A traffic safety intelligent early warning method and system based on cloud computing

Abstract

This invention discloses a cloud-based intelligent traffic safety early warning method and system, belonging to the field of intelligent transportation technology. The method includes: a cloud platform receiving real-time dynamic data from multiple traffic participants and generating a movement trajectory sequence; constructing a movement trend vector for each traffic participant based on continuous location points in the movement trajectory sequence, and generating an initial traffic conflict heatmap based on all movement trend vectors; extracting potential conflict groups from high-density areas of the heatmap, and calculating the dispersion value of the movement trend vectors of each traffic participant within the group; when the dispersion value exceeds a preset risk threshold, sending a first-level early warning instruction to the traffic participants in the high-density area. This invention also provides a system for implementing the method, achieving accurate identification and graded early warning of multi-traffic participant group conflict risks through movement trend vector heatmaps and group dispersion assessment.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, specifically to a cloud-based intelligent traffic safety early warning method and system. Background Technology

[0002] In the field of traffic safety early warning, relying on cloud platforms to collect dynamic data such as the location and speed of multiple traffic participants for conflict risk analysis is a common approach. Existing technologies generally use latitude and longitude coordinate point density to generate heat maps, directly identifying high-density areas of participant concentration as potential conflict zones. However, heat maps based solely on location density only reflect spatial occupancy and cannot reflect the movement direction and trends of each participant, easily mislabeling orderly convoys or pedestrian flows traveling in the same direction as high-risk conflict hotspots. Another typical early warning method is based on trajectory prediction to calculate collision time. This approach focuses on one-to-one conflict assessment, making it difficult to conduct overall assessment and effective early warning for group conflict scenarios involving multiple participants intertwined and exhibiting different movement trends at intersections and mixed traffic areas. For potential conflict groups composed of multiple traffic participants, existing methods lack quantitative means to assess the consistency and dispersion of movement trends within the group, failing to distinguish whether the movement directions within the group are generally convergent or highly dispersed, resulting in early warning thresholds that are difficult to adapt to the dynamic changes in group behavior. In real-world scenarios with high traffic volume and mixed directions, the inability to accurately perceive the risk of multi-party collisions caused by the dispersion of movement trends within a group often results in numerous false alarms or missed warnings of actual dangerous situations, leading to poor reliability of early warnings. How to extract the movement trend characteristics of traffic participants from real-time dynamic data and accurately assess potential conflicts caused by the dispersion of trends within a group based on these characteristics has become an urgent problem to be solved in this field. Summary of the Invention

[0003] To address the problems in existing traffic safety early warning technologies, such as the inability of location density heatmaps to reflect movement trends and the lack of quantitative assessment of the dispersion of movement trends within potential conflict groups, this invention provides a cloud-based intelligent traffic safety early warning method and system. This system accurately depicts movement trend interactions from real-time data and provides highly accurate hierarchical early warnings based on the dispersion of group trends.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a cloud computing-based intelligent traffic safety early warning method, the method comprising: a cloud platform receiving real-time dynamic data from multiple traffic participants and generating a movement trajectory sequence for each traffic participant; constructing a movement trend vector for each traffic participant based on continuous location points in the movement trajectory sequence, and then generating an initial traffic conflict heatmap using the movement trend vectors of all traffic participants; extracting potential conflict groups from high-density areas of the initial heatmap and calculating the dispersion value of the movement trend vectors of each participant within the group; when the dispersion value exceeds a preset risk threshold, the cloud platform sends a first-level early warning instruction to traffic participants located in the high-density area. Thus, traffic conflict risks can be detected in advance from the perspective of group movement trends, avoiding the lag and false alarms caused by relying solely on single location distances for judgment, and significantly improving the accuracy and timeliness of early warnings.

[0005] Preferably, the specific method for generating the movement trajectory sequence is as follows: The cloud platform receives real-time dynamic data from each traffic participant at fixed time intervals. This real-time dynamic data includes the participant's identity identifier, latitude and longitude coordinates, instantaneous speed value, and heading angle value. The latitude and longitude coordinates of the same identity identifier are arranged in chronological order to generate an original trajectory point sequence. Linear interpolation is performed on adjacent trajectory points in the original trajectory point sequence to obtain a smooth movement trajectory sequence. The linear interpolation process is based on the time interval and instantaneous speed value of adjacent trajectory points, inserting several intermediate position points between two points at equal time steps. Through interpolation compensation, trajectory jumps caused by the data acquisition cycle are effectively eliminated, improving the continuity of the movement trajectory and the reliability of subsequent calculations.

[0006] As a preferred embodiment of the present invention, when constructing the movement trend vector, three recently collected consecutive location points are extracted from the movement trajectory sequence of each traffic participant, and denoted as the first location point, the second location point, and the third location point, respectively. A first displacement vector pointing from the first location point to the second location point, and a second displacement vector pointing from the second location point to the third location point are calculated. The vector sum of the first and second displacement vectors is used as the movement trend vector of the traffic participant. This vector integrates information on the current velocity direction and acceleration direction, and can more realistically reflect the next movement tendency of the traffic participant, providing accurate directional basis for conflict prediction.

[0007] The optimal process for generating the initial traffic conflict heatmap is as follows: The cloud platform divides the covered geographical area into uniform grid cells, each with a unique grid identifier. For each traffic participant, their movement trend vector is mapped onto surrounding grid cells using their current latitude and longitude coordinates as the center, according to a preset scaling factor. Vector contribution values ​​are accumulated in the corresponding grid cells. This process iterates through all grid cells, summing the vector contribution values ​​for each cell to obtain the baseline risk value for that cell. The baseline risk values ​​of all grid cells constitute the initial traffic conflict heatmap. The preset scaling factor is directly proportional to the instantaneous speed of the traffic participant and inversely proportional to the confidence level of the heading angle. In this way, high-speed objects with clearly defined headings can generate a stronger risk response on the heatmap, highlighting individuals with genuine potential conflict tendencies and making the heatmap more closely reflect the actual risk distribution.

[0008] When extracting potential conflict groups, the process prioritizes traversing all grid cells in the initial traffic conflict heatmap, marking grid cells with a risk baseline value greater than a preset density threshold as hotspot grid cells. Adjacent hotspot grid cells are then labeled with connectivity rules using four- or eight-connectivity methods, merging multiple interconnected hotspot grid cells into a high-density area. The identities of all traffic participants within this high-density area are then collected to form a potential conflict group. Through spatial clustering, multiple objects with intertwined movement trends and concentrated risks can be automatically grouped into the same monitoring group, facilitating a comprehensive assessment of the conflict situation at the group level.

[0009] The step of calculating the dispersion value of a potential conflict group further includes: calculating the vector average of the movement trend vectors of all traffic participants within the group; for each traffic participant, calculating the magnitude of the vector difference between their movement trend vector and the vector average to obtain the individual deviation distance; and then calculating the average of all individual deviation distances, using this average as the dispersion value of the potential conflict group. A larger dispersion value indicates a more dispersed movement direction within the group, and a higher risk of mutual interference and collision.

[0010] When the dispersion value exceeds a preset risk threshold, the cloud platform compares the dispersion value of the potential conflict group with the preset risk threshold, which is dynamically adjusted based on the total number of traffic participants in the high-density area. If the dispersion value is greater than the preset risk threshold, traffic participants whose deviation distance is greater than the dispersion value are selected from the group as high-risk participants, and a first-level warning instruction containing avoidance prompts is sent to their terminal devices. This targeted warning mechanism only alerts those with significant deviations that pose prominent safety hazards, avoiding the blind distribution of warning information to all members, reducing unnecessary interference, and improving response efficiency.

[0011] As a further technical solution of the present invention, after sending the first-level warning instruction, the cloud platform continuously receives updated real-time dynamic data of traffic participants in the high-density area and recalculates the updated dispersion value. When the updated dispersion value is lower than a preset risk threshold, the cloud platform sends a warning cancellation instruction to the relevant traffic participants. When the updated dispersion value is still higher than the preset risk threshold and the number of consecutive increases reaches a preset number threshold, the cloud platform sends a second-level warning instruction to the traffic participants in the area. When sending the second-level warning instruction, the cloud platform extracts the movement trend vectors of all traffic participants in the high-density area and divides them into multiple directional intervals according to the heading angle value. It counts the number of movement trend vectors in each directional interval and takes the directional angle corresponding to the directional interval with the largest number as the main risk direction. It generates a second-level warning instruction containing the main risk direction and the recommended avoidance direction and sends it to the terminal devices of all traffic participants in the area. By dynamically tracking the changes in dispersion and directional distribution, a hierarchical warning system is achieved, which can upgrade the warning level in a timely manner according to the worsening trend of the situation and provide clear directional guidance, thereby improving the ability to deal with complex conflict scenarios.

[0012] Simultaneously, while sending the first-level warning command, the cloud platform also packages the spatial boundary coordinates and dispersion values ​​of the high-density area into a roadside broadcast message. The cloud platform then sends this roadside broadcast message to fixed roadside units located around the high-density area. The fixed roadside units convert the roadside broadcast message into visible light signals or broadcast wave signals and broadcast warnings to unconnected traffic participants entering the high-density area within a preset distance. In this way, even travelers without dedicated receiving terminals or who are offline can receive real-time risk alerts through sound and light, greatly expanding the coverage and fairness of traffic safety warnings.

[0013] This invention also provides a cloud-based intelligent traffic safety early warning system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the above-described method. This system can fully utilize cloud computing power to process real-time dynamic information from a large number of traffic participants in parallel, quickly completing movement trajectory analysis, conflict heatmap generation, group dispersion assessment, and tiered early warning instruction issuance, providing integrated safety early warning decision support for intelligent traffic managers.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0015] By extracting the three most recent consecutive location points from the movement trajectory sequence, calculating the first and second displacement vectors, and summing them to obtain the movement trend vector, this movement trend vector is then mapped onto uniform grid cells using a coefficient proportional to instantaneous velocity and inversely proportional to the heading angle confidence level. These are then accumulated to form an initial traffic conflict heatmap. This method integrates the location density and movement trend direction information of participants into the heatmap, effectively distinguishing between dense following flows in the same direction and conflict clusters with intersecting directions. It suppresses false hotspots formed by densely packed vehicles or pedestrians in the same direction, improving the accuracy of identifying real conflict areas. The mapping rules, which weight speed and adjust heading confidence level, concentrate the contributions of participants traveling at high speeds and stably, while dispersing the contributions of participants with fluctuating headings. This results in a risk baseline distribution in the heatmap that more closely reflects the actual intensity of traffic conflicts.

[0016] After extracting potential conflict groups from high-density areas of the heatmap, the average value of all movement trend vectors within the group is calculated. The magnitude of the difference between each participant's movement trend vector and the average value is taken as the individual deviation distance, and the average of all individual deviation distances is used as the dispersion value of the group. The larger the dispersion value, the more chaotic the movement directions within the group, and the higher the probability of multi-angle collisions. When the dispersion value exceeds a risk threshold dynamically adjusted based on the total number of traffic participants in the area, high-risk participants with deviation distances greater than the dispersion value are selected from the group, and a first-level warning instruction is sent. This accurately locates individuals in the group whose movement trends deviate from the overall trend and are most likely to cause conflict, avoiding blindly broadcasting warnings to all members and causing interference. The changes in the dispersion value are continuously monitored. When the dispersion value continuously increases, it is determined that the conflict risk is escalating. A second-level warning instruction, including the main risk direction and the recommended avoidance direction, is generated and sent to relevant terminals to guide participants to take timely and effective escape measures. By packaging the boundary coordinates and dispersion values ​​of high-density areas into messages, and converting them into light signals or broadcast signals through roadside units, the warning range can be extended to pedestrians or non-motorized vehicles without vehicle communication equipment, thereby improving the overall safety protection level in mixed traffic environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a flowchart of a cloud-based intelligent traffic safety early warning method.

[0019] Figure 2This is a flowchart illustrating the process of generating movement trend vectors for traffic participants.

[0020] Figure 3 This is a flowchart of a method for generating initial traffic conflict heatmaps based on grid accumulation;

[0021] Figure 4 This is a flowchart for calculating the dispersion of potential conflict groups based on traffic conflict heatmaps;

[0022] Figure 5 This is a schematic diagram illustrating the dispersion of the movement trend vector of potential conflict groups. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] See Figure 1 This invention provides a cloud-based intelligent traffic safety early warning method, comprising: a cloud platform receiving real-time dynamic data from multiple traffic participants and generating a movement trajectory sequence for each traffic participant based on the real-time dynamic data; constructing a movement trend vector for each traffic participant based on continuous location points in the movement trajectory sequence, and generating an initial traffic conflict heatmap in the cloud platform based on the movement trend vectors of all traffic participants; extracting potential conflict groups from high-density areas in the initial traffic conflict heatmap, and calculating the dispersion value of the movement trend vector of each traffic participant within the potential conflict group; when the dispersion value exceeds a preset risk threshold, the cloud platform sending a first-level early warning instruction to traffic participants located in the high-density area.

[0025] Example 1:

[0026] In specific implementation, please refer to Figure 2 The cloud platform receives real-time dynamic data from multiple traffic participants at fixed time intervals through a communication interface deployed in the cloud. Each piece of real-time dynamic data includes the traffic participant's identity, latitude and longitude coordinates, instantaneous speed value, and heading angle value. The fixed time interval can be configured to 1 second, and the cloud platform uses a message queue mechanism to cache and sort the received real-time dynamic data.

[0027] The cloud platform processes the received real-time dynamic data, grouping all latitude and longitude coordinates belonging to the same identity according to the chronological order of the timestamps carried in the real-time dynamic data, generating a sequence of original trajectory points for each traffic participant. Each trajectory point in the original trajectory point sequence corresponds to a piece of real-time dynamic data, recording latitude and longitude coordinates, instantaneous speed value, and heading angle value.

[0028] The cloud platform performs linear interpolation on the original trajectory point sequence of each traffic participant to generate a movement trajectory sequence. The linear interpolation process is as follows: for two temporally adjacent trajectory points in the original trajectory point sequence, based on the time interval between the two trajectory points and the instantaneous velocity value of the preceding trajectory point, several intermediate position points are inserted between the two trajectory points at equal time steps. The time step for inserting intermediate position points is set to 0.5 seconds. The latitude and longitude coordinates of the inserted intermediate position points are obtained through linear interpolation calculation. Let the latitude and longitude coordinates of the preceding trajectory point be... The latitude and longitude coordinates of the next trajectory point are The timestamp of the previous trajectory point is The timestamp of the next trajectory point is The instantaneous velocity value of the previous trajectory point is Then the time interval between the two trajectory points is Number of insertion points at intermediate positions , No. The latitude and longitude coordinates of the inserted intermediate position point Calculate according to the following formula: in, To insert a serial number, the value is... , It is the longitude coordinate value of the previous trajectory point. It is the latitude coordinate value of the previous trajectory point. It is the longitude coordinate value of the next trajectory point. It is the latitude coordinate value of the next trajectory point. It is the time of data collection for the previous trajectory point. It is the time of data collection for the next trajectory point. It is the time interval between adjacent trajectory points, in seconds. It is the instantaneous velocity value of the previous trajectory point, in meters per second. The number of insertion points is determined by the time interval and a fixed time step of 0.5 seconds. The trajectory points in the original trajectory point sequence and the inserted intermediate position points are merged in chronological order to form the movement trajectory sequence of each traffic participant.

[0029] The cloud platform extracts the three most recently collected consecutive location points from the movement trajectory sequence of each traffic participant. These three location points are designated as location point 1, location point 2, and location point 3 in chronological order of collection time. The latitude and longitude coordinates of location point 1 are... The latitude and longitude coordinates of the second location point are The latitude and longitude coordinates of the third location point are The cloud platform calculates the first displacement vector from the first position point to the second position point. The longitude component of the first displacement vector is... The component in the latitudinal direction is The cloud platform calculates the second displacement vector pointing from the second position point to the third position point. The longitude component of the second displacement vector is... The component in the latitudinal direction is The cloud platform adds the first displacement vector to the second displacement vector to obtain the movement trend vector for each traffic participant. The longitude component of the movement trend vector is... The component in the latitudinal direction is The movement trend vector reflects the overall displacement direction and distance of a traffic participant from the first location point to the third location point. In some embodiments, when the number of location points in the movement trajectory sequence is less than three, the cloud platform directly constructs the displacement vector as the movement trend vector using one or two existing location points; if there is only one location point, the movement trend vector is set to zero.

[0030] Example 2:

[0031] In specific implementation, please refer to Figure 3 The cloud platform acquires the spatial boundary information of the covered geographical area, using the westernmost longitude value. Longitude value of the easternmost side Southernmost latitude value and the latitude value of the northernmost side Definition. The cloud platform divides the geographical region along both the longitude and latitude directions using fixed step sizes. The step size for the longitude direction is set to... The step size for dividing in the degree and latitude directions is set to 1. The degree is used to obtain uniform grid cells. Each grid cell is represented by a longitude index value. and a latitude index value The unique identifier of a grid cell is a pair of ordered integers. Longitude index value and latitude index value The calculation method is as follows: based on the western and southernmost boundaries of the geographical region, the longitude index value of the grid cell. latitudinal index value of grid cell ,in The longitude value of the center point of the grid cell. This represents the latitude value of the center point of the grid cell.

[0032] The cloud platform assigns an accumulator to each grid cell, initially setting the accumulator to zero to store the vector contribution value. The cloud platform iterates through each traffic participant, obtaining their current latitude and longitude coordinates and movement trend vector. The current latitude and longitude coordinates are taken from the latest location point in the corresponding traffic participant's movement trajectory sequence.

[0033] For each traffic participant, the cloud platform determines a set of grid cells within a preset influence radius, centered on the participant's current latitude and longitude coordinates, as the mapping target. The preset influence radius is 50 meters, which, after conversion to latitude and longitude scales, corresponds to a range covering 5 grid cells each in the longitude and latitude directions. The cloud platform calculates the vector contribution value for each grid cell within this range and accumulates the vector contribution value into the accumulator of the corresponding grid cell.

[0034] Optionally, when mapping the movement trend vector to surrounding grid cells, the cloud platform uses a preset scaling factor to weight the movement trend vector. This preset scaling factor consists of the product of two factors, the first of which is a velocity factor. The second part of the factors is the heading confidence factor. Velocity factor The speed factor is directly proportional to the instantaneous speed of the traffic participant. The calculation method is as follows ,in The instantaneous speed of a traffic participant, expressed in meters per second. For reference speed value, The value is set to 10 m / s. Heading confidence factor. The confidence level of the heading angle value is inversely proportional to the confidence level of the traffic participant, which is calculated using the standard deviation of the heading angle value over the most recent sliding time window. Characterization, heading confidence factor The calculation method is as follows ,in The standard deviation of the heading angle value within the sliding time window is defined as 3 seconds. As a reference angle fluctuation value, The value is taken as 10 degrees. The standard deviation of the heading angle value... A smaller value indicates that the travel direction of traffic participants is stable, the confidence level of the heading angle value is high, and the heading confidence factor is [value missing]. Approaching 1; when the standard deviation of the heading angle value A larger value indicates significant fluctuations in the driving direction of traffic participants, resulting in lower confidence levels for the heading angle value and a lower heading confidence factor. The corresponding decrease.

[0035] For each grid cell located within the preset influence radius, the cloud platform calculates the vector contribution value obtained by the grid cell according to the following formula. : in, The longitude index value of the grid cell. The latitude index value of the grid cell. This represents the movement trend vector of traffic participants. For velocity factor, , It is the instantaneous speed value of traffic participants. It is a reference speed value and is set to 10. For the heading confidence factor, , It is the standard deviation of the heading angle value within the slip time window. It is a reference angle fluctuation value and is set to 10 degrees. The Euclidean distance between the center point of the grid cell and the current latitude and longitude coordinates of the traffic participant. To preset the spatial scale parameters of the influence radius, The value is 30 meters. Represented by natural constant An exponential function with base 0.5. The cloud platform will calculate the resulting vector contribution value. Accumulate them separately to the longitude index value Latitude index value In the accumulator corresponding to the grid cell.

[0036] After the cloud platform iterates through all traffic participants and accumulates their vector contributions, it iterates through all grid cells, reading the accumulated sum of vector contributions from the accumulator of each grid cell. The magnitude of this accumulated sum is then used as the base risk value for the corresponding grid cell. For longitude index values... Latitude index value The grid cells, with the risk baseline value denoted as The cloud platform sets the risk baseline values ​​for all grid cells. According to longitude index value and latitude index value The positional relationships are arranged into a two-dimensional matrix, with each row of the matrix corresponding to a latitude index value. The columns of the two-dimensional matrix correspond to longitude index values. The value of each matrix element is the base risk value of the corresponding grid cell. The cloud platform uses the resulting two-dimensional matrix as the initial traffic conflict heatmap. Areas with higher baseline risk values ​​in the two-dimensional matrix correspond to areas with a higher probability of conflict after the movement trend vectors of traffic participants are superimposed.

[0037] Example 3:

[0038] In specific implementation, please refer to Figure 4 The cloud platform iterates through all grid cells of the initial traffic conflict heatmap, obtaining the baseline risk value for each grid cell. The platform then compares this baseline risk value with a preset density threshold, set to 50. This threshold is chosen based on the following: by statistically analyzing historical traffic conflict heatmap data from numerous urban intersections and road segments, the platform extracts the distribution of baseline risk values ​​for the grid cells where actual traffic accidents occurred. Using the lower quartile of this distribution as the threshold, statistical analysis of dozens of intersections across multiple cities shows that the baseline risk values ​​corresponding to the lower quartile are concentrated around 50. Therefore, setting the baseline density threshold to 50 effectively filters out grid cells with higher risks. The platform marks grid cells with baseline risk values ​​greater than 50 as hotspot grid cells and assigns a flag bit to each hotspot grid cell with a value of 1; grid cells with baseline risk values ​​less than or equal to 50 have their flag bit set to 0.

[0039] The cloud platform performs a connected component labeling algorithm on all hotspot grid cells. This algorithm uses the eight-connectivity rule to merge adjacent hotspot grid cells. The eight-connectivity rule states that for each hotspot grid cell with a label of 1, its eight adjacent grid cells are checked. These eight adjacent grid cells include those in the eight directions: up, down, left, right, upper left, upper right, lower left, and lower right. If any of these adjacent grid cells also has a label of 1, then these two hotspot grid cells are considered connected. The cloud platform traverses all hotspot grid cells using a breadth-first search, grouping connected hotspot grid cells into the same connected component. Each connected component contains several hotspot grid cells. The cloud platform assigns a unique high-density region identifier to each connected component, classifying multiple connected hotspot grid cells into a single high-density region. If two hotspot grid cells are connected only by edge or corner adjacency, they are considered connected according to the eight-connectivity rule.

[0040] After dividing the high-density area, the cloud platform obtains a list of grid identifiers for all hotspot grid units within each high-density area. The cloud platform maintains a mapping table between grid units and traffic participant identifiers, which is established synchronously when generating the initial traffic conflict heatmap. Specifically, when a traffic participant's movement trend vector is mapped to a grid unit within a preset influence radius, the cloud platform adds the traffic participant's identifier to the identifier list of the corresponding grid unit. The cloud platform iterates through the grid identifiers of each hotspot grid unit within the current high-density area, extracts the identifier list for each hotspot grid unit from the mapping table, merges all identifier lists, removes duplicate identifiers, and obtains a set of identifiers for all traffic participants within the same high-density area. This set of identifiers is then used as a potential conflict group corresponding to that high-density area.

[0041] For each potential conflict group, the cloud platform reads the movement trend vector of each traffic participant within the stored traffic participant status data. Let there be a total of [number missing] traffic participants within the potential conflict group. The traffic participant, the first The movement trend vector of each traffic participant is denoted as... , The range of values ​​is Moving trend vector Includes longitude direction component and latitude component .

[0042] The cloud platform calculates the vector average of the movement trend vectors of all traffic participants within a potential conflict group. Longitude direction component By Sum of the longitude components of each moving trend vector and divide by To obtain, that is ; Vector average Latitude component By Sum of the latitudinal components of each moving trend vector and divide by . To obtain, that is .

[0043] For each traffic participant within a potential conflict group, the cloud platform calculates the participant's movement trend vector and vector average. The magnitude of the vector difference between the two values ​​is used to obtain the individual deviation distance. For the first... Individual traffic participants, individual deviation distance Calculate according to the following formula: in, This refers to the sequence number of traffic participants within a potential conflict group. For the first Individual deviation distance of each traffic participant For the first The longitude component of the movement trend vector of each traffic participant For the first The latitudinal component of the movement trend vector of each traffic participant. Let be the longitude component of the vector average of the movement trend vectors of all traffic participants within a potential conflict group. This represents the latitudinal component of the vector average.

[0044] The cloud platform calculates the individual deviation distances of all traffic participants within a potential conflict group. Next, the average deviation distance of all individuals is calculated, and this average is used as the dispersion value of the potential conflict group. Dispersion value The calculation method is as follows The cloud platform will calculate the dispersion value. The high-density area identifiers associated with this potential conflict group are stored for subsequent risk assessment.

[0045] See Figure 5 In the figure, the horizontal axis represents the longitude component of the movement trend vector of traffic participants, in degrees; the vertical axis represents the latitude component of the movement trend vector, also in degrees. Multiple hollow circles in the figure represent the movement trend vectors of each traffic participant within a potential conflict group. The distribution is relatively concentrated, reflecting a certain degree of spatial aggregation in the movement trends of individuals within the group. The average vector value is indicated by a solid black diamond at the center. Its longitude and latitude components are both positive, indicating that the overall movement trend vector of this potential conflict group points to the northeast.

[0046] The black dashed circle is centered on the vector average value, and its radius corresponds to the degree of dispersion. This circular pattern reflects the degree to which the movement trend vectors of all traffic participants deviate from the average. A smaller dispersion value indicates a narrower range of deviations in the movement trends of individuals within the group, a greater uniformity in overall movement, and a relatively lower risk of potential conflict.

[0047] The average value of each hollow dot in the figure The lines connecting the points are represented by thin dashed lines, clearly showing the direction and distance of each individual's deviation from the average vector. Overall, the deviation distance of each individual does not exceed the radius of the dashed circle, which conforms to the definition and calculation method of the dispersion value in Example 3.

[0048] The legend clearly marks the individual movement trend vector. Vector average and dispersion value The specific meaning of this figure facilitates understanding the spatial distribution characteristics of the movement states of traffic participants within the potential conflict group, as illustrated in this figure. This figure corresponds to the analysis steps in Example 3 regarding the dispersion of the movement trend vector of the potential conflict group. Through this diagram, the consistency and deviation magnitude of the movement trends of traffic participants can be intuitively determined, providing a quantitative basis for subsequent risk assessment and early warning based on the dispersion value.

[0049] Example 4:

[0050] In practice, the cloud platform compares the dispersion value of potential conflict groups with a preset risk threshold. The preset risk threshold is dynamically adjusted based on the total number of traffic participants in the high-density area. The cloud platform obtains the total number of traffic participants from the identity set corresponding to the high-density area, denoted as... Preset risk threshold Calculated using the following formula: in, To preset risk thresholds, The total number of traffic participants in a high-density area. Based on the number of traffic participants, The value is 10. As the benchmark risk threshold coefficient, The value is 3.2. Baseline number of traffic participants. The value of 10 is based on the following: Cluster analysis of traffic conflict data from multiple urban road intersections shows that when the number of traffic participants in a high-density area is less than 10, the dispersion value fluctuates less and the conflict incidence rate is lower; when the number of traffic participants reaches 10 or more, the mean dispersion value shows an approximately linear increasing trend with the number of participants. (Baseline risk threshold coefficient) The value of 3.2 is based on the following: Under the baseline condition of 10 traffic participants, statistical analysis of the distribution of dispersion values ​​in historical conflict events was conducted. The dispersion value corresponding to the 90th percentile was used as the discrimination threshold, and this value remained stable around 3.2. Therefore, [the value is set as 3.2]. The threshold is set to 3.2 to ensure that the risk threshold is preset when the number of traffic participants is 10 in a high-density area. It has sufficient sensitivity to identify real conflict events.

[0051] The cloud platform extracts the individual deviation distance of each traffic participant from potential conflict groups and compares the individual deviation distance with the dispersion value of the potential conflict groups. For the first... Each traffic participant, if the individual deviates from the distance Greater than the dispersion value If the traffic participant is identified, they will be marked as a high-risk participant. The cloud platform collects the identification information of all high-risk participants to form a list of high-risk participants.

[0052] The cloud platform sends a Level 1 warning command to the terminal devices of high-risk participants. This Level 1 warning command is transmitted to the high-risk participants' terminal devices via push notifications through the mobile communication network. The avoidance information carried in the Level 1 warning command includes the spatial boundary coordinates of the high-density area, a description of the deviation of the high-risk participant's current driving direction from the average driving direction of other surrounding traffic participants, and suggested deceleration or steering instructions.

[0053] After sending the first-level warning command, the cloud platform continuously receives updated real-time dynamic data from traffic participants within the high-density area. The cloud platform uses the spatial boundary coordinates of the high-density area as a spatial filter, only receiving real-time dynamic data reported by traffic participants whose current latitude and longitude coordinates fall within the spatial boundary coordinate range of the high-density area. The cloud platform receives updated real-time dynamic data at fixed time intervals, consistent with the time interval used to generate the movement trajectory sequence, both being 1 second.

[0054] The cloud platform recalculates the updated dispersion value within the high-density area based on the updated real-time dynamic data. The recalculation process involves the cloud platform using the updated real-time dynamic data to regenerate the movement trajectory sequence and movement trend vector for each traffic participant within the high-density area, reconstructing potential conflict groups, and then recalculating the updated dispersion value by calculating the average vector and individual deviation distance.

[0055] The cloud platform compares the updated dispersion value with a preset risk threshold. When the updated dispersion value is lower than the preset risk threshold, it indicates that the traffic conflict situation in the high-density area has eased, and the cloud platform sends a warning cancellation command to all traffic participants in the high-density area. The warning cancellation command instructs the terminal device to stop displaying avoidance prompts and restore the normal driving interface.

[0056] After each update of the dispersion value, the cloud platform compares the updated dispersion value with the previously calculated dispersion value and records the number of consecutive occurrences where the updated dispersion value is higher than the previous dispersion value. When the updated dispersion value is still higher than a preset risk threshold and the number of consecutive increases reaches a preset threshold, the cloud platform initiates a second-level warning process. The preset threshold is set to 3. The rationale for setting the preset threshold to 3 is as follows: When analyzing the evolution of traffic conflict situations, three consecutive increases in the dispersion value correspond to a continuously intensifying divergence trend in the movement directions of traffic participants. Furthermore, in retrospective analysis of actual traffic accidents, the time interval between three consecutive increases and the final collision is usually between 3 and 5 seconds, providing a sufficient response window for intervention while avoiding excessive warnings due to single or two fluctuations.

[0057] The cloud platform extracts the movement trend vectors of all traffic participants from a high-density area. Each participant's movement trend vector is recalculated based on updated real-time dynamic data. The cloud platform divides the movement trend vectors into multiple directional intervals according to their heading angles. With true north as 0 degrees, the 360-degree radius is divided into eight equal directional intervals clockwise, each covering a 45-degree range: 0 to 45 degrees, 45 to 90 degrees, 90 to 135 degrees, 135 to 180 degrees, 180 to 225 degrees, 225 to 270 degrees, 270 to 315 degrees, and 315 to 360 degrees. The cloud platform calculates the directional angle of each movement trend vector using the arctangent function, derived from the longitude and latitude components of the vector, and maps the directional angle to the corresponding directional interval.

[0058] The cloud platform counts the number of moving trend vectors falling within each directional interval. The number of moving trend vectors in each of the eight directional intervals is compared one by one, and the directional interval with the largest number is identified. The directional angle corresponding to this directional interval is taken as the primary risk direction. The directional angle corresponding to a directional interval is the median angle of that interval; for example, the directional angle corresponding to the 0-45 degree interval is 22.5 degrees.

[0059] The cloud platform generates a second-level warning instruction, which includes the main risk direction and a recommended avoidance direction. The recommended avoidance direction is one of two directions perpendicular to the main risk direction, causing traffic participants to move away from the spatial boundary of the high-density area. The cloud platform generates the recommended avoidance direction by calculating the normal direction of the main risk direction to obtain two perpendicular directions, and then selecting the direction that moves traffic participants away from the center of the high-density area as the recommended avoidance direction based on the positional relationship between the center point of the spatial boundary of the high-density area and each traffic participant.

[0060] The cloud platform sends the Level 2 warning instruction to the terminal devices of all traffic participants in the high-density area via broadcast message push. After receiving the Level 2 warning instruction, the terminal devices display the main risk direction and recommended avoidance direction in a strong reminder manner.

[0061] Example 5:

[0062] In practice, the cloud platform sends a Level 1 warning to the terminal devices of high-risk participants while simultaneously initiating the generation process of a roadside broadcast message. The cloud platform retrieves the spatial boundary coordinates of the high-density area from its stored high-density area information. These coordinates include the westernmost longitude, easternmost longitude, southernmost latitude, and northernmost latitude. The cloud platform also obtains the dispersion value of the potential conflict groups corresponding to the current high-density area. Finally, the cloud platform packages the spatial boundary coordinates and dispersion value of the high-density area along with the current timestamp to generate a roadside broadcast message. This message uses a structured data message format. The message header includes a message type identifier and message body length. The message body sequentially includes the westernmost longitude field, easternmost longitude field, southernmost latitude field, northernmost latitude field, dispersion value field, and timestamp field.

[0063] The cloud platform maintains a fixed roadside unit location information table in its storage. This table records the fixed roadside unit's identifier, longitude, and latitude. Using the center point's latitude and longitude coordinates of the high-density area as a reference, the cloud platform searches the fixed roadside unit location information table for all fixed roadside units whose distance from the center point's latitude and longitude coordinates falls within a preset query radius. This preset query radius is 500 meters. The reason for this 500-meter radius is that the effective direct communication distance of the visible light signal transmitter used by the fixed roadside unit is 300 meters, and the effective coverage radius of the broadcast wave signal transmitter is 400 meters. Using 500 meters as the query radius ensures coverage of all fixed roadside units capable of receiving signals, while avoiding introducing too many fixed roadside units exceeding their effective communication distance.

[0064] The cloud platform sends the generated roadside broadcast messages to each of the queried fixed roadside units via wired fiber optic network or wireless cellular network communication interfaces. The destination address of each roadside broadcast message is the network address of the corresponding fixed roadside unit.

[0065] The fixed roadside unit comprises a message receiving module, a signal conversion and processing module, and a signal transmitting module. The message receiving module receives roadside broadcast messages from the cloud platform and extracts the spatial boundary coordinates, dispersion values, and timestamps of high-density areas from these messages. The signal conversion and processing module uses the dispersion values ​​as input parameters and generates control parameters for visible light signals or modulation parameters for broadcast wave signals according to a preset conversion relationship.

[0066] Optionally, when fixed roadside units use visible light signals for broadcast warnings, the signal conversion and processing module maps the dispersion value to the flicker frequency of the visible light signal. (Flicker frequency of the visible light signal) Calculate according to the following formula: in, The flicker frequency of the visible light signal, measured in Hertz (Hz). For frequency mapping coefficients, The value is set to 0.5 Hz per unit of dispersion. The discreteness value is parsed from the roadside broadcast message. Based on the flicker frequency, The value is 2 Hz. Frequency mapping coefficient. The value of 0.5 Hz per unit of dispersion is based on the fact that the typical observation range for dispersion is between 0.5 and 8.0, and the fundamental scintillation frequency... A value of 2 Hz ensures the output flicker frequency falls within the 2.25 Hz to 6 Hz range. This range is within the frequency range where flickering is perceptible to the human eye without causing discomfort, and it is clearly distinguishable from the flickering frequency range of vehicle turn signals, allowing for accurate identification by the vehicle's photosensitive sensor. The signal transmission module uses a high-brightness LED array to flicker at the calculated frequency. It emits visible light signals, with the light band selected being the red visible light band, and the wavelength range being 620 nanometers to 630 nanometers.

[0067] Optionally, when the fixed roadside unit uses broadcast wave signals for broadcast warnings, the signal conversion and processing module encodes the spatial boundary coordinates and dispersion values ​​of the high-density area into modulated data frames of the broadcast wave signal. The modulation method adopts frequency shift keying, and the carrier frequency is selected from the 915 MHz industrial, scientific and medical frequency band.

[0068] After converting the roadside broadcast message into a visible light signal or broadcast wave signal, the signal transmitting module of the fixed roadside unit continuously broadcasts warnings to unconnected traffic participants entering the high-density area within a preset distance. The preset distance is set at 100 meters. The basis for setting the preset distance at 100 meters is as follows: a warning buffer zone is formed by extending 100 meters outward from the spatial boundary of the high-density area. The 100-meter buffer distance corresponds to approximately 12 seconds of travel time for a vehicle traveling at 30 kilometers per hour in an urban road environment. This gives drivers of unconnected traffic participants sufficient time to slow down or steer after receiving the warning signal.

[0069] The information transmitted in the broadcast warning signals of fixed roadside units includes discreteness information implied by the flashing frequency of visible light signals or spatial boundary coordinates and discreteness information carried in the broadcast wave signal data frames. After receiving the warning signals, the photosensitive sensors or broadcast wave receiving modules installed on the vehicles of unconnected traffic participants convert the warning signals into audio-visual prompts in the driver's cabin, alerting the driver to the risk of a traffic conflict ahead.

[0070] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A cloud computing-based intelligent early warning method for traffic safety, characterized in that, The method includes: The cloud platform receives real-time dynamic data from multiple traffic participants and generates a movement trajectory sequence for each traffic participant based on the real-time dynamic data. The movement trend vector of each traffic participant is constructed based on the continuous location points in the movement trajectory sequence, and an initial traffic conflict heat map is generated in the cloud platform based on the movement trend vectors of all traffic participants. Based on the high-density areas in the initial traffic conflict heatmap, potential conflict groups are extracted, and the dispersion value of the movement trend vector of each traffic participant within the potential conflict group is calculated. When the dispersion value exceeds a preset risk threshold, the cloud platform sends a first-level warning instruction to traffic participants in the high-density area.

2. The intelligent traffic safety early warning method based on cloud computing according to claim 1, characterized in that, The specific steps for generating the movement trajectory sequence of each traffic participant based on the real-time dynamic data are as follows: The cloud platform receives real-time dynamic data from each traffic participant at fixed time intervals. The real-time dynamic data includes the traffic participant's identity, latitude and longitude coordinates, instantaneous speed value, and heading angle value. Arrange the latitude and longitude coordinates of the same identity identifier in chronological order to generate the original trajectory point sequence for each traffic participant; Linear interpolation is performed on adjacent trajectory points in the original trajectory point sequence to obtain the movement trajectory sequence of each traffic participant.

3. The intelligent traffic safety early warning method based on cloud computing according to claim 2, characterized in that, The specific method of the linear interpolation process is as follows: based on the time interval and instantaneous velocity value of adjacent trajectory points, several intermediate position points are inserted between the two points at equal time steps.

4. The intelligent traffic safety early warning method based on cloud computing according to claim 2, characterized in that, The specific steps for constructing the movement trend vector of each traffic participant based on the continuous location points in the movement trajectory sequence are as follows: Extract the three most recently collected consecutive location points from the movement trajectory sequence of each traffic participant, and denot them as the first location point, the second location point, and the third location point; Calculate the first displacement vector from the first position point to the second position point, and the second displacement vector from the second position point to the third position point; The vector sum of the first displacement vector and the second displacement vector is used as the movement trend vector for each traffic participant.

5. The intelligent traffic safety early warning method based on cloud computing according to claim 1, characterized in that, The specific steps for generating the initial traffic conflict heatmap in the cloud platform based on the movement trend vectors of all traffic participants are as follows: The cloud platform divides the covered geographical area into uniform grid units, each with a unique grid identifier; For each traffic participant, their current latitude and longitude coordinates are used as the center, and their movement trend vector is mapped onto the surrounding grid cells according to a preset scaling factor. The vector contribution value is then accumulated on the corresponding grid cells. Traverse all grid cells, sum the vector contribution values ​​of each grid cell as the basic risk value of that grid cell, and use the two-dimensional matrix formed by the basic risk values ​​of all grid cells as the initial traffic conflict heat map.

6. The intelligent traffic safety early warning method based on cloud computing according to claim 5, characterized in that, The preset proportionality coefficient is directly proportional to the instantaneous speed value of the traffic participant and inversely proportional to the confidence level of its heading angle value.

7. The intelligent traffic safety early warning method based on cloud computing according to claim 5, characterized in that, The specific steps for extracting potential conflict groups from the high-density areas in the initial traffic conflict heatmap are as follows: Traverse all grid cells of the initial traffic conflict heatmap and mark grid cells with a risk baseline value greater than a preset density threshold as hotspot grid cells; Perform a connected component labeling algorithm on all adjacent hot spot grid cells to divide multiple interconnected hot spot grid cells into a high-density region; Collecting the identities of all traffic participants located in the same high-density area creates a potential conflict group.

8. The intelligent traffic safety early warning method based on cloud computing according to claim 7, characterized in that, The connected component labeling algorithm uses four-connectivity or eight-connectivity rules to merge adjacent hot spot grid cells.

9. The intelligent traffic safety early warning method based on cloud computing according to claim 7, characterized in that, The specific steps for calculating the dispersion value of the movement trend vector of each traffic participant within the potential conflict group are as follows: Calculate the vector average of the movement trend vectors of all traffic participants within the potential conflict group; For each traffic participant within the potential conflict group, the magnitude of the vector difference between their movement trend vector and the average value of the vector is calculated to obtain the individual deviation distance of each traffic participant; Calculate the average individual deviation distance of all traffic participants within the potential conflict group, and use the average value as the dispersion value of the potential conflict group.

10. A cloud-based intelligent traffic safety early warning system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the cloud computing-based intelligent traffic safety early warning method according to any one of claims 1 to 9.

Images