Pedestrian-vehicle conflict early warning method for traffic field

Abstract

The invention discloses a pedestrian-vehicle conflict early warning method for the traffic field, relates to the technical field of intelligent traffic, and aims to identify and track pedestrians, non-motor vehicles and motor vehicles in real time through target detection and multi-target tracking technologies and obtain accurate positions and continuous movement tracks of the pedestrians, the non-motor vehicles and the motor vehicles. Then, correlation analysis is carried out on the motion trails, target pairs with interaction risks are screened out, and a collision time sequence of the target pairs is calculated; and finally, constructing a conflict risk quantitative index based on the sequence, generating early warning information when the risk exceeds a threshold value, and prompting a motor vehicle driver to take deceleration or avoidance measures in time.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, specifically to a method for early warning of pedestrian-vehicle conflicts in the transportation sector. Background Technology

[0002] With the deepening application of artificial intelligence technology in complex scenarios, pose estimation, as a key research direction in machine vision, has demonstrated significant value in target localization and behavior recognition. The continuous evolution of this technology provides core support for the intelligent upgrading of fields such as smart parks, smart transportation, and drones.

[0003] By the end of 2023, the number of motor vehicles in China had exceeded 435 million. The continued growth in the number of motor vehicles, coupled with the uneven distribution of citizens' commutes in time and space, has exacerbated traffic pressure and congestion on urban roads. Against this backdrop, the concept of "smart cities" has driven various data-driven road management methods. However, most existing congestion relief solutions rely primarily on motor vehicle traffic data to optimize traffic light timing, failing to fully incorporate the travel needs of pedestrians, non-motorized vehicles, and other "vulnerable road users," resulting in blind spots in intersection resource allocation.

[0004] Taking Jinan City as an example, according to traffic police statistics, there are currently 4.04 million motor vehicles, while electric bicycles have reached 4.698 million. The actual volume of pedestrian and non-motorized vehicle traffic has surpassed that of motor vehicles, becoming an indispensable part of urban traffic. However, on the one hand, frequent conflicts between pedestrians, non-motorized vehicles, and motor vehicles have become a contributing factor to traffic congestion at many intersections; on the other hand, some dangerous traffic behaviors are also highly likely to cause accidents, threatening the lives and property of citizens. Therefore, strengthening the perception and understanding of pedestrian and non-motorized vehicle information is of urgent practical significance for improving the overall traffic efficiency at intersections and ensuring traffic safety. Summary of the Invention

[0005] In order to overcome the shortcomings of the above technologies, the present invention provides a method for generating early warning information when the risk exceeds a threshold, prompting motor vehicle drivers to take timely measures to slow down or avoid the risk.

[0006] The technical solution adopted by this invention to overcome its technical problems is: A method for early warning of pedestrian-vehicle conflicts in the transportation sector includes: S1. Obtain the intersection data from the electronic police cameras at the intersection. RGB image at any given time; S2. Label the RGB image for vehicle detection area and zebra crossing detection area to obtain the vehicle detection area. and pedestrian and non-motorized vehicle passage areas ; S3. Perform object detection on the RGB image to obtain the set of detected objects. The targets include: pedestrians, non-motorized vehicles, and motorized vehicles; S4. Based on the motor vehicle inspection area Pedestrian and non-motorized vehicle traffic areas Detection target set Construct the first Detection target set for each lane line area Detection target set for pedestrian and non-motorized vehicle traffic areas ; S5. Calculate the... Perspective transformation matrix for each lane area Perspective transformation matrix of pedestrian and non-motorized vehicle passage areas ; S6. Through the first Perspective transformation matrix for each lane area Perspective transformation matrix of pedestrian and non-motorized vehicle traffic areas calculate The world coordinates of the target being detected at any given moment; S7. Calculate the time interval based on the world coordinates of the detected target. Physical displacement within ; S8. Based on physical displacement Calculated Filtered velocity value at time 1 ; S9. Based on the filtered velocity value Calculate the TTC value for reaching the conflict point; S10. Obtain the risk level based on the TTC value and output the risk level to the early warning or control system.

[0007] Furthermore, step S2 includes the following steps: S2-1. Manually mark each lane line area in the RGB image to obtain the vehicle detection area. , ,in For the first The lane line area, the first Each lane line area The coordinates of the top left corner are The coordinates of its lower left corner are The coordinates of its upper right corner are The coordinates of its lower right corner are , , The number of lane marking areas. , The x-coordinate of the top-left corner point. The ordinate of the top-left corner point. , The x-coordinate of the bottom left corner point The ordinate of the bottom left corner point. , The x-coordinate of the top right corner point. The ordinate of the top right corner point. , The x-coordinate of the bottom right corner point. The ordinate of the bottom right corner point; S2-2. Manually mark the zebra crossing detection area in the RGB image. Motor vehicle inspection area The outermost lane lines on both the left and right sides are geometrically extended towards the zebra crossing, and the two extended lines intersect the zebra crossing detection area. The intersecting areas are designated for pedestrian and non-motorized vehicle traffic. pedestrian and non-motorized vehicle traffic areas The coordinates of the top left corner are The coordinates of its lower left corner are The coordinates of its upper right corner are The coordinates of its lower right corner are , , The x-coordinate of the top-left corner point. The ordinate of the top-left corner point. , The x-coordinate of the bottom left corner point The ordinate of the bottom left corner point. , The x-coordinate of the top right corner point. The ordinate of the top right corner point. , The x-coordinate of the bottom right corner point. The ordinate of the bottom right corner point.

[0008] Furthermore, in step S3, the RGB image is input into the YOLOv11 model to identify pedestrians, non-motorized vehicles, and motorized vehicles in the RGB image, and the output is the set of detected targets. , ,in For the first One detection target, , To detect the number of targets, , For the first Individual detection targets The x-coordinate of the top left corner of the rectangular detection box. For the first Individual detection targets The ordinate of the top left corner of the rectangular detection box. For the first Individual detection targets The width of the rectangular detection box, For the first Individual detection targets The height of the rectangular detection frame.

[0009] Furthermore, step S4 includes the following steps: S4-1. According to the... Individual detection targets The x-coordinate of the top left corner of the rectangular detection box y-axis Width of the rectangular detection frame ,high Calculate the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame; S4-2. If the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame are located at the th Each lane line area Inside, then the first Individual detection targets Add to the Detection target set for each lane line area middle; S4-3. If the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame are located in the pedestrian and non-motorized vehicle passage area. Inside, then the first Individual detection targets Added to the detection target set of pedestrian and non-motorized vehicle traffic areas middle.

[0010] Furthermore, step S5 includes the following steps: S5-1. The measurement of the first [item] in a real-world intersection scenario is obtained. Each lane line area physical length and physical width , No. Each lane line area The physical coordinates of the top left corner are Its lower left corner physical coordinates are Its upper right corner physical coordinates are Its physical coordinates in the lower right corner are ; S5-2. Measure the pedestrian and non-motorized vehicle passage areas in a real-world intersection scenario. physical length and physical width pedestrian and non-motorized vehicle traffic areas The physical coordinates of the top left corner are Its lower left corner physical coordinates are Its upper right corner physical coordinates are Its physical coordinates in the lower right corner are ; S5-3. Constructing the first Each lane line area The coordinates of the top left corner The coordinates of the lower left corner The coordinates of the upper right corner The coordinates of the bottom right corner Using coordinate points Coordinate point pair Coordinate point pair Coordinate point pair The solution for the 3rd row and 3rd column is obtained by the direct linear transformation algorithm. Perspective transformation matrix for each lane area , ,in Perspective transformation matrix The Middle Line number Column parameters, , ; S5-4. Construct pedestrian and non-motorized vehicle access areas The coordinates of the top left corner The coordinates of the lower left corner The coordinates of the upper right corner The coordinates of the bottom right corner Using coordinate points Coordinate point pair Coordinate point pair Coordinate point pair The perspective transformation matrix of the 3x3 pedestrian and non-motorized vehicle passage area was obtained by solving the direct linear transformation algorithm. , ,in Perspective transformation matrix The Middle Line number Column parameters, , .

[0011] Furthermore, step S6 includes the following steps: S6-1. For the first Detection target set for each lane line area The first in Individual detection targets Based on the x-coordinate of the top left corner of its rectangular detection box The ordinate of the top left corner ,width ,high The x-coordinate of the bottom center point of the rectangular detection box was calculated. and the center point pixel ordinate Through formula Calculated Time of the first Individual detection targets x-coordinate of world coordinates Through formula Calculated Time of the first Individual detection targets The ordinate of the world coordinates ; S6-2. Detection target set for pedestrian and non-motorized vehicle traffic areas The first in Individual detection targets Based on the x-coordinate of the top left corner of its rectangular detection box The ordinate of the top left corner ,width ,high The x-coordinate of the bottom center point of the rectangular detection box was calculated. and the center point pixel ordinate Through formula Calculated Time of the first Individual detection targets x-coordinate of world coordinates Through formula Calculated Time of the first Individual detection targets The ordinate of the world coordinates .

[0012] Furthermore, step S7 includes the following steps: S7-1. Through formula Calculations were made at the time interval The first Individual detection targets physical displacement In the formula, the time interval is... The reciprocal of the video frame rate per second. for Time of the first Individual detection targets The x-coordinate of the world coordinates, for Time of the first Individual detection targets The ordinate of the world coordinates.

[0013] Furthermore, step S8 includes the following steps: S8-1. Through formula Calculate the first Individual detection targets exist instantaneous velocity at a moment ; S8-2. Through formula Calculated Filtered velocity value at time 1 In the formula, As a smoothing factor, , for The velocity value after filtering at any given time. .

[0014] Furthermore, step S9 includes the following steps: S9-1. The first Each lane line area The intersection of the central axis and the zebra crossing is used as the point of conflict. ; S9-2. For the first Detection target set for each lane line area The first in Individual detection targets Based on the filtered velocity value Calculate the time it takes for the motor vehicle to arrive at the conflict point ; S9-3. Detection target set for pedestrian and non-motorized vehicle traffic areas The first in Individual detection targets Based on the filtered velocity value Calculate the time it takes for pedestrians or non-motorized vehicles to arrive at the conflict point. ; S9-4. Pass The TTC value of the point of conflict is calculated.

[0015] Furthermore, in step S10, when the TTC value is less than or equal to 3 seconds, it is determined to be high risk and a high-level warning is triggered; when the TTC value is greater than 3 seconds and less than or equal to 5 seconds, it is determined to be medium risk and a low-level warning is triggered; when the TTC value is greater than 5 seconds, it is determined to be low risk and no warning is triggered.

[0016] The beneficial effects of this invention are: It overcomes deployment limitations: by utilizing the inherent lane lines and zebra crossings at intersections as calibration references, it overcomes the dependence of traditional methods on precise camera parameters and additional calibration objects, enabling the system to be deployed quickly and economically in existing electronic police camera networks.

[0017] Data accuracy is guaranteed: The joint perspective transformation method based on inherent traffic signs can robustly establish the mapping relationship between image space and world space, providing a high-precision data source for subsequent distance, speed and TTC calculations, and ensuring the reliability of the early warning.

[0018] Active safety has been achieved: by integrating target detection, tracking and dynamic TTC calculation, quantitative assessment and early warning of the risk of "human-vehicle" conflict have been realized, shifting traffic safety management from post-event handling to pre-event prevention, effectively improving the traffic safety and efficiency at intersections. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] The following is in conjunction with the appendix Figure 1 The present invention will be further described below.

[0021] A method for early warning of pedestrian-vehicle conflicts in the transportation sector includes: S1. Obtain the intersection data from the electronic police cameras at the intersection. RGB image at any given time.

[0022] S2. Label the RGB image for vehicle detection area and zebra crossing detection area to obtain the vehicle detection area. and pedestrian and non-motorized vehicle passage areas .

[0023] S3. Perform object detection on the RGB image to obtain the set of detected objects. The targets include: pedestrians, non-motorized vehicles, and motorized vehicles.

[0024] S4. Based on the motor vehicle inspection area Pedestrian and non-motorized vehicle traffic areas Detection target set Construct the first Detection target set for each lane line area Detection target set for pedestrian and non-motorized vehicle traffic areas .

[0025] S5. Calculate the... Perspective transformation matrix for each lane area Perspective transformation matrix of pedestrian and non-motorized vehicle passage areas .

[0026] S6. Through the first Perspective transformation matrix for each lane area Perspective transformation matrix of pedestrian and non-motorized vehicle traffic areas calculate The world coordinates of the target being detected at any given moment.

[0027] S7. Calculate the time interval based on the world coordinates of the detected target. Physical displacement within .

[0028] S8. Based on physical displacement Calculated Filtered velocity value at time 1 .

[0029] S9. Based on the filtered velocity value Calculate the TTC value for reaching the conflict point.

[0030] S10. Obtain the risk level based on the TTC value and output the risk level to the early warning or control system.

[0031] This system eliminates the need for complex external camera parameter calibration and dedicated calibration objects in the scene. Instead, it directly utilizes the inherent geometric features of traffic scenarios, such as lane lines and zebra crossings, to complete system calibration, significantly reducing deployment costs and technical barriers. By fusing a joint perspective transformation model with multiple features, a high-precision coordinate mapping relationship from image pixels to the real physical world is established, providing a reliable data foundation for real-time distance and speed calculations. Based on this high-precision coordinate mapping and dynamic speed information, a complete "perception-computation-early warning" closed-loop system is constructed. Through real-time calculation of the time-to-collision (TTC) sequence between motor vehicles and pedestrians / non-motorized vehicles, and based on this, dynamic risk classification and early warning are performed, achieving a leap from passive perception to proactive intervention and significantly improving the proactive safety of intersection traffic. This invention can effectively improve the proactive safety of intersection traffic, reduce the probability of pedestrian-vehicle conflicts, and provide intelligent and forward-looking decision support for urban traffic management.

[0032] In one embodiment of the present invention, step S2 includes the following steps: S2-1. Manually mark each lane line area in the RGB image to obtain the vehicle detection area. , ,in For the first The lane line area, the first Each lane line area The coordinates of the top left corner are The coordinates of its lower left corner are The coordinates of its upper right corner are The coordinates of its lower right corner are , , The number of lane marking areas. , The x-coordinate of the top-left corner point. The ordinate of the top-left corner point. , The x-coordinate of the bottom left corner point The ordinate of the bottom left corner point. , The x-coordinate of the top right corner point. The ordinate of the top right corner point. , The x-coordinate of the bottom right corner point. The ordinate of the bottom right corner point.

[0033] S2-2. Manually mark the zebra crossing detection area in the RGB image. Motor vehicle inspection area The outermost lane lines on both the left and right sides are geometrically extended towards the zebra crossing, and the two extended lines intersect the zebra crossing detection area. The intersecting areas are designated for pedestrian and non-motorized vehicle traffic. pedestrian and non-motorized vehicle traffic areas The coordinates of the top left corner are The coordinates of its lower left corner are The coordinates of its upper right corner are The coordinates of its lower right corner are , , The x-coordinate of the top-left corner point. The ordinate of the top-left corner point. , The x-coordinate of the bottom left corner point The ordinate of the bottom left corner point. , The x-coordinate of the top right corner point. The ordinate of the top right corner point. , The x-coordinate of the bottom right corner point. The ordinate of the bottom right corner point.

[0034] In one embodiment of the present invention, step S3 involves inputting an RGB image into a YOLOv11 model to identify pedestrians, non-motorized vehicles, and motorized vehicles in the RGB image, and outputting a set of detected targets. , ,in For the first One detection target, , To detect the number of targets, , For the first Individual detection targets The x-coordinate of the top left corner of the rectangular detection box. For the first Individual detection targets The ordinate of the top left corner of the rectangular detection box. For the first Individual detection targets The width of the rectangular detection box, For the first Individual detection targets The height of the rectangular detection frame.

[0035] In one embodiment of the present invention, step S4 includes the following steps: S4-1. According to the... Individual detection targets The x-coordinate of the top left corner of the rectangular detection box y-axis Width of the rectangular detection frame ,high Calculate the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame.

[0036] S4-2. If the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame are located at the th Each lane line area Inside, then the first Individual detection targets Add to the Detection target set for each lane line area middle.

[0037] S4-3. If the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame are located in the pedestrian and non-motorized vehicle passage area. Inside, then the first Individual detection targets Added to the detection target set of pedestrian and non-motorized vehicle traffic areas middle.

[0038] In one embodiment of the present invention, step S5 includes the following steps: S5-1. The measurement of the first [item] in a real-world intersection scenario is obtained. Each lane line area physical length and physical width , No. Each lane line area The physical coordinates of the top left corner are Its lower left corner physical coordinates are Its upper right corner physical coordinates are Its physical coordinates in the lower right corner are .

[0039] S5-2. Measure the pedestrian and non-motorized vehicle passage areas in a real-world intersection scenario. physical length and physical width pedestrian and non-motorized vehicle traffic areas The physical coordinates of the top left corner are Its lower left corner physical coordinates are Its upper right corner physical coordinates are Its physical coordinates in the lower right corner are .

[0040] S5-3. Constructing the first Each lane line area The coordinates of the top left corner The coordinates of the lower left corner The coordinates of the upper right corner The coordinates of the bottom right corner Using coordinate points Coordinate point pair Coordinate point pair Coordinate point pair The solution for the 3rd row and 3rd column is obtained by the direct linear transformation algorithm. Perspective transformation matrix for each lane area , ,in Perspective transformation matrix The Middle Line number Column parameters, , .

[0041] S5-4. Construct pedestrian and non-motorized vehicle access areas The coordinates of the top left corner The coordinates of the lower left corner The coordinates of the upper right corner The coordinates of the bottom right corner Using coordinate points Coordinate point pair Coordinate point pair Coordinate point pair The perspective transformation matrix of the 3x3 pedestrian and non-motorized vehicle passage area was obtained by solving the direct linear transformation algorithm. , ,in Perspective transformation matrix The Middle Line number Column parameters, , .

[0042] Based on this mapping relationship, by calculating the scale ratio of the reference benchmark (such as the selected lane line) in the image and the real world, the conversion relationship from pixels to physical size (such as meters) can be accurately established, thus providing a unified benchmark for subsequent distance and speed calculations.

[0043] In one embodiment of the present invention, step S6 includes the following steps: S6-1. For the first Detection target set for each lane line area The first in Individual detection targets Based on the x-coordinate of the top left corner of its rectangular detection box The ordinate of the top left corner ,width ,high The x-coordinate of the bottom center point of the rectangular detection box was calculated. and the center point pixel ordinate Through formula Calculated Time of the first Individual detection targets x-coordinate of world coordinates Through formula Calculated Time of the first Individual detection targets The ordinate of the world coordinates .

[0044] S6-2. Detection target set for pedestrian and non-motorized vehicle traffic areas The first in Individual detection targets Based on the x-coordinate of the top left corner of its rectangular detection box The ordinate of the top left corner ,width ,high The x-coordinate of the bottom center point of the rectangular detection box was calculated. and the center point pixel ordinate Through formula Calculated Time of the first Individual detection targets x-coordinate of world coordinates Through formula Calculated Time of the first Individual detection targets The ordinate of the world coordinates .

[0045] In one embodiment of the present invention, step S7 includes the following steps: S7-1. Through formula Calculations were made at the time interval The first Individual detection targets physical displacement (Unit: meters), where the time interval is... The reciprocal of the video frame rate per second. for Time of the first Individual detection targets The x-coordinate of the world coordinates, for Time of the first Individual detection targets The ordinate of the world coordinates.

[0046] In one embodiment of the present invention, step S8 includes the following steps: S8-1. Through formula Calculate the first Individual detection targets exist instantaneous velocity at a moment (Unit: meters per second).

[0047] S8-2. Through formula Calculated Filtered velocity value at time 1 In the formula, As a smoothing factor, , for The velocity value after filtering at any given time. In this embodiment, preferably, The value is 0.3.

[0048] In one embodiment of the present invention, step S9 includes the following steps: S9-1. The first Each lane line area The intersection of the central axis and the zebra crossing is used as the point of conflict. .

[0049] S9-2. For the first Detection target set for each lane line area The first in Individual detection targets Based on the filtered velocity value Calculate the time it takes for the motor vehicle to arrive at the conflict point .

[0050] S9-3. Detection target set for pedestrian and non-motorized vehicle traffic areas The first in Individual detection targets Based on the filtered velocity value Calculate the time it takes for pedestrians or non-motorized vehicles to arrive at the conflict point. .

[0051] S9-4. Pass The TTC value of the point of conflict is calculated.

[0052] In one embodiment of the present invention, in step S10, when the TTC value is less than or equal to 3 seconds, it is determined to be high risk and a high-level warning is triggered; when the TTC value is greater than 3 seconds and less than or equal to 5 seconds, it is determined to be medium risk and a low-level warning is triggered; when the TTC value is greater than 5 seconds, it is determined to be low risk and no warning is triggered. The final risk level information is output to the relevant warning or control system to prompt the driver to take evasive action or to activate traffic indication equipment.

[0053] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for early warning of pedestrian-vehicle conflicts in the transportation sector, characterized in that, include: S1. Obtain the intersection data from the electronic police cameras at the intersection. RGB image at any given time; S2. Label the RGB image for vehicle detection area and zebra crossing detection area to obtain the vehicle detection area. and pedestrian and non-motorized vehicle passage areas ; S3. Perform object detection on the RGB image to obtain the set of detected objects. The targets include: pedestrians, non-motorized vehicles, and motorized vehicles; S4. Based on the motor vehicle inspection area Pedestrian and non-motorized vehicle traffic areas Detection target set Construct the first Detection target set for each lane line area Detection target set for pedestrian and non-motorized vehicle traffic areas ; S5. Calculate the... Perspective transformation matrix for each lane area Perspective transformation matrix of pedestrian and non-motorized vehicle passage areas ; S6. Through the first Perspective transformation matrix for each lane area Perspective transformation matrix of pedestrian and non-motorized vehicle traffic areas calculate The world coordinates of the target being detected at any given moment; S7. Calculate the time interval based on the world coordinates of the detected target. Physical displacement within ; S8. Based on physical displacement Calculated Filtered velocity value at time 1 ; S9. Based on the filtered velocity value Calculate the TTC value for reaching the conflict point; S10. Obtain the risk level based on the TTC value and output the risk level to the early warning or control system.

2. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 1, characterized in that, Step S2 includes the following steps: S2-1. Manually mark each lane line area in the RGB image to obtain the vehicle detection area. , ,in For the first The lane line area, the first Each lane line area The coordinates of the top left corner are The coordinates of its lower left corner are The coordinates of its upper right corner are The coordinates of its lower right corner are , , The number of lane marking areas. , The x-coordinate of the top-left corner point. The ordinate of the top-left corner point. , The x-coordinate of the bottom left corner point The ordinate of the bottom left corner point. , The x-coordinate of the top right corner point. The ordinate of the top right corner point. , The x-coordinate of the bottom right corner point. The ordinate of the bottom right corner point; S2-2. Manually mark the zebra crossing detection area in the RGB image. Motor vehicle inspection area The outermost lane lines on both the left and right sides are geometrically extended towards the zebra crossing, and the two extended lines intersect the zebra crossing detection area. The intersecting areas are designated for pedestrian and non-motorized vehicle traffic. pedestrian and non-motorized vehicle traffic areas The coordinates of the top left corner are The coordinates of its lower left corner are The coordinates of its upper right corner are The coordinates of its lower right corner are , , The x-coordinate of the top-left corner point. The ordinate of the top-left corner point. , The x-coordinate of the bottom left corner point The ordinate of the bottom left corner point. , The x-coordinate of the top right corner point. The ordinate of the top right corner point. , The x-coordinate of the bottom right corner point. The ordinate of the bottom right corner point.

3. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 2, characterized in that: In step S3, the RGB image is input into the YOLOv11 model to identify pedestrians, non-motorized vehicles, and motorized vehicles in the RGB image, and the output is the set of detected targets. , ,in For the first One detection target, , To detect the number of targets, , For the first Individual detection targets The x-coordinate of the top left corner of the rectangular detection box. For the first Individual detection targets The ordinate of the top left corner of the rectangular detection box. For the first Individual detection targets The width of the rectangular detection box, For the first Individual detection targets The height of the rectangular detection frame.

4. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 3, characterized in that, Step S4 includes the following steps: S4-1. According to the... Individual detection targets The x-coordinate of the top left corner of the rectangular detection box y-axis Width of the rectangular detection frame ,high Calculate the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame; S4-2. If the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame are located at the th Each lane line area Inside, then the first Individual detection targets Add to the Detection target set for each lane line area middle; S4-3. If the first Individual detection targets The x and y coordinates of the center point of the rectangular detection frame are located in the pedestrian and non-motorized vehicle passage area. Inside, then the first Individual detection targets Added to the detection target set of pedestrian and non-motorized vehicle traffic areas middle.

5. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 2, characterized in that, Step S5 includes the following steps: S5-1. The measurement of the first [item] in a real-world intersection scenario is obtained. Each lane line area physical length and physical width , No. Each lane line area The physical coordinates of the top left corner are Its lower left corner physical coordinates are Its upper right corner physical coordinates are Its physical coordinates in the lower right corner are ; S5-2. Measure the pedestrian and non-motorized vehicle passage areas in a real-world intersection scenario. physical length and physical width pedestrian and non-motorized vehicle traffic areas The physical coordinates of the top left corner are Its lower left corner physical coordinates are Its upper right corner physical coordinates are Its physical coordinates in the lower right corner are ; S5-3. Constructing the first Each lane line area The coordinates of the top left corner The coordinates of the lower left corner The coordinates of the upper right corner The coordinates of the bottom right corner Using coordinate points Coordinate point pair Coordinate point pair Coordinate point pair The solution for the 3rd row and 3rd column is obtained by the direct linear transformation algorithm. Perspective transformation matrix for each lane area , ,in Perspective transformation matrix The Middle Line number Column parameters, , ; S5-4. Construct pedestrian and non-motorized vehicle access areas The coordinates of the top left corner The coordinates of the lower left corner The coordinates of the upper right corner The coordinates of the bottom right corner Using coordinate points Coordinate point pair Coordinate point pair Coordinate point pair The perspective transformation matrix of the 3x3 pedestrian and non-motorized vehicle passage area was obtained by solving the direct linear transformation algorithm. , ,in Perspective transformation matrix The Middle Line number Column parameters, , .

6. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 5, characterized in that, Step S6 includes the following steps: S6-1. For the first Detection target set for each lane line area The first in Individual detection targets Based on the x-coordinate of the top left corner of its rectangular detection box The ordinate of the top left corner ,width ,high The x-coordinate of the bottom center point of the rectangular detection box was calculated. and the center point pixel ordinate Through formula Calculated Time of the first Individual detection targets x-coordinate of world coordinates Through formula Calculated Time of the first Individual detection targets The ordinate of the world coordinates ; S6-2. Detection target set for pedestrian and non-motorized vehicle traffic areas The first in Individual detection targets Based on the x-coordinate of the top left corner of its rectangular detection box The ordinate of the top left corner ,width ,high The x-coordinate of the bottom center point of the rectangular detection box was calculated. and the center point pixel ordinate Through formula Calculated Time of the first Individual detection targets x-coordinate of world coordinates Through formula Calculated Time of the first Individual detection targets The ordinate of the world coordinates .

7. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 6, characterized in that, Step S7 includes the following steps: S7-1. Through formula Calculations were made at the time interval The first Individual detection targets physical displacement In the formula, the time interval is... The reciprocal of the video frame rate per second. for Time of the first Individual detection targets The x-coordinate of the world coordinates, for Time of the first Individual detection targets The ordinate of the world coordinates.

8. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 7, characterized in that, Step S8 includes the following steps: S8-1. Through formula Calculate the first Individual detection targets exist instantaneous velocity at a moment ; S8-2. Through formula Calculated Filtered velocity value at time 1 In the formula, As a smoothing factor, , for The velocity value after filtering at any given time. .

9. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 1, characterized in that, Step S9 includes the following steps: S9-1. The first Each lane line area The intersection of the central axis and the zebra crossing is used as the point of conflict. ; S9-2. For the first Detection target set for each lane line area The first in Individual detection targets Based on the filtered velocity value Calculate the time it takes for the motor vehicle to arrive at the conflict point ; S9-3. Detection target set for pedestrian and non-motorized vehicle traffic areas The first in Individual detection targets Based on the filtered velocity value Calculate the time it takes for pedestrians or non-motorized vehicles to arrive at the conflict point. ; S9-4. Pass The TTC value of the point of conflict is calculated.

10. The method for early warning of pedestrian-vehicle conflicts in the transportation field according to claim 1, characterized in that: In step S10, when the TTC value is less than or equal to 3 seconds, it is determined to be high risk and a high-level warning is triggered; when the TTC value is greater than 3 seconds and less than or equal to 5 seconds, it is determined to be medium risk and a low-level warning is triggered. When the TTC value is greater than 5 seconds, it is judged as low risk and no warning is triggered.

Images