Traffic safety monitoring and early warning method and system

Abstract

The application discloses a traffic safety monitoring and early warning method and system, relates to the technical field of intelligent transportation, and discloses a traffic safety monitoring and early warning method and system, which generates a joint perception blind area map through an edge perception cooperative network, dynamically predicts the path of a high-risk object after the high-risk object enters a blind area, and cooperates with a secondary perception node to perform directional detection, generates a risk propagation early warning in combination with traffic density and historical data, solves the problems of insufficient blind area coverage and weak prediction capability of a traditional system, has the advantages of dynamic blind area identification, cooperative perception resources and active risk prediction, and can improve the reliability and real-time performance of traffic safety monitoring at a low cost.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, and in particular to a traffic safety monitoring and early warning method and system. Background Technology

[0002] Existing traffic safety monitoring systems heavily rely on fixed-location sensing devices such as cameras, radar, and inductive loops. These devices have several limitations in practical applications: First, due to cost constraints and the physical characteristics of the equipment (such as detection angle and distance limitations), they cannot achieve full-area coverage without blind spots; second, in complex road environments, static obstacles such as buildings and green belts, as well as dynamic obstacles such as large vehicles, create numerous monitoring blind spots; third, traditional systems have weak predictive capabilities for the behavior of traffic participants within blind spots, especially in critical scenarios such as turning at intersections and merging onto highway ramps, where blind spots can easily lead to traffic accidents.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide a traffic safety monitoring and early warning method and system, which aims to improve the reliability and real-time performance of traffic safety monitoring.

[0005] To achieve the above objectives, this application proposes a traffic safety monitoring and early warning method, which includes:

[0006] By deploying an edge sensing collaborative network in the target area, real-time monitoring data of the main sensing node of the edge sensing collaborative network and device location data of the auxiliary sensing node of the edge sensing collaborative network are obtained, and combined with pre-stored map data, a joint sensing blind spot map is generated.

[0007] The main sensing node acquires the category data and motion state data of traffic participants based on the real-time monitoring data. When it determines that a traffic participant meets the preset high-risk conditions based on the category data and motion state data, it marks the traffic participant as a high-risk object and predicts the initial location range of the participant after entering the joint sensing blind zone based on the motion state data.

[0008] The primary sensing node broadcasts the high-risk object's marking information, motion status data, and initial location range to the secondary sensing node corresponding to the initial location range.

[0009] Upon receiving the broadcast, the auxiliary sensing node activates its sensing module, adjusts its sensing direction to focus on the initial position range for directional detection, and generates blind zone detection results.

[0010] If the blind spot detection results do not identify the high-risk object within a preset time, at least one risk migration path and its corresponding risk propagation area range and time window are generated based on the motion state data of the traffic participants, the road structure rules of the joint perception blind spot map, the traffic density data reported by the auxiliary perception nodes, and the historical trajectory database.

[0011] Early warning instructions are generated based on the scope of the risk propagation zone and the time window, and the early warning information is released through at least one of roadside equipment, vehicle-mounted terminals, and mobile terminals.

[0012] In one embodiment, the step of obtaining real-time monitoring data of the main sensing node and device location data of the auxiliary sensing nodes of the edge sensing collaborative network deployed in the target area, and generating a joint sensing blind spot map by combining it with pre-stored map data, includes:

[0013] The main sensing node collects the first monitoring data through fixed sensing devices, and at the same time receives the device location status data sent by each auxiliary sensing node;

[0014] Based on the coordinates of static obstacles in the pre-stored map data, spatial overlay calculations are performed on the coverage area of ​​the first monitoring data to generate a blind spot map of the main node.

[0015] By integrating the device location status data and sensing capability parameters of the auxiliary sensing nodes, their coverage area is calculated, and an auxiliary node supplementary map is generated.

[0016] The blind spot map of the main node and the supplementary map of the auxiliary node are topologically stitched together to output the joint sensing blind spot map.

[0017] In one embodiment, the method further includes:

[0018] Identify moving obstacles from the real-time monitoring data of the main sensing node, and extract the position coordinates, velocity vector and external dimensions of the moving obstacles;

[0019] The trajectory of the moving obstacle is predicted within a preset time period based on its position coordinates and velocity vector, and the area continuously occluded during its movement is calculated based on its shape and size to generate a dynamically supplemented map.

[0020] The blind spot map of the main node, the supplementary map of the auxiliary node, and the dynamically supplemented map are topologically stitched together to update the joint sensing blind spot map.

[0021] In one embodiment, the motion state data of the traffic participant includes its corresponding position coordinates, velocity vector, and acceleration vector;

[0022] A traffic participant is deemed to meet the high-risk criteria when all of the following conditions are met:

[0023] The distance from the location coordinates of a traffic participant to the boundary of the joint sensing blind zone is less than the movable distance within a preset time period calculated based on the traffic participant's velocity vector.

[0024] The angle between the velocity vector direction of traffic participants and the normal vector of the joint perception blind zone boundary is less than a preset critical angle;

[0025] The magnitude of the velocity vector or acceleration vector of a traffic participant exceeds the preset safety threshold corresponding to the category of the traffic participant.

[0026] In one embodiment, the step of generating at least one risk migration path and its corresponding risk propagation area and time window based on the motion state data of the traffic participants, the road structure rules of the joint sensing blind spot map, the traffic density data reported by the auxiliary sensing nodes, and the historical trajectory database includes:

[0027] Extract the velocity vector and direction angle data of high-risk objects before they enter the blind zone as motion basis data;

[0028] The blind spot exit directions, lane connections, and traffic rule constraints are analyzed from the joint perception blind spot map to generate a path topology network.

[0029] Calculate the traffic resistance coefficient of each route branch based on surrounding traffic density data;

[0030] Based on historical trajectory databases, the probability of path selection in similar scenarios is matched;

[0031] The motion baseline data, traffic resistance coefficient, and path selection probability are weighted and fused to calculate the generation probability of each migration path;

[0032] Select the first preset number of paths with the highest generation probability, calculate the spatial reachability along each path based on the maximum and minimum speeds of high-risk objects, and generate a risk propagation zone polygon and time window.

[0033] In one embodiment, the step of weightedly fusing motion base data, traffic resistance coefficient, and path selection probability to calculate the generation probability of each migration path includes:

[0034] Assign velocity direction weights to the motion basis data and calculate the probability of the path continuing as a straight line;

[0035] Assign density influence weights to the traffic resistance coefficient and calculate the path avoidance probability;

[0036] Assign scene similarity weights to the historical path selection probabilities and calculate the weighted historical path selection probabilities.

[0037] The probability of straight path continuation, the probability of path avoidance, and the weighted historical path selection probability are superimposed according to a preset proportional coefficient and then normalized and sorted.

[0038] In one embodiment, the step of generating an early warning instruction based on the risk propagation zone range and time window includes:

[0039] Based on the polygon of the risk propagation zone, the name of the road in which it is located and its position relative to the road network node are analyzed.

[0040] Based on the high-risk object category, a pre-set warning template library is matched to generate descriptive text;

[0041] By binding the descriptive text with the spatial location code of the risk propagation zone polygon, a tiered early warning instruction is generated.

[0042] In one embodiment, the method further includes:

[0043] Within the time window, the scope of the risk spread area is continuously monitored. If no high-risk objects are identified and the maximum window time has been exceeded, the scope of the risk spread area is gradually reduced until the warning is lifted.

[0044] In one embodiment, the step of continuously monitoring the risk propagation area within a time window, and gradually narrowing the risk propagation area until the warning is lifted if no high-risk object is identified and the maximum window time has been exceeded, includes:

[0045] Based on the initial risk propagation zone, time slices are generated at preset time intervals;

[0046] At the beginning of each time slice, if no high-risk object is identified, the shrinkage boundary of the propagation zone is calculated based on the object's minimum velocity, and at the end of the time slice, the shrinkage boundary is used as the new range of the risk propagation zone.

[0047] The alert will be terminated when the number of time slices in which no high-risk objects are identified consecutively exceeds a preset threshold, or when the cumulative time exceeds the maximum window time.

[0048] In addition, to achieve the above objectives, this application also proposes a traffic safety monitoring and early warning system, the system comprising: a memory, a processor, and a traffic safety monitoring and early warning program stored in the memory and executable on the processor, the traffic safety monitoring and early warning program being configured to implement the steps of the traffic safety monitoring and early warning method.

[0049] The traffic safety monitoring and early warning method and system proposed in this application generate a joint sensing blind zone map through an edge sensing collaborative network, dynamically predict the path of high-risk objects after entering the blind zone, and coordinate with auxiliary sensing nodes for directional detection. Combined with traffic density and historical data, it generates risk propagation early warning, which solves the problems of insufficient blind zone coverage and weak prediction capability of traditional systems. It has the advantages of dynamically identifying blind zones, coordinating sensing resources, and actively predicting risks, and can improve the reliability and real-time performance of traffic safety monitoring in a low-cost manner. Attached Figure Description

[0050] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart illustrating an embodiment of the traffic safety monitoring and early warning method of this application.

[0053] Figure 2 For this application Figure 1 A detailed flowchart of step S100 is provided in one embodiment.

[0054] Figure 3 This is a flowchart illustrating another embodiment of the traffic safety monitoring and early warning method of this application.

[0055] Figure 4 For this application Figure 1 A detailed flowchart of step S500 is provided in one embodiment.

[0056] Figure 5 For this application Figure 1 A detailed flowchart of step S600 is provided in one embodiment.

[0057] Figure 6 This is a flowchart illustrating yet another embodiment of the traffic safety monitoring and early warning method of this application.

[0058] Figure 7 This is a schematic diagram of a traffic safety monitoring and early warning system according to an embodiment of the present application.

[0059] Explanation of icon numbers:

[0060] 10. Memory; 20. Processor.

[0061] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0062] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of this application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0063] It should be understood that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0064] In existing technologies, traffic safety monitoring systems generally rely on fixed-location sensing devices, such as cameras, radar, and inductive loop detectors. These devices are limited by their installation location and physical characteristics; for example, visual obstructions behind curves or insufficient detection angles can lead to blind spots. Particularly in intersection turning scenarios, vehicles cannot detect pedestrians suddenly appearing in the blind spot; in highway ramp merging areas, vehicles on the main road struggle to detect slow-moving merging vehicles within the blind spot. Existing solutions typically achieve coverage by increasing device density or relying on vehicle-mounted terminals, but the former is costly and difficult to deploy comprehensively, while the latter is limited by device availability and cannot form an effective proactive warning mechanism.

[0065] To address the aforementioned issues, the inventors discovered that the core contradiction of traditional methods lies in the inability of static perception networks to adapt to dynamic traffic environments. By analyzing the risk propagation patterns in blind spots, they realized that risks are not static but migrate to visible areas along with the movement trajectories of traffic participants. Based on this, they proposed constructing a dynamic collaborative perception network, achieving relay-style blind spot detection through the linkage of primary and secondary nodes. The specific approach is as follows: First, the primary node identifies high-risk objects and predicts their initial position entering the blind spot, triggering directional detection by secondary nodes. When the secondary nodes fail to detect the target, they combine road structure, real-time traffic density, and historical trajectories to deduce the risk propagation path, thereby generating early warning information before the target is actually exposed.

[0066] Based on this, the embodiments of this application provide a traffic safety monitoring and early warning method, referring to... Figure 1The traffic safety monitoring and early warning method includes steps S100 to S600, wherein:

[0067] Step S100: By deploying an edge sensing collaborative network in the target area, real-time monitoring data of the main sensing node of the edge sensing collaborative network and device location data of the auxiliary sensing node of the edge sensing collaborative network are obtained, and combined with the pre-stored map data, a joint sensing blind spot map is generated.

[0068] Step S200: The main sensing node acquires the category data and motion state data of traffic participants based on the real-time monitoring data, and when it determines that a traffic participant meets the preset high-risk conditions based on the category data and motion state data, it marks the traffic participant as a high-risk object, and predicts the initial location range of the participant after entering the joint sensing blind zone based on its motion state data.

[0069] In step S300, the main sensing node broadcasts the marking information, motion state data, and initial position range of the high-risk object to the auxiliary sensing node corresponding to the initial position range.

[0070] In step S400, the auxiliary sensing node that receives the broadcast activates its sensing module, adjusts the sensing direction to focus on the initial position range for directional detection, and generates blind zone detection results.

[0071] Step S500: If the blind spot detection results do not identify the high-risk object within a preset time, based on the motion state data of the traffic participants, the road structure rules of the joint perception blind spot map, the traffic density data reported by the auxiliary perception nodes, and the historical trajectory database, at least one risk migration path and its corresponding risk propagation area range and time window are generated.

[0072] Step S600: Generate an early warning instruction based on the risk propagation zone and time window, and release the early warning information through at least one of roadside equipment, vehicle-mounted terminals, and mobile terminals.

[0073] In this embodiment, the edge-aware collaborative network refers to a distributed sensing system composed of master nodes and auxiliary nodes. Master nodes can use fixed cameras or radar for continuous monitoring, while auxiliary nodes can use mobile millimeter-wave radar or acoustic sensors for on-demand activation. The joint sensing blind spot map is generated by fusing the blind spots of master nodes, the coverage areas of auxiliary nodes, and the occlusion areas of dynamic obstacles, and is used to characterize the spatial distribution of real-time monitoring blind spots. High-risk object marking refers to the system marking traffic participants as potential risk sources when they approach the blind spot and their movement is abnormal, such as exceeding a speed threshold or pointing directly at the boundary of the blind spot. Risk migration path inference refers to predicting the possible spatiotemporal range of the target by combining the target's movement trend, road topology constraints, and historical behavior patterns, for example, by determining the most likely route by weighted calculation of path selection probability.

[0074] In this embodiment, the primary sensing node continuously collects data on the categories and motion states of traffic participants. When a target is detected rapidly approaching the blind zone boundary, a risk marking mechanism is triggered. The system predicts the target's initial position after entering the blind zone based on its speed and sends an activation command to the auxiliary nodes within that area. After activation, the auxiliary nodes adjust their detection angle to focus on the predicted area. If detection fails, the path deduction module is activated. During the deduction process, the system analyzes the blind zone exit direction and lane connection relationship, calculates the path resistance based on real-time traffic density, and matches similar scenarios in historical trajectories to generate multiple possible paths and their spatiotemporal impact ranges. Finally, based on the spatial location and time window of the risk propagation zone, tiered early warning information is generated, such as displaying risks in specific directions on roadside screens or sending location warnings to vehicle terminals.

[0075] In this embodiment, blind zone dynamic detection is achieved through the collaboration of primary and secondary nodes. The secondary node is activated only when a high-risk object is detected, reducing energy consumption and computational load. A risk propagation path is predicted using a simulation model, issuing an early warning before the target actually appears. Simultaneously, the early warning information is bound to a dynamic spatiotemporal range, improving the accuracy of information transmission. Thus, this application can proactively identify high-risk targets approaching the blind zone and trigger directional detection, preventing targets from completely losing tracking after escaping monitoring. When a target is not detected, its possible migration path is predicted through multi-dimensional data analysis, the risk impact range is pre-defined, and a tiered early warning is generated. Therefore, without the need for a large-scale increase in equipment, the risk of traffic accidents caused by blind zones is effectively reduced, while also minimizing the interference of invalid early warnings on traffic participants.

[0076] In one feasible implementation, refer to Figure 2 Step S100 includes steps S110 to S140, wherein:

[0077] In step S110, the main sensing node collects the first monitoring data through the fixed sensing device, and at the same time receives the device location status data sent by each auxiliary sensing node.

[0078] Step S120: Based on the static obstacle coordinates in the pre-stored map data, perform spatial overlay calculation on the coverage area of ​​the first monitoring data to generate a blind spot map of the main node.

[0079] Step S130: Integrate the device location status data and sensing capability parameters of the auxiliary sensing nodes, calculate their coverage area, and generate an auxiliary node supplementary map.

[0080] Step S140: Topologically stitch the blind zone map of the main node and the supplementary map of the auxiliary node to output the joint sensing blind zone map.

[0081] In this embodiment, spatial overlay calculation refers to performing geometric calculations on the monitoring coverage area of ​​the main sensing node based on static obstacle coordinates to eliminate undetectable areas obscured by obstacles. Specifically, this can be achieved using vector map overlay and Boolean operations, used to accurately identify inherent blind spots formed by physical obstructions of the main node's fixed sensing equipment. Equipment location status data refers to the geographical location, installation height, and operational status information of the auxiliary sensing node. This can be collected through a GPS module and status sensors, used to dynamically track the effective detection position of the auxiliary node. Sensing capability parameters refer to the detection angle, effective distance, and resolution of the auxiliary sensing node. These can be obtained from the equipment technical manual or calibration experiments, used to calculate the actual coverage area of ​​the auxiliary node. Topology stitching refers to spatially aligning and merging map data from different coordinate systems. This can be achieved using coordinate transformation algorithms and polygon merging algorithms, used to construct a complete blind spot map that complements the sensing areas of the main and auxiliary nodes.

[0082] In this embodiment, the main sensing node collects real-time monitoring data through a fixedly installed camera or radar. Combined with the coordinate information of buildings and bridges in a pre-stored map, a spatial overlay algorithm is used to subtract obscured areas, generating a blind spot map for the main node. The positional status data and sensing capability parameters of the auxiliary sensing nodes are fused, and their detection coverage is calculated through geometric projection to generate a supplementary map for the auxiliary nodes. After coordinate transformation, the blind spot map of the main node and the supplementary map of the auxiliary nodes are used to eliminate overlapping areas and connect adjacent boundaries using a polygon merging algorithm, ultimately forming a joint sensing blind spot map. Thus, the inherent blind spot of the main node is covered by the dynamically supplemented area of ​​the auxiliary nodes, while avoiding blind spot calculation errors caused by ignoring changes in device position or differences in sensing capabilities in traditional solutions.

[0083] Understandably, existing technologies generally rely on monitoring data from a single master node to calculate blind spots, or simply overlay the fixed coverage area of ​​auxiliary nodes. This results in blind spot maps being unable to dynamically adapt to changes in device location and environmental occlusion. In contrast, this solution dynamically integrates blind spot calculations from the master node with areas supplemented by auxiliary nodes, combined with real-time device location and sensing capability parameters. This allows the blind spot map to adjust its coverage area according to the deployment status of auxiliary nodes. At the same time, topology stitching is used to eliminate map stitching errors, improving the accuracy and real-time performance of blind spot identification.

[0084] Through the above technical solution, this application solves the problem of insufficient dynamic coverage of monitoring blind spots caused by physical limitations of fixed sensing devices. It generates a high-precision joint blind spot map by fusing collaborative sensing from main and auxiliary nodes with map data. Specifically, the main node's blind spot calculation eliminates permanent blind spots caused by static obstacles, while the auxiliary nodes dynamically compensate for the main node's detection blind spots by supplementing the area. A topology stitching algorithm achieves seamless integration of heterogeneous maps. Therefore, in the scenario of monitoring turning vehicles at intersections, the blind spot range obscured by buildings can be accurately identified, and the coverage area can be expanded by using auxiliary nodes to move equipment, avoiding the omission of warnings caused by incomplete blind spot calculations in traditional solutions. In the merging area of ​​highway ramps, by dynamically adjusting the map supplemented by auxiliary nodes, temporary blind spots caused by vehicle obstruction are covered in real time, improving the tracking continuity of high-risk objects.

[0085] In one feasible implementation, refer to Figure 3 The implementation method also includes steps S150 to S170, wherein:

[0086] Step S150: Identify moving obstacles from the real-time monitoring data of the main sensing node, and extract the position coordinates, velocity vector and external dimensions of the moving obstacles;

[0087] Step S160: Based on the position coordinates and velocity vector, predict the movement trajectory of the moving obstacle within a preset time period, and calculate the area continuously occluded during its movement based on its shape and size, and generate a dynamically supplemented map.

[0088] Step S170: Topologically stitch the main node blind zone map, the auxiliary node supplementary map, and the dynamically supplemented map to update the joint sensing blind zone map.

[0089] In this embodiment, a moving obstacle refers to an entity with autonomous movement capabilities in a traffic environment, such as a freight vehicle or construction machinery, whose presence can be identified through a target detection algorithm. Position coordinates refer to the obstacle's location data in three-dimensional space, specifically obtained using GPS and LiDAR point cloud matching technology. The velocity vector contains movement speed and direction information, which can be calculated through differential calculation of multi-frame point cloud data. The external dimensions refer to the projected dimensions of the obstacle in the direction of movement, which can be extracted from the bounding box parameters after point cloud clustering. The dynamically supplemented map refers to a time-varying occlusion region model, specifically calculated using a trajectory prediction algorithm to determine the obstacle's movement path within a future time period, and combined with the external dimensions to generate a set of spatial coordinates for continuously occluded regions.

[0090] In this embodiment, the main sensing node identifies moving obstacles through real-time monitoring data and extracts their spatial location, movement trend, and physical size parameters. Based on motion trajectory prediction algorithms, such as the Kalman filter model, the moving path of the obstacle within the next 10 seconds is calculated. The continuous occlusion area caused by the obstacle on this path is calculated based on the shape and size parameters, for example, by generating a fan-shaped or rectangular occlusion area using a geometric projection algorithm. The dynamic occlusion area is spatially overlaid with the static blind spot map to form a composite blind spot model that integrates the coverage capabilities of fixed obstacles, auxiliary nodes, and the influence of moving obstacles. This model achieves data fusion in different coordinate systems through a topology stitching algorithm; for example, rasterization is used to map the dynamic occlusion area to the coordinate system of the main node's blind spot map.

[0091] Compared to existing technologies, traditional blind spot monitoring systems only consider the coverage area of ​​fixed obstacles and fixed sensing devices, failing to address time-varying blind spots caused by moving obstacles. For example, when a large truck turns at an intersection, its body obstructs the monitoring field of view in different areas over time. This solution, however, establishes a dynamic occlusion model by tracking the spatiotemporal characteristics of moving obstacles in real time, enabling the blind spot map to reflect real-time changes in the traffic environment. This effectively solves the problem of blind spot monitoring failure caused by dynamic occlusion from moving obstacles. By identifying and predicting the movement trajectory of moving obstacles in real time and accurately calculating the continuous occlusion areas they cause, the joint sensing blind spot map has dynamic update capabilities. This avoids sudden changes in monitoring blind spots caused by changes in the position of moving obstacles, ensuring the continuity of tracking after high-risk objects enter the blind spot, and improving the environmental adaptability and monitoring reliability of the traffic safety early warning system.

[0092] In one feasible implementation, the motion state data of the traffic participant includes its corresponding position coordinates, velocity vector, and acceleration vector. A traffic participant is deemed to meet a high-risk condition when it simultaneously meets the following conditions: the distance from the traffic participant's position coordinates to the boundary of the joint perception blind zone is less than the movable distance within a preset time period calculated based on the traffic participant's velocity vector; the angle between the direction of the traffic participant's velocity vector and the normal vector of the joint perception blind zone boundary is less than a preset critical angle; and the magnitude of the traffic participant's velocity vector or acceleration vector exceeds a preset safety threshold corresponding to the traffic participant's category.

[0093] In this embodiment, location coordinates refer to the real-time positioning data of traffic participants in three-dimensional space, which can be implemented using a Global Positioning System (GPS) or visual positioning algorithms to establish the spatial relationship between the object and the blind spot boundary. Velocity vectors are motion parameters containing speed and direction information, which can be implemented through multi-frame image tracking or radar echo analysis to predict the object's movement capability within a preset time period. Acceleration vectors are vector parameters of velocity change, which can be implemented using an inertial measurement unit (IMU) or differential velocity calculation to assess the risk of sudden changes in motion state. The preset time period is a dynamically adjusted time window, which can be set to, for example, 3 to 5 seconds depending on the road type to match reaction time requirements in different scenarios. The preset critical angle is the maximum permissible deviation angle between the velocity direction and the normal vector of the blind spot boundary, which can be set to, for example, 30 degrees, to exclude motion trajectories that do not intersect with the blind spot. The preset safety threshold is a speed or acceleration limit set according to the type of traffic participant, specifically 5 m / s for pedestrians and 15 m / s for motor vehicles, used to establish classification and evaluation standards.

[0094] In this embodiment, a dynamic risk assessment model is constructed by acquiring the three-dimensional motion parameters of traffic participants in real time. First, the Euclidean distance from the object's current position to the blind zone boundary is calculated. Combined with its current speed, the maximum achievable displacement distance within a preset time period is calculated. When the actual distance is less than this displacement distance, the first condition is triggered. Second, the cosine of the angle between the velocity direction and the normal vector of the blind zone boundary is calculated using vector dot product operations. When this value exceeds a preset cosine threshold for a critical angle, the second condition is triggered. Finally, a preset speed or acceleration safety threshold is matched according to the object type. When the actual parameter exceeds the corresponding threshold, the third condition is triggered. These three conditions form a progressive screening mechanism: the first condition ensures the object has the physical possibility of entering the blind zone; the second condition verifies the spatial correlation between its motion trajectory and the blind zone; and the third condition assesses the degree of abnormality in its motion state. Only when all three conditions are met simultaneously is the object deemed a high-risk object and enters the warning process.

[0095] In this embodiment, through the collaborative analysis of multi-dimensional dynamic parameters, misjudgments of laterally moving objects are eliminated, and invalid warnings for low-speed compliant vehicles are avoided. Furthermore, classification threshold settings adapt to the different motion characteristics of various traffic participants. By establishing a triple verification mechanism based on spatial distance, direction of movement, and dynamic parameters, high-risk objects that are truly likely to enter the blind spot are accurately identified, avoiding interference with warnings for unrelated traffic participants. The classification threshold settings enable the system to intelligently distinguish the different risk characteristics of pedestrians, non-motorized vehicles, and motorized vehicles, improving the accuracy of warnings.

[0096] In one feasible implementation, refer to Figure 4 Step S500 includes steps S510 to S560, wherein:

[0097] Step S510: Extract the velocity vector and direction angle data of the high-risk object before it enters the blind zone as motion basis data;

[0098] Step S520: Analyze the blind spot exit direction, lane connection relationship and traffic rule constraints from the joint perception blind spot map to generate a path topology network;

[0099] Step S530: Calculate the traffic resistance coefficient of each path branch based on the surrounding traffic density data;

[0100] Step S540: Match path selection probabilities for similar scenarios based on the historical trajectory database;

[0101] Step S550: The motion base data, traffic resistance coefficient and path selection probability are weighted and fused to calculate the generation probability of each migration path;

[0102] Step S560: Select the preset number of paths with the highest generation probability, calculate the spatial reachability along each path based on the maximum and minimum speeds of high-risk objects, and generate a risk propagation zone polygon and time window.

[0103] In this embodiment, motion base data refers to the velocity vector and direction angle data of a high-risk object before entering the blind zone. Specifically, this can be implemented using real-time trajectory data collected by radar or cameras, reflecting the motion inertial characteristics of the object before entering the blind zone. The path topology network refers to a node connection graph generated based on the road structure. Specifically, it can be implemented by parsing lane lines, traffic signs, and blind zone exit coordinates from map data, constraining the physical feasibility of path prediction. The traffic resistance coefficient is a quantitative indicator characterizing the ease of passage through path branches. Specifically, it can be calculated by combining vehicle density data reported by auxiliary sensing nodes with lane width, reflecting the impact of real-time traffic flow on path selection. The historical trajectory database refers to a database storing the behavioral patterns of traffic participants in similar scenarios. Specifically, it can be implemented by clustering historical trajectory data using data mining algorithms, providing a probability distribution reference for path selection. Weighted fusion refers to assigning weights to different data sources and probabilistically superimposing them. Specifically, it can be implemented using linear weighting or neural network models, used to comprehensively optimize path prediction results by integrating multiple dimensions of factors. The risk propagation zone polygon and time window refer to the spatial reach and time range of high-risk objects along the path. Specifically, it can be calculated by taking the spatial boundary at the maximum and minimum speeds and combining it with the path length to calculate the time threshold, which is used to determine the spatiotemporal dimensions of the early warning coverage.

[0104] In this embodiment, when a high-risk object is not detected in the blind spot detection, its velocity vector and direction angle before entering the blind spot are first extracted as the basis for predicting its subsequent movement trend. For example, when a vehicle enters the blind spot at a speed of 30 km / h traveling eastward, its motion base data will include this speed and direction information. Next, the road structure in the blind spot map is analyzed, for example, the exit direction of the blind spot is identified as the northeast ramp, and the lane connection relationship and lane change prohibition rules of this area are obtained to generate a topology network containing multiple possible path branches. Further, based on the traffic density data reported by the auxiliary sensing nodes, for example, the current vehicle density of the northeast ramp is 85 vehicles / km, the traffic resistance coefficient of this path branch is calculated to reflect the difficulty of vehicle merging. At the same time, similar scenarios are matched from the historical trajectory database, for example, 70% of vehicles at the same location and similar speed in the past 30 days choose to go straight into the main road, generating the path selection probability. Subsequently, motion basis data are weighted by velocity direction (e.g., consistency between the current direction and the path branch direction accounts for 40%); traffic resistance coefficient is weighted by density influence (e.g., high-density path avoidance accounts for 30%); and historical path selection probability is weighted by scene similarity (e.g., matching degree above 80% accounts for 30%). The generation probability of each path is calculated through weighted fusion. Finally, the three paths with the highest probabilities are selected, and the spatial reachability of each path within the next 30 seconds is calculated at a maximum speed of 50 km / h and a minimum speed of 20 km / h, generating polygonal regions covering the three paths and corresponding time windows.

[0105] In this embodiment, a multi-dimensional prediction model is constructed by integrating road topology, real-time traffic density, and historical behavior patterns. This model can dynamically generate risk propagation areas that conform to road rules and reflect real-time traffic conditions, thus solving the problem of inaccurate path prediction caused by blind spot occlusion. For example, in the scenario of blind spots on highway ramps, the system can accurately predict the range of main road lanes that slow-moving vehicles may merge into and issue targeted warnings within the corresponding time window, avoiding the warning information from being invalid due to an excessively large prediction range or the warning failing due to a prediction time lag.

[0106] In one feasible implementation, step S550 includes assigning velocity direction weights to the motion base data and calculating the probability of straight path continuation; assigning density influence weights to the traffic resistance coefficient and calculating the probability of path avoidance; assigning scene similarity weights to the historical path selection probability and calculating the weighted historical path selection probability; and superimposing the probability of straight path continuation, the probability of path avoidance, and the weighted historical path selection probability according to a preset proportional coefficient and then normalizing and sorting them.

[0107] In this embodiment, the speed direction weight refers to an influence factor set based on the consistency between the current speed vector and the path direction of traffic participants. Specifically, it can be calculated using a cosine similarity algorithm to calculate the cosine value of the angle between the speed vector and the path direction, and then multiplied by a preset coefficient. The density influence weight refers to a quantitative indicator of the resistance to path selection based on traffic density data. Specifically, it can be implemented by performing an exponential function transformation on the ratio of the density value to a preset threshold. The scene similarity weight refers to a correction factor set based on the matching degree between the current traffic scene and the historical trajectory database. Specifically, it can be implemented by calculating the Euclidean distance between the current scene and historical scenes using a spatiotemporal trajectory similarity algorithm, and then transforming it using a Gaussian kernel function.

[0108] In this embodiment, the probability of straight-line continuation is calculated based on the consistency between the velocity vector and the candidate path direction. For example, when the angle between the velocity vector and a certain path direction is less than 15 degrees, a higher weight is assigned. The path avoidance probability is dynamically adjusted based on real-time traffic density data. For example, when the traffic density of a certain path branch exceeds a threshold, the generation probability is reduced. The weighted historical path selection probability is calculated by matching path selection patterns of similar scenarios in the historical trajectory database. For example, under historical data where the probability of a left turn at the same intersection is 70%, the probability of the current left-turn path is weighted and increased. After the three types of probabilities are superimposed according to preset proportional coefficients, the generation probability ranking of each migration path is obtained through normalization processing. For example, the probability of straight-line continuation accounts for 40%, the probability of path avoidance accounts for 35%, and the probability of historical path selection accounts for 25%.

[0109] In this embodiment, the above-mentioned technical solution addresses the problem of insufficient accuracy in predicting the migration path of risks in blind spots. It can generate a multi-dimensional fusion path probability distribution by comprehensively considering the movement trend, real-time traffic interference, and historical behavior patterns of traffic participants after they enter the blind spot. This solution improves the ability to predict potential risks in complex scenarios such as curves and ramps, providing a reliable basis for the generation of subsequent warning instructions.

[0110] In one feasible implementation, refer to Figure 5 Step S600 includes steps S610 to S630, wherein:

[0111] Step S610: Based on the risk propagation zone polygon, parse the name of the road in which it is located and its position relative to the road network node;

[0112] Step S620: Match the pre-set warning template library according to the high-risk object category to generate descriptive text;

[0113] Step S630: Bind the descriptive text to the spatial location code of the risk propagation zone polygon to generate a graded early warning instruction.

[0114] In this embodiment, the risk propagation zone polygon refers to the geometric boundary of the region calculated based on the path reachability. Specifically, it can be spatially represented using a sequence of polygon vertex coordinates to characterize the geographical area that a high-risk object may move and cover. The road name and its position relative to road network nodes refer to mapping the polygon coordinates to the actual road network topology using a geographic information system. Specifically, it can be described using a combination of road network node numbers and offset distances to associate abstract coordinates with the location of traffic management entities. The preset warning template library refers to a set of text generation rules stored according to object type. Specifically, it can be constructed using natural language processing technology to include a standard sentence structure library containing object characteristic verbs and risk level adverbs to ensure semantic standardization. Spatial location encoding refers to converting geographic coordinates into a machine-recognizable structured data format. Specifically, it can use geographic coding standards or custom area identifiers to achieve cross-platform data interoperability.

[0115] In this embodiment, after the risk propagation zone polygon is defined spatially using a sequence of vertex coordinates, the name of the road it covers and adjacent road network nodes are reverse-analyzed using a geographic information system. For example, the area covered by the polygon is identified as "50 meters south of the intersection of XX Road and YY Road". Corresponding templates are called based on the high-risk object category; for example, pedestrians are matched with the "Caution: Pedestrians" template, and vehicles are matched with the "Caution: Slow-moving Vehicles Ahead" template. The descriptive text is bound to the spatial location code to form a composite data structure. For example, the text "There is a risk of pedestrians crossing XX Road from south to north" is associated with a geocode, allowing roadside equipment to match electronic map coordinates via geocode, and vehicle-mounted terminals to display the risk description via text.

[0116] Compared to existing technologies, traditional early warning systems can only send general location alerts, such as "risk 200 meters ahead," without associating them with actual road names and road network structures, making it difficult for drivers to quickly locate risk points. Existing solutions lack standardized templates for generated warning text, leading to semantic ambiguity when parsed by different devices. This solution integrates geographic information parsing with templated text generation to achieve precise correspondence between warning content and road network entities, while ensuring semantic consistency. It resolves data compatibility issues between heterogeneous devices and achieves a two-way mapping between geographic information of the risk propagation area and natural language descriptions, ensuring that warning instructions contain both machine-resolvable spatial encoding and human-understandable textual prompts.

[0117] In one feasible implementation, the method further includes: continuously monitoring the range of the risk propagation zone within a time window; if no high-risk object is identified and the maximum window time is exceeded, gradually narrowing the range of the risk propagation zone until the warning is lifted.

[0118] In this embodiment, gradually narrowing the risk propagation zone refers to a propagation zone shrinkage mechanism based on the minimum movement speed constraint of high-risk objects. Specifically, this can be achieved using a spatial polygon boundary layer-by-layer reduction algorithm, which dynamically adjusts the monitoring coverage area by calculating the shortest distance the object may move at its minimum speed. The maximum window time refers to the longest duration threshold for monitoring the risk propagation zone, which can be implemented using a dynamic time parameter configuration table that matches traffic scenario types, setting differentiated termination conditions based on road grade and traffic flow characteristics.

[0119] In this embodiment, within the polygonal coverage area of ​​the initial risk propagation zone, multiple consecutive time slices are generated at preset time intervals. At the beginning of each time slice, if no high-risk object is detected, the minimum displacement that the object might move within the current time slice is calculated based on the minimum motion speed parameter corresponding to that object category, thereby deriving the contraction range of the propagation zone boundary. At the end of the time slice, the contracted boundary is updated as the new risk propagation zone range. When the number of consecutive time slices in which no high-risk object is identified exceeds a preset threshold, or the cumulative monitoring time reaches the maximum window time, a warning termination command is triggered, and the monitoring of the risk propagation zone and the issuance of warning information are lifted.

[0120] In one specific implementation, reference is made to... Figure 6 The steps of continuously monitoring the risk propagation area within the time window, and gradually narrowing the risk propagation area until the warning is lifted if no high-risk object is identified and the maximum window time has been exceeded, include steps S710 to S730, where:

[0121] Step S710: Based on the initial risk propagation zone, generate time slices at preset time intervals;

[0122] Step S720: At the beginning of each time slice, if no high-risk object is identified, the shrinkage boundary of the propagation zone is calculated based on the minimum speed of the object, and at the end of the time slice, the shrinkage boundary is used as the new range of the risk propagation zone.

[0123] Step S730: When the number of time slices in which no high-risk objects are continuously identified exceeds a preset threshold, or the cumulative time exceeds the maximum window time, the warning is terminated.

[0124] In this embodiment, time slicing refers to dividing the warning duration into multiple monitoring periods of equal or unequal length. This can be implemented using a periodically triggered time window segmentation algorithm to establish a phased monitoring mechanism and reduce system resource consumption. The shrinking boundary refers to the reduced warning area calculated based on the target's minimum movement speed. Specifically, it can be calculated using a speed-time product model to determine the spatial reachability, preventing excessive expansion of the warning area while ensuring safety. The continuous unidentified threshold refers to the number of consecutive monitoring failures required to trigger warning termination. This can be implemented using a sliding window counter to prevent false alarms caused by brief monitoring failures. The maximum window time refers to the time range of a high-risk object on the path. This can be achieved by calculating the spatial boundary using the maximum and minimum speeds and combining this with the path length to calculate the time threshold.

[0125] In this embodiment, the initial risk propagation zone is divided into multiple time slices, with each slice's starting point triggering a monitoring and judgment. If no high-risk object is detected, the shortest possible distance it can travel within that time period is calculated based on its minimum movement speed, generating a corresponding shrinkage boundary. This boundary serves as the warning range for the next stage, and the area is gradually reduced through iterative updates. When no target is detected in multiple consecutive slices, or when the total warning time exceeds a preset upper limit, the system automatically terminates the warning process. This mechanism, through spatiotemporally coupled shrinkage logic, avoids missed warnings while suppressing the spread of invalid warnings.

[0126] In some specific implementations, the time interval can be dynamically adjusted based on the traffic scenario; for example, shorter intervals are used for urban roads, while longer intervals are used for highways. The calculation of the contraction boundary can be constrained by the road topology, for example, using an arc length calculation model in curved areas.

[0127] In this embodiment, a dynamic shrinkage mechanism and dual termination conditions are used to optimize the allocation of early warning resources while maintaining security, avoid excessive early warnings caused by fixed thresholds, and solve the problem of continuous expansion of the early warning area caused by blind spot monitoring failure. By dynamically adjusting the early warning range through phased shrinkage and termination conditions, the system's computing load and communication resource consumption are reduced, while also reducing the interference of false alarms on traffic participants.

[0128] This application also provides a traffic safety monitoring and early warning system, for reference. Figure 7 The system includes: a memory 10, a processor 20, and a traffic safety monitoring and early warning program stored on the memory 10 and executable on the processor 20, wherein the traffic safety monitoring and early warning program is configured to implement the steps of the traffic safety monitoring and early warning method.

[0129] The traffic safety monitoring and early warning system provided in this application, employing the traffic safety monitoring and early warning method in the above embodiments, can improve the reliability and real-time performance of traffic safety monitoring. Compared with the prior art, the beneficial effects of the traffic safety monitoring and early warning system provided in this application are the same as those of the traffic safety monitoring and early warning method provided in the above embodiments, and other technical features of the traffic safety monitoring and early warning system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0130] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. All equivalent structural transformations made under the technical concept of this application using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included within the scope of patent protection of this application.

Claims

1. A traffic safety monitoring and early warning method, characterized in that, The traffic safety monitoring and early warning methods include: By deploying an edge sensing collaborative network in the target area, real-time monitoring data of the main sensing node of the edge sensing collaborative network and device location data of the auxiliary sensing node of the edge sensing collaborative network are obtained, and combined with pre-stored map data, a joint sensing blind spot map is generated. The main sensing node acquires the category data and motion state data of traffic participants based on the real-time monitoring data. When it determines that a traffic participant meets the preset high-risk conditions based on the category data and motion state data, it marks the traffic participant as a high-risk object and predicts the initial location range of the participant after entering the joint sensing blind zone based on the motion state data. The primary sensing node broadcasts the high-risk object's marking information, motion status data, and initial location range to the secondary sensing node corresponding to the initial location range. Upon receiving the broadcast, the auxiliary sensing node activates its sensing module, adjusts its sensing direction to focus on the initial position range for directional detection, and generates blind zone detection results. If the blind spot detection results do not identify the high-risk object within a preset time, at least one risk migration path and its corresponding risk propagation area range and time window are generated based on the motion state data of the traffic participants, the road structure rules of the joint perception blind spot map, the traffic density data reported by the auxiliary perception nodes, and the historical trajectory database. Early warning instructions are generated based on the scope of the risk propagation zone and the time window, and the early warning information is released through at least one of roadside equipment, vehicle-mounted terminals, and mobile terminals.

2. The traffic safety monitoring and early warning method as described in claim 1, characterized in that, The step of generating a joint sensing blind spot map by acquiring real-time monitoring data of the main sensing node and device location data of the auxiliary sensing nodes of the edge sensing collaborative network deployed in the target area, and combining this data with pre-stored map data, includes: The main sensing node collects the first monitoring data through fixed sensing devices, and at the same time receives the device location status data sent by each auxiliary sensing node; Based on the coordinates of static obstacles in the pre-stored map data, spatial overlay calculations are performed on the coverage area of ​​the first monitoring data to generate a blind spot map of the main node. By integrating the device location status data and sensing capability parameters of the auxiliary sensing nodes, their coverage area is calculated, and an auxiliary node supplementary map is generated. The blind spot map of the main node and the supplementary map of the auxiliary node are topologically stitched together to output the joint sensing blind spot map.

3. The traffic safety monitoring and early warning method as described in claim 2, characterized in that, The method further includes: Identify moving obstacles from the real-time monitoring data of the main sensing node, and extract the position coordinates, velocity vector and external dimensions of the moving obstacles; The trajectory of the moving obstacle is predicted within a preset time period based on its position coordinates and velocity vector, and the area continuously occluded during its movement is calculated based on its shape and size to generate a dynamically supplemented map. The blind spot map of the main node, the supplementary map of the auxiliary node, and the dynamically supplemented map are topologically stitched together to update the joint sensing blind spot map.

4. The traffic safety monitoring and early warning method as described in claim 1, characterized in that, The motion state data of the traffic participants includes their corresponding position coordinates, velocity vector, and acceleration vector; A traffic participant is deemed to meet the high-risk criteria when all of the following conditions are met: The distance from the location coordinates of a traffic participant to the boundary of the joint sensing blind zone is less than the movable distance within a preset time period calculated based on the traffic participant's velocity vector. The angle between the velocity vector direction of traffic participants and the normal vector of the joint perception blind zone boundary is less than a preset critical angle; The magnitude of the velocity vector or acceleration vector of a traffic participant exceeds the preset safety threshold corresponding to the category of the traffic participant.

5. The traffic safety monitoring and early warning method as described in claim 1, characterized in that, The step of generating at least one risk migration path and its corresponding risk propagation area and time window based on the motion state data of the traffic participants, the road structure rules of the joint sensing blind spot map, the traffic density data reported by the auxiliary sensing nodes, and the historical trajectory database includes: Extract the velocity vector and direction angle data of high-risk objects before they enter the blind zone as motion basis data; The blind spot exit directions, lane connections, and traffic rule constraints are analyzed from the joint perception blind spot map to generate a path topology network. Calculate the traffic resistance coefficient of each route branch based on surrounding traffic density data; Based on historical trajectory databases, the probability of path selection in similar scenarios is matched; The motion baseline data, traffic resistance coefficient, and path selection probability are weighted and fused to calculate the generation probability of each migration path; Select the first preset number of paths with the highest generation probability, calculate the spatial reachability along each path based on the maximum and minimum speeds of high-risk objects, and generate a risk propagation zone polygon and time window.

6. The traffic safety monitoring and early warning method as described in claim 5, characterized in that, The step of weightedly fusing motion baseline data, traffic resistance coefficient, and path selection probability to calculate the generation probability of each migration path includes: Assign velocity direction weights to the motion basis data and calculate the probability of the path continuing as a straight line; Assign density influence weights to the traffic resistance coefficient and calculate the path avoidance probability; Assign scene similarity weights to the historical path selection probabilities and calculate the weighted historical path selection probabilities. The probability of straight path continuation, the probability of path avoidance, and the weighted historical path selection probability are superimposed according to a preset proportional coefficient and then normalized and sorted.

7. The traffic safety monitoring and early warning method as described in claim 5, characterized in that, The step of generating early warning instructions based on the risk propagation zone and time window includes: Based on the polygon of the risk propagation zone, the name of the road in which it is located and its position relative to the road network node are analyzed. Based on the high-risk object category, a pre-set warning template library is matched to generate descriptive text; By binding the descriptive text with the spatial location code of the risk propagation zone polygon, a tiered early warning instruction is generated.

8. The traffic safety monitoring and early warning method as described in claim 1, characterized in that, The method further includes: Within the time window, the scope of the risk spread area is continuously monitored. If no high-risk objects are identified and the maximum window time is exceeded, the scope of the risk spread area is gradually reduced until the warning is lifted.

9. The traffic safety monitoring and early warning method as described in claim 8, characterized in that, The steps of continuously monitoring the risk propagation area within the time window, and gradually narrowing the risk propagation area until the warning is lifted if no high-risk object is identified and the maximum window time has been exceeded, include: Based on the initial risk propagation zone, time slices are generated at preset time intervals; At the beginning of each time slice, if no high-risk object is identified, the shrinkage boundary of the propagation zone is calculated based on the object's minimum speed, and at the end of the time slice, the shrinkage boundary is used as the new range of the risk propagation zone. The alert will be terminated when the number of time slices in which no high-risk objects are identified consecutively exceeds a preset threshold, or when the cumulative time exceeds the maximum window time.

10. A traffic safety monitoring and early warning system, characterized in that, The system includes: a memory, a processor, and a traffic safety monitoring and early warning program stored in the memory and executable on the processor, the traffic safety monitoring and early warning program being configured to implement the steps of the traffic safety monitoring and early warning method as described in any one of claims 1 to 9.

Images