Pedestrian safety warning method and device combined with high-precision perception technology

Abstract

The application provides a pedestrian safety warning method and device combined with high-precision perception technology, and relates to the technical field of intelligent transportation. The method comprises the following steps: positioning a target warning area; presetting a warning intervention window, and performing intermittent environment perception; performing timing risk feature identification based on information perception timing; performing risk trajectory prediction on dynamic risk factor distribution to obtain a warning intervention risk distribution; performing risk level evaluation according to the warning intervention risk distribution, and outputting a real-time risk level; matching a traffic flow cooperative warning strategy; and taking the target warning area as a warning boundary, and performing traffic flow cooperative warning optimization on pedestrian safety by using the traffic flow cooperative warning strategy. The application solves the technical problem that it is difficult to accurately identify the sudden behavior of pedestrians due to the limitation of perception, and the warning effect is poor. Through timing risk feature identification and trajectory prediction, early warning intervention is performed, the accuracy of pedestrian safety warning is improved, and the overall traffic flow is optimized.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, and in particular to a pedestrian safety early warning method and device that combines high-precision sensing technology. Background Technology

[0002] With the continuous development of urban transportation systems, pedestrian safety has become an increasingly important concern. As a vulnerable group within the transportation system, pedestrians are easily threatened by motor vehicles in complex traffic environments. To reduce traffic accidents and improve road traffic safety, pedestrian safety early warning systems have gradually become an important direction in intelligent transportation research. Pedestrian safety early warning methods mainly rely on technologies such as visual perception, radar detection, and vehicle-to-infrastructure (V2I) communication, achieving early warning through real-time perception of pedestrian positions and environmental factors.

[0003] However, existing methods typically rely on fixed sensing points, making it difficult to identify sudden pedestrian behaviors in advance and affecting the timeliness and accuracy of warnings. The lack of sufficient integration with temporal risk characteristic analysis results in high false alarm rates and significant missed alarms in pedestrian warning systems, impacting the response efficiency of both drivers and pedestrians.

[0004] In summary, existing technologies suffer from limitations in perception, making it difficult to accurately identify sudden pedestrian behaviors, resulting in poor early warning effects and impacting overall traffic safety. Summary of the Invention

[0005] The purpose of this application is to provide a pedestrian safety early warning method and device that combines high-precision sensing technology, in order to solve the technical problem in the prior art that due to the limitation of sensing, it is difficult to accurately identify sudden pedestrian behavior, resulting in poor early warning effect and thus affecting the overall traffic safety.

[0006] In view of the above problems, this application provides a pedestrian safety early warning method and device that combines high-precision sensing technology.

[0007] Firstly, this application provides a pedestrian safety early warning method combining high-precision sensing technology. This method is implemented using a pedestrian safety early warning device incorporating high-precision sensing technology. The method includes: locating a target early warning area by performing early warning intervention analysis on the target risk area; pre-setting an early warning intervention window and performing intermittent environmental sensing of the target early warning area based on the early warning intervention window to obtain multiple dynamic information sets of areas; identifying the temporal risk characteristics of the multiple dynamic information sets of areas based on the information sensing time sequence to locate the distribution of dynamic risk factors; predicting the risk trajectory of the dynamic risk factor distribution using the target risk area as the risk boundary to obtain an early warning intervention risk distribution; evaluating the risk level based on the early warning intervention risk distribution and outputting a real-time risk level; matching early warning strategies based on the real-time risk level and outputting a traffic flow collaborative early warning strategy; and optimizing pedestrian safety using the traffic flow collaborative early warning strategy with the target early warning area as the early warning boundary.

[0008] Optionally, network data is retrieved based on the target risk area to obtain driving speed characteristics; the warning intervention response time is obtained interactively; the warning intervention response distance is calculated based on the warning intervention response time and acceleration speed characteristics; and the target warning area is obtained by extending the road network selection using the target risk area as the starting point and the warning intervention response distance as the road network selection scale.

[0009] Optionally, network data is retrieved based on the target early warning area to obtain the vehicle flow characteristics and pedestrian flow characteristics of the extended road network; the vehicle flow characteristics of the extended road network are compared with the perception function table to obtain a first perception configuration strategy; the pedestrian flow characteristics of the extended road network are compared with the perception function table to obtain a second perception configuration strategy; the target perception configuration strategy is output by fusing the first and second perception configuration strategies; and a high-precision perception array is configured for the target early warning area based on the target perception configuration strategy, wherein the high-precision perception array includes multiple perception nodes configured in multiple perception areas.

[0010] Optionally, the effective duration of early warning intervention is obtained by fusing and analyzing the vehicle flow characteristics and pedestrian flow characteristics of the extended road network; the early warning intervention window is localized based on the effective duration of early warning intervention; multiple sensing nodes in the high-precision sensing array are synchronously and intermittently operated with the early warning intervention window as a constraint to obtain multiple regional dynamic sensing sequences of the multiple sensing areas; after time-series fusion of the multiple regional dynamic sensing sequences, risk feature detection is performed based on the fusion result to obtain the multiple regional dynamic information sets.

[0011] Optionally, the multiple regional dynamic perception sequences are spatiotemporally synchronized according to the early warning intervention window to obtain multiple sets of regional dynamic perception sequences; the multi-view image sequences of the multiple sets of regional dynamic perception sequences are synchronously stitched together according to the mapping relationship between the multiple perception regions and the target early warning region to output multiple regional stitched perception images; traffic flow feature detection is performed on the first regional stitched perception image to obtain K vehicle dynamic features of K real-time vehicles, wherein the vehicle dynamic features include vehicle identification features, driving speed features, and driving direction features; pedestrian flow feature detection is performed on the first regional stitched perception image to obtain F pedestrian dynamic features of F real-time pedestrians, wherein the pedestrian dynamic features include pedestrian identification features, walking speed features, and walking direction features; the K vehicle dynamic features and F pedestrian dynamic features are associated and stored using the first perception time of the first regional stitched perception image to obtain a first regional dynamic information set; and so on, the multiple regional dynamic information sets are obtained by performing risk feature detection on the multiple regional stitched perception images.

[0012] Optionally, based on the consistency of vehicle identification features, the driving change features of the multiple regional dynamic information sets are aggregated in a time series to obtain K first feature change sequences of the K real-time vehicles; based on the consistency of pedestrian identification features, the walking change features of the multiple regional dynamic information sets are aggregated in a time series to obtain F second feature change sequences of the F real-time pedestrians; a driving feature change scale is preset, wherein the driving feature change scale includes a driving speed change scale and a driving direction change scale; the driving feature change scale is used to traverse the K first feature change sequences to perform risk assessment of the K real-time vehicles, and output M risky vehicles; similarly, a walking feature change scale is preset to perform risk assessment of the F real-time pedestrians, and output N risky pedestrians; M vehicle position features of the M risky vehicles and N pedestrian position features of the N risky pedestrians are extracted as the dynamic risk factor distribution.

[0013] Optionally, extract M first feature change sequences from the M risky vehicles; extract N second feature change sequences from the N risky pedestrians; use the target risk area as the risk boundary and the target warning area as the warning boundary, and perform driving prediction for the M risky vehicles based on the M first feature change sequences to obtain the vehicle intervention risk distribution; use the target risk area as the risk boundary and the target warning area as the warning boundary, and perform walking prediction for the N risky pedestrians based on the N second feature change sequences to obtain the pedestrian intervention risk distribution; perform cross-fusion of the vehicle intervention risk distribution and the pedestrian intervention risk distribution based on the risk location distribution to output the warning intervention risk distribution.

[0014] Secondly, this application also provides a pedestrian safety early warning device combining high-precision perception technology, used to execute the pedestrian safety early warning method combining high-precision perception technology as described in the first aspect. The pedestrian safety early warning device combining high-precision perception technology includes: an early warning intervention analysis module, used to locate a target early warning area by performing early warning intervention analysis on the target risk area; an intermittent environmental perception module, used to preset an early warning intervention window and perform intermittent environmental perception of the target early warning area based on the early warning intervention window to obtain multiple regional dynamic information sets; a risk factor positioning module, used to identify the temporal risk characteristics of the multiple regional dynamic information sets based on the information perception time sequence to locate the dynamic risk factor distribution; a risk trajectory prediction module, used to predict the risk trajectory of the dynamic risk factor distribution with the target risk area as the risk boundary to obtain an early warning intervention risk distribution; a risk level determination module, used to evaluate the risk level according to the early warning intervention risk distribution and output a real-time risk level; an early warning strategy matching module, used to match early warning strategies according to the real-time risk level and output a traffic flow collaborative early warning strategy; and a collaborative early warning module, used to optimize pedestrian safety through traffic flow collaborative early warning using the traffic flow collaborative early warning strategy with the target early warning area as the early warning boundary.

[0015] One or more technical solutions provided in this application have at least the following beneficial effects:

[0016] By performing early warning intervention analysis on target risk areas, the target early warning area is located; an early warning intervention window is preset, and environmental intermittent perception of the target early warning area is performed based on the early warning intervention window to obtain multiple regional dynamic information sets; based on the information perception time sequence, the temporal risk characteristics of the multiple regional dynamic information sets are identified to locate the distribution of dynamic risk factors; using the target risk area as the risk boundary, the risk trajectory of the dynamic risk factor distribution is predicted to obtain the early warning intervention risk distribution; risk level evaluation is performed based on the early warning intervention risk distribution to output the real-time risk level; early warning strategy matching is performed based on the real-time risk level to output the traffic flow collaborative early warning strategy; using the target early warning area as the early warning boundary, the traffic flow collaborative early warning strategy is used to optimize traffic flow collaborative early warning for pedestrian safety. In other words, by locating the early warning area through early warning intervention analysis, early perception is achieved, risk factors in multiple areas are identified and risk trajectory prediction is performed, the predicted risks are evaluated, and appropriate early warning strategies are selected for early warning intervention, forming an intelligent collaborative early warning system, improving the accuracy of pedestrian behavior prediction, and thus enhancing the level of pedestrian safety in complex traffic environments.

[0017] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in 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, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the pedestrian safety early warning method combining high-precision sensing technology as described in this application.

[0020] Figure 2 This is a schematic diagram of the pedestrian safety early warning device that incorporates high-precision sensing technology, as described in this application.

[0021] Figure labeling: 11 Early warning intervention analysis module, 12 Intermittent environmental perception module, 13 Risk factor location module, 14 Risk trajectory prediction module, 15 Risk level determination module, 16 Early warning strategy matching module, 17 Collaborative early warning module. Detailed Implementation

[0022] This application provides a pedestrian safety early warning method and device that combines high-precision sensing technology. It solves the technical problem in existing technologies where limited sensing capabilities make it difficult to accurately identify sudden pedestrian behaviors, leading to poor early warning effects and impacting overall traffic safety. By analyzing and locating the early warning area through early warning intervention, it achieves advance perception, identifies risk factors in multiple areas, predicts risk trajectories, assesses the predicted risks, selects appropriate early warning strategies for intervention, and forms an intelligent collaborative early warning system. This improves the accuracy of pedestrian behavior prediction and thus enhances pedestrian safety in complex traffic environments.

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0024] Example 1, please refer to the appendix. Figure 1 This application provides a pedestrian safety early warning method combining high-precision sensing technology. The method is executed by a pedestrian safety early warning device combining high-precision sensing technology, and specifically includes the following steps:

[0025] S100: Locate the target warning area by conducting early warning intervention analysis on the target risk area.

[0026] Furthermore, this application S100 includes:

[0027] Based on the target risk area, network data is retrieved to obtain driving speed characteristics; the warning intervention response time is obtained interactively; the warning intervention response distance is calculated based on the warning intervention response time and acceleration speed characteristics; starting from the target risk area and using the warning intervention response distance as the road network selection scale, an extended road network selection is performed to obtain the target warning area.

[0028] Specifically, this involves conducting early warning and intervention analysis for target risk areas. This means pre-assessing areas requiring pedestrian safety risk warnings and identifying key warning areas—those requiring safety intervention measures. By accessing networked data from the vehicle-to-everything (V2X) system, the driving speed characteristics of the target risk area are obtained, including vehicle speed distribution such as average speed, maximum speed, acceleration, and braking deceleration. Networked data access refers to retrieving real-time data, including vehicle speed, pedestrian flow, and road congestion conditions, through vehicle-to-everything (V2X), cloud computing, or Intelligent Transportation Systems (ITS).

[0029] Interacting with traffic management centers or other relevant systems, the system obtains the warning intervention response time, which is the time from issuing a warning signal to the driver / autonomous driving system taking action. This includes the driver's reaction time (typically 0.5 to 1.5 seconds), system calculation time, and the time it takes for the vehicle to perform braking or evasive maneuvers. Based on the warning intervention response time and acceleration characteristics, the system calculates the minimum distance required from when the vehicle receives the warning signal to when it completes braking or evasive maneuvers. Simply put, response distance = speed × response time. However, determining the actual response distance can be more complex, involving braking deceleration, the vehicle's current speed, etc., and the specific distance depends on the actual situation. For example, if a vehicle is traveling at 50 km / h, at least 38 meters are needed to safely stop after the driver has fully reacted and braked.

[0030] Starting from the identified target risk area and using the calculated early warning intervention response distance as the radius, an extended selection is made within the road grid to form a complete target early warning area. Extended road network selection refers to selecting surrounding road grids within a certain distance range (early warning intervention response distance) centered on the target risk area to form a larger target early warning area for comprehensive detection of pedestrians, vehicles, and other moving entities. For example, at a school entrance during peak hours, all roads within a 2km radius might be included in the early warning area to ensure safety. By accurately calculating the early warning intervention response distance, areas requiring early warning can be more precisely identified, reducing false alarms and missed alarms.

[0031] S200: A pre-set early warning intervention window is established, and the environment of the target early warning area is intermittently perceived based on the early warning intervention window to obtain multiple regional dynamic information sets.

[0032] Furthermore, this application S200 includes:

[0033] Based on the target early warning area, network data is retrieved to obtain the vehicle flow characteristics and pedestrian flow characteristics of the extended road network; the vehicle flow characteristics of the extended road network are compared with the perception function table to obtain a first perception configuration strategy; the pedestrian flow characteristics of the extended road network are compared with the perception function table to obtain a second perception configuration strategy; by fusing the first and second perception configuration strategies, a target perception configuration strategy is output; a high-precision perception array is configured for the target early warning area using the target perception configuration strategy as a constraint, wherein the high-precision perception array includes multiple perception nodes configured in multiple perception areas.

[0034] Specifically, the target warning area identified in the previous analysis requires focused monitoring of pedestrian and vehicle traffic. This involves accessing network data via V2X (vehicle-to-everything), traffic management cloud platforms, cameras, and radar to acquire real-time data, including vehicle and pedestrian flow characteristics. Vehicle flow characteristics of the extended road network refer to the characteristics of vehicle flow in and around the target warning area, such as traffic volume, speed, and vehicle density. Pedestrian flow characteristics of the extended road network refer to the characteristics of pedestrian flow in and around the target warning area, such as the number of pedestrians, their direction of movement, and their speed.

[0035] The first perception configuration strategy is derived by comparing the traffic flow characteristics of the extended road network with the perception function table. For example, based on traffic flow characteristics, millimeter-wave radar sensors capable of covering a large area are selected. The perception function table consists of predefined matching rules between perception devices and different traffic scenarios, used to automatically select the optimal perception device configuration. Similarly, the second perception configuration strategy is derived by comparing the pedestrian flow characteristics of the extended road network with the perception function table. For example, if the area has high pedestrian traffic and long dwell time, a panoramic camera and infrared sensors are selected.

[0036] The first sensing configuration strategy is a sensor configuration scheme suitable for vehicle flow monitoring, while the second sensing configuration strategy is a sensor configuration scheme suitable for pedestrian monitoring. Combining the characteristics of vehicle and pedestrian flow, the first and second sensing configuration strategies are integrated to output a target sensing configuration strategy, ensuring monitoring accuracy. Using the target sensing configuration strategy as a constraint, sensing devices are deployed at multiple sensing nodes in different sub-regions of the target warning area to achieve three-dimensional monitoring and form a high-precision sensing array. By comprehensively considering the characteristics of vehicle and pedestrian flow and selecting appropriate sensing devices, the accuracy of environmental monitoring is improved.

[0037] Furthermore, this application also includes the following steps:

[0038] By fusing and analyzing the vehicle flow characteristics and pedestrian flow characteristics of the extended road network, the effective duration of early warning intervention is obtained; the early warning intervention window is localized based on the effective duration of early warning intervention; with the early warning intervention window as a constraint, multiple sensing nodes in the high-precision sensing array are synchronously and intermittently operated to obtain multiple regional dynamic sensing sequences of the multiple sensing areas; after time-series fusion of the multiple regional dynamic sensing sequences, risk feature detection is performed based on the fusion result to obtain the multiple regional dynamic information sets.

[0039] Specifically, a fusion analysis of traffic flow characteristics and pedestrian flow characteristics of the extended road network is performed to determine the effective duration of early warning intervention. This duration is calculated as the time required for a vehicle to decelerate from its current speed to a safe stop, ensuring timely warnings. The effective duration of early warning intervention refers to the time interval from detecting a potential traffic risk to taking intervention measures that effectively prevent an accident. It depends on factors such as traffic flow characteristics, pedestrian flow characteristics, and traffic light control strategies. Due to the complexity of the road network, the early warning window needs to be dynamically adjusted based on real-time conditions. Therefore, the effective duration of early warning intervention is used for adjustment to align with actual road conditions, ensuring that warnings are neither too early (leading to false alarms) nor too late (leading to accidents).

[0040] Using the localized early warning and intervention window as a constraint, multiple sensing nodes in the high-precision sensing array are operated synchronously and intermittently. These nodes are started and stopped synchronously at specific time intervals to reduce power consumption and optimize data acquisition. For example, a camera and radar can alternately collect data for 0.5 seconds per second, avoiding unnecessary redundant calculations. All sensing nodes operate collaboratively, forming multiple dynamic sensing sequences for different areas—a set of sensing data that changes over time.

[0041] Multiple dynamic sensing sequences from different regions are fused temporally and aligned spatially. Based on their mapping to the target warning area, multiple sets of synchronized image sequences from various regions are merged into a stitched sensing image for the entire region. Vehicle and pedestrian features are detected in each region's stitched sensing image, and these dynamic features are stored according to the sensing time of the stitched sensing image, resulting in multiple regional dynamic information sets. For example, the risk features detected in each region are stored together with the corresponding vehicle and pedestrian dynamic information to form a regional dynamic information set. These multiple regional dynamic information sets include a collection of vehicle, pedestrian, weather, and lighting information collected from different regions at different time points, used to analyze dynamic changes within the region, such as the changing trends of vehicle and pedestrian flow.

[0042] In summary, a suitable time window is pre-defined, typically generated based on historical data, to obtain the early warning intervention window. Based on the traffic and pedestrian flow characteristics of the extended road network accessed through the network for the target early warning area, the effective duration of the early warning intervention is set, and the early warning intervention window is localized to better reflect current road conditions. Intermittent environmental sensing is performed on the target early warning area, with a set sensing frequency. Monitoring is conducted according to this frequency to reduce computational load and energy consumption, while avoiding redundant data. Ultimately, through the above specific processes, multiple regional dynamic information sets are obtained, enabling a comprehensive analysis of the risk status of the entire early warning area and improving the intelligence level of safety early warning.

[0043] Furthermore, this application also includes the following steps:

[0044] Based on the early warning intervention window, the multiple regional dynamic perception sequences are spatiotemporally synchronized to obtain multiple sets of regional dynamic perception sequences. Based on the mapping relationship between the multiple perception regions and the target early warning region, the multi-view image sequences of the multiple sets of regional dynamic perception sequences are synchronously stitched together to output multiple regional stitched perception images. Traffic flow feature detection is performed on the first regional stitched perception image to obtain K real-time vehicle dynamic features, where the vehicle dynamic features include vehicle identification features, driving speed features, and driving direction features. Pedestrian flow feature detection is performed on the first regional stitched perception image to obtain F real-time pedestrian dynamic features, where the pedestrian dynamic features include pedestrian identification features, walking speed features, and walking direction features. The K vehicle dynamic features and F pedestrian dynamic features are associated and stored using the first perception time of the first regional stitched perception image to obtain a first regional dynamic information set. Similarly, by performing risk feature detection on the multiple regional stitched perception images, the multiple regional dynamic information sets are obtained.

[0045] Specifically, based on the localized early warning and intervention window, multiple regional dynamic sensing sequences are spatiotemporally synchronized. The sensing data from multiple regions are uniformly aligned in terms of both time axis and spatial scope to ensure the accuracy and timeliness of the analysis. Timestamp alignment is used to ensure all data is stored according to the same time scale. When different sensing devices are out of sync, a clock synchronization algorithm is used to ensure that all sensors are time-consistent. A spatial mapping method is employed to match data from different sensing devices according to geographic coordinates. For example, if A, B, and C are located at different positions within the same road segment, they are associated according to the road topology. After spatiotemporal synchronization, multiple sets of regional dynamic sensing sequences are obtained, in which the data from different sensing nodes are aligned in both time and space.

[0046] Based on the mapping relationship between multiple perception areas and target warning areas, image sequences from different perspectives are merged into a coherent image sequence to provide more comprehensive visual information. For example, image processing techniques, such as image stitching algorithms, are used to merge image sequences from different perspectives into a coherent image sequence. Feature extraction algorithms such as SIFT, ORB, and SURF are employed to find key points in the images, and fast nearest neighbor matching algorithms are used to calculate matching points from different visual images for image stitching. After image stitching, a stitched perception image of multiple regions is obtained.

[0047] From multiple stitched sensing images, one region is randomly selected as the first region for traffic flow feature monitoring. In other words, based on the stitched sensing image of this region, traffic flow characteristics are detected in real time to identify specific vehicles, including vehicle identification features (such as license plate, vehicle type, speed, etc.), driving speed (average speed), and driving direction. For example, at an intersection, three vehicles are detected: Vehicle 1 is an SUV traveling at 45 km / h from east to west; Vehicle 2 is a sedan traveling at 60 km / h from south to north; and Vehicle 3 is a bus traveling at 35 km / h from north to south. Depending on the specific road conditions, the direction can be specified down to the street name.

[0048] Based on the first region's stitched perception image, pedestrian flow feature detection is performed, and pedestrians passing by are detected in real time, including pedestrian identification features (such as pedestrian gender, clothing color, distinctive features, etc.), walking speed features (average speed), and walking direction features.

[0049] The dynamic features of K real-time vehicles and F real-time pedestrians are associated and stored, linked according to the first perception time of the first region stitched perception image. For example, traffic and pedestrian flow data at 12:00 are one set. Then, the situation at the next moment is recorded and stored in the database. The dynamic features of each vehicle and pedestrian are associated with the specific timestamp of their appearance in the image. This process is repeated for multiple regions of stitched perception images to obtain multiple sets of dynamic information for each region, including the dynamic features of vehicles and pedestrians in each region. By synchronously stitching multi-view image sequences and detecting multiple features, the perception accuracy of traffic and pedestrian flow dynamics is improved.

[0050] S300: Based on the information perception time series, identify the time series risk characteristics of the dynamic information sets of the multiple regions and locate the distribution of dynamic risk factors.

[0051] Furthermore, this application S300 includes:

[0052] Based on the consistency of vehicle identification features, the driving change features of the multiple regional dynamic information sets are aggregated in a time series to obtain K first feature change sequences of the K real-time vehicles; based on the consistency of pedestrian identification features, the walking change features of the multiple regional dynamic information sets are aggregated in a time series to obtain F second feature change sequences of the F real-time pedestrians; a preset driving feature change scale is used, which includes a driving speed change scale and a driving direction change scale; the driving feature change scale is used to traverse the K first feature change sequences to perform risk assessment of the K real-time vehicles, outputting M risky vehicles; similarly, a preset walking feature change scale is used to perform risk assessment of the F real-time pedestrians, outputting N risky pedestrians; M vehicle position features of the M risky vehicles and N pedestrian position features of the N risky pedestrians are extracted as the dynamic risk factor distribution.

[0053] Specifically, dynamic information from multiple regions is aggregated based on the driving change characteristics of the same vehicle over time. Data on vehicle driving characteristics (such as speed and direction) at different time points are merged and analyzed to track changes in vehicle behavior, resulting in K real-time vehicle first feature change sequences, including driving speed and driving direction change characteristics for each vehicle. Vehicle identification feature consistency refers to the ability to identify and confirm the same vehicle across different times and perception areas using information such as license plate number, vehicle color, model, and body features.

[0054] Based on traffic safety regulations and actual conditions, pre-defined driving characteristic change scales, including driving speed change scales and driving direction change scales, are used to assess the degree of change in vehicle behavior. For example, a risk threshold is set for a vehicle speed change exceeding 20 km / h or a direction change exceeding 25 degrees within 1 second. Based on these driving characteristic change scales, the K first characteristic change sequences of K real-time vehicles are traversed to determine whether they meet the risk criteria. All vehicles exceeding the pre-defined scales are classified as risk vehicles, resulting in M ​​risk vehicles that may exhibit dangerous driving behavior.

[0055] Similar to vehicle driving, dynamic information from multiple areas is aggregated by focusing on the pedestrian movement characteristics of a single individual over time. This involves merging and analyzing pedestrian movement characteristic data from different time periods to track behavioral changes and obtain F real-time pedestrian second-feature change sequences, including changes in each individual's speed and direction. Pedestrian feature consistency is similar to vehicle recognition, utilizing methods such as clothing color, body shape, gait characteristics, and AI facial recognition to confirm the trajectory of the same pedestrian at different times.

[0056] Based on traffic safety regulations and actual conditions, pre-set scales for changes in walking characteristics are established, including scales for changes in walking speed and walking direction, to assess the degree of change in pedestrian behavior. For example, a sudden increase in walking speed of 0.8 m / s, with the possibility of suddenly crossing the road within 1 second. All pedestrian data is traversed to check if they meet the risk criteria. All pedestrians who do not meet the pre-set scales are identified as high-risk pedestrians, resulting in N high-risk pedestrians who may engage in dangerous behaviors, such as jaywalking.

[0057] M vehicle location features from M high-risk vehicles and N pedestrian location features from N high-risk pedestrians are extracted to form a dynamic risk factor distribution, indicating where and at what time high-risk vehicles and pedestrians exist, which is used for risk area modeling. Through temporal aggregation and preset feature change scales, the behavior of vehicles and pedestrians that may lead to traffic accidents can be accurately identified, allowing for early response to potential dangers and timely warnings.

[0058] S400: Using the target risk area as the risk boundary, predict the risk trajectory of the dynamic risk factor distribution to obtain the early warning intervention risk distribution.

[0059] Furthermore, this application S400 includes:

[0060] Extract M first feature change sequences from the M risky vehicles; extract N second feature change sequences from the N risky pedestrians; using the target risk area as the risk boundary and the target warning area as the warning boundary, predict the driving of the M risky vehicles based on the M first feature change sequences to obtain the vehicle intervention risk distribution; using the target risk area as the risk boundary and the target warning area as the warning boundary, predict the walking of the N risky pedestrians based on the N second feature change sequences to obtain the pedestrian intervention risk distribution; perform cross-fusion of the vehicle intervention risk distribution and the pedestrian intervention risk distribution based on the risk location distribution to output the warning intervention risk distribution.

[0061] Specifically, M sequences of changes in the primary characteristics of M risky vehicles are randomly extracted from the dynamic risk factor distribution; these sequences represent data sequences showing the changes in vehicle driving characteristics (such as speed and direction) over time. The target risk area is used as the risk boundary, and the target warning area as the warning boundary. The risk boundary defines the area where the risk exists, and the warning boundary defines the area where a warning is issued. Machine learning algorithms, such as time series prediction models, are used to analyze the driving characteristic change sequences of each risky vehicle to predict its future driving behavior. For example, by analyzing changes in vehicle speed and direction, it can be predicted whether the vehicle is likely to speed or suddenly turn.

[0062] Choose an appropriate time series prediction model, such as an ARIMA model or an LSTM network, to analyze the driving characteristic change sequence of each risk vehicle. Predict the trajectory of each vehicle for the next two seconds using data from the past two seconds. For example, use an ARIMA model to analyze changes in vehicle speed and direction over time to predict future speed and direction. Based on the analysis results of the time series prediction model, predict the future driving behavior of each risk vehicle. For example, predict changes in vehicle speed and direction in the next few seconds. Generate a vehicle risk distribution based on the predicted driving behavior. For example, if a vehicle's speed decreases from 20 m / s to 5 m / s and its direction angle shifts by 15 degrees in the past 5 seconds, it is predicted that it may change lanes or stop at the next moment. The vehicle intervention risk distribution refers to the risk distribution map obtained by spatially analyzing the driving prediction results of all M risk vehicles.

[0063] Similarly, the same steps are performed for at-risk pedestrians, which will not be detailed here. N second-feature change sequences of N at-risk pedestrians are randomly extracted from the dynamic risk factor distribution; these are data sequences showing the changes in pedestrian driving characteristics (such as speed and direction) over time. Using the target risk area as the risk boundary and the target warning area as the warning boundary, pedestrian movement predictions are made for the N at-risk pedestrians based on the N second-feature change sequences, resulting in a pedestrian intervention risk distribution. For example, if a pedestrian is found to linger at an intersection for more than 10 seconds and exhibits random directional changes, it may be due to obstructed vision or distraction, and it is predicted that they may enter the lane within the next 3 seconds. The pedestrian intervention risk distribution refers to the spatial analysis of the pedestrian movement prediction results for all N at-risk pedestrians, resulting in a risk distribution map.

[0064] Based on the determined risk location distribution—that is, the location of each risk—the risk distributions of vehicles and pedestrians are cross-fused, considering the positional relationship between risky vehicles and pedestrians, and their impact on traffic flow, to comprehensively obtain the early warning and intervention risk distribution for the target area. This early warning and intervention risk distribution is the final calculated risk distribution map of the entire target area, which can help traffic management systems or intelligent driving systems to conduct proactive early warnings and interventions. By extracting the characteristic change sequences of risky vehicles and pedestrians, performing driving and walking predictions, and cross-fusion based on the risk location distribution, more efficient and accurate risk assessment and early warning strategies are provided for intelligent transportation systems, effectively improving traffic safety.

[0065] S500: Evaluate the risk level based on the risk distribution of the early warning intervention and output the real-time risk level.

[0066] S600: Match early warning strategies based on the real-time risk level and output a traffic flow collaborative early warning strategy.

[0067] Specifically, based on the aforementioned calculated risk distribution for early warning and intervention, i.e., the traffic risk distribution map of the target area, the risk situation of each area at the current moment is displayed. A risk hierarchy evaluation is performed on the risk distribution for early warning and intervention, classifying and grading risk areas to quantify the risk level. Risk assessment methods, such as the Analytic Hierarchy Process (AHP) or fuzzy comprehensive evaluation, are used to analyze the risk distribution for early warning and intervention, dividing the risks into different levels. Based on the risk distribution, risks are divided into three levels: high, medium, and low. Based on the results of the risk hierarchy evaluation, the real-time risk level of each area is output. For example, if the risk of a certain area is assessed as high risk, a high-risk level is output.

[0068] Based on real-time risk levels, appropriate early warning strategies are selected, such as adjusting traffic light durations, issuing driver alerts, and deploying traffic police patrols. Traffic flow collaborative early warning strategies are optimization strategies that comprehensively consider factors such as vehicles, pedestrians, traffic lights, and road control, aiming to reduce traffic congestion and accident risks. Based on the matching results of the early warning strategies, a traffic flow collaborative early warning strategy is output. For example, if an emergency warning is selected, it may be necessary to coordinate traffic lights and vehicle communication systems to ensure timely delivery of warning information. For instance, if the current vehicle flow at an intersection is 200 vehicles / min and the current pedestrian flow is 1320 people / min, the calculated risk value is R = 0.88 (high risk). The matched early warning strategy is: in-vehicle navigation prompts a high-risk area, advising drivers to drive cautiously, recommending low speeds, adjusting traffic lights, increasing traffic police dispatch and robot patrols, and issuing voice broadcasts to remind pedestrians to pay attention to safety. By dynamically matching the optimal early warning strategy through real-time traffic risk levels and forming a traffic flow collaborative optimization scheme, a more efficient and accurate early warning strategy is provided for intelligent transportation systems, effectively improving traffic safety.

[0069] S700: Using the target warning area as the warning boundary, the traffic flow collaborative warning strategy is used to optimize pedestrian safety through traffic flow collaborative warning.

[0070] Specifically, the target warning area serves as the warning boundary, representing key road areas requiring monitoring and intervention, such as school entrances, intersections, and shopping mall entrances / exits. These areas experience high pedestrian traffic and pose significant traffic risks. Within the target area, a traffic flow collaborative warning strategy is implemented, comprehensively optimizing factors such as pedestrians, vehicles, traffic lights, and traffic management systems to improve road safety and reduce traffic accidents. For example, when vehicle traffic is low but pedestrian traffic is high, the green light duration for pedestrian crossings is extended. Electronic fences are set up, automatically triggering voice alerts when pedestrians jaywalk or enter dangerous areas. High-density pedestrian area alerts are sent to connected vehicles (V2X) to remind drivers to reduce speed. By setting target warning areas and utilizing traffic flow collaborative warning strategies, pedestrian and vehicle traffic is optimized, improving road safety and traffic efficiency.

[0071] In summary, the pedestrian safety early warning method combining high-precision sensing technology provided in this application has the following beneficial effects:

[0072] By performing early warning intervention analysis on target risk areas, the target early warning area is located; an early warning intervention window is preset, and environmental intermittent perception of the target early warning area is performed based on the early warning intervention window to obtain multiple regional dynamic information sets; based on the information perception time sequence, the temporal risk characteristics of the multiple regional dynamic information sets are identified to locate the distribution of dynamic risk factors; using the target risk area as the risk boundary, the risk trajectory of the dynamic risk factor distribution is predicted to obtain the early warning intervention risk distribution; risk level evaluation is performed based on the early warning intervention risk distribution to output the real-time risk level; early warning strategy matching is performed based on the real-time risk level to output the traffic flow collaborative early warning strategy; using the target early warning area as the early warning boundary, the traffic flow collaborative early warning strategy is used to optimize traffic flow collaborative early warning for pedestrian safety. In other words, by locating the early warning area through early warning intervention analysis, early perception is achieved, risk factors in multiple areas are identified and risk trajectory prediction is performed, the predicted risks are evaluated, and appropriate early warning strategies are selected for early warning intervention, forming an intelligent collaborative early warning system, improving the accuracy of pedestrian behavior prediction, and thus enhancing the level of pedestrian safety in complex traffic environments.

[0073] Example 2: Based on the same inventive concept as the pedestrian safety early warning method combining high-precision sensing technology in Example 1, this application also provides a pedestrian safety early warning device combining high-precision sensing technology. Please refer to the appendix. Figure 2 The pedestrian safety early warning device combining high-precision sensing technology includes:

[0074] The system comprises the following modules: Early Warning Intervention Analysis Module 11, for locating a target early warning area by performing early warning intervention analysis on the target risk area; Intermittent Environmental Perception Module 12, for presetting an early warning intervention window and performing intermittent environmental perception of the target early warning area based on the early warning intervention window to obtain multiple regional dynamic information sets; Risk Factor Location Module 13, for identifying the temporal risk characteristics of the multiple regional dynamic information sets based on the information perception time sequence to locate the dynamic risk factor distribution; Risk Trajectory Prediction Module 14, for predicting the risk trajectory of the dynamic risk factor distribution using the target risk area as the risk boundary to obtain the early warning intervention risk distribution; Risk Level Determination Module 15, for evaluating the risk level based on the early warning intervention risk distribution and outputting a real-time risk level; Early Warning Strategy Matching Module 16, for matching early warning strategies based on the real-time risk level and outputting a traffic flow collaborative early warning strategy; and Collaborative Early Warning Module 17, for optimizing pedestrian safety through traffic flow collaborative early warning using the traffic flow collaborative early warning strategy with the target early warning area as the early warning boundary.

[0075] Furthermore, the early warning intervention analysis module 11 in the pedestrian safety early warning device incorporating high-precision sensing technology is also used for:

[0076] Based on the target risk area, network data is retrieved to obtain driving speed characteristics; the warning intervention response time is obtained interactively; the warning intervention response distance is calculated based on the warning intervention response time and acceleration speed characteristics; starting from the target risk area and using the warning intervention response distance as the road network selection scale, an extended road network selection is performed to obtain the target warning area.

[0077] Furthermore, the intermittent environmental sensing module 12 in the pedestrian safety early warning device incorporating high-precision sensing technology is also used for:

[0078] Based on the target early warning area, network data is retrieved to obtain the vehicle flow characteristics and pedestrian flow characteristics of the extended road network; the vehicle flow characteristics of the extended road network are compared with the perception function table to obtain a first perception configuration strategy; the pedestrian flow characteristics of the extended road network are compared with the perception function table to obtain a second perception configuration strategy; by fusing the first and second perception configuration strategies, a target perception configuration strategy is output; a high-precision perception array is configured for the target early warning area using the target perception configuration strategy as a constraint, wherein the high-precision perception array includes multiple perception nodes configured in multiple perception areas.

[0079] Furthermore, the intermittent environmental sensing module 12 in the pedestrian safety early warning device incorporating high-precision sensing technology is also used for:

[0080] By fusing and analyzing the vehicle flow characteristics and pedestrian flow characteristics of the extended road network, the effective duration of early warning intervention is obtained; the early warning intervention window is localized based on the effective duration of early warning intervention; with the early warning intervention window as a constraint, multiple sensing nodes in the high-precision sensing array are synchronously and intermittently operated to obtain multiple regional dynamic sensing sequences of the multiple sensing areas; after time-series fusion of the multiple regional dynamic sensing sequences, risk feature detection is performed based on the fusion result to obtain the multiple regional dynamic information sets.

[0081] Furthermore, the intermittent environmental sensing module 12 in the pedestrian safety early warning device incorporating high-precision sensing technology is also used for:

[0082] Based on the early warning intervention window, the multiple regional dynamic perception sequences are spatiotemporally synchronized to obtain multiple sets of regional dynamic perception sequences. Based on the mapping relationship between the multiple perception regions and the target early warning region, the multi-view image sequences of the multiple sets of regional dynamic perception sequences are synchronously stitched together to output multiple regional stitched perception images. Traffic flow feature detection is performed on the first regional stitched perception image to obtain K real-time vehicle dynamic features, where the vehicle dynamic features include vehicle identification features, driving speed features, and driving direction features. Pedestrian flow feature detection is performed on the first regional stitched perception image to obtain F real-time pedestrian dynamic features, where the pedestrian dynamic features include pedestrian identification features, walking speed features, and walking direction features. The K vehicle dynamic features and F pedestrian dynamic features are associated and stored using the first perception time of the first regional stitched perception image to obtain a first regional dynamic information set. Similarly, by performing risk feature detection on the multiple regional stitched perception images, the multiple regional dynamic information sets are obtained.

[0083] Furthermore, the risk factor positioning module 13 in the pedestrian safety early warning device incorporating high-precision sensing technology is also used for:

[0084] Based on the consistency of vehicle identification features, the driving change features of the multiple regional dynamic information sets are aggregated in a time series to obtain K first feature change sequences of the K real-time vehicles; based on the consistency of pedestrian identification features, the walking change features of the multiple regional dynamic information sets are aggregated in a time series to obtain F second feature change sequences of the F real-time pedestrians; a preset driving feature change scale is used, which includes a driving speed change scale and a driving direction change scale; the driving feature change scale is used to traverse the K first feature change sequences to perform risk assessment of the K real-time vehicles, outputting M risky vehicles; similarly, a preset walking feature change scale is used to perform risk assessment of the F real-time pedestrians, outputting N risky pedestrians; M vehicle position features of the M risky vehicles and N pedestrian position features of the N risky pedestrians are extracted as the dynamic risk factor distribution.

[0085] Furthermore, the risk trajectory prediction module 14 in the pedestrian safety early warning device incorporating high-precision sensing technology is also used for:

[0086] Extract M first feature change sequences from the M risky vehicles; extract N second feature change sequences from the N risky pedestrians; using the target risk area as the risk boundary and the target warning area as the warning boundary, predict the driving of the M risky vehicles based on the M first feature change sequences to obtain the vehicle intervention risk distribution; using the target risk area as the risk boundary and the target warning area as the warning boundary, predict the walking of the N risky pedestrians based on the N second feature change sequences to obtain the pedestrian intervention risk distribution; perform cross-fusion of the vehicle intervention risk distribution and the pedestrian intervention risk distribution based on the risk location distribution to output the warning intervention risk distribution.

[0087] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Figure 1 The pedestrian safety warning method and specific examples combining high-precision sensing technology in Embodiment 1 are also applicable to the pedestrian safety warning device combining high-precision sensing technology in this embodiment. Through the foregoing detailed description of the pedestrian safety warning method combining high-precision sensing technology, those skilled in the art can clearly understand the pedestrian safety warning device combining high-precision sensing technology in this embodiment; therefore, for the sake of brevity, it will not be described in detail here. As for the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant details can be found in the method section.

[0088] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0089] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A pedestrian safety early warning method combining high-precision sensing technology, characterized in that, include: By conducting early warning intervention analysis on the target risk area, the target early warning area can be located; A preset early warning intervention window is established, and environmental intermittent sensing of the target early warning area is performed based on the early warning intervention window to obtain multiple regional dynamic information sets; Based on information perception time series, the temporal risk characteristics of the dynamic information sets of the multiple regions are identified, and the distribution of dynamic risk factors is located. Using the target risk area as the risk boundary, the risk trajectory of the dynamic risk factor distribution is predicted to obtain the early warning intervention risk distribution; Based on the risk distribution of the early warning intervention, a risk level assessment is performed, and the real-time risk level is output. Based on the real-time risk level, a warning strategy is matched, and a traffic flow collaborative warning strategy is output. Using the target warning area as the warning boundary, the traffic flow collaborative warning strategy is adopted to optimize the traffic flow collaborative warning for pedestrian safety. Based on the aforementioned early warning intervention window, intermittent environmental sensing of the target early warning area is performed to obtain multiple sets of dynamic information for the area. Prior to this, the process also includes: Based on the target early warning area, network data is retrieved to obtain the vehicle flow characteristics and pedestrian flow characteristics of the extended road network; The first perception configuration strategy is obtained by comparing the vehicle flow characteristics of the extended road network with the perception function table. The second perception configuration strategy is obtained by comparing the pedestrian flow characteristics of the extended road network with the perception function table. By integrating the first perception configuration strategy and the second perception configuration strategy, the target perception configuration strategy is output. The high-precision sensing array of the target early warning area is configured with the target perception configuration strategy as a constraint, wherein the high-precision sensing array includes multiple sensing nodes configured in multiple sensing areas; Based on the aforementioned early warning intervention window, intermittent environmental sensing of the target early warning area is performed to obtain multiple sets of dynamic information for the area, including: By integrating and analyzing the vehicle flow characteristics and pedestrian flow characteristics of the extended road network, the effective duration of early warning intervention can be obtained; The early warning intervention window is localized based on the effective duration of the early warning intervention. With the early warning intervention window as a constraint, multiple sensing nodes in the high-precision sensing array are operated synchronously and intermittently to obtain multiple dynamic sensing sequences of the multiple sensing areas. After temporal fusion of the dynamic sensing sequences of the multiple regions, risk feature detection is performed based on the fusion result to obtain the dynamic information set of the multiple regions. After temporal fusion of the multiple region dynamic sensing sequences, risk feature detection is performed based on the fusion result to obtain the dynamic information set of the multiple regions, which also includes: Based on the early warning intervention window, the spatiotemporal synchronization of the multiple regional dynamic sensing sequences is performed to obtain multiple sets of regional dynamic sensing sequences. Based on the mapping relationship between the multiple sensing areas and the target warning area, the multi-view image sequences of the multiple sets of dynamic sensing sequences of the regions are synchronously stitched together to output multiple region stitched sensing images. Traffic flow features are detected in the first region stitched perception image to obtain K vehicle dynamic features of K real-time vehicles, wherein the vehicle dynamic features include vehicle identification features, driving speed features and driving direction features. The first region's stitched perception image is subjected to pedestrian flow feature detection to obtain F real-time pedestrian dynamic features, wherein the pedestrian dynamic features include pedestrian recognition features, walking speed features, and walking direction features; K vehicle dynamic features and F pedestrian dynamic features are associated and stored using the first perception time of the first region stitched perception image to obtain the first region dynamic information set; Similarly, by performing risk feature detection on the stitched perception images of the multiple regions, a dynamic information set of the multiple regions is obtained; Based on information-aware time series analysis, the system identifies the time-series risk characteristics of the dynamic information sets of the multiple regions, locates the distribution of dynamic risk factors, and further includes: Based on the consistency of vehicle identification features, the driving change features of the multiple regional dynamic information sets are aggregated in time to obtain the K first feature change sequences of the K real-time vehicles. Based on the consistency of pedestrian recognition features, the walking change features of the multiple regional dynamic information sets are aggregated in time to obtain the F second feature change sequences of the F real-time pedestrians; A preset driving characteristic change scale is defined, wherein the driving characteristic change scale includes a driving speed change scale and a driving direction change scale; The driving feature change scale is used to traverse the K first feature change sequences to perform risk assessment on the K real-time vehicles, and output M risk vehicles; Similarly, the risk assessment of the F real-time pedestrians is performed based on the preset walking feature change scale, and N risk pedestrians are output; M vehicle location features of the M risky vehicles and N pedestrian location features of the N risky pedestrians are extracted as the dynamic risk factor distribution.

2. The pedestrian safety early warning method combining high-precision sensing technology as described in claim 1, characterized in that, By conducting early warning intervention analysis on target risk areas, the target early warning areas are located, including: Based on the target risk area, network data is retrieved to obtain driving speed characteristics; Interactively obtain the early warning and intervention response time; Based on the aforementioned early warning and intervention response duration and acceleration characteristics, the early warning and intervention response distance is calculated. Starting from the target risk area and using the early warning intervention response distance as the road network selection scale, an extended road network selection is performed to obtain the target early warning area.

3. The pedestrian safety early warning method combining high-precision sensing technology as described in claim 1, characterized in that, Using the target risk area as the risk boundary, the risk trajectory of the dynamic risk factor distribution is predicted to obtain the early warning intervention risk distribution, which also includes: Extract the M first feature change sequences from the M risk vehicles; Extract N second feature change sequences from the N risky pedestrians; Using the target risk area as the risk boundary and the target warning area as the warning boundary, driving predictions for the M risk vehicles are made based on the M first feature change sequences to obtain the vehicle intervention risk distribution. Using the target risk area as the risk boundary and the target warning area as the warning boundary, the walking prediction of the N risky pedestrians is performed based on the N second feature change sequences to obtain the pedestrian intervention risk distribution; The risk distribution of vehicle intervention and pedestrian intervention are cross-fused based on the risk location distribution to output the early warning intervention risk distribution.

4. A pedestrian safety early warning device incorporating high-precision sensing technology, characterized in that: The step of implementing the pedestrian safety early warning method combining high-precision sensing technology according to any one of claims 1 to 3, wherein the pedestrian safety early warning device combining high-precision sensing technology comprises: The early warning intervention analysis module is used to locate the target early warning area by performing early warning intervention analysis on the target risk area; An intermittent environmental sensing module is used to preset an early warning intervention window and perform intermittent environmental sensing of the target early warning area based on the early warning intervention window to obtain multiple regional dynamic information sets. The risk factor localization module is used to identify the temporal risk characteristics of the dynamic information sets of the multiple regions based on the information perception time series, and to locate the distribution of dynamic risk factors. The risk trajectory prediction module is used to predict the risk trajectory of the dynamic risk factor distribution with the target risk area as the risk boundary, so as to obtain the early warning intervention risk distribution. The risk level determination module is used to evaluate the risk level based on the risk distribution of the early warning intervention and output the real-time risk level. The early warning strategy matching module is used to match early warning strategies according to the real-time risk level and output a traffic flow collaborative early warning strategy. The collaborative early warning module is used to optimize pedestrian safety by using the target early warning area as the early warning boundary and the traffic flow collaborative early warning strategy.

Images