Rail transit platform intrusion detection and intelligent reminding method and system

Abstract

The invention discloses a rail transit platform intrusion detection and intelligent reminding method and system, and the method comprises the following steps: carrying out the overhead scanning of a safety yellow line based on a single-line laser radar; processing the obtained laser radar data to obtain a center coordinate of the intrusion target; performing coordinate conversion to obtain Cartesian coordinates of the intrusion target; unifying a coordinate system, and carrying out the linkage of a holder camera and a follow spotlight, and obtaining an intrusion target video stream; an intrusion target and behaviors thereof are described, and acousto-optic reminding and warning are carried out; the system comprises a safety yellow line intrusion detection device located in a platform rail running area, and the safety yellow line intrusion detection device comprises a single line laser radar, a holder camera, a voice module, a follow spotlight, an edge calculation box and a router. According to the invention, real-time detection of line crossing personnel and articles of the safe yellow line of the platform rail-mounted area is realized. According to the invention, the problem of low reflectivity of a laser radar to a black or transparent object is particularly solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of rail operation safety detection, and specifically relates to a rail transit platform intrusion detection and intelligent reminder method and system. Background Art

[0002] As an efficient and fast urban public transportation mode, rail transit has been widely promoted and developed rapidly around the world. The scale of China's rail transit has also grown rapidly. By the end of 2023, the national railway operating mileage has reached 159,000 kilometers, and the main channels of the "eight horizontal and eight vertical" high-speed railway network have been built for about 80%. The conventional railway network is constantly improving, and the operation demand is gradually increasing. However, rail transit faces huge challenges in safe operation. The safety of the platform waiting area is one of the key issues. The current platform safety protection mainly includes isolation doors, platform edge signs and other facilities, but many high-speed trains currently do not have platform isolation doors in the platform waiting area, and some rail transit platform doors that are difficult to reconstruct are low in height. In objective reality, passengers and their belongings may cross the safety yellow line and invade the platform track area. In fact, a large number of high-speed rail platforms and subway platforms across the country have invested a large number of special security guards on duty to carry out intrusion safety risk prevention and control, but the intrusion of passengers and items is still common. These situations not only affect the driving order, but also pose safety hazards, leading to frequent safety accidents.

[0003] The current intrusion detection mode of rail transit platforms is not only time-consuming, labor-intensive and inefficient, but also sometimes leads to conflicts between security officers and passengers, and cannot meet the safety requirements of modern rail transit systems. Summary of the invention

[0004] The present invention is proposed to solve the problems existing in the prior art, and its purpose is to provide a rail transit platform intrusion detection and intelligent reminder method and system.

[0005] The technical solution of the present invention is: a rail transit platform intrusion detection and intelligent reminder method, comprising the following steps:

[0006] A. Based on a single-line laser radar, a bird’s-eye view scan of the yellow safety line is performed;

[0007] B. Process the obtained laser radar data to obtain the center coordinates of the intrusion target;

[0008] C. Perform coordinate conversion to obtain the Cartesian coordinates of the intrusion target;

[0009] D. Unify the coordinate system, link the PTZ camera and the spotlight, and obtain the intrusion target video stream;

[0010] E. Describe the intrusion target and its behavior, and provide sound and light reminders and warnings.

[0011] Furthermore, step A also includes supplementary intrusion detection using a pan-tilt camera, and the specific process is as follows:

[0012] First, when a single-line LiDAR scans black or transparent objects with low reflectivity, it is difficult to reflect back LiDAR data;

[0013] Then, the entire detection area is divided into several areas according to the field of view of the PTZ camera;

[0014] Then, when black objects or transparent objects with low reflectivity invade, the boundary of the benchmark model will be blocked and the radar data will be lost. The intrusion area can be determined by comparing with the benchmark model.

[0015] Finally, the PTZ camera is used to collect the video stream of the intrusion area.

[0016] Furthermore, the video stream of the change area of ​​the above-mentioned baseline model is transmitted back, and the YOLO intrusion detection method is used to supplement the detection of intrusion behavior to ensure that the intrusion target is not missed.

[0017] Furthermore, step A performs a top-down scan of the safety yellow line based on a single-line laser radar, and the specific process is as follows:

[0018] First, use a single-line laser radar to scan the yellow safety line environment of the platform track area from above;

[0019] Then, a benchmark model is constructed;

[0020] Finally, the collected lidar data is obtained.

[0021] Furthermore, step B processes the obtained laser radar data to obtain the center coordinates of the intrusion target;

[0022] First, the benchmark model is used as a comparison example, the lidar data outside the benchmark model is filtered, and the lidar data within the benchmark model is retained;

[0023] Then, the points in the benchmark model are filtered and density clustered to filter out noise points and reduce the redundancy of intrusion points;

[0024] Finally, the center coordinates of the intrusion target are obtained through the laser radar data processed as above.

[0025] Furthermore, step C performs coordinate transformation to obtain the Cartesian coordinates of the intrusion target. The specific process is as follows:

[0026] The polar coordinate points detected by the single-line laser radar are transformed to obtain the Cartesian coordinates of the intrusion target.

[0027] Furthermore, step D unifies the coordinate system, links the PTZ camera and the spotlight, and obtains the intrusion target video stream. The specific process is as follows:

[0028] First, unify the coordinate systems of single-line laser radar, gimbal camera, and spotlight;

[0029] Then, the coordinates of the intrusion target are converted into points in the coordinate system of the PTZ camera and the spotlight ball;

[0030] Finally, control the pan and tilt of the PTZ camera and the follow-spot light so that the PTZ platform of the PTZ camera and the follow-spot light is aimed at the intrusion target.

[0031] Furthermore, step E describes the intrusion target and its behavior, and performs sound and light reminders and warnings. The specific process is as follows:

[0032] First, after collecting the video stream of the intrusion target, the deep learning model is used in the edge computing box to identify, classify and understand the semantics of the intrusion target;

[0033] Then, use the big model to accurately describe the intrusion target and behavior, and give polite prompts and warnings;

[0034] Finally, the voice module is used to broadcast information and the light beam of the chase light is used to drive away the invading target.

[0035] A system for rail transit platform intrusion detection and intelligent reminder method, comprising a safety yellow line intrusion detection device located in a platform track area, wherein the safety yellow line intrusion detection device comprises the following units:

[0036] Single-line laser radar for cross-line intrusion detection;

[0037] PTZ camera, used to capture images of intrusion targets;

[0038] Voice module, used for playing and outputting pre-stored voice;

[0039] Follow-up light, used for tracking and locating intrusion targets;

[0040] An edge computing box connected to the above-mentioned sound, light and image unit;

[0041] Routers provide network communications.

[0042] The beneficial effects of the present invention are as follows:

[0043] The present invention realizes real-time detection of people and objects crossing the yellow safety line in the platform track area. The system includes a single-line laser radar, a pan-tilt camera, a spotlight, a voice module, an edge computing box and a router, which are composed of a track platform track area yellow safety line intrusion detection device and a prompt alarm device.

[0044] The present invention introduces multi-sensor fusion and edge computing boxes, combined with deep learning models, to achieve high-precision and high-efficiency intrusion detection, and especially solves the problem of low reflectivity of laser radar to black or transparent objects.

[0045] The present invention realizes all-weather real-time monitoring of the perimeter of rail transit platforms, effectively improving the accuracy and efficiency of intrusion detection. Through automated alarm and data processing, the system greatly reduces the cost of manual monitoring, while improving platform safety and operational management efficiency, and ensuring driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a layout diagram of the system in the present invention;

[0047] Figure 2 is a flow chart of the method of the present invention;

[0048] in:

[0049] 1Safety yellow line intrusion detection device 2Single line laser radar

[0050] 3 PTZ cameras 4 spotlights

[0051] 5 Voice Module 6 Edge Computing Box

[0052] 7 routers. DETAILED DESCRIPTION

[0053] Hereinafter, the present invention will be described in detail with reference to the accompanying drawings and embodiments:

[0054] like Figure 1 to Figure 2 As shown, a rail transit platform intrusion detection and intelligent reminder method includes the following steps:

[0055] A. Based on the single-line laser radar 2, the yellow safety line is scanned from above;

[0056] B. Process the obtained laser radar data to obtain the center coordinates of the intrusion target;

[0057] C. Perform coordinate conversion to obtain the Cartesian coordinates of the intrusion target;

[0058] D. Unify the coordinate system, link the pan / tilt camera 3 and the spotlight 4, and obtain the intrusion target video stream;

[0059] E. Describe the intrusion target and its behavior, and provide sound and light reminders and warnings.

[0060] Step A also includes supplementary intrusion detection using the pan-tilt camera 3, and the specific process is as follows:

[0061] First, when the single-line LiDAR 2 scans black objects or transparent objects with low reflectivity, it is difficult to reflect back LiDAR data;

[0062] Then, the entire detection area is divided into several areas according to the field of view of the PTZ camera;

[0063] Then, when black objects or transparent objects with low reflectivity invade, the boundary of the benchmark model will be blocked and the radar data will be lost. The intrusion area can be determined by comparing with the benchmark model.

[0064] Finally, the PTZ camera is used to collect the video stream of the intrusion area.

[0065] The video stream transmitted back to the change area of ​​the above-mentioned baseline model uses the YOLO intrusion detection method to supplement the detection of intrusion behavior to ensure that the intrusion target is not missed.

[0066] Step A is based on the single-line laser radar 2 to scan the yellow safety line from above. The specific process is as follows:

[0067] First, the single-line laser radar 2 is used to scan the yellow safety line environment of the platform track area from above;

[0068] Then, a benchmark model is constructed;

[0069] Finally, the collected lidar data is obtained.

[0070] Step B processes the obtained laser radar data to obtain the center coordinates of the intrusion target;

[0071] First, the benchmark model is used as a comparison example, the lidar data outside the benchmark model is filtered, and the lidar data within the benchmark model is retained;

[0072] Then, the points in the benchmark model are filtered and density clustered to filter out noise points and reduce the redundancy of intrusion points;

[0073] Finally, the center coordinates of the intrusion target are obtained through the laser radar data processed as above.

[0074] Step C performs coordinate conversion to obtain the Cartesian coordinates of the intrusion target. The specific process is as follows:

[0075] The polar coordinate points detected by the single-line laser radar 2 are transformed to obtain the Cartesian coordinates of the intrusion target, which are expressed as follows:

[0076] x=cosθ·l

[0077] z=sinθ·l.

[0078] Step D unifies the coordinate system, links the pan / tilt camera 3 and the spotlight 4, and obtains the intrusion target video stream. The specific process is as follows:

[0079] First, unify the coordinate systems of the single-line laser radar 2, the gimbal camera 3, and the spotlight 4;

[0080] Then, the coordinates of the intrusion target are converted into points in the spherical coordinate system of the pan / tilt camera 3 and the spotlight 4;

[0081] Finally, the pan and tilt of the PTZ camera 3 and the spotlight 4 are controlled so that the PTZ platform of the PTZ camera 3 and the spotlight 4 is aimed at the intrusion target. The expression is as follows:

[0082]

[0083]

[0084] Step E describes the intrusion target and its behavior, and provides sound and light reminders and warnings. The specific process is as follows:

[0085] First, after collecting the video stream of the intrusion target, the deep learning model is used in the edge computing box 6 to identify, classify and semantically understand the intrusion target;

[0086] Then, use the big model to accurately describe the intrusion target and behavior, and give polite prompts and warnings;

[0087] Finally, the voice module 5 broadcasts the message and cooperates with the light beam of the spotlight 4 to drive away the intruding target.

[0088] Specifically, the processed intrusion pictures are uploaded to the central server and the train cab for the driver to make a judgment on whether the train can be driven.

[0089] A system for rail transit platform intrusion detection and intelligent reminder method, comprising a safety yellow line intrusion detection device 1 located in the platform track area, the safety yellow line intrusion detection device 1 comprising the following units:

[0090] Single-line laser radar 2, used for cross-line intrusion detection;

[0091] PTZ camera 3, used to capture images of intrusion targets;

[0092] Voice module 5, used for playing and outputting pre-stored voice;

[0093] Follow-up light 4, used for follow-up and positioning of intrusion targets;

[0094] Edge computing box 6, connected to the above-mentioned sound, light and image unit;

[0095] Router 7 provides network communication.

[0096] Specifically, the single-line laser radar 2 is installed just above the yellow safety line of the platform, so that the laser of the single-line laser radar 2 scans the yellow safety line to perform cross-line intrusion detection.

[0097] Specifically, the pan-tilt camera 3, voice module 5 and follow-up spot 4 are installed on the horizontal plane of the single-line laser radar 2 and away from the safety yellow line, so that the pan-tilt camera 3 can clearly shoot the side of the intrusion target, provide a basis for accurately identifying the intrusion target, and provide prompts and alarms through the light beam of the follow-up spot 4. The voice module 5 accurately and politely describes the intruder, and combined with the follow-up spot 4 to keep the intruder away from the track area, also provides a basis for the staff to quickly locate the intrusion target.

[0098] Specifically, the edge computing box 6 is used for some data processing and analysis tasks, reducing the delay of data transmission and processing, improving real-time performance and efficiency, combining sensors, and using deep learning models to identify and classify intrusion targets. When the intrusion target moves, the target tracking algorithm is used to track the target position in real time, and the direction of the pan-tilt camera 3 and the follow-up spot 4 is dynamically adjusted, and the voice module 5 is controlled to play the behavior and courtesy prompts described by the large model.

[0099] Specifically, the router 7 provides a local area network and realizes information transfer, so that each module works in coordination. Through this system and method, people and foreign objects in the waiting area can be detected and monitored with high precision, in real time, and around the clock, and target identification and sound and light warning of intruders can be achieved, thereby improving the safety and management efficiency of rail transit platforms.

[0100] Embodiment 1

[0101] The single-line laser radar 2 is installed just above the yellow safety line of the platform track area. The scanning surface of the single-line laser radar 2 is orthogonal to the ground, and the intersection line is the yellow safety line of the track area. Through a reasonable installation position and angle and the configuration of the single-line laser radar 2, the entire platform track area can be covered to achieve comprehensive and accurate intrusion detection.

[0102] The single-line laser radar 2 emits laser and receives echo data, using the high precision and light-independent characteristics of the laser to obtain accurate distance and angle information of the platform perimeter. The minimum object size that can be detected is calculated based on the angular resolution as follows:

[0103]

[0104] Among them, x is the detection accuracy, that is, the minimum detection object size, p is the measurement range, and α is the angular resolution.

[0105] The edge computing box 6 is connected to the single-line laser radar 2 and has data processing and analysis functions. The edge computing box 6 receives the data collected by the single-line laser radar 2 scanning the yellow safety line environment of the platform track area, builds a benchmark model, and connects the points obtained by the laser radar scanning and filtering in the direction of the polar angle growth of the polar coordinates and connects them end to end to obtain the benchmark coordinates of the environment. For easy viewing and calculation, the points in the benchmark model are converted to points in the Cartesian coordinate system.

[0106] When detecting intrusion on the platform safety yellow line, the radar data scanned by the single-line laser radar 2 is preprocessed, including noise filtering and threshold filtering. Then, the preprocessed data is converted from polar coordinates to Cartesian coordinates and compared with the benchmark model. Through comparison, points outside the benchmark model are filtered out to reduce the false alarm rate.

[0107] In the Cartesian coordinate system of the two-dimensional plane scanned by the laser radar, vertical downward is set as the positive direction of the z-axis, the laser radar scanning range is set to 0° to 180°, the direction when the polar angle is 0° is set to the positive direction of the x-axis, θ is the polar angle of the radar data point, l is the polar diameter, and the coordinate transformation relationship is as follows.

[0108] x=cosθ·l (2)

[0109] x=sinθ·l (3)

[0110] Use the clustering algorithm (DBSCAN) to perform cluster analysis on the remaining points and identify the cluster center of the possible intrusion target. Based on the clustering results, return whether there is an intrusion behavior and the center position of the intrusion target. This can reduce the redundant radar data when large targets invade and the noise data generated by small objects.

[0111] When the coordinates of the intrusion target are returned, the coordinates of the intrusion target are converted into points in the spherical coordinate system of the pan-tilt camera 3 and the follow-up light 4 by unifying the coordinate systems of the single-line laser radar 2, the pan-tilt camera 3 and the follow-up light 4. The pan and pitch of the pan-tilt camera 3 and the follow-up light 4 are controlled so that the ptz platform of the pan-tilt camera 3 and the follow-up light 4 is aimed at the intrusion target.

[0112] In the three-dimensional coordinate system of the gimbal camera 3 and the spotlight 4, the positive direction of the z-axis is set vertically downward, the positive direction of the x-axis is parallel to the direction when the laser radar polar angle is 0°, and the y-axis is perpendicular to the xOz plane and takes the side where the laser radar is located as the positive direction.

[0113] In this coordinate system, the coordinates of the intrusion target are converted into points in the spherical coordinate system of the PTZ camera 3 and the follow-up light 4. The pan and pitch of the PTZ camera 3 and the follow-up light 4 are controlled so that the PTZ platform of the PTZ camera 3 and the follow-up light 4 is aimed at the intrusion target. Pan is the pan angle of the PTZ camera 3 and the follow-up light 4, and tilt is the pitch angle of the PTZ camera 3 and the follow-up light 4.

[0114]

[0115]

[0116] As a continuous action, in order to make the PTZ camera 3 and the follow-up light 4 continuously pan and tilt, when there is a new intrusion target, the PTZ camera 3 and the follow-up light 4 continue to move based on the current angle, and only need to move the changed angle. pan and tilt are the pan and tilt angles of the intrusion point coordinate conversion. cur 、tilt cur is the current position of roll and pitch, △pan and △tilt are the angles to be adjusted.

[0117] △pan=pan―pan cur (6)

[0118] △tilt=tilt―tilt cur (7).

[0119] After the pan-tilt camera 3 and the follow-spot light 4 rotate to the designated position, the follow-spot light 4 will illuminate the intruding target to give a prompt, and the pan-tilt camera 3 transmits the video stream to the edge computing box for target classification and identification, and performs target tracking.

[0120] After receiving the video stream transmitted by the PTZ camera 3, the edge computing box 6 uses YOLOv8 to identify, classify and track the intrusion target, uses the large model to describe the intrusion behavior, reminds the intrusion target by dynamically adjusting the direction of the camera and the spotlight, and uses the voice module 5 to play the behavior and courtesy prompts described by the large model to drive away the intrusion target.

[0121] The image frames after the intrusion target is identified are transmitted to the train cab through the edge computing box for the driver to make a judgment on whether the train can be driven. When an intrusion is detected, the image after the target is identified and the coordinates of the intrusion are uploaded to the central server, and the staff is notified via SMS to suspend driving and check the intrusion image in time to ensure driving safety.

[0122] In view of the inherent problems of low reflectivity of black objects and no reflective points of transparent objects when collecting data by single-line LiDAR 2, YOLOv8 is used as a supplement for intrusion detection algorithm. When the intruder is a black object or a transparent object with low reflectivity, although the LiDAR cannot reflect the intrusion point, the points originally hit on the boundary of the benchmark model will be blocked and produce differences. By comparing this feature with the benchmark model, it is possible to infer the presence of intrusion targets. The boundary of the benchmark model is divided into several areas according to the field of view of the gimbal camera. When the LiDAR data points on the boundary of the benchmark model change significantly, the gimbal camera is controlled to rotate to this area. The video stream is used for intrusion detection algorithm and the frame is cut and transmitted to the train cab and the central server for staff to make decisions. Ensure that the intrusion target is not missed.

[0123] Based on the intrusion detection technology of the present invention, the rail transit system can provide timely prompts and warnings when people or objects invade the track area, greatly reducing the accident rate, which has positive significance for improving the overall performance of the rail transit system and ensuring the safe operation of the rail transit system.

[0124] The application features of the present invention are as follows:

[0125] First, the high-precision and high-efficiency measurement capability of intrusion detection can achieve high-precision detection of intrusion targets on the platform safety yellow line through scanning and data processing of a single-line laser radar. Under the condition of meeting the accuracy of intrusion detection on rail transit platforms, that is, the detection accuracy is 2 cm, it can cover a 40m detection range and the accuracy rate can reach 99%, truly realizing low-cost, high-efficiency and high-precision intrusion detection. And whether it is day or night, in bright or dim light, the laser radar can work stably and detect intrusion behavior in real time, ensuring the real-time nature of detection.

[0126] Second, accurate description of intrusion behavior and polite reminders and warnings. After the collected intrusion target video stream is transmitted to the edge computing box, the intrusion target is identified, classified and semantically understood using a deep learning model. The intrusion target and behavior are accurately described using a large model, and polite reminders and warnings are given, which are broadcast through the voice module to prevent intrusion behavior in time. Third, high accuracy is achieved by establishing a benchmark model in the actual environment, filtering the lidar data outside the benchmark model, accurately limiting the monitoring range, identifying the center point of the object through a clustering algorithm, and extracting the coordinate information of the abnormal object. The location and number of intrusion targets can be determined more accurately, avoiding false alarms of irrelevant targets.

[0127] Third, the inherent problem of low reflectivity of black objects and no reflective points of transparent objects when the laser radar collects data can also be solved by comparing with the benchmark model. When the intruder is a black object or a transparent object with high reflectivity, although the intrusion point cannot be reflected, the points originally hitting the boundary of the benchmark model will be blocked and produce differences. By comparing this feature with the benchmark model, it can be inferred that the intrusion target may appear. At this time, the gimbal camera is used to transmit the video stream in the half area where the benchmark model changes, and the intrusion detection method of YOLO is used to supplement the monitoring of intrusion behavior to ensure that the intrusion target is not missed. That is, when the laser radar is doing intrusion detection, the gimbal camera mainly performs target recognition. In special cases, when the laser radar can recognize the intrusion behavior but cannot locate it, the gimbal camera can also realize intrusion detection. Ensure the high accuracy of intrusion detection.

[0128] Fourth, intelligent alarm and data processing, through edge computing boxes and data processing algorithms, realize automated intrusion detection and alarm. When an intrusion is detected, the system can automatically trigger an alarm and transmit the intrusion information to the central server in a timely manner, so as to achieve rapid response and processing of intrusion events and ensure the safe operation of public transportation.

[0129] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A rail transit platform intrusion detection and intelligent reminder method, characterized in that: The following steps are involved: A. Based on the single-line laser radar (2), a bird’s-eye view scan of the yellow safety line is performed; B. Process the obtained laser radar data to obtain the center coordinates of the intrusion target; C. Perform coordinate conversion to obtain the Cartesian coordinates of the intrusion target; D. Unify the coordinate system, link the pan / tilt camera (3) and the spotlight (4) to obtain the intrusion target video stream; E. Describe the intrusion target and its behavior, and provide sound and light reminders and warnings.

2. A rail transit platform intrusion detection and intelligent reminder method according to claim 1, characterized in that: Step A also includes supplementary intrusion detection using a pan-tilt camera (3), the specific process is as follows: First, when a single-line LiDAR (2) scans black or transparent objects with low reflectivity, it is difficult to reflect back LiDAR data; Then, the entire detection area is divided into several areas according to the field of view of the PTZ camera; Then, when black objects or transparent objects with low reflectivity invade, the boundary of the benchmark model will be blocked and the radar data will be lost. The intrusion area can be determined by comparing with the benchmark model. Finally, the PTZ camera is used to collect the video stream of the intrusion area.

3. A rail transit platform intrusion detection and intelligent reminder method according to claim 2, characterized in that: The video stream transmitted back to the change area of ​​the above-mentioned baseline model uses the YOLO intrusion detection method to supplement the detection of intrusion behavior to ensure that the intrusion target is not missed.

4. A rail transit platform intrusion detection and intelligent reminder method according to claim 1, characterized in that: Step A uses a single-line laser radar (2) to scan the yellow safety line from above. The specific process is as follows: First, a single-line laser radar (2) is used to scan the yellow safety line environment of the platform track area from above; Then, a benchmark model is constructed; Finally, the collected lidar data is obtained.

5. A rail transit platform intrusion detection and intelligent reminder method according to claim 1, characterized in that: Step B processes the obtained laser radar data to obtain the center coordinates of the intrusion target; First, the benchmark model is used as a comparison example, the lidar data outside the benchmark model is filtered, and the lidar data within the benchmark model is retained; Then, the points in the benchmark model are filtered and density clustered to filter out noise points and reduce the redundancy of intrusion points; Finally, the center coordinates of the intrusion target are obtained through the laser radar data processed as above.

6. A rail transit platform intrusion detection and intelligent reminder method according to claim 1, characterized in that: Step C performs coordinate conversion to obtain the Cartesian coordinates of the intrusion target. The specific process is as follows: The polar coordinate points detected by the single-line laser radar (2) are transformed into coordinates to obtain the Cartesian coordinates of the intrusion target.

7. A rail transit platform intrusion detection and intelligent reminder method according to claim 1, characterized in that: Step D unifies the coordinate system, and performs linkage between the pan / tilt camera (3) and the spotlight (4) to obtain the intrusion target video stream. The specific process is as follows: First, unify the coordinate systems of the single-line laser radar (2), the gimbal camera (3), and the spotlight (4); Then, the coordinates of the intrusion target are converted into points in the spherical coordinate system of the pan-tilt camera (3) and the spotlight (4); Finally, the pan and tilt of the pan / tilt camera (3) and the follow-up light (4) are controlled so that the PTZ platform of the pan / tilt camera (3) and the follow-up light (4) is aimed at the intrusion target.

8. A rail transit platform intrusion detection and intelligent reminder method according to claim 1, characterized in that: Step E describes the intrusion target and its behavior, and provides sound and light reminders and warnings. The specific process is as follows: First, after collecting the video stream of the intrusion target, a deep learning model is used in the edge computing box (6) to identify, classify and semantically understand the intrusion target; Then, use the big model to accurately describe the intrusion target and behavior, and give polite prompts and warnings; Finally, the voice module (5) broadcasts the message and cooperates with the light beam of the spotlight (4) to drive away the intruding target.

9. The system of a rail transit platform intrusion detection and intelligent reminder method according to claim 1, characterized in that: It comprises a safety yellow line intrusion detection device (1) located in a platform track area, the safety yellow line intrusion detection device comprising the following units: Single-line laser radar (2) for cross-line intrusion detection; A pan / tilt camera (3), used for capturing images of intrusion targets; A voice module (5), used for playing and outputting pre-stored voice; A spotlight (4) is used to spot and locate intruders; An edge computing box (6), connected to the above-mentioned sound, light and imaging unit; Router (7), providing network communication.

Images