constructed, allowing data to enter the processing flow in priority order. This queue employs a dynamic scheduling mechanism, prioritizing computing resources for high-priority data to ensure timely processing of data related to emergency events. For example, when high-priority data on deployed vehicles enters the queue, the processing order is automatically adjusted, allowing it to skip the waiting queue of some low-priority data and directly enter the local policy engine for processing, reducing response latency.

[0055] The vehicle basic data in the hierarchical processing queue is input into the local policy engine one by one for rapid deployment or passage verification. The real-time comparison algorithm built into the local policy engine uses hash indexing and feature vector matching technology to perform millisecond-level retrieval on the input vehicle basic data: During deployment, the hash table is used to quickly compare with the deployment blacklist database, storing feature data such as vehicles involved in cases and vehicles that have fled the scene. At the same time, the violation record database is called, and regular expressions are used to match unprocessed serious violation records, such as speeding more than 50% and running red lights. During passage verification, the rule base stores traffic restriction rules (such as date-area mapping table for tail number restrictions) and highway passage permit rules (such as whitelist of dangerous goods vehicle passage time periods) according to scenario classification. The algorithm uses a time-space dual indexing mechanism to retrieve vehicle passage time and checkpoint location data in real time for rule matching. For example, when verifying whether a vehicle complies with the rule that yellow-plate trucks are prohibited from passing through a certain highway from 7 to 9 am, the passage time period is first determined by the timestamp, and then the area to which the checkpoint belongs is verified based on the GIS coordinates. The local policy processing result is generated by combining the matching results.

[0056] Based on the processing results of the local policy engine, target vehicles that meet the control conditions or exhibit abnormal traffic are selected. For example, vehicles that match the blacklist in the control processing results, or vehicles that have repeatedly violated traffic restrictions during the passage verification process, will be marked as target vehicles and enter the subsequent cross-scenario trajectory analysis process.

[0057] By prioritizing based on factors such as traffic frequency and vehicle type, a dynamic hierarchical processing queue is constructed. Combined with the rapid deployment and passage verification mechanism of the local policy engine, efficient screening of basic vehicle data in the cross-scenario full-domain data pool is achieved. The local identification time of target vehicles is shortened from minutes to seconds in traditional technology, significantly improving the real-time detection capability of abnormal vehicles and the response efficiency of local policies. This provides a precise target screening foundation for subsequent cross-scenario trajectory stitching and violation identification.

[0058] Furthermore, step A300 in the method provided in this application embodiment includes:

[0059] A310: Extract historical trajectories, frequently accessed nodes, and cross-regional travel patterns associated with the target vehicle from the cross-scenario global data pool to obtain the global data extraction results.

[0060] A320: The local policy processing result is fused with the global data extraction result to generate a fused dataset.

[0061] A330: Input the fused dataset into the map generation module. The map generation module performs vehicle spatiotemporal trajectory stitching and feature association based on the fused dataset to construct a complete cross-scene driving map of the target vehicle.

[0062] In this embodiment of the application, the map generation module is a core functional module used to construct a complete driving map of the target vehicle across scenarios.

[0063] Specifically, firstly, historical trajectories, frequently accessed nodes, and cross-regional travel patterns associated with the target vehicle are extracted from a cross-scenario, full-domain data pool. For example, for a monitored suspicious vehicle, the data pool can be used to obtain its passage records at 20 city checkpoints and 5 highway toll stations within the past 7 days, as well as information on high-frequency stopping points in areas such as logistics parks, forming a full-domain data extraction result that provides multi-scenario data support for trajectory stitching.

[0064] The results of local policy processing are merged with the results of global data extraction to generate a fused dataset. The local policy processing results include information such as the target vehicle's deployment attributes (e.g., whether it is a vehicle involved in a case) and abnormal traffic records (e.g., the time of speeding capture). Taking a hazardous materials transport vehicle as an example, the fused dataset integrates its speeding records at highway checkpoints (local policy results) with its cross-provincial travel trajectories from the global data, forming a comprehensive data set that includes both violations and cross-regional routes. This expands the data dimension from local information of a single scenario to three-dimensional information related to multiple scenarios.

[0065] The fused dataset is input into the map generation module. The map generation module uses the fused dataset as the stitching anchor point to perform preliminary trajectory stitching on the multi-checkpoint passage records. The graph structure matching method is used to fuse the passage paths to construct a preliminary driving trajectory. The accuracy of the trajectory is corrected and the confidence is enhanced by comparing vehicle color, vehicle size and similarity of adjacent checkpoint images. Finally, a complete cross-scene driving map of the target vehicle is generated. The specific steps are explained in detail in A331-A334.

[0066] By extracting historical trajectories and traffic patterns from a global data pool across multiple scenarios and integrating local strategies to generate a comprehensive dataset, and utilizing a graph generation module to stitch together spatiotemporal trajectories and associate cross-scenario features, the fragmented single-checkpoint data in traditional technologies is transformed into a complete driving graph covering urban, highway, and other scenarios. This improves the accuracy of vehicle trajectory reconstruction across scenarios and provides a three-dimensional and accurate trajectory data foundation for subsequent cross-scenario violation identification, effectively solving the problem of blind spots in control caused by trajectory discontinuity in traditional technologies.

[0067] Furthermore, step A330 in the method provided in this application embodiment includes:

[0068] A331: Using the fused dataset as the stitching anchor point, perform preliminary trajectory stitching on the passage records of multiple checkpoints.

[0069] A332: Based on the splicing anchor points, a graph structure matching method is used to merge the passage paths between multiple checkpoints and construct a preliminary driving trajectory.

[0070] A333: Perform feature association processing on the constructed preliminary driving trajectory, wherein the feature association processing includes comparing the image similarity of vehicle color, vehicle size and adjacent checkpoints, correcting the accuracy and enhancing the confidence of the preliminary driving trajectory, and obtaining the feature association result.

[0071] A334: Based on the preliminary driving trajectory and the feature association results, generate a complete cross-scene driving map of the target vehicle.

[0072] Specifically, firstly, the fused dataset is used as the stitching anchor. This dataset integrates the target vehicle's local policy processing results (such as deployment attributes and violation records) with historical trajectories and passage nodes from the global data pool. Taking a vehicle involved in a case as an example, the fused dataset contains its speeding record at city checkpoint A (timestamp: 2025-06-10-10:00:15, spatial coordinates X1, Y1) and its passage record at highway checkpoint B (timestamp: 2025-06-10-11:30:20, spatial coordinates X2, Y2). Using this as the anchor, the passage records from multiple checkpoints are initially stitched together to form an initial path framework based on spatiotemporal anchors.

[0073] Based on splicing anchor points, a graph structure matching algorithm combining trajectory overlap and traffic patterns is used for path fusion. The algorithm first abstracts the passage records of each checkpoint into nodes in a graph structure. Node attributes integrate multi-dimensional information such as timestamps, spatial locations, vehicle colors, and vehicle dimensions. Then, the edge weights are defined by both trajectory overlap and traffic patterns. Trajectory overlap is calculated by comparing the time difference between adjacent checkpoints with the standard travel time. For example, a deviation rate of 1.6 hours between the time difference of a target vehicle at checkpoints A and B and the normal travel time of 1.5 hours for similar vehicles corresponds to a weight of 0.85. Traffic patterns are assigned weight coefficients based on scenario characteristics such as the 90% proportion of truck traffic on this road segment in historical data. Finally, the edge weights are obtained through weighted summation, such as 0.85 × 0.5 + 0.9 × 0.5 = 0.875. This is used to fuse multi-checkpoint passage paths, constructing a preliminary driving trajectory network that conforms to spatiotemporal logic from isolated nodes.

[0074] Next, when performing feature association processing on the initial driving trajectory, computer vision technology is first used to extract features from vehicle images at adjacent checkpoints: the HOG (Histogram of Oriented Gradients) algorithm is used to calculate the gradient direction distribution of local image regions to capture the contour and shape features of the vehicles. Simultaneously, deep learning models (such as CNNs) are used to extract semantic features, accurately identifying high-level semantic information such as vehicle color and vehicle size. Taking checkpoints C and D as examples, the feature vector of a white SUV is extracted from the image of checkpoint C. After extracting the corresponding feature vector from the image of checkpoint D, a cosine similarity algorithm is used for comparison: the cosine value of the color feature vector reaches 0.92 (92% similarity), and the cosine value of the vehicle size feature vector reaches 0.95 (95% similarity). Then, the feature matching degree of the overall image is calculated using an image semantic segmentation model. When the overall matching degree exceeds 90%, it is determined to be the same vehicle. Subsequently, the preliminary trajectory is corrected based on the matching result: if the difference between the theoretical travel time and the actual passage time at the two checkpoints exceeds a preset threshold (such as 15 minutes), the timestamp deviation is adjusted, and the travel route is remapped in conjunction with the GIS map. Finally, the trajectory confidence is increased to over 90% by updating the model with Bayesian confidence, forming a feature association result, which provides accurate trajectory data support for the subsequent generation of a complete driving map.

[0075] Finally, based on the preliminary driving trajectory and feature association results, a complete cross-scenario driving map of the target vehicle is generated. First, the preliminary driving trajectory is used as the basic framework, which has been fused with multiple checkpoint paths using graph structure matching methods to form a preliminary path network across scenarios. Then, the correction information obtained from feature association processing (such as feature matching results for vehicle color, vehicle size, etc., and trajectory confidence improvement data) is embedded into the spatiotemporal nodes of the preliminary trajectory to finely calibrate the trajectory's timestamp deviation and spatial location error. Next, a GIS geographic information system is used to map the calibrated trajectory to a unified spatial coordinate system, integrating checkpoint data from different scenarios such as urban main roads, highway entrances, and toll stations to form a continuous spatiotemporal trajectory chain. Finally, through the map visualization module, the complete driving path of the vehicle across multiple scenarios is dynamically presented in time axis order and spatial location coordinates, while simultaneously annotating the confidence parameters after feature association. For example, trajectory segments with a confidence level of over 90% are highlighted. Ultimately, a complete cross-scenario driving map containing the vehicle's full-scenario traffic records, behavioral characteristics, and risk levels is generated, providing intuitive and accurate trajectory data support for cross-scenario violation identification.

[0076] By using the fused dataset as an anchor point for initial trajectory stitching, path fusion based on graph structure matching algorithm, and multi-feature association verification combining vehicle color, vehicle type and image similarity, the generated cross-scene complete driving map can accurately restore the real driving path of vehicles in multiple scenarios such as urban roads and highway networks, providing high-precision trajectory data support for subsequent cross-scene violation recognition.

[0077] Furthermore, step A400 in the method provided in this application embodiment includes:

[0078] A410: Based on the preset cross-scenario linkage strategy, and based on the cross-scenario complete driving map of the target vehicle, combined with the cross-scenario full-domain data pool, extract the traffic behavior features associated with the target vehicle.

[0079] A420: Perform indicator modeling and risk assessment on the aforementioned traffic behavior characteristics, and obtain risk assessment results by combining different data dimensions.

[0080] A430: Based on the risk assessment results, assess the risk level of the target vehicle and identify whether it has cross-scenario violations. If so, obtain the cross-scenario violation identification result.

[0081] In this embodiment, the cross-scenario linkage strategy is a pre-defined rule for achieving linked analysis of checkpoint data in different scenarios. The passage behavior characteristics are behavioral features associated with the target vehicle extracted from the target vehicle's complete cross-scenario driving map and combined with the cross-scenario full-domain data pool.

[0082] Specifically, firstly, based on preset cross-scenario linkage strategies, such as triggering highway overload tracking rules when toll station load data exceeds 55 tons, or linking highway checkpoint path prediction when a vehicle enters or exits a certain area ≥3 times within 24 hours, the system extracts associated traffic behavior features from the cross-scenario full-domain data pool based on the target vehicle's complete cross-scenario driving map. These features include path sparseness, checkpoint skipping frequency, high-frequency nighttime traffic patterns, and similar trajectory collision situations. For example, if a heavy truck is detected at a toll station with a load of 60 tons, exceeding the threshold of 55 tons, its highway driving trajectory is automatically linked to subsequent city checkpoint passage records to extract composite behavioral features of overload, highway speeding, and passage through prohibited urban areas.

[0083] Next, when modeling and assessing the risks of the extracted traffic behavior features, the features such as path sparseness, checkpoint skipping frequency, high-frequency traffic patterns at night, and collisions of similar trajectories are first quantified into computable indicators. The specific steps are as follows:

[0084] Step a: Path rarity quantification method. Based on a cross-scenario, full-domain data pool, the driving route data of the target vehicle is extracted, and the frequency of the route's occurrence in historical trajectories is statistically analyzed. A piecewise linear mapping method is used for score quantification. If the route's occurrence frequency is <5%, it is judged as high rarity, corresponding to 80 points; a frequency between 5% and 10% receives 60 points; and a frequency exceeding 10% receives 40 points. This process achieves a quantitative assessment of the rarity of vehicle driving routes through statistical analysis of historical data.

[0085] Step b: Quantification method for checkpoint skipping frequency. First, for adjacent checkpoint sections, establish a baseline of the mean and standard deviation of historical passage times based on vehicle type. For example, the normal travel time for a truck is 15±3 minutes. When the time difference of the target vehicle passing through the checkpoint exceeds 30% of the upper limit of the baseline time, it is judged as a skip. If the cumulative number of skips is ≥2 within 24 hours, a score is calculated based on the number of skips (75 points for 2 skips, 85 points for 3 skips). By comparing the time difference with the baseline data, the abnormal travel frequency of vehicles between checkpoints is quantified.

[0086] Step c: Quantification method for high-frequency nighttime traffic patterns. The period from 22:00 to 6:00 is defined as the nighttime period. The number of times a vehicle passes through a checkpoint during this period is counted weekly. A step function is used to convert scores: ≤2 nighttime passes per week = 40 points, 3-4 passes = 60 points, ≥5 passes = 80 points. Statistical analysis of nighttime traffic frequency identifies abnormal patterns of high-frequency nighttime traffic.

[0087] Step d: Quantification method for similar trajectory collisions. Utilizing the Dynamic Time Warping (DTW) algorithm, the spatiotemporal trajectory point sequence of the target vehicle is matched with the trajectory database of vehicles involved in the case, and the trajectory overlap is calculated. An overlap greater than 70% is considered high-risk, corresponding to a score of 85. This process, through trajectory feature extraction and algorithm matching, achieves quantitative analysis of vehicle trajectory similarity, assisting in the identification of potential risks.

[0088] Then, a multi-dimensional linear weighted model is used for indicator modeling. First, the weight parameters corresponding to the traffic behavior features are obtained: First, a historical violation dataset containing features such as path rarity, checkpoint jumping frequency, high-frequency traffic patterns at night, and similar trajectory collisions needs to be constructed. Each data point needs to be labeled as whether it is a cross-scene violation, with a label value of 0 or 1. Then, the features are standardized to eliminate the influence of units, for example, the path rarity score and checkpoint jumping frequency score are uniformly mapped to the [0,1] interval. Next, a supervised learning algorithm (such as logistic regression, random forest, or gradient boosting tree) is selected to build the model. The standardized features are used as input and the violation labels are used as output. The weight parameters are optimized by minimizing the cross-entropy loss function or maximizing the AUC value. During training, the k-fold cross-validation method is used to divide the training set and the validation set, and the weights of each feature are dynamically adjusted (such as initial weights of path rarity × 0.3, checkpoint jumping frequency × 0.2, etc. can be used as prior values) to maximize the model's recognition accuracy for high-risk behaviors (such as samples marked as 1 in historical data). Ultimately, the weight parameters obtained through training enable the risk score calculation model to achieve a preset threshold (such as an accuracy rate ≥ 90%) in recognizing cross-scenario violations on the test set, ensuring that the weight allocation conforms to the risk distribution characteristics of actual violation scenarios.

[0089] Furthermore, an indicator model is constructed based on the obtained weight parameters. For example, the risk score is set as: Risk Score = Path Rarity × 0.3 + Checkpoint Jump Frequency × 0.2 + Nighttime High-Frequency Traffic × 0.2 + Similar Trajectory Collision × 0.3. The weight parameters are obtained by training historical violation data using machine learning algorithms to ensure the model's sensitivity to high-risk behaviors. Taking a hazardous materials transport vehicle as an example, its path rarity score is 80, its checkpoint jump frequency score is 75, its nighttime high-frequency traffic and similar trajectory collision scores are 70 and 85 respectively. Substituting these into the formula, the comprehensive risk assessment result is 80 × 0.3 + 75 × 0.2 + 70 × 0.2 + 85 × 0.3 = 24 + 15 + 14 + 25.5 = 78.5 points. This score exceeds the preset risk threshold (e.g., 70 points), thus identifying a risk of cross-scenario violations and providing a quantitative basis for subsequent early warning and handling.

[0090] Based on the risk assessment results, a risk level threshold is set (e.g., 70 points is high risk). When the target vehicle's risk score is ≥70 points, cross-scenario violations are automatically identified, and cross-scenario violation identification results are generated. For example, the above-mentioned hazardous materials transport vehicle has a comprehensive score of 78.5 points, triggering cross-scenario violation identification for high-frequency nighttime restricted areas and abnormal routes. Compared with traditional single load detection, it can simultaneously detect its violations during restricted hours in the city.

[0091] By extracting complex traffic features through a pre-set cross-scenario linkage strategy, a multi-dimensional indicator modeling and risk assessment system is constructed to effectively identify complex violations. This provides accurate risk judgment basis for subsequent cross-scenario early warning and handling, solving the problem of missed violations caused by fragmented multi-scenario data in traditional technologies.

[0092] Furthermore, step A500 in the method provided in this application embodiment includes:

[0093] A510: Based on the cross-scenario violation recognition results, trigger the cross-scenario warning command.

[0094] A520: Combining the cross-scenario complete driving map of the target vehicle and the cross-scenario early warning command, predict the potential passage checkpoints of the target vehicle according to the multi-checkpoint strategy linkage module, and determine the early warning priority of the checkpoint.

[0095] A530: Issue linkage strategies based on the distributed edge nodes corresponding to the potential access checkpoints, wherein the linkage strategies include real-time camera tracking instructions, deployment prompts, and blacklist temporary storage instructions.

[0096] A540: Based on the aforementioned linkage strategy, cross-checkpoint warning is completed before the target vehicle arrives at the potential checkpoint.

[0097] In one embodiment, a cross-scenario warning instruction indicating potential violations by the target vehicle is first triggered based on the cross-scenario violation identification results. For example, when a cross-scenario violation is identified involving an overloaded truck traveling in a restricted urban area, a warning instruction containing the violation type and vehicle characteristics is automatically generated, providing a basis for subsequent predictions.

[0098] Next, combining the target vehicle's complete cross-scenario driving map and warning instructions, the spatiotemporal prediction algorithm in the multi-checkpoint strategy linkage module is used to analyze the vehicle's historical traffic patterns, current driving direction, and real-time road conditions to predict potential checkpoints. The trajectory prediction model based on the LSTM neural network is constructed and analyzed as follows:

[0099] Step e: Extract the target vehicle's historical passage records from the complete cross-scene driving map, including sequence data such as timestamps, checkpoint spatial coordinates, driving speed, and stopping points. Transform this data into spatiotemporal feature vectors, such as mapping latitude and longitude to Cartesian coordinates, converting timestamps into time interval features, and performing standardization to eliminate the influence of dimensions. The model employs a multi-layer LSTM network structure. The input layer receives spatiotemporal sequences within a fixed time window (e.g., the past 24 hours), the hidden layer captures the temporal dependence and long-term dependence patterns of the trajectory through a gating mechanism, and the output layer predicts the probability distribution of potential passage locations within a future period (e.g., the next 1-3 hours).

[0100] Step f: During the model training phase, supervised learning is performed using historical trajectory data (covering checkpoints in various scenarios such as cities and highways). Real checkpoint records are used as labels, and the network parameters, such as the weight matrix and bias terms, are optimized through backpropagation to minimize the mean square error between the predicted and actual trajectories. Simultaneously, an attention mechanism is introduced to enhance the model's ability to identify key checkpoints (such as traffic hubs and high-frequency traffic nodes). During analysis, the prediction parameters are dynamically adjusted by combining the current driving direction (calculated by the heading angle from continuous checkpoint coordinates) and real-time road conditions (calculated by traffic flow data to determine road segment travel time). For example, if real-time traffic conditions show congestion on a certain highway segment, the weight of the passage probability for that segment is reduced.

[0101] Step g: Embed cross-scenario early warning instructions as constraints into the model. For example, when an early warning instruction indicates that overloaded vehicles need to be closely monitored at highway entrances, the model will prioritize predicting the passage probability of highway checkpoints and increase the early warning priority of such checkpoints. Finally, the passage probability of each potential checkpoint is output through the softmax function, and a priority ranking is generated by combining the control importance of checkpoints (e.g., the weight of checkpoints on main roads is higher than that of branch roads), providing accurate predictive support for the subsequent issuance of linkage strategies.

[0102] Next, the warning priority index for each checkpoint is calculated, such as priority = trajectory matching degree × 0.5 + estimated arrival time urgency × 0.3 + checkpoint control importance × 0.2, and the prediction results are ranked. For example, a vehicle involved in a case is predicted to have an 85% probability of passing through city checkpoint X within the next hour, and this checkpoint is a transportation hub. Its warning priority index reaches 8.2 (out of 10), and it is judged as a high-priority warning checkpoint.

[0103] Then, according to the distributed edge nodes corresponding to the potential passing checkpoints, linkage strategies such as real-time tracking instructions for cameras, deployment prompt messages, and blacklist temporary storage instructions are issued. Exemplarily, an instruction is sent to the edge node of checkpoint X to start real-time tracking of the target vehicle by a high-definition camera. At the same time, deployment prompt messages (including vehicle photos and violation types) are pushed to the terminal of checkpoint staff, and the vehicle information is temporarily stored in the checkpoint blacklist database, realizing the linkage response between front-end devices and back-end management and control.

[0104] Finally, based on the linkage strategy, cross-checkpoint early warning is completed before the target vehicle arrives at the potential passing checkpoint. By calculating the vehicle driving speed and estimated arrival time in real time, when a certain over-limit truck is expected to arrive at checkpoint Y in 30 minutes, the camera tracking deployment, deployment prompt push, and blacklist entry of this checkpoint are completed 25 minutes in advance, making the early warning response time advance from the post-arrival processing of traditional technologies to active deployment 30 minutes before arrival, and improving the interception preparation time.

[0105] By triggering early warning instructions through cross-scenario violation recognition, determining potential checkpoints and priorities by combining driving maps and spatio-temporal prediction algorithms, and issuing multi-dimensional linkage strategies using edge nodes, the lagging passive early warning mode in traditional technologies is upgraded to an active early warning system of violation recognition - trajectory prediction - early deployment, constructing a closed-loop investigation system of accurate perception at a single checkpoint - data interconnection among multiple checkpoints - strategy linkage, effectively solving the problem of management and control failure caused by lagging traditional early warnings.

[0106] In summary, the vehicle investigation method and electronic device provided by the embodiments of the present application under the distribution of traffic checkpoints have the following technical effects:

[0107] In this application, vehicle basic data is collected through the collaborative network of a single checkpoint and multiple checkpoints at traffic checkpoints. After operations such as spatio-temporal calibration and hierarchical processing, vehicle trajectory-related information is obtained. A complete cross-scenario driving map and risk assessment model are constructed, and early warnings are issued in combination with preset cross-scenario linkage strategies and prediction results, so as to accurately identify various situations of illegal vehicles in a complex traffic network, making the vehicle investigation results at traffic checkpoints more accurate and efficient, achieving data interconnection and strategy linkage among multiple traffic checkpoints, forming a closed-loop investigation system, and improving the accuracy and efficiency of vehicle management and control in a complex traffic network.

[0108] Embodiment 2, as Figure 2 shown, based on the same inventive concept as in the foregoing Embodiment 1, the embodiments of the present application provide an electronic device, and the electronic device includes:

[0109] A memory 303 for storing executable instructions; a processor 302 for implementing a vehicle investigation method under the distribution of traffic checkpoints when executing the executable instructions stored in the memory 303.

[0110] Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, showing a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present invention. Figure 2 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention. This electronic device is in the form of a general-purpose computing device, and its components may include, but are not limited to, an input device 301, a processor 302, a memory 303, and an output device 304. The processor 302 may be one or more; the memory 303 may include a computer-readable medium and at least one program product having a set (at least one) of program modules configured to perform the functions of the embodiments of this application.

[0111] The memory 303 shown in this embodiment of the invention can be any combination of one or more computer-readable media. The computer-readable storage medium can be, but is not limited to, infrared, semiconductor systems, devices or components, or any combination thereof, for storing software programs, computer-executable programs and modules, such as the program instructions / modules corresponding to a vehicle screening method under traffic checkpoint distribution in this embodiment of the invention. The processor 302 executes various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 303, thereby realizing the above-mentioned vehicle screening method under traffic checkpoint distribution.

[0112] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0113] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A vehicle screening method under a traffic checkpoint distribution, characterized in that, include: Construct a multi-checkpoint collaborative network, traverse any single checkpoint in the multi-checkpoint collaborative network, collect vehicle basic data in real time, and build a cross-scenario full-domain data pool; The cross-scenario global data pool is processed hierarchically, and local strategy processing is performed on the basic data of each vehicle in the cross-scenario global data pool to determine the target vehicle. Combining the target vehicle, the vehicle's spatiotemporal trajectory is spliced ​​and features are associated based on the cross-scenario full-domain data pool to generate a complete cross-scenario driving map; Based on the preset cross-scenario linkage strategy, the cross-scenario full-domain data pool is analyzed in multiple dimensions based on the cross-scenario complete driving map to obtain the cross-scenario violation behavior recognition results. Based on the cross-scenario violation recognition results, a cross-scenario warning command is triggered. Based on the cross-scenario warning command and the complete cross-scenario driving map, the potential access checkpoints of the target vehicle are predicted. The target vehicle is then given a cross-scenario warning through the multi-checkpoint strategy linkage module.

2. The vehicle searching method under the traffic kiosk distribution according to claim 1, characterized in that, Construct a multi-checkpoint collaborative network, traverse any single checkpoint in the multi-checkpoint collaborative network, collect basic vehicle data in real time, and build a cross-scenario, full-domain data pool, including: Distributed edge nodes are deployed at multiple traffic checkpoints to construct the multi-checkpoint collaborative network; Traverse any single checkpoint in the multi-checkpoint collaborative network, call the edge node of the corresponding checkpoint, and obtain the video stream of the checkpoint; The video stream is analyzed in real time to extract dynamic information of the vehicle as it passes through the checkpoint and generate the vehicle's basic data. The vehicle basic data is transmitted to the central node through the edge node according to a preset synchronous communication protocol; The central node performs unified timestamp calibration and spatial location mapping on the vehicle basic data to generate standardized multi-source data. The standardized multi-source data is fused to construct the cross-scenario, full-domain data pool.

3. The vehicle screening method under traffic checkpoint distribution as described in claim 1, characterized in that, The cross-scenario global data pool is processed hierarchically, and local policy processing is performed on the basic data of each vehicle in the cross-scenario global data pool to determine the target vehicle, including: Prioritize the vehicle basic data in the cross-scenario global data pool and construct a hierarchical processing queue; The vehicle basic data in the hierarchical processing queue is input into the local policy engine one by one to perform rapid deployment processing or passage verification processing and generate local policy processing results. The target vehicles are selected based on the results of the local strategy processing.

4. The vehicle screening method under traffic checkpoint distribution as described in claim 1, characterized in that, Combining the target vehicle, and based on the cross-scenario full-domain data pool, the vehicle's spatiotemporal trajectory is stitched together and features are associated to generate a complete cross-scenario driving map, including: The historical trajectory, frequently passing nodes, and cross-regional passage mode information associated with the target vehicle are extracted from the cross-scenario full-domain data pool to obtain the full-domain data extraction results. The local strategy processing results are fused with the global data extraction results to generate a fused dataset; The fused dataset is input into the map generation module, which performs vehicle spatiotemporal trajectory stitching and feature association based on the fused dataset to construct a complete cross-scene driving map of the target vehicle.

5. The vehicle screening method under traffic checkpoint distribution as described in claim 4, characterized in that, The fused dataset is input into the map generation module, which performs vehicle spatiotemporal trajectory stitching and feature association based on the fused dataset to construct a complete cross-scene driving map of the target vehicle, including: Using the fused dataset as a stitching anchor, preliminary trajectory stitching is performed on the passage records of multiple checkpoints; Based on the splicing anchor points, a graph structure matching method is used to merge the passage paths between multiple checkpoints and construct a preliminary driving trajectory. The constructed preliminary driving trajectory is subjected to feature association processing, which includes comparing the image similarity of vehicle color, vehicle size and adjacent checkpoints to correct the accuracy and enhance the confidence of the preliminary driving trajectory, and obtain the feature association result. Based on the correlation results between the preliminary driving trajectory and the features, a complete cross-scene driving map of the target vehicle is generated.

6. The vehicle screening method under traffic checkpoint distribution as described in claim 1, characterized in that, Based on a preset cross-scenario linkage strategy, a multi-dimensional analysis is performed on the cross-scenario full-domain data pool using the complete cross-scenario driving map to obtain cross-scenario violation behavior recognition results, including: Based on the preset cross-scenario linkage strategy, and combined with the cross-scenario full-domain data pool, the traffic behavior features associated with the target vehicle are extracted. The aforementioned traffic behavior characteristics are modeled and risk assessed using indicators, and the risk assessment results are obtained by combining different data dimensions. Based on the risk assessment results, the risk level of the target vehicle is assessed, and it is identified whether there are cross-scenario violations. If so, the cross-scenario violation identification result is obtained.

7. The method according to claim 1, characterized in that, Based on the cross-scenario violation recognition results, a cross-scenario early warning command is triggered. Based on the cross-scenario early warning command and the complete cross-scenario driving map, potential checkpoints for the target vehicle are predicted. A multi-checkpoint strategy linkage module then provides a cross-scenario early warning for the target vehicle, including: Based on the cross-scenario violation recognition results, the cross-scenario early warning command is triggered; Combining the cross-scenario complete driving map of the target vehicle and the cross-scenario early warning command, the potential passage checkpoints of the target vehicle are predicted according to the multi-checkpoint strategy linkage module, and the early warning priority of the checkpoint is determined. The linkage strategy is issued according to the distributed edge node corresponding to the potential access checkpoint, wherein the linkage strategy includes real-time camera tracking instructions, deployment prompt information and blacklist temporary storage instructions; Based on the aforementioned linkage strategy, cross-checkpoint warnings are issued before the target vehicle arrives at the potential checkpoint.

8. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the vehicle screening method under traffic checkpoint distribution as described in any one of claims 1 to 7.

Images