Intelligent college vehicle safety monitoring and control processing system

Abstract

The invention relates to the field of intelligent traffic management, and discloses an intelligent college vehicle safety monitoring and management and control processing system, and the system comprises a cloud control platform which generates a space-time passing authority based on a vehicle identity and dynamic reputation, and carries out the credit evaluation management of a vehicle; the edge computing node is used for executing fusion processing of multi-source sensing data, dynamic scheduling of sensing resources and real-time compliance monitoring of a vehicle driving track; the roadside sensing and executing layer is used for collecting real-time state data of a road environment and responding to the control instruction to execute a physical control action; and the cloud control platform, the edge computing node and the roadside sensing and execution layer establish data connection through a network. Differentiated resource allocation is realized by combining a reputation and environment dynamic scheduling sensor mode; while key target supervision is ensured, the computing power and bandwidth bottlenecks caused by all-time monitoring are effectively relieved, and the computing efficiency and the operation stability of the system are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent traffic management, in particular to a smart college vehicle safety monitoring and management processing system. BACKGROUND

[0002] With the improvement of the opening degree of colleges and universities, the number of social vehicles entering the campus is increasing, and the campus traffic environment presents the characteristics of mixed walking and driving and complex road conditions. At present, the management of college vehicles mainly relies on the gate control system of the entrance and exit and the video monitoring network inside the campus. The conventional technical means include using license plate recognition technology to verify the identity of the entering vehicle and release it, and setting up fixed-point monitoring cameras and physical speed reduction facilities on major road sections to regulate vehicle driving. Some colleges have established a digital security management platform, which records the entry and exit time of vehicles and access control records through the background, and cooperates with the daily patrol of offline security personnel to maintain the traffic order and safety inside the campus.

[0003] However, the existing vehicle management method still has deficiencies in the fineness of process supervision and the utilization efficiency of sensing resources. The current monitoring system usually adopts a unified monitoring intensity for all passing vehicles, and cannot dynamically adjust the sampling frequency and algorithm mode of the sensing device according to the credit status of the vehicle, resulting in that when facing massive data on the whole road section, the computing resources are difficult to concentrate on real-time tracking of high-risk targets. At the same time, the existing management means focuses on static entry qualification audit, lacks dynamic constraint mechanism for the driving path and stay time of the vehicle after entering the campus, and is difficult to automatically identify and intervene in the illegal behaviors such as vehicle bypassing non-open areas or overtime retention. In addition, there are a large number of pedestrians and bicycles and other vulnerable traffic participants in the campus who do not carry communication terminals, and the entering vehicle cannot perceive the obstacles outside the visual range through the communication link in advance, and this lack of information interaction is easy to cause safety accidents in the limited visibility section. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a smart college vehicle safety monitoring and management processing system, which solves the problems that it is difficult to finely and dynamically control the driving track and stay time of the vehicle after entering the campus in the existing college vehicle management, and the utilization efficiency of computing resources is low due to full-quantity high-frequency monitoring, and there is a safety blind area for non-networked vulnerable traffic participants.

[0005] To achieve the above purpose, the present application is realized by the following technical scheme: a smart college vehicle safety monitoring and management processing system, comprising:

[0006] a cloud control platform, which generates space-time access rights based on vehicle identity and dynamic credit, and performs credit evaluation management on the vehicle;

[0007] An edge computing node for performing fusion processing of multi-source perception data, dynamic scheduling of perception resources, and real-time compliance monitoring of vehicle driving trajectories;

[0008] A roadside perception and enforcement layer for collecting real-time state data of road environment and performing physical control actions in response to control instructions;

[0009] The cloud control platform, edge computing node, and roadside perception and enforcement layer are connected through a network.

[0010] Preferably, the cloud control platform comprises:

[0011] An identity information database for storing identity attributes of vehicle drivers, including faculty, students, and visitors;

[0012] A reputation evaluation engine for maintaining a campus traffic reputation score for each registered vehicle and performing score deduction operations based on received violation records;

[0013] A path planning module for generating corresponding electronic escort fences based on the identity attributes and destination requests of vehicles; the electronic escort fence is a four-dimensional space-time traffic pipeline that defines the geographical area that the vehicle is allowed to pass through within a specific time period.

[0014] Preferably, the edge computing node comprises:

[0015] A data fusion module for receiving raw data from the roadside perception and enforcement layer, performing space-time synchronization and feature-level fusion algorithms, and outputting traffic element states in a unified coordinate system;

[0016] A resource scheduling module for sending work parameter adjustment instructions, including sampling frequency and algorithm running mode, to the roadside perception and enforcement layer based on the vehicle reputation score issued by the cloud control platform;

[0017] A closed-loop monitoring module for receiving vehicle state and electronic escort fence data output by the data fusion module, calculating the space-time deviation of the trajectory in real time, and generating an abnormality determination signal.

[0018] Preferably, the roadside perception and enforcement layer comprises:

[0019] A roadside perception device for collecting road environment information, including visual sensors and radar sensors;

[0020] A control and enforcement device for responding to control instructions issued by the edge computing node.

[0021] Preferably, the campus traffic credit score of each registered vehicle is maintained, and score deduction is performed according to the received violation records, specifically including:

[0022] An initial credit upper limit value is set, and a discrete-time event trigger mechanism is used to update the credit score;

[0023] When receiving a violation record, a deduction is performed from the current credit score according to the weight value corresponding to the type of violation;

[0024] In the period without violation records, automatic recovery of the credit score is performed according to a preset linear growth model until the initial credit upper limit value is reached.

[0025] Preferably, the corresponding electronic escort fence is generated, specifically including:

[0026] For each discrete node in the planned path, the earliest allowable time and the latest allowable time are calculated respectively to form a time constraint window;

[0027] The earliest allowable time is calculated based on the highest physical speed limit of the road section, and is used to constrain speeding behavior;

[0028] The latest allowable time is calculated based on the reference travel speed of the road section and the real-time congestion index, and superimposes a preset fault tolerance buffer time, and is used to constrain the retention behavior.

[0029] Preferably, according to the vehicle credit score issued by the cloud control platform, the work parameter adjustment instruction is sent to the roadside perception and execution layer, specifically including:

[0030] The current campus traffic credit score of the vehicle and the environmental sensitivity coefficient of the area where the vehicle is located are obtained;

[0031] The monitoring weight coefficient of the vehicle is calculated, and the monitoring weight coefficient is configured as:

[0032] It is negatively correlated with the campus traffic credit score and positively correlated with the environmental sensitivity coefficient;

[0033] The monitoring weight coefficient is compared with a preset grading threshold, and a corresponding perception mode instruction is selected according to the comparison result, the perception mode including a basic tracking mode focusing on position tracking, a behavior analysis mode focusing on feature recognition, and a lock monitoring mode focusing on high-frequency data fusion.

[0034] Preferably, the traffic element state in the unified coordinate system further includes virtualization processing for non-networked targets, specifically including:

[0035] By comparing the spatial position of the sensor detection target with the vehicle networking broadcast message received by the roadside, non-networked targets are screened out.

[0036] allocating a virtual identity for the non-networking target, and encapsulating a virtual basic safety message containing the virtual identity, fusion coordinates and a velocity vector;

[0037] sending the virtual basic safety message to the roadside perception and execution layer, and proxying the non-networking target to broadcast a state packet.

[0038] Preferably, the real-time trajectory is calculated to obtain a spatial and temporal deviation degree, and an abnormality determination signal is generated, specifically including:

[0039] calculating a vertical distance of a vehicle actual position projected to a center line of a planned path to obtain a lateral spatial deviation degree;

[0040] based on an actual time of a vehicle reaching a current position, difference calculation is performed with an earliest allowed time and a latest allowed time set in the electronic escort fence respectively to obtain a longitudinal time deviation degree;

[0041] when the lateral spatial deviation degree or the longitudinal time deviation degree exceeds a preset fault tolerance threshold, the abnormality determination signal is triggered.

[0042] Preferably, after the score deduction operation is performed according to the received illegal record, the system further includes a permission reevaluation logic, specifically including:

[0043] comparing the deducted reputation score with a minimum access threshold set by the system;

[0044] if the updated reputation score is lower than the minimum access threshold, the permission state of the vehicle is marked as rejected, and is broadcast to all network nodes to reject the generation of the electronic escort fence at the next request.

[0045] The present application provides a smart college vehicle safety monitoring and management processing system. It has the following beneficial effects:

[0046] 1. The present application dynamically adjusts the sensor monitoring mode by combining the reputation score and the environmental coefficient through the resource scheduling module, solves the algorithm bottleneck and bandwidth pressure caused by full-section high-precision monitoring, and improves the system calculation efficiency and operation stability under the premise of ensuring effective supervision of key targets.

[0047] 2. The present application solves the drawbacks of traditional static management that cannot effectively control the vehicle driving process and relies on manual escort by constructing a four-dimensional electronic escort fence containing space-time double constraints and calculating the space-time deviation degree in real time, realizes the full-process dynamic management of off-campus vehicles without manual escort, and effectively eliminates the safety hazards caused by illegal parking and passing.

[0048] 3. This invention solves the problems of information barriers between roadside perception and vehicle-mounted terminals and the lack of perception in blind spots by converting non-networked vulnerable traffic participants into virtual V2X messages and injecting them into the vehicle network channel. This enables vehicles entering the school to perceive potential obstacles in advance and reduces the risk of collisions in mixed traffic scenarios on campus. Attached Figure Description

[0049] Figure 1 This is a system architecture diagram of the present invention;

[0050] Figure 2 This is a diagram of the cloud control platform architecture of the present invention;

[0051] Figure 3 This is a diagram of the edge computing node architecture of the present invention;

[0052] Figure 4 This is a diagram of the roadside perception and execution layer architecture of the present invention. Detailed Implementation

[0053] See attached document Figure 1 - Appendix Figure 4 This invention provides an intelligent vehicle safety monitoring and control system for universities, comprising: a cloud control platform 10, an edge computing node 20, a roadside perception and execution layer 30, and a user terminal 40. These components establish data connections through a wired fiber optic network or a dedicated wireless network, collectively forming an end-edge-cloud collaborative control architecture.

[0054] The cloud control platform 10, serving as the core data processing center of the system, is equipped with an identity information database 11, a credit assessment engine 12, and a route planning module 13. The identity information database 11 connects to the university's internal personnel management system, student registration management system, and visitor reservation system via data interfaces. It stores and updates the identity attributes of vehicle drivers, which include at least faculty, students, support staff, and visitors. The credit assessment engine 12 maintains the campus traffic credit score for each registered vehicle and performs score deductions based on received traffic violation records. The route planning module 13 is configured to generate a corresponding four-dimensional spatiotemporal passage based on the vehicle's identity attributes and destination request. This four-dimensional spatiotemporal passage defines the geographical area that the vehicle is allowed to travel within a specific time period.

[0055] The edge computing node 20 is deployed at the key nodes or regional centers of the campus road, and is configured with a data fusion module 21, a resource scheduling module 22, and a closed-loop monitoring module 23. The edge computing node 20 is connected to the roadside perception and execution layer 30 downwardly and to the cloud control platform 10 upwardly. The data fusion module 21 is used to receive raw data from different sensors, perform spatio-temporal synchronization and feature-level fusion algorithm, and output the traffic element state in a unified coordinate system. The resource scheduling module 22 is used to send work parameter adjustment instructions to the roadside perception device according to the vehicle reputation score issued by the cloud, and the work parameters include sampling frequency and algorithm running mode. The closed-loop monitoring module 23 is used to receive the vehicle state output by the data fusion module 21 and the electronic escort fence data issued by the cloud, calculate the spatio-temporal deviation of the trajectory in real time, and generate an abnormality judgment signal.

[0056] The roadside perception and execution layer 30 is distributed in the whole campus road, and includes a visual sensor 31, a radar sensor 32, a roadside communication unit 33, and a management and control execution device 34. The visual sensor 31 and the radar sensor 32 are used to collect dynamic information of vehicles, pedestrians and non-motor vehicles in the road environment. The roadside communication unit 33 is used to execute the V2X communication protocol and broadcast traffic safety messages. The management and control execution device 34 includes an electronic display screen, an audible and light alarm, and an intelligent barrier set at the intersection, and is used to respond to the control instructions issued by the edge computing node 20.

[0057] The above is used when the vehicle applies to enter the campus, and the cloud control platform 10 calls the data in the identity information database 11 for authentication. The path planning module 13 calculates the optimal driving path according to the identity permission of the vehicle and the real-time road network state, and generates an electronic escort fence containing a spatial coordinate sequence and a time window constraint. The electronic escort fence data is issued to the edge computing node 20 along the vehicle driving path together with the reputation score of the vehicle.

[0058] After the resource scheduling module 22 in the edge computing node 20 receives the vehicle information, the monitoring intensity of the roadside perception device is dynamically set according to the reputation score. For the vehicle with a reputation score lower than a preset threshold, the resource scheduling module 22 instructs the visual sensor 31 to increase the sampling frame rate and activates the fine behavior analysis algorithm to perform high-frequency feature extraction on the vehicle. For the vehicle with a reputation score higher than the preset threshold, the system maintains the basic trajectory tracking mode, so as to realize differentiated allocation of computing resources.

[0059] During the driving process of the vehicle, the visual sensor 31 and the radar sensor 32 continuously collect environmental data. The data fusion module 21 fuses the visual image data with the radar point cloud data to generate real-time position coordinates and a speed vector of the vehicle. At the same time, for a target such as a pedestrian or a non-motor vehicle that does not carry a communication terminal and is identified by the sensor, the data fusion module 21 assigns a virtual identity to the target and generates a virtual basic safety message. The roadside communication unit 33 injects the virtual basic safety message into a vehicle networking communication channel, so that a vehicle with communication capability can perceive the target that does not carry a communication terminal, and realize the cooperative interaction of heterogeneous traffic elements.

[0060] The edge computing node 20 compares the fused vehicle actual trajectory data with the preset electronic escort fence in real time. The data fusion module 21 calculates the spatial deviation degree of the current position of the vehicle from the center line of the planned path, and the time deviation degree of the time at which the vehicle reaches the current position from the allowed time window. When the spatial deviation degree or the time deviation degree exceeds the preset fault tolerance range, the edge computing node 20 determines that the vehicle has a violation behavior.

[0061] Once the violation is determined, the system triggers a closed-loop management process. The edge computing node 20 sends an interception or induction instruction to the management and execution device 34 in front of the vehicle to limit the vehicle's passing right. At the same time, the edge computing node 20 uploads the violation evidence to the cloud control platform 10. The reputation evaluation engine 12 deducts the corresponding reputation score according to the type of violation, and updates the record in the identity information database 11. The updated reputation score will be used as the basis for path planning and resource allocation for the vehicle in the next application for entry into the school, thereby forming a dynamic closed-loop management mechanism based on data driving.

[0062] The identity information database 11 and the reputation evaluation engine 12 establish a communication connection with the existing personnel management system, student status management system and visitor reservation system of the university through an encrypted data interface, for building and maintaining a multi-dimensional identity reputation model, i.e. a digital identity file of the campus vehicle.

[0063] In the specific implementation process, the building and maintaining process of the multi-dimensional identity reputation model includes the following steps:

[0064] Step S201, building a vehicle state vector.

[0065] The system cleans and matches the data obtained from different sources, and generates a unique identity state vector for each registered vehicle The state vector is defined as:

[0066]

[0067] In the formula, ​For state vectors, It represents the vehicle's physical license plate number and is used as an index key for the visual recognition system; This indicates the identity category of the user to whom the vehicle belongs, and its value set includes at least faculty and staff, students, visitors, and support staff; Indicates at time The vehicle's real-time campus traffic credit score; It is a linked list structure that records historical traffic violations, storing the types of violations and timestamps that have occurred to the vehicle in the past.

[0068] For identity categories The system has a pre-defined basic permission matrix. This matrix defines the default set of passable areas, authorized parking areas, and passable time periods for different identity categories with a perfect reputation score. For example, the basic permission matrix for faculty and staff includes parking on all campus roads and access during all times, while the basic permission matrix for students only includes parking on dormitory roads and access during non-teaching hours. This static basic permission matrix serves as the initial constraint for subsequent dynamic path planning.

[0069] Step S202: Perform dynamic update calculation of the credit score.

[0070] The reputation assessment engine 12 uses a discrete-time event triggering mechanism to... An update will be performed. Initially, the system assigns a uniform initial credit cap to all newly registered vehicles. When the system receives the violation determination result uploaded by edge computing node 20 or after a preset time period, it calculates a new reputation score based on the following reputation update model:

[0071]

[0072] In the formula, The updated reputation score; The credit score at the time of the last update; The total number of violation types defined by the system; For the first Indicator functions for traffic violations, when the received violation data contains the first type of violation. When class behavior, The value is 1 if it is set to 1, and 0 otherwise. For the first The point deduction weight value corresponding to a type of traffic violation, such as serious speeding; For reputation recovery functions; The time interval since the last violation occurred; This is the highest credit score.

[0073] The above reputation recovery function The specific configuration is a linear growth model, meaning that during periods without traffic violations, the credit score automatically recovers at a preset rate until it reaches a certain level. This mechanism is used to quantify drivers' compliant driving behavior, thereby enabling dynamic adjustments to vehicle control levels over a long period.

[0074] Step S203: Generate reputation rating labels.

[0075] The system is based on real-time calculations. Vehicles are mapped to different credit rating ranges. A first threshold is set. Second threshold ,in .when At that time, the vehicle was marked as having a high credit rating; when At that time, the vehicle was marked as having a normal credit rating; when At that time, the vehicle is marked as having a low reputation level. This reputation level label is synchronized in real time to the path planning module 13 and the resource scheduling module 22, serving as a direct input parameter for subsequently generating the width of the electronic fence constraints and the amount of sensing resources allocated.

[0076] For encrypted data transmission through data interfaces and routine CRUD operations on databases, those skilled in the art can use existing SSL / TLS encryption protocols and SQL database management technologies, which are well-known technologies in the field and will not be elaborated upon here.

[0077] The route planning module 13 generates a passage constraint pipeline with four-dimensional spatiotemporal characteristics, i.e., an electronic escort fence, based on the vehicle's identity and permissions and the real-time road network status. This process transforms abstract management permissions into specific spatiotemporal geometric constraints.

[0078] Specifically, the process of generating a dynamic electronic security fence includes the following steps:

[0079] Step S204: Determine the optimal spatial path sequence.

[0080] The system first retrieves high-precision campus map data stored in the database to construct a road network topology map. .in, Represents the set of road network nodes. This represents the set of edges connecting road segments. For each edge... It is associated with physical attribute data, including road segment length, maximum speed limit, and current road congestion index.

[0081] When the vehicle When applying for admission, the route planning module 13 filters out the subgraphs that allow the vehicle to pass based on the basic permission matrix corresponding to the vehicle's identity. In the subgraph In this process, starting from the vehicle's current position and ending at the permitted parking area parsed in step S201, the optimal spatial path is calculated using Dijkstra's algorithm. This optimal spatial path is represented as an ordered sequence of discrete road network nodes. In the formula, As the starting point for entering the school, The destination is the final destination. This represents the total number of nodes contained in the path.

[0082] Step S205: Calculate the spatiotemporal constraint window of the path node.

[0083] The system not only constrains the spatial routes of vehicle travel but also dynamically controls vehicle speed and stopping behavior by calculating the allowable time range to reach each path node. For path sequences... Each node in The system calculates the earliest allowed time for a vehicle to arrive at the node. and the latest permitted time .

[0084] Earliest allowed time This calculation is based on the maximum physical speed limit of a road segment and is used to prevent vehicles from speeding. The calculation formula is as follows:

[0085]

[0086] In the formula, The time when the vehicle passes through the school gate or the starting point; For nodes With nodes The physical path length between them; For connecting nodes and section of road The maximum speed limit.

[0087] Latest permitted time Based on the minimum traffic efficiency of a road segment and real-time congestion conditions, a calculation method is used to identify abnormal parking, lingering, and deliberate detours by vehicles. The calculation formula is as follows:

[0088]

[0089] In the formula, For road section The baseline speed for traffic is usually set as a preset percentage of the maximum speed limit. This represents the current congestion index for this road segment, with a value ranging from [0,1). A larger value indicates greater congestion. The system has a preset fault tolerance buffer time to compensate for sensor errors and minor driving operation delays; For nodes With nodes The physical path length between them.

[0090] Step S206: Construct a four-dimensional spatiotemporal pipeline data structure.

[0091] Based on the spatial coordinates determined in step S204 and the time window calculated in step S205, the system generates the final set of electronic escort fences. This set defines the legal four-dimensional existence space of the vehicle throughout its entire driving process:

[0092]

[0093] In the formula, For the final electronic security fence assembly; Path node Geodetic coordinates; This refers to the allowable lateral deviation width corresponding to this road segment. This width value is dynamically set according to the road grade, such as that for arterial roads. The value is greater than that of the auxiliary road. value.

[0094] Generated electronic escort fence data The data is then distributed via the network to edge computing nodes 20 along the vehicle's travel path, serving as a baseline template for subsequent real-time trajectory deviation monitoring. Through these steps, the system refines the single permission to enter the school into precise control requirements for vehicles at specific times and locations.

[0095] The edge computing node 20 is equipped with a resource scheduling module 22. This module receives vehicle reputation data from the cloud and, in conjunction with the real-time environmental conditions perceived by the roadside, calculates the monitoring weight coefficient for each vehicle on the road. This monitoring weight coefficient quantifies the system's focus on a specific target and is the core basis for subsequent allocation of computing resources for the sensing hardware.

[0096] In the specific implementation process, the dynamic calculation model for monitoring weight coefficients includes the following steps:

[0097] Step S301: Obtain vehicle reputation and regional attribute parameters.

[0098] When the vehicle When a vehicle enters the jurisdiction of an edge computing node 20, the resource scheduling module 22 synchronously obtains the vehicle's current campus traffic credit score from the cloud. Simultaneously, the resource scheduling module 22 reads the environmental sensitivity coefficient of the current geographical area where the vehicle is located. .

[0099] Environmental sensitivity coefficient These are predefined static attribute values ​​in the digital map, ranging from [0,1]. This value reflects the stringency of traffic safety requirements in different areas. For example, higher values ​​are set around school buildings and in densely populated areas. Values ​​(e.g., 0.8 to 1.0) are set lower for remote, off-the-beaten-path sections or closed freight lanes. Values ​​(e.g., 0.2 to 0.4).

[0100] Step S302: Perform normalization calculation of the monitoring weight coefficient.

[0101] Resource scheduling module 22 calculates vehicle allocation based on reputation score and environmental sensitivity coefficient using a weighted summation model. Monitoring weight This calculation process aims to combine the vehicle's inherent unreliability with the vulnerability of the current environment to form a comprehensive risk measure. The calculation formula is as follows:

[0102]

[0103] In the formula, The monitoring weight coefficient has a value range of [0,1]. The larger the value, the higher the potential risk of the vehicle and the more monitoring resources are required. This refers to the credit factor weighting coefficient. These are the environmental factor weighting coefficients; and satisfy . The upper limit for the credit score set by the system; This refers to the vehicle's current credit score. This represents the environmental sensitivity coefficient of the geographical area where the vehicle is currently located. This represents the highest credit score.

[0104] The above formula shows that when the vehicle credit score... The lower the temperature, or the more sensitive the area (the vehicle is in a high-sensitivity area). When the value is larger, the calculated monitoring weight coefficient is... The higher the risk level, the better. This calculation mechanism ensures that system resources are allocated to high-risk targets and high-risk areas.

[0105] Step S303: Correct the monitoring weight coefficient.

[0106] To further improve the robustness of the system, the resource scheduling module 22 also adjusts the scheduling based on the vehicle's historical violation types. Dynamic correction will be performed. This includes addressing the vehicle's historical traffic violation records. The system introduces penalty factors for specific serious violations (such as hit-and-run or serious speeding). Corrected monitoring weight coefficient The calculation is as follows:

[0107]

[0108] In the formula, The default penalty gain coefficient is greater than 0. The corrected monitoring weight coefficient; This is the monitoring weighting coefficient before correction. This step ensures that vehicles with a bad history will still be listed as key monitoring targets by the system even if their current credit score has not been completely depleted.

[0109] The final monitoring weight coefficients obtained from the calculation The parameters will be directly transmitted to the control unit of the roadside sensing device to determine the specific operating modes of the vision sensor 31 and the radar sensor 32. The specific implementation of parameter storage and floating-point operations can be accomplished by those skilled in the art using conventional memory and microprocessor instruction sets, and will not be elaborated upon here.

[0110] The resource scheduling module 22 of the edge computing node 20 sends specific operating parameter instructions to the roadside visual sensor 31 and radar sensor 32 based on the monitoring weight coefficients calculated in the previous steps, thereby realizing differentiated configuration of hardware resources among different vehicle targets. This hierarchical control strategy solves the computing power bottleneck and bandwidth pressure problems caused by high-frequency monitoring across the entire road segment.

[0111] Specifically, the hierarchical control process of the operating mode of sensing devices includes the following steps:

[0112] Step S304: Determine the target monitoring mode.

[0113] Resource scheduling module 22 has a preset first-level threshold. Second grade threshold And satisfy The system assigns a monitoring weight coefficient to each vehicle on the road. The comparison is performed against the aforementioned thresholds, and the corresponding working mode is selected based on the comparison results. :

[0114]

[0115] In the formula, The basic tracking mode is suitable for vehicles with high credit ratings and located in low-sensitivity areas. This is a behavioral analysis model, suitable for vehicles with average credit or located in moderately sensitive areas; This is a lock-on monitoring mode, suitable for key vehicles with low reputation or located in highly sensitive areas.

[0116] Step S305: Perform basic tracking mode configuration. When it is determined that the vehicle is in... At this time, the system mainly relies on radar sensor 32 for position updates to reduce the computational power consumption of the visual algorithm. Resource scheduling module 22 sets the radar sampling frequency to the reference frequency. At the same time, the image acquisition frequency of the vision sensor 31 is set to low frequency mode. Furthermore, only the vehicle detection algorithm is executed, without activating the license plate recognition and feature analysis algorithms. At this time, the system only maintains coarse-grained tracking of the vehicle's center coordinates for basic path node check-in recording.

[0117] Step S306: Configure the behavior analysis mode.

[0118] When it is determined that the vehicle is in At this time, the system enhances the importance of visual perception. Resource scheduling module 22 increases the image acquisition frequency of visual sensor 31 to a mid-frequency mode. ( The system then activates a lightweight convolutional neural network model. In this mode, while maintaining trajectory tracking, the system identifies in real time whether the driver is using a mobile phone or not wearing a seatbelt. If a violation is detected, the system immediately generates an event trigger signal, increases the monitoring weight coefficient of the vehicle, and prompts it to enter a higher-level locking monitoring mode.

[0119] Step S307: Configure the lock monitoring mode.

[0120] When it is determined that the vehicle is in When the system determines that the target is high-risk, it allocates maximum sensing resources for relay monitoring. The resource scheduling module 22 sets the sampling frequency of both the radar sensor 32 and the vision sensor 31 to high-frequency mode. Furthermore, it activates pixel-level data fusion algorithms. In this mode, edge computing nodes 20 calculate the relative distance between the vehicle outline and the electronic escort fence boundary in real time, with their position update cycle shortened to the millisecond level, ensuring that any slight path deviation can be detected instantly.

[0121] For the calling of the sensor driver interface and the configuration of the underlying registers in the above modes, those skilled in the art can implement them based on the SDK (Software Development Kit) and API documentation provided by the specific hardware manufacturer. These are well-known technologies in the field of embedded development and will not be elaborated here.

[0122] The data fusion module 21 configured within the edge computing node 20 processes the raw data streams from roadside heterogeneous sensors. Through spatiotemporal alignment and weighted fusion algorithms, it outputs the traffic element status in a unified world coordinate system. This process aims to eliminate the perception blind spots of single sensors and environmental interference, constructing a high-precision holographic traffic scene.

[0123] Specifically, the spatiotemporal synchronization fusion process of visual and radar data includes the following steps:

[0124] Step S401: Perform time base synchronization and frame rate alignment. Since the sampling frequencies of the visual sensor 31 and the radar sensor 32 are inconsistent (e.g., visual at 30fps, radar at 20Hz), the data fusion module 21 first assigns a unified timestamp to all acquired frames based on a unified Network Time Protocol (NTP) or Precision Time Protocol (PTP) clock source. The system uses the acquisition time of the visual image frame... Based on this, the two adjacent data frames with the closest timestamps are searched in the radar data stream, and the time is reconstructed using linear interpolation. The radar point cloud state is obtained, thereby eliminating the position error of high-speed moving targets caused by asynchronous sensor acquisition.

[0125] Step S402: Establish spatial coordinate mapping relationship.

[0126] The system pre-calibrates the intrinsic and extrinsic parameter matrices of the visual sensor 31 and the rigid body transformation matrix of the radar sensor 32 relative to the world coordinate system. The data fusion module 21 projects the 3D point cloud targets detected by the radar onto the visual image plane and uses the intersection-over-union (IoU) algorithm to associate and match the radar targets with the visual detection boxes.

[0127] For a successfully matched target, the system calculates two sets of position estimates in the world coordinate system. The visual position estimate relies on back-projection of depth information provided by the radar, while the radar position estimate is obtained directly through coordinate transformation.

[0128] Step S403: Perform weighted fusion calculation.

[0129] To comprehensively utilize the advantages of the visual sensor 31 in lateral orientation resolution and the radar sensor 32 in longitudinal distance measurement, the system employs an adaptive weighted algorithm to calculate the final fused coordinates of the target. The calculation formula is as follows:

[0130]

[0131] In the formula, The final fused 3D coordinate vector of the target in the world coordinate system; This is a visual inverse projection transformation function used to convert image pixel coordinates into world space coordinates; The pixel coordinates of the center point of the visual inspection box ; The depth value of the radar target after association and matching is used to compensate for the depth information missing in monocular vision; This is a rigid body transformation function from the radar coordinate system to the world coordinate system, which includes a rotation matrix and a translation vector. These are the original local coordinates of the target detected by radar; This is the confidence weight matrix for visual data; This is the confidence weight matrix for radar data.

[0132] Step S404: Dynamically adjust the confidence weights.

[0133] The system updates the aforementioned weights in real time based on the covariance matrix of the Kalman filter. and In well-lit scenarios where the target is moving laterally significantly, the system automatically increases the size. The system automatically increases the value when insufficient light (such as at night) or rainy / foggy weather causes increased noise in the visual sensor 31. The numerical value relies on radar data to maintain perception accuracy.

[0134] The camera calibration method and the specific iterative process of Kalman filtering involved in the above steps can be implemented by those skilled in the art using the Zhang Zhengyou calibration method and the Extended Kalman Filter (EKF) standard algorithm, which are well-known technologies in the fields of computer vision and signal processing, and will not be elaborated here.

[0135] For pedestrians, bicycles, and ordinary vehicles without onboard units that are widely present in the campus environment (collectively referred to as non-networked vulnerable traffic participants), the system uses an edge-side proxy mechanism to transform them into virtual nodes with communication capabilities, thereby enabling vehicles with autonomous driving or assisted driving functions to perceive these hard-to-identify obstacles.

[0136] Specifically, the virtual identity generation and message injection process for non-networked targets includes the following steps:

[0137] Step S405: Perform non-networked target filtering and deduplication.

[0138] Data fusion module 21 obtains a list of all traffic targets fused from sensors at the current moment. Simultaneously, the roadside communication unit 33 monitors all active vehicle-to-everything (V2X) broadcast messages within the current communication range. The system uses a spatial location matching algorithm to compare targets detected by sensors with vehicles actually broadcasting BMS messages, eliminating vehicles that already have communication capabilities and filtering out the true set of non-networked targets. .

[0139] This screening process is based on the following distance threshold determination formula:

[0140]

[0141] In the formula, For the first Sensor fusion target; The complete set of targets detected by the sensor; For the first A networked vehicle that reports its status via a V2X channel; For the currently received Message set; For the goal The sensor measures the coordinates; For connected vehicles Reported GPS coordinates; This is for Euclidean distance calculation; A location matching tolerance threshold is set; when the distance between two objects exceeds this threshold, the system determines the target. For non-networked targets, virtualization mapping is required.

[0142] Step S406: Assign a virtual identity.

[0143] The system provides each selected non-networked target with... A unique virtual MAC address and a temporary application layer ID are assigned. To prevent conflicts with real vehicle IDs, the system reserves specific ID ranges (e.g., ranges starting with a specific hexadecimal digit) specifically for identifying virtual targets. This virtual ID remains unchanged as the target continues to appear in the sensor's field of view, and inter-frame correlation is performed using the state prediction values ​​of a Kalman filter to ensure the continuity of the target's identity.

[0144] Step S407: Construct a virtual basic security message.

[0145] Edge computing node 20 encapsulates the motion state data acquired by the sensor into a standard BSM message structure according to the SAE J2735 or CSAE 53-2017 communication standard protocol stack. The generated virtual message content... The definition is as follows:

[0146]

[0147] In the formula, The virtual identity identifier assigned in step S406; This is the current system timestamp; The fused three-dimensional coordinates (longitude, latitude, and elevation) are obtained through step S403. The velocity vector of the target, including the magnitude of the velocity and the heading angle; The three-dimensional bounding box information (length, width, height) of the target; The target classification attribute is used to indicate whether the target is a pedestrian, a non-motorized vehicle, or a motorized vehicle; This is the content of a virtual message.

[0148] Step S408: Perform virtual message injection and broadcast.

[0149] Edge computing node 20 will be built The data packet is sent to the roadside communication unit 33 (RSU). The roadside communication unit 33 adds the data packet to the queue to be sent and broadcasts it to the outside at the vehicle-road cooperative standard frequency (e.g., 10Hz) when the physical layer channel is idle.

[0150] Through the above steps, pedestrians or vehicles that originally lacked communication capabilities can broadcast their status messages. When a vehicle equipped with an Onboard Unit (OBU) enters the area, the data stream it receives contains not only real messages from other connected vehicles in the vicinity but also virtual messages sent by roadside equipment. After parsing these messages, the OBU can render the positions of all traffic participants on a local high-precision map and treat real and virtual targets equally when calculating collision warnings, thereby effectively eliminating safety hazards caused by blind spots.

[0151] For the specific encoding format and ASN.1 encoding / decoding rules of the SAE J2735 protocol, those skilled in the art can refer to relevant communication standard documents for implementation. These are well-known technologies in the field of vehicle networking communication and will not be elaborated here.

[0152] Edge computing node 20 is equipped with a closed-loop monitoring module 23, which receives the actual vehicle state vector output by data fusion module 21 in real time and compares it with the electronic escort fence data sent from the cloud. By calculating the dual deviation indicators in the spatial and temporal dimensions, the system can accurately identify various abnormal behaviors, including path deviation, illegal parking, speeding, and excessive delay.

[0153] Specifically, the real-time calculation and anomaly determination process of trajectory spatiotemporal deviation includes the following steps:

[0154] Step S501: Perform trajectory projection and road network matching.

[0155] Closed-loop monitoring module 23 acquires the vehicle's current fused position coordinates. Simultaneously, the system retrieves the electronic escort fence set corresponding to the vehicle. Using a nearest neighbor search algorithm, the system finds the path node in the planned path sequence that has the closest Euclidean distance to the current vehicle position. and its adjacent nodes Determine the current road segment where the vehicle is located. The system will display the vehicle's actual coordinates. By orthogonally projecting the projection onto the centerline of the road segment, the projection point is obtained. And calculate the cumulative mileage position of the projection point along the entire path.

[0156] Step S502: Calculate lateral spatial deviation. The system calculates the vertical distance between the vehicle's actual center position and the centerline of the planned path to quantify compliance in the spatial dimension. Lateral spatial deviation. The calculation formula is as follows:

[0157]

[0158] In the formula, For a moment Lateral spatial deviation (unit: meters); , These are the vehicle's current actual planar coordinates; For the current road section The coefficients of the general equation of the line containing the centerline satisfy the following conditions: .

[0159] Step S503: Calculate the longitudinal time deviation.

[0160] The system determines the location based on the vehicle's arrival at the current projection point. actual moment The time frame is compared with a preset time window in the electronic security fence. Since the path nodes are discrete, the system uses linear interpolation to calculate the lower limit of the allowable time corresponding to the current position. and the maximum allowed time Longitudinal time deviation includes overspeed deviation. and the deviation of retention The calculation formula is as follows:

[0161]

[0162] In the formula, The earliest legal time to reach the current position, calculated based on the maximum speed limit; The latest legal time to reach the current location, calculated based on minimum traffic efficiency and congestion index; This is the current actual timestamp of the system. When, it indicates that the vehicle arrived at excessive speed; when This indicates that the vehicle has illegally stopped or is deliberately lingering at low speed along the way.

[0163] Step S504: Execute multi-level exception handling logic. The system has a preset spatial fault tolerance threshold. and time tolerance threshold The closed-loop monitoring module 23 generates an anomaly detection signal based on the above calculation results. .

[0164] The decision logic is configured as follows: If If it is determined to be a violation of spatial boundaries, then in the formula, The permissible passage width defined in the electronic fence for this road segment; This is the spatial fault tolerance threshold; if If so, it will be judged as a serious speeding violation; if If so, it will be judged as an overstaying violation.

[0165] once Once triggered, the edge computing node 20 immediately locks the current abnormal data frame and transmits it as evidence of violation to the subsequent control and execution unit and the cloud reputation assessment engine 12, thereby triggering the linkage between physical interception and reputation deduction.

[0166] The derivation of the point-to-line distance formula and the linear interpolation calculation method involved in the above steps can be implemented by those skilled in the art based on the fundamental theories of analytical geometry and numerical analysis. These are well-known techniques in the field of mathematical computation and will not be elaborated here.

[0167] When the closed-loop monitoring module 23 of the edge computing node 20 determines that the vehicle has an abnormal spatiotemporal trajectory, the system immediately activates the hierarchical response mechanism, performs on-site control through the actions of roadside physical equipment, and transmits the violation data back to the cloud control platform 10, triggering real-time deduction of the credit value, thereby forming a closed-loop link from monitoring to handling to management.

[0168] Specifically, the violation-based tiered response and credit-based closed-loop feedback mechanism includes the following steps:

[0169] Step S505: Generate hierarchical control instructions.

[0170] The edge computing node 20 outputs an anomaly detection signal based on the previous steps. The system determines the current response level based on the violation type and its corresponding violation type, and then generates the corresponding equipment control command vector. The control command vector is defined as follows:

[0171]

[0172] In the formula, Text-encoded instructions sent to the LED display screen to display warning messages or guidance messages; This is the trigger level command sent to the audible and visual alarm, used to control the alarm volume and flashing frequency; This is a relay switching signal sent to the smart barrier gate or bollard to execute a forced interception operation.

[0173] The system has a pre-set violation level mapping logic: when a violation is determined to be a spatial boundary overstepping violation or a minor overstay, the system outputs a level one response command, setting... To request immediate return to the main road or prohibit parking, set... Set to a short notification sound. To maintain the current state;

[0174] When a serious speeding violation or entry into a restricted area is detected, the system outputs a level-two response command and sets... This is a serious violation; the account has been locked and set. Set as a high-frequency continuous alarm. When in the closed / raised state, it physically blocks the vehicle's passage.

[0175] Step S506: Perform on-site physical control.

[0176] The control execution device 34 receives the data sent by the edge computing node 20. After processing the vector, parse and execute the corresponding action. For During command execution, the system employs anti-collision logic for safety verification. Before lowering the barrier or raising the bollard, the system again calls the obstacle detection interface of radar sensor 32 to confirm that there are no objects obstructing the view under the gate arm or above the bollard. If an obstruction is detected, the system temporarily suspends physical interception and instead raises the bollard. The alarm level will remain in effect until the obstruction is removed.

[0177] Step S507: Perform data feedback and reputation closed-loop update.

[0178] Upon triggering physical control, edge computing node 20 packages and uploads evidence of the violation to cloud control platform 10. The uploaded data packet... The structure is as follows:

[0179]

[0180] In the formula, I For vehicle license plate number; Code for traffic violations; The precise timestamp of the violation; The location of the violation is its geographical coordinates and the ID of the road segment where it occurred. The image or short video clip captured by the visual sensor 31 at the moment the violation occurred.

[0181] The reputation assessment engine 12 of the cloud control platform 10 receives Then, the corresponding credit score for the vehicle will be immediately deducted based on the credit update model. Subsequently, the cloud system executes the permission reassessment logic:

[0182]

[0183] In the formula, This indicates the vehicle's access permission status on campus. This is the latest reputation score after the points were deducted; The minimum admission threshold set for the system; An authorized status that allows vehicles to enter or continue to pass; This is a denial status indicating that a vehicle may be prohibited from entering or added to a blacklist.

[0184] If the updated reputation value falls below the minimum access threshold, the cloud control platform 10 immediately adds the vehicle to the blacklist and broadcasts the update information to all edge nodes in the network. When the vehicle attempts to pass through the school gate or apply for route planning again, the system will directly reject its request, thus realizing a closed-loop management system that uses real-time data to drive changes in physical permissions.

[0185] The underlying hardware communication interfaces (such as RS485 and GPIO control) and network transmission protocols (such as MQTT and HTTP / HTTPS) involved in the above steps can be configured by those skilled in the art according to the specific hardware selection and network environment. These are well-known technologies in the field of industrial control and Internet of Things communication, and will not be elaborated here.

Claims

1. A smart college vehicle safety monitoring and management processing system, characterized in that, The cloud control platform generates space-time passing permissions based on vehicle identity and dynamic reputation, and performs credit evaluation management on vehicles. The edge computing node is used for fusion processing of multi-source perception data, dynamic scheduling of perception resources, and real-time compliance monitoring of vehicle driving trajectory. The roadside perception and execution layer is used for collecting real-time state data of road environment and executing physical control actions in response to control instructions. The cloud control platform, edge computing node and roadside perception and execution layer are connected through a network. The cloud control platform comprises: 2.The intelligent high school vehicle safety monitoring and management processing system according to claim 1, characterized in that, An identity information database is used to store the identity attributes of vehicle drivers, including faculty, students and visitors. A reputation evaluation engine is used to maintain the campus traffic reputation score of each registered vehicle and perform score deduction operation according to the received violation records. A path planning module is used to generate corresponding electronic escort fences according to the identity attributes and destination requests of vehicles. The electronic escort fence is a four-dimensional space-time passing pipeline that limits the geographical area that the vehicle is allowed to pass within a certain time period. The edge computing node comprises: 3.The intelligent high school vehicle safety monitoring and management processing system according to claim 1, characterized in that, A data fusion module is used to receive raw data from the roadside perception and execution layer, perform space-time synchronization and feature-level fusion algorithm, and output traffic element state in a unified coordinate system. A resource scheduling module is used to send work parameter adjustment instructions to the roadside perception and execution layer according to the vehicle reputation score issued by the cloud control platform. The work parameters include sampling frequency and algorithm running mode. A closed-loop monitoring module is used to receive vehicle state and electronic escort fence data output by the data fusion module, calculate the space-time deviation of the trajectory in real time, and generate an abnormality determination signal. The roadside perception and execution layer comprises: 4.The intelligent high school vehicle safety monitoring and management processing system according to claim 1, characterized in that, A roadside perception device is used to collect road environment information, including visual sensors and radar sensors. A control execution device is used to respond to control instructions issued by the edge computing node. The maintenance of the campus traffic reputation score of each registered vehicle and the performance of score deduction operation according to the received violation records specifically includes: 5.The intelligent high school vehicle safety monitoring and management processing system according to claim 2, characterized in that, Set an initial reputation upper limit value, and update the reputation score using a discrete time event trigger mechanism; When a violation record is received, the current reputation score is deducted according to the weight value corresponding to the type of violation; In the period without violation records, the reputation score is automatically restored according to a preset linear growth model until the initial reputation upper limit value is reached. The generation of corresponding electronic escort fences specifically includes: 6.The intelligent high school vehicle safety monitoring and management processing system according to claim 2, characterized in that, For each discrete node in the planned path, calculate the earliest allowed time and the latest allowed time to form a time constraint window; The earliest allowed time is calculated based on the highest physical speed limit of the road section, which is used to constrain speeding behavior; The latest allowed time is calculated based on the reference passing speed of the road section and the real-time congestion index, and a preset fault tolerance buffer time is added, which is used to constrain the stagnation behavior. According to the vehicle reputation score issued by the cloud control platform, the work parameter adjustment instructions are sent to the roadside perception and execution layer, specifically including: 7.The intelligent high school vehicle safety monitoring and management processing system according to claim 3, characterized in that, ​ Acquire a current campus traffic credit score of the vehicle and an environmental sensitivity coefficient of an area where the vehicle is located; Calculate a monitoring weight coefficient of the vehicle, which is configured to be negatively correlated with the campus traffic credit score and positively correlated with the environmental sensitivity coefficient; Compare the monitoring weight coefficient with a preset grading threshold, and select a corresponding perception mode instruction according to a comparison result, wherein the perception mode includes a basic tracking mode focusing on position tracking, a behavior analysis mode focusing on feature recognition, and a lock monitoring mode focusing on high-frequency data fusion. The output of the traffic element state in the unified coordinate system further includes virtualization processing for non-networked targets, specifically including: 8.The intelligent high school vehicle safety monitoring and management processing system according to claim 3, characterized in that, Filtering non-networked targets by comparing the spatial positions of sensor-detecting targets and vehicle networking broadcast messages received by the roadside; Assigning a virtual identity to the non-networked target and encapsulating a virtual basic safety message containing the virtual identity, a fusion coordinate and a speed vector; Sending the virtual basic safety message to the roadside perception and execution layer to proxy the non-networked target to broadcast a state message. The real-time calculation of the spatio-temporal deviation degree of the trajectory and the generation of the abnormality judgment signal specifically include: 9.The intelligent high school vehicle safety monitoring and management processing system according to claim 3, characterized in that, Calculating the vertical distance of the actual position of the vehicle projected to the center line of the planned path to obtain the lateral spatial deviation degree; Based on the actual time when the vehicle reaches the current position, the earliest allowable time and the latest allowable time set in the electronic escort fence are respectively calculated by difference to obtain the longitudinal time deviation degree; When the lateral spatial deviation degree or the longitudinal time deviation degree exceeds the preset fault tolerance threshold, the abnormality judgment signal is triggered. After performing the score deduction operation according to the received illegal record, the permission reevaluation logic is further executed, specifically including: 10.The intelligent high school vehicle safety monitoring and management processing system according to claim 7, characterized in that, Comparing the credit score after deduction with the minimum access threshold set by the system; If the updated credit score is lower than the minimum access threshold, the permission state of the vehicle is marked as rejected, and is broadcast to all network nodes to reject the generation of the electronic escort fence at the next request. ​

Images