A dynamic monitoring system and method for the in-transit status of an oversize load transport vehicle

Abstract

The application is suitable for the technical field of intelligent traffic supervision, and provides a dynamic supervision system and method for the in-transit state of large-transportation vehicles, which comprises the following steps: obtaining approval data and synchronously collecting multi-source real-time monitoring data; identifying large-transportation vehicles through the algorithm of fusing three-dimensional target detection and visual target detection, and comparing and matching the identification result with the approval data; comparing and matching the real-time cargo image with the approval recorded cargo image through a preset cargo comparison neural network to generate a cargo compliance state; generating an adaptive electronic fence according to the approval route, and detecting the off-route, overspeed and illegal parking of the vehicle according to the real-time track data of the vehicle to generate a route compliance state; and generating early warning information according to the cargo compliance state and the route compliance state. The application can realize full-section, high-precision and automatic vehicle parameter verification, and focuses on supervising the illegal behaviors such as cargo replacement in the middle of the journey and off-route driving of large-transportation vehicles.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation monitoring technology, and in particular relates to a dynamic monitoring system and method for the status of oversized transport vehicles en route. Background Technology

[0002] With the continuous growth in demand for cross-regional transportation of oversized and heavy-duty goods such as wind power equipment, heavy chemical equipment, and rail transit equipment, the transportation of oversized and heavy-duty goods, as a core link in the circulation of major production factors, directly impacts the stability of the industrial and supply chains in terms of transportation safety and regulatory efficiency. Currently, the supervision of oversized and heavy-duty transportation has formed a basic framework of "pre-approval - in-process supervision - post-enforcement," but in-transit supervision remains a core weakness. Existing technologies and regulatory models can no longer meet the regulatory needs of oversized and heavy-duty transportation, which involves high frequency, long distances, and high risks. Specifically, existing in-transit supervision technologies for oversized and heavy-duty transportation mainly consist of two core modules, and their mainstream implementation methods are as follows:

[0003] Current mainstream solutions for identifying and verifying oversized transport vehicles rely on fixed monitoring equipment at overload control stations, highway gantries, and toll stations. These systems utilize non-stop weighing, fixed-point video capture, and single-point LiDAR detection to achieve vehicle weighing, license plate recognition, and basic external dimension measurement. Some solutions incorporate AI target detection algorithms such as YOLO to identify vehicle types within fixed-point videos. For manual verification, existing technologies primarily employ a model where law enforcement personnel take photos on-site and manually compare them with approved parameters to complete the verification of goods and vehicles. Currently, there is no effective solution for accurately detecting violations such as changing cargo mid-journey in oversized transport vehicles.

[0004] Currently, the monitoring of the trajectory of oversized transport mainly relies on comparing vehicle trajectory data with approved routes. When the vehicle trajectory deviates from the predetermined route, an alarm is triggered. Some solutions have added basic abnormal behavior recognition functions such as speeding and long-term parking. However, when the vehicle positioning equipment is turned off or malfunctioning, it is impossible to detect vehicle violations in a timely manner, resulting in delayed law enforcement.

[0005] In addition, the existing patent CN120997592A discloses a method and system for detecting violations of large transport vehicles based on an improved Mask RCNN model. It focuses on the detection of violations of large transport vehicles in occluded scenes. The core improvement is to the traditional Mask RCNN model by improving the original FPN network, adding a bottom-up feature enhancement path and a horizontal connection path between layers, removing the bounding box and confidence output unit, adding a violation judgment and calculation module, removing the bounding box loss, adding edge loss to optimize the mask segmentation accuracy, training the dataset with the improved model and saving the weights, and finally using it for violation detection and result visualization of images that did not participate in the training.

[0006] Existing patent CN119515238A discloses a method and system for monitoring the transportation of oversized cargo on highways. This patent focuses on the detection of anomalies during the transportation of oversized cargo on highways, and its core is based on speed and vibration data to achieve monitoring. It acquires vehicle transportation speed data and transportation vibration data, extracts features from the two types of data respectively, and constructs transportation speed and vibration feature map data. By collecting abrupt change features in the feature map and extracting the degree of change, it obtains transportation abrupt change feature data and transportation change degree data, thereby completing the detection of anomalies in the transportation of oversized cargo and improving transportation safety and management level.

[0007] Existing patent CN120318776A discloses a method and system for detecting violations in heavy-duty transport vehicles based on multi-dimensional data matching. This patent focuses on detecting overloading and illegal loading in heavy-duty transport vehicles. It collects axle load and wheelbase data of multi-axle heavy-duty vehicles through a dynamic weighing system, and forms a one-dimensional discrete data map through data segmentation to establish a mapping relationship between axle load, wheelbase, and total load to determine whether the vehicle is overloaded. At the same time, it collects vehicle images through highway cameras, uses the SLAM algorithm to generate a vehicle point cloud model, extracts geometric parameters and matches them with a preset contour template to determine whether there is illegal loading, and finally realizes the detection of violations in heavy-duty transport vehicles.

[0008] However, existing literature and patents have not yet presented a comprehensive solution that integrates multi-source data from "fixed monitoring + drone blind spot filling + vehicle-mounted terminals" to achieve full-segment status monitoring of oversized transport vehicles. Furthermore, the existing technologies only address a single violation status, failing to effectively identify instances of oversized transport vehicles changing cargo en route or deviating from approved routes when positioning equipment is off or malfunctioning. This lack of a complete closed-loop monitoring solution hinders the development of a robust technical solution. In summary, existing technologies suffer from the following problems that urgently need to be addressed: blind spots in monitoring coverage and insufficient identification accuracy and automation. Current oversized transport vehicle identification relies entirely on fixed monitoring points, lacking effective monitoring coverage in remote sections, branch lines of national and provincial highways, and blind spots in the road network. This makes it impossible to achieve full-segment on-the-go verification, and prevents the timely detection of hidden violations such as unauthorized cargo changes en route or deviations from approved transport routes when positioning equipment is off or malfunctioning. Summary of the Invention

[0009] The purpose of this invention is to provide a dynamic monitoring system and method for the status of oversized transport vehicles en route, in order to solve the aforementioned technical problems.

[0010] This invention is implemented as follows: a method for dynamic monitoring of the status of oversized transport vehicles en route, comprising the following steps:

[0011] The system acquires approval data and simultaneously collects multi-source real-time monitoring data from fixed monitoring units, drone monitoring units, and vehicle-mounted terminal units. This data is then pre-processed to form a full-process data archive indexed by vehicle identification.

[0012] Based on the full-process data archive, the algorithm that integrates 3D target detection and visual target detection is used to identify oversized transport vehicles, and the identification results are compared and matched with the approval data to obtain the matched information of vehicles on the way.

[0013] Based on the information of the vehicles in transit, real-time vehicle parameters and cargo images are obtained at fixed monitoring points or through drones to fill blind spots. The real-time parameters are verified with the approval data, and the real-time cargo images are compared with the approved and filed cargo images through a preset cargo comparison neural network to generate the cargo compliance status.

[0014] Based on the information of vehicles on the road, an adaptive electronic fence is generated according to the approved route, and the deviation from the route, speeding and illegal parking are detected by combining the real-time trajectory data of the vehicles, and the route compliance status is generated.

[0015] Early warning information is generated based on the compliance status of the goods and the route, and the early warning information and vehicle location are pushed to the execution terminal to form a regulatory closed loop.

[0016] Furthermore, the approval data includes a list L of approved large-item transportation tasks awaiting delivery; L = {l1, l2, ..., l...} n} indicates that there are a total of n transportation tasks, where l i Let l represent the i-th transportation task in L; i =[Task ID, License Plate Number, Vehicle Parameters, Cargo Parameters, Transportation Route, Transportation Time Limit, Vehicle Image].

[0017] Furthermore, the fixed monitoring unit is used to monitor one or more of the following: vehicle passing video data, lidar size data, non-stop weighing data, passing timestamp, and location latitude and longitude data;

[0018] The UAV monitoring unit is used to supplement the monitoring of one or more of the following: all-round high-definition aerial images of the vehicle, three-dimensional point cloud data, latitude and longitude coordinates and timestamp data of the monitoring point;

[0019] The vehicle-mounted terminal unit is used to collect one or more of the following: real-time vehicle trajectory data, driving speed, heading angle, and timestamp data.

[0020] Furthermore, the preprocessing method includes one or more of the following: normalization, deduplication and cleaning, noise suppression, and image enhancement.

[0021] Furthermore, the steps for identifying oversized transport vehicles using algorithms that integrate 3D object detection and visual object detection specifically include:

[0022] The system takes images and 3D point cloud data acquired by fixed monitoring units and UAV monitoring units as inputs. For 3D point cloud data, PointRCNN is used to extract 3D targets and output point cloud geometric size features. For images, deep convolutional neural networks are used to extract image visual features. The point cloud geometric features and image visual features are fused to output a unified feature vector, and the recognition result of heavy transport vehicles is obtained through a fully connected neural network.

[0023] Furthermore, the parameter verification method employs a difference comparison mode:

[0024] Determine the approval benchmark parameters;

[0025] The parameter deviation rate is calculated based on the vehicle's real-time parameters and the approval benchmark parameters.

[0026] Set a differential deviation threshold, compare the parameter deviation rate with the differential deviation threshold, and determine whether the parameter violates the rules.

[0027] Furthermore, the step of comparing the real-time cargo images with the approved and registered cargo images using a pre-set cargo comparison neural network specifically includes:

[0028] A cargo comparison neural network is constructed using a dual-branch structure of CNN and Transformer, and then trained and optimized. The CNN branch is used to extract shallow visual features of cargo images, while the Transformer branch is used to extract deep semantic features.

[0029] The real-time cargo image and the approved and registered cargo image are respectively input into the trained cargo comparison neural network, and two sets of feature vectors are output.

[0030] The cosine similarity algorithm is used to calculate the similarity between two sets of feature vectors;

[0031] The compliance status of goods is determined based on the similarity between the two sets of feature vectors.

[0032] Furthermore, the step of generating an adaptive electronic fence based on the approved route specifically includes:

[0033] The system retrieves and approves delivery routes containing baseline information, including the route's origin, destination, route segments, permitted driving time periods, and speed limits for each segment.

[0034] Retrieve the latitude and longitude coordinates and monitoring coverage data of the fixed monitoring units to construct a coordinate library of fixed monitoring points;

[0035] The latitude and longitude coordinate sequence of the approval route is thinned out, retaining the key coordinates of the route turning points and intersections, and eliminating redundant coordinates to obtain the standardized approval route.

[0036] An electronic fence is generated based on the standardized approval route. The electronic fence covers the area that the heavy transport vehicle must travel in. A Gaussian buffer zone with a preset range on one side is generated as the core control area, with the approval route as the center.

[0037] Spatially match the coordinates of fixed monitoring points with the electronic fence, mark the fixed monitoring points inside and outside the fence, and clarify that the monitoring points inside the fence are used for routine verification, while the monitoring points outside the fence are used for route deviation judgment.

[0038] The system receives real-time information on approved route adjustments, road network construction, and changes in fixed monitoring points. Through an incremental update algorithm, it verifies and updates the electronic fence parameters to obtain an adaptive electronic fence.

[0039] Another objective of this invention is to provide a dynamic monitoring system for the en route status of heavy-duty transport vehicles, used to implement the aforementioned dynamic monitoring method for the en route status of heavy-duty transport vehicles, specifically including:

[0040] The data layer is used to acquire approval data and simultaneously collect multi-source real-time monitoring data from fixed monitoring units, drone monitoring units, and vehicle-mounted terminal units. It performs preprocessing to form a full-process data archive indexed by vehicle identification.

[0041] The on-the-way vehicle identification and matching module is used to identify oversized transport vehicles based on the full-process data archive by using an algorithm that integrates 3D target detection and visual target detection, and compares and matches the identification results with the approval data to obtain the matched on-the-way vehicle information.

[0042] The cargo compliance verification module is used to acquire real-time vehicle parameters and cargo images at fixed monitoring points or through drones based on the information of the vehicles in transit. The real-time parameters are verified against the approval data, and the real-time cargo images are compared with the approved and filed cargo images through a preset cargo comparison neural network to generate the cargo compliance status.

[0043] The route compliance detection module is used to generate an adaptive electronic fence based on the information of the vehicles on the road and the approved route, and to detect deviation from the route, speeding and illegal parking by combining the real-time trajectory data of the vehicles, and generate a route compliance status.

[0044] The violation warning module is used to generate warning information based on the compliance status of the goods and the route, and push the warning information and vehicle location to the execution terminal to form a regulatory closed loop.

[0045] The present invention provides a dynamic monitoring method for the status of oversized transport vehicles en route, which integrates real-time monitoring data from multiple sources such as fixed monitoring, UAV blind spot filling, and vehicle-mounted terminals, and combines computer vision, 3D point cloud reconstruction and target detection algorithms to achieve high-precision and automated vehicle parameter verification across the entire road segment. It can focus on monitoring violations such as changing cargo midway and deviating from the approved route by oversized transport vehicles. Attached Figure Description

[0046] Figure 1 A flowchart illustrating the dynamic monitoring method for the status of oversized transport vehicles en route, provided in an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0048] like Figure 1 As shown, in one embodiment of the present invention, a dynamic monitoring method for the en route status of oversized transport vehicles is provided to solve the technical problems in the prior art, such as low efficiency in the identification and verification of oversized transport vehicles and the monitoring of their en route status, inability to achieve linkage between intelligent identification of violations and control upgrades, lack of a closed-loop monitoring system, and poor monitoring coordination. The method specifically includes the following steps:

[0049] S1. Obtain approval data and simultaneously collect multi-source real-time monitoring data from fixed monitoring units, drone monitoring units, and vehicle-mounted terminal units, perform preprocessing, and form a full-process data archive indexed by vehicle identification.

[0050] S2. Based on the full-process data archive, identify the oversized transport vehicle by integrating three-dimensional target detection and visual target detection algorithms, and compare and match the identification results with the approval data to obtain the matched on-the-way vehicle information.

[0051] S3. Based on the information of the vehicles in transit, real-time vehicle parameters and cargo images are obtained at fixed monitoring points or through drones to fill blind spots. The real-time parameters are verified with the approval data, and the real-time cargo images are compared with the approved and filed cargo images through a preset cargo comparison neural network to generate the cargo compliance status.

[0052] S4. Based on the information of the vehicles on the way, generate an adaptive electronic fence according to the approved route, and combine it with the real-time trajectory data of the vehicles to detect deviation from the route, speeding and illegal parking, and generate the route compliance status.

[0053] S5. Generate early warning information based on the compliance status of the goods and the route, and push the early warning information and vehicle location to the execution terminal to form a regulatory closed loop.

[0054] In another embodiment of the present invention, a dynamic monitoring system for the status of oversized transport vehicles en route is also provided to implement the above method, specifically including a data layer, an algorithm layer (core execution layer) and an application layer;

[0055] The data layer is used to acquire approval data and simultaneously collect multi-source real-time monitoring data from fixed monitoring units, UAV monitoring units, and vehicle-mounted terminal units. This data is preprocessed to form a complete data archive indexed by vehicle identification. The data layer is responsible for the collection, preprocessing, and storage of all data throughout the process, serving as the system's fundamental support. Its main components include:

[0056] Fixed monitoring unit: Connects to existing equipment at overload control stations, highway gantries, and toll stations to monitor in real time vehicle passing video data, lidar size data, non-stop weighing data, passing timestamps, and location latitude and longitude data;

[0057] Drone monitoring unit: A low-altitude drone equipped with a high-definition zoom camera and a lightweight lidar is responsible for filling in the blind spots of remote road sections, road network blind spots, and key monitored vehicles. It is used to monitor the target vehicle in real time with all-round high-definition aerial images, three-dimensional point cloud data, latitude and longitude and timestamp data of monitoring points.

[0058] Vehicle-mounted terminal unit: Interacts with the vehicle's Beidou / GPS vehicle-mounted terminal to collect real-time vehicle trajectory data (latitude and longitude), driving speed, heading angle, and timestamp data;

[0059] Approval data storage subunit: Connects to the oversized cargo transportation permit approval system to store the approval and filing data of the target vehicle, including basic information such as permitted vehicle parameters, cargo parameters, permitted driving routes, transportation time limits, and compliance records;

[0060] Data preprocessing unit: unifies the timestamps (UTC format) and coordinate system (WGS-84 latitude and longitude) of all data, completes preprocessing such as data deduplication and cleaning, noise suppression, format standardization, and image enhancement, and constructs a full-process data archive with "vehicle ID + transportation task ID" as the unique identifier.

[0061] The algorithm layer mainly includes:

[0062] The on-the-way vehicle identification and matching module is used to identify oversized transport vehicles based on the full-process data archive by using an algorithm that integrates 3D target detection and visual target detection, and compares and matches the identification results with the approval data to obtain the matched on-the-way vehicle information.

[0063] The cargo compliance verification module is used to acquire real-time vehicle parameters and cargo images at fixed monitoring points or through drones based on the information of the vehicles in transit. The real-time parameters are verified against the approval data, and the real-time cargo images are compared with the approved and filed cargo images through a preset cargo comparison neural network to generate the cargo compliance status.

[0064] The route compliance detection module is used to generate an adaptive electronic fence based on the information of the vehicles on the road and the approved route, and to detect deviation from the route, speeding and illegal parking by combining the real-time trajectory data of the vehicles, and generate a route compliance status.

[0065] The violation warning module is used to generate warning information based on the compliance status of the goods and the route, and push the warning information and vehicle location to the execution terminal to form a regulatory closed loop.

[0066] The application layer serves as the system's execution terminal, comprising a regulatory visualization platform and a tiered early warning subunit. The regulatory visualization platform displays real-time vehicle locations, verification results, compliance status, electronic fence warning information, and enforcement personnel distribution, supporting full-process data backtracking, violation statistics, and report generation. The tiered early warning subunit, based on vehicle compliance status and abnormal behavior, establishes a three-tiered early warning mechanism, pushing early warning information, information on violating vehicles, and precise locations to frontline execution terminals in real time. This supports law enforcement personnel in conducting precise interception and handling, with the handling results synchronously transmitted back to the system, updating vehicle compliance files and forming a closed-loop management system.

[0067] In a preferred embodiment of the present invention, the approval data includes a list L of approved large-item transportation tasks awaiting delivery; L = {l1, l2, ..., l...} n} indicates that there are a total of n transportation tasks, where l i Let l represent the i-th transportation task in L; i =[Task ID, License Plate Number, Vehicle Parameters, Cargo Parameters, Transportation Route, Transportation Time Limit, Vehicle Image]. In practical applications, vehicle parameters include detailed parameters such as the model, length, width, height, wheelbase, torque, load capacity, and turning radius of the transport vehicle; cargo parameters include cargo type, cargo weight, cargo quantity, cargo dimensions (length, width, height), and hazardous attributes; the transportation route is the data of the road segments that need to be traversed during delivery; the transportation time limit is the time range within which the transportation task is approved for execution; and the vehicle image refers to the image of the loaded vehicle that needs to be uploaded when applying for a transportation task for heavy-duty transport vehicles.

[0068] In a preferred embodiment of the present invention, the step of identifying oversized transport vehicles using an algorithm that integrates 3D target detection and visual target detection specifically includes:

[0069] The system takes images and 3D point cloud data acquired by fixed monitoring units and UAV monitoring units as input. For the 3D point cloud data, PointRCNN is used to extract 3D targets and output the geometric dimensions (e.g., length, width, and height) of the point cloud. For the images, a deep convolutional neural network is used to extract visual features. The geometric features of the point cloud and the visual features of the images are fused to output a unified feature vector, which is then used to obtain the identification results of oversized transport vehicles through a fully connected neural network. The identified oversized transport vehicles are compared with the aforementioned L list to determine whether they are approved transport vehicles. If so, they are matched and associated; otherwise, they are judged as unapproved and illegal vehicles.

[0070] In a preferred embodiment of the present invention, step S3 is implemented as follows:

[0071] For fixed-point verification scenarios, when a vehicle passes through fixed monitoring points such as overload control stations, gantries, and toll stations, video, weighing, and lidar size data are integrated, and the core parameters of the vehicle and cargo are extracted through AI recognition models. These parameters are then compared with the approval and filing data to complete the compliance verification and output a vehicle compliance status label.

[0072] For drone-based blind spot verification scenarios, targeting remote road sections, fixed monitoring blind spots, or high-risk key monitored vehicles, based on historical monitoring data, current transportation task lists, and road network topology data, an improved version of the existing A* algorithm is used to achieve intelligent planning of drone patrol paths, ensuring optimal paths, comprehensive coverage, and controllable endurance. Drones are dispatched to conduct low-altitude cruise scanning and fixed-point hovering monitoring, acquiring comprehensive 3D point cloud data and high-definition images of vehicles and cargo, completing parameter extraction and compliance verification, and filling coverage gaps in fixed monitoring.

[0073] Compliance verification consists of two core stages: parameter verification and cargo matching. Parameter verification uses a direct comparison method, while cargo matching uses a dedicated neural network comparison algorithm to ensure verification accuracy. Specific technical details are as follows:

[0074] The parameter verification method uses a difference comparison mode between "real-time parameters and approval parameters":

[0075] First, determine the approval baseline parameters; retrieve the approval parameters for the corresponding transportation task from the approval data storage sub-unit, including baseline information data such as vehicle rated length, width and height, cargo rated length, width and height, permitted total load, and axle load limit;

[0076] Next, based on the vehicle's real-time parameters and the approval benchmark parameters, the parameter deviation rate is calculated; the formula is as follows:

[0077] ;

[0078] In the formula, For parameter deviation rate, For real-time vehicle parameters, These are the approval benchmark parameters;

[0079] Then, a differential deviation threshold is set to adapt to the characteristics of large-item transportation. The parameter deviation rate is compared with the differential deviation threshold to determine whether the parameter is in violation.

[0080] For example, for length, width, and height: if the parameter deviation rate is ≤5%, it is considered compliant; if the parameter deviation rate is >5%, it is considered a parameter violation.

[0081] Total load capacity: If the parameter deviation rate is ≤3%, it is considered compliant; if the parameter deviation rate is >3%, it is considered a parameter violation.

[0082] Furthermore, images from the approval data are compared with data acquired from fixed locations or drones. A dedicated cargo comparison neural network is constructed, and through feature vector cosine similarity calculation, accurate matching of approved and registered cargo with actual transported cargo is achieved, identifying hidden violations such as unauthorized cargo replacement en route. The specific algorithm implementation is as follows:

[0083] First, a cargo comparison neural network is constructed using a dual-branch structure of CNN and Transformer, and then trained and optimized. The CNN branch is used to extract shallow visual features (such as color, texture, and contour) of cargo images, while the Transformer branch is used to extract deep semantic features (such as cargo shape, structure, and surface markings). The training and optimization process is as follows: using image data of common large-item transportation cargo as the training set, cargo features are manually labeled, the Adam optimizer is used, the number of training iterations is ≥100 rounds, and after the accuracy of the validation set is ≥97%, the training weights are saved for real-time comparison.

[0084] Next, real-time cargo images (captured from fixed locations or by drones) and images of cargo undergoing approval and registration (uploaded during the approval process) are input into the trained cargo comparison neural network, which outputs two sets of feature vectors; the feature vector generated from the images of cargo undergoing approval and registration is denoted as... The feature vector generated from the real-time cargo image is denoted as . ;

[0085] Then, the cosine similarity algorithm is used to calculate the similarity between the two sets of 1024-dimensional feature vectors to measure the consistency of the goods. The calculation formula is as follows:

[0086] ;

[0087] In the formula, These represent the lengths of the two feature vectors respectively; similarity is the cosine similarity, with a value range of [0,1]. The closer the value is to 1, the more similar the two sets of goods features are, and the higher the consistency of the goods.

[0088] Finally, the compliance status of the goods is determined based on the similarity between the two sets of feature vectors; for example, the violation criteria for determining whether the goods are compliant or non-compliant are as follows:

[0089] Similarity ≥ 0.85: The goods are considered to have been matched successfully, and no violations such as changing goods midway have occurred;

[0090] 0.7 ≤ similarity < 0.85: This is considered a suspected violation, an early warning message is generated and pushed to the regulatory platform for manual review.

[0091] Similarity < 0.7: The goods are deemed to be in violation of regulations, specifically due to unauthorized replacement of goods midway through the journey. A violation label is issued, and subsequent tiered warnings and law enforcement coordination are triggered simultaneously.

[0092] In a preferred embodiment of the present invention, step S4 is implemented as follows:

[0093] First, an adaptive electronic fence is generated based on the approved route. Specifically, using the approved delivery route of oversized transport vehicles as a benchmark, and integrating fixed monitoring unit point data and road network topology data, a dual-layer adaptive electronic fence of "core control area + dynamic early warning area" is generated to ensure that the fence is accurately adapted to the actual road network and monitoring points. Specific technical details are as follows:

[0094] S41. Data Retrieval: Retrieve approved delivery routes containing baseline information; baseline information includes route start point, end point, road segments, permitted driving time periods, and speed limits for road segments, in WGS-84 latitude and longitude coordinate sequence (sampling interval ≤ 100m); retrieve latitude and longitude coordinates and monitoring coverage data of fixed monitoring units (such as overload control stations, gantries, toll stations, etc.) to construct a fixed monitoring point coordinate database to ensure accurate association between fences and monitoring points;

[0095] S42. Route standardization: The latitude and longitude coordinate sequence of the approved route is thinned out, key coordinate points such as route turning points and intersections are retained, redundant coordinates are eliminated to reduce the amount of calculation, while ensuring that the route outline remains unchanged, resulting in a standardized approved route.

[0096] S43. Electronic Fence Generation: Based on the standardized approval route, an electronic fence is generated. The electronic fence covers the area that the oversized transport vehicle must travel in. Centered on the approval route, a Gaussian buffer zone with a preset range on one side (e.g., within 50m, which can be dynamically adjusted according to the road network risk level and vehicle compliance status) is generated as the core control area to ensure that the fence covers the core passage area of ​​the approval route.

[0097] S44. Fence and monitoring point association: Spatially match the coordinate library of fixed monitoring points with the electronic fence, mark the fixed monitoring points inside and outside the fence, and clarify that the monitoring points inside the fence are used for routine verification, and the monitoring points outside the fence are used for route deviation judgment.

[0098] S45. Fence Update Mechanism: Receives real-time information on approved route adjustments, road network construction, and changes in fixed monitoring points. Using an incremental update algorithm, the electronic fence parameters are verified and updated every 30 minutes to ensure consistency between the fence and the actual road network and monitoring conditions. This avoids misjudgments caused by changes in the road network and results in an adaptive electronic fence.

[0099] Then, the violations of transportation routes are verified, specifically including:

[0100] 1. Route deviation detection (core violation):

[0101] Detection algorithm: The algorithm adopts the "ray method + shortest distance method" and combines the electronic fence boundary to determine the spatial relationship between the vehicle's current position and the fence in real time; at the same time, it is associated with the data of fixed monitoring units. If the vehicle is detected by a fixed monitoring unit outside the fence (and there is no route adjustment approval), it is directly judged as deviating from the route.

[0102] Deviation Degree Judgment: Calculate the shortest distance between the vehicle's current position and the electronic fence boundary, and classify it into two levels: slight deviation (distance ≤ 100m) and serious deviation (distance > 100m). Among them, serious deviation is judged as a major violation, and slight deviation is judged as a minor violation, triggering different levels of warnings.

[0103] 2. Overspeed detection:

[0104] The vehicle's real-time speed collected by the vehicle terminal unit is compared in real time with the speed limit of the road segment corresponding to the approved route (retrieving approval data) and the actual speed limit of the current road segment (road network data). The shortest travel time is calculated by combining the monitoring unit location information of the vehicle identified in the previous two instances. If the actual time of the vehicle arriving at the monitoring unit is more than 10% earlier than the maximum speed limit, it is judged as suspected speeding, and the real-time speed is further compared with the speed limit standard.

[0105] Speeding criteria: If the real-time speed is 10% greater than the speed limit of the road section but not more than 20% greater than the speed limit of the road section, it is judged as a minor speeding; if the real-time speed is 20% greater than the speed limit of the road section, it is judged as a serious speeding; the duration of speeding is recorded simultaneously, and if it lasts for more than 30 seconds, the warning level is upgraded.

[0106] 3. Unauthorized parking for inspection (routine violation):

[0107] The system employs "LSTM time series prediction + isolated forest anomaly detection" to analyze real-time speed and position changes of vehicles. If a vehicle's speed is ≤5km / h and its position remains unchanged (latitude and longitude deviation ≤5m) for more than 10 minutes, it is judged as illegal parking.

[0108] Exclusion criteria: If the vehicle stops at permitted stops (service areas, maintenance points) along the approved route and the stopping time does not exceed the approved time limit, it will not be considered a violation; if the stopping location exceeds the permitted range or time limit, it will be considered a violation.

[0109] In a preferred embodiment of the present invention, step S5 is implemented as follows:

[0110] 1. Early warning levels:

[0111] Level 1 Warning (General Violation): Minor deviation from the route, minor speeding, illegal parking (lasting 10-30 minutes); push regular warning information to the monitoring platform and vehicle terminal to remind the driver to correct and check the cargo status in time, without triggering law enforcement linkage;

[0112] Level 2 Warning (Serious Violations): Serious speeding, illegal parking (lasting more than 30 minutes), suspected cargo violations (cargo feature similarity 0.7 ≤ similarity < 0.85); push emergency warning information to the regulatory platform and execution terminal, simultaneously mark vehicle location, upload evidence of cargo violations (real-time images, feature comparison reports), and notify nearby law enforcement personnel to prepare for interception;

[0113] Level 3 Warning (Major Violation): Serious deviation from the route (distance from the core control area boundary > 100m) or clear violation involving cargo (similarity of cargo characteristics < 0.7, i.e. unauthorized replacement of cargo midway) will trigger the highest level warning information; the execution terminal will be immediately linked to push the vehicle's real-time location and evidence of violation (trajectory data, monitoring images, cargo comparison records, etc.), dispatch law enforcement personnel to carry out precise interception, and suspend the vehicle's subsequent approval authority.

[0114] 2. Retention of evidence of violations: All violations will be subject to a complete chain of evidence, including real-time vehicle trajectory data, fixed monitoring / drone monitoring images, speed records, location prediction results, and violation judgment logs. The evidence data will be retained for ≥1 year for subsequent law enforcement investigations and liability determination.

[0115] 3. Closed-loop supervision: After a violation warning is triggered, the results of law enforcement personnel's handling (such as interception and investigation, ordering rectification) are simultaneously transmitted back to the system, updating the vehicle's violation record and compliance status label, and feeding back to the approval system, forming a closed-loop supervision route of prediction-detection-warning-handling-feedback.

[0116] In summary, the method provided by the embodiments of the present invention has the following beneficial technical effects compared with the prior art:

[0117] I. The method provided in this invention can achieve full-segment coverage without blind spots, significantly improving recognition accuracy and automation level. Specifically, this invention, through a multi-source data fusion monitoring system, completely solves the problem of insufficient coverage by traditional fixed equipment, achieving full-segment and full-process verification coverage for oversized cargo transportation. By integrating multi-source data with oversized cargo vehicle recognition algorithms, it achieves automated and high-precision extraction of vehicle parameters and cargo characteristics, completely replacing the traditional manual verification mode. It can automatically identify hidden violations such as mid-journey reloading and parameter discrepancies, filling the regulatory gaps in the existing technology.

[0118] Second, the embodiments of the present invention can achieve differentiated and precise supervision by constructing a dynamic adaptive control system. Specifically, the embodiments of the present invention replace the traditional static fence with an adaptive electronic fence with a core control area and a dynamic early warning area, thereby achieving early warning and dynamic control. Combined with the vehicle compliance status, it realizes graded early warning and differentiated control, minimizing intervention for compliant vehicles, avoiding ineffective interception and traffic congestion, and strengthening full-process control for non-compliant vehicles. It achieves the goal of low-risk, low-intervention and high-risk, high-control, completely solving the core pain point of traditional one-size-fits-all supervision, and taking into account both supervision safety and transportation efficiency.

[0119] Third, the method provided by the embodiments of the present invention has strong adaptability, low implementation cost, and extremely high practicality and replicability. Specifically, the method provided by the embodiments of the present invention can directly connect to the hardware and data resources of existing overload control stations, gantries, and approval systems without large-scale replacement of existing equipment. It has low implementation cost, extremely high replicability and promotion value, and can effectively serve the efficient and safe circulation of major strategic production factors in the region.

[0120] It should be noted that each of the above modules can be implemented as a computer program, which can run on a computer device. The computer device's memory can store the computer program that makes up each module, enabling the processor to execute each step of the above method.

[0121] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0122] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.

[0123] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for dynamic monitoring of the en route status of heavy-duty transport vehicles, characterized in that, Includes the following steps: The system acquires approval data and simultaneously collects multi-source real-time monitoring data from fixed monitoring units, drone monitoring units, and vehicle-mounted terminal units. This data is then pre-processed to form a full-process data archive indexed by vehicle identification. Based on the full-process data archive, the algorithm that integrates 3D target detection and visual target detection is used to identify oversized transport vehicles, and the identification results are compared and matched with the approval data to obtain the matched information of vehicles on the way. Based on the information of the vehicles in transit, real-time vehicle parameters and cargo images are obtained at fixed monitoring points or through drones to fill blind spots. The real-time parameters are verified with the approval data, and the real-time cargo images are compared with the approved and filed cargo images through a preset cargo comparison neural network to generate the cargo compliance status. Based on the information of vehicles on the road, an adaptive electronic fence is generated according to the approved route, and the deviation from the route, speeding and illegal parking are detected by combining the real-time trajectory data of the vehicles, and the route compliance status is generated. Early warning information is generated based on the compliance status of the goods and the route, and the early warning information and vehicle location are pushed to the execution terminal to form a regulatory closed loop.

2. The method for dynamic monitoring of the en route status of oversized transport vehicles according to claim 1, characterized in that, The approval data includes a list L of approved large-item transportation tasks awaiting delivery; L = {l1, l2, ..., l...} n } indicates that there are a total of n transportation tasks, where l i Let l represent the i-th transportation task in L; i =[Task ID, License Plate Number, Vehicle Parameters, Cargo Parameters, Transportation Route, Transportation Time Limit, Vehicle Image].

3. The method for dynamic monitoring of the en route status of oversized transport vehicles according to claim 1, characterized in that, The fixed monitoring unit is used to monitor one or more of the following: vehicle passing video data, lidar size data, non-stop weighing data, passing timestamp, and location latitude and longitude data. The UAV monitoring unit is used to supplement the monitoring of one or more of the following: all-round high-definition aerial images of the vehicle, three-dimensional point cloud data, latitude and longitude coordinates and timestamp data of the monitoring point; The vehicle-mounted terminal unit is used to collect one or more of the following: real-time vehicle trajectory data, driving speed, heading angle, and timestamp data.

4. The method for dynamic monitoring of the en route status of oversized transport vehicles according to claim 1, characterized in that, The preprocessing methods include one or more of the following: normalization, deduplication and cleaning, noise suppression, and image enhancement.

5. The method for dynamic monitoring of the en route status of oversized transport vehicles according to claim 3, characterized in that, The steps for identifying oversized transport vehicles using an algorithm that integrates 3D object detection and visual object detection specifically include: The system takes images and 3D point cloud data acquired by fixed monitoring units and UAV monitoring units as inputs. For 3D point cloud data, PointRCNN is used to extract 3D targets and output point cloud geometric size features. For images, deep convolutional neural networks are used to extract image visual features. The point cloud geometric features and image visual features are fused to output a unified feature vector, and the recognition result of heavy transport vehicles is obtained through a fully connected neural network.

6. The method for dynamic monitoring of the en route status of oversized transport vehicles according to claim 1, characterized in that, The parameter verification method uses a difference comparison mode: Determine the approval benchmark parameters; The parameter deviation rate is calculated based on the vehicle's real-time parameters and the approval benchmark parameters. Set a differential deviation threshold, compare the parameter deviation rate with the differential deviation threshold, and determine whether the parameter violates the rules.

7. The method for dynamic monitoring of the en route status of oversized transport vehicles according to claim 1, characterized in that, The steps of comparing real-time cargo images with approved and registered cargo images using a pre-set cargo comparison neural network specifically include: A cargo comparison neural network is constructed using a dual-branch structure of CNN and Transformer, and then trained and optimized. The CNN branch is used to extract shallow visual features of cargo images, while the Transformer branch is used to extract deep semantic features. The real-time cargo image and the approved and registered cargo image are respectively input into the trained cargo comparison neural network, and two sets of feature vectors are output. The cosine similarity algorithm is used to calculate the similarity between two sets of feature vectors; The compliance status of goods is determined based on the similarity between the two sets of feature vectors.

8. The method for dynamic monitoring of the en route status of oversized transport vehicles according to claim 7, characterized in that, The steps for generating an adaptive electronic fence based on the approved route include: The system retrieves and approves delivery routes containing baseline information, including the route's origin, destination, route segments, permitted driving time periods, and speed limits for each segment. Retrieve the latitude and longitude coordinates and monitoring coverage data of the fixed monitoring units to construct a coordinate library of fixed monitoring points; The latitude and longitude coordinate sequence of the approval route is thinned out, retaining the key coordinates of the route turning points and intersections, and eliminating redundant coordinates to obtain the standardized approval route. An electronic fence is generated based on the standardized approval route. The electronic fence covers the area that the heavy transport vehicle must travel in. A Gaussian buffer zone with a preset range on one side is generated as the core control area, with the approval route as the center. Spatially match the coordinates of fixed monitoring points with the electronic fence, mark the fixed monitoring points inside and outside the fence, and clarify that the monitoring points inside the fence are used for routine verification, while the monitoring points outside the fence are used for route deviation judgment. The system receives real-time information on approved route adjustments, road network construction, and changes in fixed monitoring points. Through an incremental update algorithm, it verifies and updates the electronic fence parameters to obtain an adaptive electronic fence.

9. A dynamic monitoring system for the en route status of heavy-duty transport vehicles, used to implement the dynamic monitoring method for the en route status of heavy-duty transport vehicles as described in any one of claims 1-8, characterized in that, include: The data layer is used to acquire approval data and simultaneously collect multi-source real-time monitoring data from fixed monitoring units, drone monitoring units, and vehicle-mounted terminal units. It performs preprocessing to form a full-process data archive indexed by vehicle identification. The on-the-way vehicle identification and matching module is used to identify oversized transport vehicles based on the full-process data archive by using an algorithm that integrates 3D target detection and visual target detection, and compares and matches the identification results with the approval data to obtain the matched on-the-way vehicle information. The cargo compliance verification module is used to acquire real-time vehicle parameters and cargo images at fixed monitoring points or through drones based on the information of the vehicles in transit. The real-time parameters are verified against the approval data, and the real-time cargo images are compared with the approved and filed cargo images through a preset cargo comparison neural network to generate the cargo compliance status. The route compliance detection module is used to generate an adaptive electronic fence based on the information of the vehicles on the road and the approved route, and to detect deviation from the route, speeding and illegal parking by combining the real-time trajectory data of the vehicles, and generate a route compliance status. The violation warning module is used to generate warning information based on the compliance status of the goods and the route, and push the warning information and vehicle location to the execution terminal to form a regulatory closed loop.

Images