Air monitoring method, device and equipment for vehicle pressure speed driving

Abstract

The invention relates to the technical field of intelligent traffic management, in particular to an air monitoring method, device and equipment for vehicle pressure-speed driving, and the method comprises the steps: carrying out the panoramic road condition perception and lane-level target tracking of a video stream, recognizing a road region, a lane line and a target vehicle through a deep learning model, and carrying out the recognition of the target vehicle; generating structured state data including the vehicle ID, the driving track and the lane attribution; constructing a vehicle microdynamics model, and calculating key discrimination parameters in real time; the calculated key judgment parameters are input into a preset pressure speed behavior judgment model for logical reasoning, the model comprises a single-lane pressure speed sub-model and a parallel pressure speed sub-model, and whether illegal pressure speed behaviors exist or not is recognized by jointly judging the forward clearance distance, the driving speed, the parallel duration and the rear congestion degree; according to the invention, closed-loop management from passive supervision to active intervention can be realized, and the road traffic efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic management technology, specifically to an aerial monitoring method, device, and equipment for vehicle speed control. Background Technology

[0002] With economic development and the continuous expansion of highway and urban expressway networks, traffic flow density has increased significantly, making the optimization of traffic efficiency a core requirement of intelligent traffic management. However, in actual operation, the behavior of vehicles driving at a snail's pace or slow speeds has evolved into a major safety hazard that induces traffic congestion, reduces lane turnover, and leads to serious rear-end collisions. Unlike traditional speeding violations, slow speeding is highly concealed and dynamic, often occurring under conditions of good road conditions and clear visibility, artificially creating "moving roadblocks."

[0003] Existing monitoring methods based on fixed equipment have significant limitations in addressing such issues:

[0004] First, there is a lack of large-scale contextual awareness. Traditional fixed video checkpoints or fixed-point speed radar can only capture the instantaneous speed of a vehicle passing by, which is a "point-like" perception. Because it cannot obtain road condition information over a longer distance in front of the target vehicle, the system has difficulty distinguishing whether the vehicle's "passive deceleration" is due to an accident ahead or traffic saturation, or "active speed reduction" due to subjective factors. Without the ability to determine forward clearance distance, it is highly susceptible to misjudgments in law enforcement.

[0005] Secondly, there is a serious lack of real-time intervention mechanisms. Currently, the supervision of slow-moving behavior relies mainly on post-event non-on-site penalties, lacking an immediate warning mechanism for ongoing traffic obstruction. Once a congestion cluster forms behind slow-moving vehicles, the resulting "traffic wave effect" will cause a sharp drop in the overall road network efficiency, and simple post-event penalty points cannot recover the lost travel time and safety margin.

[0006] Finally, constructing a dynamic chain of evidence is extremely difficult. Determining speed-delaying behavior requires rigorous proof of temporal and spatial continuity, especially in scenarios involving multiple vehicles blocking lanes side-by-side, necessitating coordinated observation data across lanes. Single-dimensional captured images are insufficient to reconstruct the complex longitudinal overlap relationships between vehicles, leading to inadequate evidence for law enforcement.

[0007] Therefore, there is an urgent need for an aerial monitoring solution that can break through the limitations of fixed perspectives, possess long-distance spatiotemporal correlation perception capabilities, and achieve closed-loop management of "identification-early warning-advice-evidence collection" to ensure the traffic efficiency and safety of highways and urban expressways. Summary of the Invention

[0008] To improve traffic efficiency on highways and urban expressways, this invention proposes an aerial monitoring method for vehicle speed control, which specifically includes the following steps:

[0009] The aerial monitoring data collected by the aerial monitoring component is acquired and preprocessed with spatiotemporal alignment and coordinate mapping. The aerial monitoring data includes road overhead video stream, camera attitude parameters and positioning data.

[0010] The system performs panoramic road condition perception and lane-level target tracking on the video stream, uses a deep learning model to identify road areas, lane lines and target vehicles, and generates structured status data including vehicle ID, driving trajectory and lane affiliation.

[0011] A vehicle micro-dynamics model is constructed, and key discrimination parameters are calculated in real time. The key discrimination parameters include at least the absolute speed and acceleration of the target vehicle, as well as the forward clearance distance relative to the nearest vehicle in the current lane and the longitudinal overlap with vehicles in adjacent lanes.

[0012] The calculated key discrimination parameters are input into the preset speed-restriction behavior judgment model for logical reasoning. The speed-restriction behavior judgment model includes a single-lane speed-restriction sub-model and a parallel speed-restriction sub-model. As long as either sub-model determines that speed-restriction exists, the vehicle is judged to have speed-restriction behavior. The two sub-models identify whether there is illegal speed-restriction behavior by jointly judging the forward clearance distance, driving speed, parallel duration and the degree of congestion behind.

[0013] It performs real-time intervention and closed-loop management in coordination with air and ground. When it is determined to be a speed-reducing behavior, it automatically generates control commands to dispatch monitoring equipment to provide voice guidance and simultaneously triggers the intelligent evidence collection process.

[0014] The system generates a structured package of evidence of violations and sends it back to the law enforcement platform. It automatically captures short videos and keyframes containing the entire violation process, overlays dynamic annotation information to generate the evidence package, and uploads it in real time.

[0015] This invention also proposes an aerial monitoring device for vehicle speed control, the device being used in an aerial monitoring method for vehicle speed control on highways and urban expressways, comprising:

[0016] The data acquisition and preprocessing module is used to acquire aerial monitoring data collected by the aerial monitoring component and perform spatiotemporal alignment and coordinate mapping preprocessing.

[0017] The panoramic perception and tracking module is used to perform panoramic road condition perception and lane-level target tracking on the video stream, and generate structured status data including vehicle ID, driving trajectory and lane affiliation.

[0018] The dynamic parameter calculation module is used to construct a vehicle micro-dynamic model and calculate key discrimination parameters in real time. The key discrimination parameters include at least the absolute speed and acceleration of the target vehicle, as well as the forward clearance distance relative to the nearest vehicle in the current lane and the longitudinal overlap relative to vehicles in adjacent lanes.

[0019] The pressure speed behavior determination module is used to input the calculated key discrimination parameters into the preset pressure speed behavior determination model for logical reasoning to identify whether there is any illegal pressure speed behavior.

[0020] The air-ground collaborative intervention module is used to perform real-time intervention and closed-loop management. When it is determined to be a speed-reducing behavior, it automatically generates control commands to dispatch monitoring equipment for voice persuasion and simultaneously triggers the intelligent evidence collection process.

[0021] The evidence generation and transmission module is used to generate structured packages of evidence of illegality and transmit them back to the law enforcement platform.

[0022] This invention also proposes an aerial monitoring device for vehicle speed control, comprising a receiver, a transmitter, a memory, a processor, and an aerial monitoring component. The aerial monitoring component includes a high-definition camera mounted on a drone platform or fixed bracket, a directional loudspeaker, and a high-precision positioning module, wherein:

[0023] The receiver is used to receive video streams and sensor data from the airborne monitoring component, as well as control commands from the ground command center;

[0024] The transmitter is used to send flight or gimbal control commands to the airborne monitoring component and to transmit a structured package of evidence of violations back to the ground command center.

[0025] The memory stores computer-executed instructions;

[0026] The processor executes computer execution instructions stored in the memory to implement an aerial monitoring method for vehicle speed control.

[0027] The present invention also proposes a computer storage medium, characterized in that the computer storage medium stores computer execution instructions, which, when executed by a processor, are used to implement an aerial monitoring method for vehicle speed control.

[0028] The present invention also proposes a computer program product, characterized in that it includes a computer program, which, when executed by a processor, is used to implement an aerial monitoring method for vehicle speed control.

[0029] This invention acquires video streams and location data containing forward road conditions through an aerial monitoring component and performs coordinate mapping preprocessing; it extracts vehicle driving status using deep learning and multi-target tracking technology and calculates the forward clearance distance of the target vehicle relative to the vehicle in front in its lane; based on the forward clearance distance, driving speed, and parallel duration, it determines speed-restriction behavior through a pre-set single-lane speed-restriction sub-model and a parallel speed-restriction sub-model; after identifying speed-restriction violations, it coordinates with drones or loudspeakers to perform directional voice persuasion, and simultaneously captures and transmits a structured evidence package containing panoramic road conditions, license plate close-ups, and dynamic parameter annotations in real time. This achieves accurate identification, immediate intervention, and closed-loop evidence collection of "mobile roadblocks" on highways and urban expressways, significantly improving the traffic efficiency and regulatory automation level of highways and urban expressways. Attached Figure Description

[0030] Figure 1 A schematic flowchart of an aerial monitoring method for vehicle speed control provided in an embodiment of this application;

[0031] Figure 2 A schematic diagram of the structure of an aerial monitoring device for vehicle speed control provided in an embodiment of this application;

[0032] Figure 3 This is a schematic diagram of an aerial device structure for vehicle speed control provided in an embodiment of this application. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] This invention proposes an aerial monitoring method for vehicle speed control, which specifically includes the following steps:

[0035] The aerial monitoring data collected by the aerial monitoring component is acquired and preprocessed with spatiotemporal alignment and coordinate mapping. The aerial monitoring data includes road overhead video stream, camera attitude parameters and positioning data.

[0036] The system performs panoramic road condition perception and lane-level target tracking on the video stream, uses a deep learning model to identify road areas, lane lines and target vehicles, and generates structured status data including vehicle ID, driving trajectory and lane affiliation.

[0037] A vehicle micro-dynamics model is constructed, and key discrimination parameters are calculated in real time. The key discrimination parameters include at least the absolute speed and acceleration of the target vehicle, as well as the forward clearance distance relative to the nearest vehicle in the current lane and the longitudinal overlap with vehicles in adjacent lanes.

[0038] The calculated key discrimination parameters are input into a pre-set speed-restriction behavior judgment model for logical reasoning. The speed-restriction behavior judgment model includes a single-lane speed-restriction sub-model and a parallel speed-restriction sub-model. If either sub-model determines that speed-restriction exists, the vehicle is judged to have speed-restriction behavior. The two sub-models identify whether there is illegal speed-restriction behavior by jointly judging the forward clearance distance, driving speed, parallel duration, and rear congestion level. Real-time intervention and closed-loop management are performed in air-ground coordination. When speed-restriction behavior is judged, control commands are automatically generated to dispatch monitoring equipment for voice persuasion and simultaneously trigger the intelligent evidence collection process.

[0039] The system generates a structured package of evidence of violations and sends it back to the law enforcement platform. It automatically captures short videos and keyframes containing the entire violation process, overlays dynamic annotation information to generate the evidence package, and uploads it in real time.

[0040] Figure 1 This is a schematic flowchart of an aerial monitoring method for vehicle speed control provided in an embodiment of this application. Figure 1 As shown in the figure, this embodiment provides an aerial monitoring method for vehicle speed control on highways and urban expressways, including:

[0041] To achieve the above objectives, this application adopts the following technical solution:

[0042] An aerial monitoring method for vehicle speed control on highways and urban expressways includes the following steps:

[0043] S1: Acquire aerial monitoring data collected by drones or elevated monitoring equipment and perform spatiotemporal benchmark preprocessing.

[0044] Specifically, it receives real-time road-view video streams and synchronous sensor telemetry data collected by drones or elevated monitoring equipment. Based on the high-precision positioning information of the acquisition equipment and the gimbal attitude parameters (such as pitch angle and yaw angle), it constructs a homography transformation matrix from the image pixel coordinate system to the road surface world coordinate system. The matrix is ​​used to extract the pixel coordinates of the video image frame by frame. Inverse projection mapping is used to obtain road surface geographic coordinates with physical metric significance. ,Right now:

[0045]

[0046] For the dynamic perspective of drones, the length of standard lane lines on the road can be used as a reference point for the matrix. Real-time adaptive calibration is performed to eliminate measurement errors caused by platform jitter and wide-angle perspective distortion, thus establishing a unified and robust spatial benchmark for subsequent accurate calculation of the vehicle's forward clearance distance.

[0047] S2: Perform target detection and multi-target tracking on the video stream, and extract vehicle driving trajectory, lane affiliation and motion state parameters.

[0048] This step aims to identify all vehicle targets on the road in real time and accurately from continuous video streams collected by drones or elevated monitoring equipment, track their movement trajectories, and determine the lane affiliation and movement status of each vehicle by combining lane line information.

[0049] To simultaneously acquire vehicle location and road structure information, a multi-task learning network is used to process each frame of video image.

[0050] (1) Vehicle target detection

[0051] Model selection: Improved real-time target detection algorithms such as YOLOv8 or RT-DETR are adopted. To address the large scale variation of vehicles from an aerial perspective (larger at close range, smaller at distance), a multi-scale feature fusion module (such as PANet or BiFPN) is introduced to enhance the recall rate of small, distant target vehicles.

[0052] Output information: For each detected vehicle target Output its bounding box. ,category (Cars, trucks, buses, etc.) and confidence level ,in For vehicle target The center pixel coordinates of the bounding box , For vehicle target The width and height of the bounding box.

[0053] (2) Lane line semantic segmentation

[0054] Model selection: Use lightweight semantic segmentation networks (such as BiSeNet or SegFormer) or dedicated lane detection networks (such as CLRNet).

[0055] Output information: Output lane line mask or lane line key point sequence. Through polynomial fitting, the pixel-level lane lines are transformed into a mathematical equation description. ,in Representing the In this embodiment, the output y represents the vertical coordinate of the k-th lane line at pixel position x in the input image. A function representing the mapping from the image pixel position to the longitudinal coordinates of the lane line.

[0056] Lane area division: Based on the detected lane lines, the road surface is divided into several lane areas, where the k-th lane area is represented as... (For example: first lane / fast lane, second lane / slow lane, emergency lane, etc.).

[0057] (3) Multi-target tracking and trajectory generation

[0058] To obtain the continuous motion state of the vehicle, it is necessary to correlate the detection results in different frames to form a complete driving trajectory.

[0059] Tracking algorithm: A tracking algorithm based on the fusion of motion features and appearance features (such as DeepSORT or ByteTrack) is adopted.

[0060] Motion matching: The Kalman filter is used to predict the vehicle's position in the current frame, and the intersection-over-union ratio (IoU) or Mahalanobis distance between the predicted box and the detected box is calculated.

[0061] Appearance matching: Extract vehicle appearance feature vectors (ReID Feature), calculate cosine similarity, and solve the problem of re-identification after the vehicle is occluded or temporarily disappears.

[0062] Trajectory smoothing:

[0063] To address trajectory noise caused by drone jitter, sliding window filtering or Gaussian process regression is used to analyze the original trajectory point sequence. Perform smoothing to obtain the smoothed trajectory. .

[0064] (4) Lane ownership determination

[0065] Determining which lane each vehicle is currently traveling in is a prerequisite for judging "lane occupation and speed control".

[0066] Geometric inclusion criteria:

[0067] To eliminate the projection offset caused by vehicle height, the contact point between the vehicle and the road surface is uniformly extracted as a reference to obtain the bottom center point of the vehicle detection frame. This point approximately represents the contact point between the vehicle and the road surface.

[0068] Judgment point Which lane area did it land in? Inside.

[0069] The formula means: If Then the vehicle is determined to belong to the first There are 10 lanes (assuming the lane line equations have been standardized to the image coordinate system).

[0070] Lane change behavior recognition:

[0071] Monitor the lateral distance between the vehicle's center point and the lane center line. The rate of change.

[0072] like If the speed continues to decrease, it is determined to be in "lane keeping" mode; if If the lane continues to widen and crosses the lane boundary, it is marked as "changing lanes," and a certain grace period is given when determining the speed limit.

[0073] (5) Extraction of motion state parameters

[0074] Based on the smoothed trajectory and timestamp, the vehicle's micro-motion parameters are calculated.

[0075] Pixel speed calculation:

[0076] Calculate the pixel displacement vector of the vehicle center point between adjacent frames. , This represents the pixel displacement of the vehicle's center point between adjacent frames.

[0077] Pixel speed ,in The frame interval.

[0078] (6) Physical state mapping

[0079] Using the homography matrix constructed in S1 , pixel speed Mapped to road surface physical velocity .

[0080] in This refers to the vehicle's speed (km / h) and acceleration. It is used to determine whether a vehicle is accelerating or decelerating.

[0081] Through the above steps, the system can output a structured vehicle status list in real time: {Vehicle_ID,Class,Position(x,y),Velocity,Lane_ID,Trajectory_History}, where Vehicle_ID represents the unique identifier of the vehicle, Class is the vehicle category label, indicating the vehicle type (such as sedan, truck, bus, motorcycle, etc.); Position(x,y) is the spatial coordinates of the vehicle, representing the pixel coordinates of the vehicle in the current frame image; Velocity represents the physical speed of the vehicle; Lane_ID represents the lane affiliation identifier, indicating the lane number the vehicle is currently in; and Trajectory_History represents the trajectory history sequence, i.e., the previously mentioned... In the process, a set of vehicle position points is recorded over several past frames. This list will serve as the basis for calculating clearance distance and determining pressure-speed behavior in subsequent steps S3 and S4.

[0082] S3: Using perspective transformation, calculate the forward clearance distance between the target vehicle and the nearest vehicle in front of it in the same lane, as well as the longitudinal overlap with vehicles in adjacent lanes.

[0083] This step uses inverse perspective mapping (IPM) to map the nonlinear image pixel space into a linear geophysical space, thereby accurately quantifying the spatiotemporal topological relationships between vehicles.

[0084] (1) Projection mapping of key feature points of vehicles

[0085] Extract target vehicle Bottom center point of the detection frame And the vehicle in front in the same lane bottom center point Then, the homography matrix is ​​used to convert the pixels into world coordinates:

[0086]

[0087]

[0088] in, Indicates the target vehicle World coordinates Indicates the target vehicle The car in front in the same lane World coordinates; An axis is defined as the longitudinal direction extending along the lane lines. The axis is the lateral direction perpendicular to the lane lines.

[0089] (2) Forward clearance distance

[0090] Calculating the forward clearance distance refers to the physical gap between a target vehicle and the vehicle directly in front of it, and is a core indicator for determining "good road conditions ahead". Target vehicle The car in front of it in the same lane Forward clearance distance The calculation formula is:

[0091]

[0092] in, This is the compensation value for the length of the vehicle itself. If ( If a preset forward safety clearance threshold is set (e.g., 200 meters), the system will mark the current target vehicle as having the conditions to accelerate. If there are no vehicles in front of the target vehicle within the detection area, it can be set... It is a large value that is greater than the safe following distance threshold.

[0093] (3) Longitudinal overlap of adjacent lanes

[0094] Calculating longitudinal overlap is used to quantify the target vehicle. Vehicles in adjacent lanes The degree of parallelism in the direction of travel is key to identifying "parallel speed reduction". The target vehicle is... The projection range on the axis is The projection range of vehicles in adjacent lanes is Overlap ratio is defined using a one-dimensional intersection ratio. :

[0095]

[0096] in, ,when At that time, it was determined that the two vehicles were in a highly parallel state, thus forming a substantial "moving barrier".

[0097] (4) Spatiotemporal parameter fusion and smoothing

[0098] Considering the detection fluctuations between video frames, a first-order hysteresis filter or Kalman filter is introduced. and The timing smoothing process can be represented as follows: ,in For weight parameters, This represents the smoothed result of the input data at time t. The input data is at time t-1. This smoothing operation ensures the stability of the subsequent pressure-speed determination logic.

[0099] S4: Construct a pressure speed behavior determination model, which includes a single-lane pressure speed sub-model and a parallel pressure speed sub-model.

[0100] Pressure-induced velocity is not an instantaneous state, but rather an action with spatial exclusivity and temporal continuity. The model introduces a time-sliding window. And a confidence accumulation operator to eliminate false alarms caused by brief avoidance or sensor noise.

[0101] (1) Single-lane pressure velocity sub-model

[0102] If the target vehicle meets the following conditions: the forward clearance distance is greater than the preset safety threshold, the driving speed is lower than the minimum speed limit of the road or lower than the average speed of the traffic flow by a certain percentage, and the continuous driving time in the fast lane exceeds the threshold, it is judged as single-lane speed control.

[0103] Define the single-lane pressure-speed state function If and only if the following criterion applies during the duration When Neiheng was established, it was deemed to be in violation of regulations:

[0104]

[0105] in, For high-speed lanes, The identifier for the lane currently occupied by the target vehicle; The preset forward safety clearance threshold (e.g., 200m) indicates that there are no obstacles ahead for a long distance. Let be the instantaneous speed of the target vehicle at time t. The legally mandated minimum speed limit for roads, The average flow rate of traffic along the entire line. The proportionality factor for the average flow speed of the entire line ( Typically, a value of 0.7 to 0.8 is used to trigger a judgment when the flow rate is significantly lower than the speed limit.

[0106] (2) Parallel pressure-velocity sub-model

[0107] If a target vehicle meets the following conditions: the forward clearance distance is greater than the preset safety threshold, and the duration of longitudinal overlap with vehicles in adjacent lanes exceeds the parallel threshold, causing queuing congestion behind it, it is judged as parallel speed suppression, forming a substantial "mobile roadblock" and blocking the flow of traffic behind it.

[0108] For vehicles in adjacent lanes Its parallel pressure-velocity state function Must meet:

[0109]

[0110] Furthermore, it needs to be coupled with subsequent congestion feedback conditions:

[0111]

[0112] in, For vertical overlap, The threshold for longitudinal overlap is set to 0.8 in this embodiment. When the longitudinal overlap exceeds the set threshold, it means that the two vehicles are physically side by side. : Indicates the vehicles in adjacent lanes The absolute value of the difference in driving speed. The threshold for the vehicle speed in adjacent lanes (preferably set to 5 km / h in this embodiment). This indicates that the relative speed difference between the two vehicles is extremely small; For the rear congestion factor, The threshold for the rear congestion factor is used in this embodiment when the rear... The average actual following distance of a vehicle is much smaller than the standard safe distance. This proved that queue congestion had already occurred. This indicates the actual following distance of the k-th vehicle behind the target vehicle.

[0113] S5: When the behavior is determined to be slowing down, control the drone to fly over the target vehicle or adjust the focus of the overhead camera and play a warning voice through the directional loudspeaker;

[0114] To ensure that the warning voice can be clearly perceived by the driver of the target vehicle without interfering with the normal driving of adjacent lanes, the drone is controlled to fly to a preset offset position above and to the left rear of the target vehicle, and maintains speed synchronization with the target vehicle to obtain the real-time speed parameters of the target vehicle, calculate the sound pressure compensation gain, and use the nonlinear acoustic principle of the directional loudspeaker to modulate the voice signal onto a high-frequency ultrasonic carrier. The nonlinear effect of air is used to generate a highly directional audible sound beam to ensure that the sound pressure level reaches the effective perception range at the target vehicle window.

[0115] Based on the speed type, a targeted voice message is played, such as "XA·XXXXX, the road ahead is clear, please accelerate" or "Please do not drive side-by-side for an extended period, please change lanes to the right." Within a preset time window (e.g., 30 seconds) after the voice message is issued, the speed change or lane-changing intention of the target vehicle is continuously monitored. If the target vehicle accelerates or changes lanes, it is marked as "persuasion successful"; otherwise, it is marked as "persuasion ineffective" and penalty evidence is generated.

[0116] S6: Simultaneously generate an evidence package containing license plate close-ups, panoramic road conditions, and dynamic annotations (speed, distance), and upload it to the traffic control platform via 5G / 4G network.

[0117] The system employs a circular buffer storage mechanism. When a violation determination signal is output in step S4, the video playback function is automatically triggered. The system retrieves the video from the buffer up to the time preceding the violation trigger. After the judgment is completed in seconds seconds (total duration) The original video stream (30-60 seconds) is used to ensure that the video includes the entire process of the violation, its occurrence, and the feedback after intervention.

[0118] Using the parameters calculated in step S3, augmented reality annotations are rendered frame by frame in the edited video stream, such as target bounding boxes, real-time velocity values ​​(km / h), instantaneous acceleration, forward clearance values, and duration of violations.

[0119] The system then automatically filters and extracts three types of key feature frames from the video stream: recognition frames, which capture the clearest moment of license plate information during the drone's dive or camera zoom, and use OCR technology to extract the license plate number; environmental frames, which capture panoramic images including road speed limit signs, geographical reference objects (such as milestones), and the sparse state of surrounding vehicles to prove the "good road conditions" of the external environment; and intervention frames, which record the spatial positional relationship diagram when the drone performs voice broadcasts to prove the compliance of the law enforcement process.

[0120] Finally, the system containerizes and encapsulates the aforementioned media files and metadata, establishes a high-priority communication channel, and ensures that evidence packages containing key violation information can still be uploaded to the traffic command center platform in real time and completely even in a network congestion environment.

[0121] This application provides an aerial monitoring method for vehicle speed control on highways and urban expressways. It acquires raw video streams and pose data from aerial monitoring components and performs spatiotemporal alignment preprocessing. The video streams are then subjected to panoramic perception and multi-target tracking to extract vehicle trajectory, lane affiliation, and motion state parameters. Based on inverse perspective transformation, the forward clearance distance and longitudinal overlap of the target vehicle are calculated in real time. This leads to the construction of a judgment model that includes single-lane speed control and parallel speed control sub-models, accurately identifying violations and triggering air-to-ground collaborative intervention and the transmission of dynamically labeled evidence packages. This invention not only solves the problem of ranging failure during UAV altitude and zoom flight by dynamically updating the scale by extracting known physical size features such as lane lines in real time, but also decouples the perspective view into mutually perpendicular physical criteria using IPM transformation. This achieves non-interference between single-lane and parallel speed control judgments, significantly improving the algorithm's evaluation accuracy, real-time performance, and versatility in complex curves and dynamic perspective scenarios.

[0122] Figure 2 This is a schematic diagram of the structure of an aerial monitoring device for vehicle speed control on highways and urban expressways, provided as an embodiment of this application. Figure 2As shown, the vehicle speed monitoring aerial device 200 for highways and urban expressways provided in this embodiment includes:

[0123] The data acquisition and preprocessing module 201 is used to acquire aerial monitoring data collected by UAVs or elevated monitoring equipment, and perform spatiotemporal alignment and coordinate mapping preprocessing. The aerial monitoring data includes road overhead video streams, camera attitude parameters and positioning data.

[0124] The panoramic perception and tracking module 202 is used to perform panoramic road condition perception and lane-level target tracking on the video stream. It uses a deep learning model to identify road areas, lane lines and vehicle targets, and generates structured status data containing vehicle ID, driving trajectory and lane affiliation.

[0125] The dynamic parameter calculation module 203 is used to construct a vehicle micro-dynamic model and calculate key discrimination parameters in real time.

[0126] The dynamic parameter calculation module 203 is specifically used to map the pixel coordinates of the vehicle's grounding point to the geographic plane based on the inverse perspective transformation matrix, and to calculate the absolute speed and acceleration of the target vehicle.

[0127] The dynamic parameter calculation module 203 is also used to retrieve the target vehicle in the target lane and calculate the forward clearance distance between the target vehicle and the nearest vehicle in front of the current lane.

[0128] The dynamic parameter calculation module 203 is also used to calculate the projection overlap range of the target vehicle and the vehicle in the adjacent lane in the longitudinal direction of the road, so as to obtain the longitudinal overlap degree.

[0129] The pressure speed behavior determination module 204 is used to input the calculated key discrimination parameters into the preset pressure speed behavior determination model for logical reasoning;

[0130] The speed reduction behavior determination module 204 includes a single-lane speed reduction sub-model, which is used to determine single-lane speed reduction behavior when the forward clearance distance is greater than a preset safety threshold and the driving speed is lower than the speed limit threshold.

[0131] The pressure speed behavior determination module 204 also includes a parallel pressure speed sub-model, which is used to determine parallel pressure speed behavior when the longitudinal overlap and parallel duration exceed the threshold and cause queuing congestion in the back.

[0132] The air-ground coordinated response module 205 is used to perform real-time intervention and closed-loop management.

[0133] The air-ground collaborative handling module 205 is used to automatically generate control commands to dispatch airborne directional loudspeakers for targeted voice persuasion when the judgment result is a violation.

[0134] The air-ground collaborative handling module 205 is also used to monitor the changes in the vehicle's movement status after the warning in real time and to trigger the intelligent evidence collection process simultaneously.

[0135] The evidence package generation module 206 is used to generate a structured illegal evidence package and send it back to the law enforcement platform. The module is configured to automatically capture short videos and key frames containing the entire process of the violation, and dynamically overlay annotation information such as speed scales, distance lines and timestamps on the video screen.

[0136] The vehicle speed monitoring device for highways and urban expressways provided in this embodiment can execute the vehicle speed monitoring method for highways and urban expressways provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0137] Figure 3 This is a schematic diagram of the structure of an aerial monitoring device for vehicle speed control on highways and urban expressways, provided as an embodiment of this application. Figure 3 As shown, the vehicle speed monitoring device 300 for highways and urban expressways provided in this application includes: a receiver 301, a transmitter 302, a processor 303, and a memory 304.

[0138] Receiver 301 is used to receive commands and data. Specifically, in implementation, it is used to receive road-view video streams collected by front-end perception components such as drones and elevated cameras, sensor telemetry data (such as GPS coordinates, altitude, and gimbal angle), and control commands from the ground command center.

[0139] Transmitter 302 is used to send commands and data. Specifically, in implementation, it is used to send control commands to aerial monitoring components (such as UAV flight control systems, gimbals, and directional loudspeakers), and to send generated structured violation evidence packages and real-time early warning signals to traffic control platforms.

[0140] Memory 304 is used to store computer-executed instructions;

[0141] The processor 303 is used to execute computer execution instructions stored in the memory 304 to implement the various steps of the vehicle speed monitoring method for highways and urban expressways in the above embodiments.

[0142] For details, please refer to the relevant descriptions in the aforementioned embodiments of the aerial monitoring method for vehicle speed control on highways and urban expressways, such as: performing spatiotemporal alignment and coordinate mapping of video streams, calculating the forward clearance distance and longitudinal overlap of the target vehicle, using the speed control judgment model to identify violations, generating air-ground collaborative intervention instructions, and synthesizing augmented reality evidence videos, etc.

[0143] Optionally, the memory 304 can be either standalone or integrated with the processor 303.

[0144] When the memory 304 is set up independently, the electronic device also includes a bus for connecting the memory 304 and the processor 303.

[0145] This application embodiment also provides a computer storage medium storing computer execution instructions. When the processor executes the computer execution instructions, it implements the multi-dimensional evaluation method for driving behavior of resource-constrained terminals as described above by the multi-dimensional evaluation device for driving behavior of resource-constrained terminals.

[0146] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An aerial monitoring method for vehicle speed control, characterized in that, Specifically, the following steps are included: The aerial monitoring data collected by the aerial monitoring component is acquired and preprocessed with spatiotemporal alignment and coordinate mapping. The aerial monitoring data includes road overhead video stream, camera attitude parameters and positioning data. The system performs panoramic road condition perception and lane-level target tracking on the video stream, uses a deep learning model to identify road areas, lane lines and target vehicles, and generates structured status data including vehicle ID, driving trajectory and lane affiliation. A vehicle micro-dynamics model is constructed, and key discrimination parameters are calculated in real time. The key discrimination parameters include at least the absolute speed and acceleration of the target vehicle, as well as the forward clearance distance relative to the nearest vehicle in the current lane and the longitudinal overlap with vehicles in adjacent lanes. The calculated key discrimination parameters are input into the preset speed-restriction behavior judgment model for logical reasoning. The speed-restriction behavior judgment model includes a single-lane speed-restriction sub-model and a parallel speed-restriction sub-model. As long as either sub-model determines that speed-restriction exists, the vehicle is judged to have speed-restriction behavior. The two sub-models identify whether there is illegal speed-restriction behavior by jointly judging the forward clearance distance, driving speed, parallel duration and the degree of congestion behind. It performs real-time intervention and closed-loop management in coordination with air and ground. When it is determined to be a speed-reducing behavior, it automatically generates control commands to dispatch monitoring equipment to provide voice guidance and simultaneously triggers the intelligent evidence collection process. The system generates a structured package of evidence of violations and sends it back to the law enforcement platform. It automatically captures short videos and keyframes containing the entire violation process, overlays dynamic annotation information to generate the evidence package, and uploads it in real time.

2. The aerial monitoring method for vehicle speed control according to claim 1, characterized in that, The spatiotemporal alignment and coordinate mapping preprocessing of aerial monitoring data specifically includes the following steps: Using real-time positioning data and gimbal angle data provided by the aerial monitoring component, a homography matrix from the world coordinate system to the image pixel coordinate system is constructed; For each frame of video image, the homography matrix is ​​used to map the image pixel coordinates to the road surface geographic coordinates, so as to eliminate the measurement error caused by the posture jitter of the acquisition device or the perspective effect, and to provide a unified spatial reference for the calculation of the forward clearance distance.

3. The aerial monitoring method for vehicle speed control according to claim 1, characterized in that, The real-time calculation of key discrimination parameters by constructing a vehicle microdynamics model includes the following steps: The geographic coordinate trajectory of the target vehicle is smoothed using Kalman filtering, and the vector velocity of the target vehicle is calculated. The system retrieves the nearest vehicle ahead of the target vehicle in its lane and calculates the Euclidean distance between the two vehicles as the forward clearance distance. If no vehicle is detected ahead of the target vehicle in the current lane within the monitoring field of view, the forward clearance distance is set to the maximum effective monitoring distance of the current field of view. Extract the projection intervals of the target vehicle and vehicles in adjacent lanes along the longitudinal direction of the road, calculate the intersection-to-overlap ratio or overlap length of the projection intervals of the two vehicles, and obtain the longitudinal overlap degree.

4. The aerial monitoring method for vehicle speed control according to claim 1, characterized in that, In the single-lane pressure-velocity sub-model, the single-lane pressure-velocity sub-model determines the existence of pressure velocity if the following conditions are met simultaneously: Condition 1: The forward clearance distance is greater than the preset long-distance clearance threshold; Condition 2: The driving speed is lower than the minimum speed limit of the road, or lower than a preset percentage of the current average speed of the traffic flow; Condition 3: The duration of conditions 1 and 2 exceeds the single-lane speed control time threshold; Condition 4: The adjacent lane on the right is detected to be empty, but the target vehicle does not change lanes.

5. The aerial monitoring method for vehicle speed control according to claim 1, characterized in that, In the parallel pressure-velocity sub-model, the parallel pressure-velocity sub-model determines the existence of pressure velocity if the following conditions are met simultaneously: Condition 1: The forward clearance distance between each lane of the two vehicles is greater than the preset long-distance clearance threshold; Condition 2: The speed difference between the two vehicles is less than the relative speed threshold, and the longitudinal position difference is less than the position buffer threshold; Condition 3: The duration of the parallel state exceeds the parallel time threshold; Condition 4: A vehicle queue signal is detected within a preset range behind the target vehicle.

6. The aerial monitoring method for vehicle speed control according to claim 1, characterized in that, The generation of structured evidence packages of illegal activities and their transmission back to the law enforcement platform includes: Automatically backtrack video streams of preset duration before and after the violation determination time, and perform intelligent editing; Using augmented reality technology, a red surround frame of the target vehicle, real-time speed value, a visual line connecting the forward clearance distance, and a timer for the duration of the violation are dynamically overlaid on the video footage. Automatically extract close-up frames containing clear license plate information and panoramic frames containing surrounding road signs; The video files, image files, and metadata including time, geographic coordinates, violation codes, average vehicle speed, and clearance distance are packaged and uploaded via the wireless network priority channel.

7. An aerial monitoring device for vehicle speed control, characterized in that, The device is used to implement the aerial monitoring method for vehicle speed control on highways and urban expressways as described in claim 1, comprising: The data acquisition and preprocessing module is used to acquire aerial monitoring data collected by the aerial monitoring component and perform spatiotemporal alignment and coordinate mapping preprocessing. The panoramic perception and tracking module is used to perform panoramic road condition perception and lane-level target tracking on the video stream, and generate structured status data including vehicle ID, driving trajectory and lane affiliation. The dynamic parameter calculation module is used to construct a vehicle micro-dynamic model and calculate key discrimination parameters in real time. The key discrimination parameters include at least the absolute speed and acceleration of the target vehicle, as well as the forward clearance distance relative to the nearest vehicle in the current lane and the longitudinal overlap relative to vehicles in adjacent lanes. The pressure speed behavior determination module is used to input the calculated key discrimination parameters into the preset pressure speed behavior determination model for logical reasoning to identify whether there is any illegal pressure speed behavior. The air-ground collaborative intervention module is used to perform real-time intervention and closed-loop management. When it is determined to be a speed-reducing behavior, it automatically generates control commands to dispatch monitoring equipment for voice persuasion and simultaneously triggers the intelligent evidence collection process. The evidence generation and transmission module is used to generate structured packages of evidence of illegality and transmit them back to the law enforcement platform.

8. An aerial monitoring device for vehicle speed control, characterized in that, The system includes a receiver, transmitter, memory, processor, and aerial monitoring components. The aerial monitoring components include a high-definition camera mounted on a drone platform or fixed bracket, a directional loudspeaker, and a high-precision positioning module. The receiver is used to receive video streams and sensor data from the airborne monitoring component, as well as control commands from the ground command center; The transmitter is used to send flight or gimbal control commands to the airborne monitoring component and to transmit a structured package of evidence of violations back to the ground command center. The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement an aerial monitoring method for vehicle speed control as described in any one of claims 1-6.

9. A computer storage medium, characterized in that, The computer storage medium stores computer execution instructions, which, when executed by a processor, are used to implement an aerial monitoring method for vehicle speed control as described in any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, is used to implement an aerial monitoring method for vehicle speed control as described in any one of claims 1-6.

Images