Incoming vehicle studying and judging and safety early warning method in train inspection on-track operation process

Abstract

The invention discloses an incoming vehicle research and judgment and safety early warning method in a train inspection on-track operation process, and the method comprises the steps: collecting the interface information of an SAM or CIPS software system in real time through a signal recognition module, and recognizing the states of a station track and a signal lamp through a machine vision algorithm; the coming vehicle studying and judging module judges whether a coming vehicle exists or not according to the divided states of station tracks and adjacent signal lamps in the personnel operation area, the operation adjacent area and the safety early warning area; the safety early warning module is combined with an operation plan, personnel bracelet positioning information and a coming vehicle signal to accurately screen out maintenance personnel with safety risks and push an early warning message to a bracelet to remind the operation personnel to avoid dangers; according to the method, a machine vision algorithm, rule research and judgment, the Internet of Things technology and train inspection on-track operation coming vehicle safety reminding service are fused, early warning is timely, the accuracy rate is high, and a technical guarantee is provided for safety of train inspection on-track operation personnel.

Description

Technical Field

[0001] The present invention relates to the field of safety early warning for track inspection operations on the track. Specifically, through visual algorithms, oncoming train rule judgment, and Internet of Things technology, it can real-time monitor the track lines with oncoming trains, quickly locate the operating personnel on the oncoming track lines, and accurately push the oncoming train information to the operating personnel to notify them to leave in advance, so as to provide safety protection for track inspection operating personnel on the track. Background Art

[0002] When a train stops at a track inspection operation yard, track inspection operating personnel need to use technical means to inspect the core components of the train to ensure the safe operation of the train. During the operation process of operating personnel on the track, there may be situations where vehicles pass by on adjacent tracks or hump shunting occurs. Currently, operating personnel lack effective means to obtain the above information in advance. Therefore, there are very large safety hazards during the process of crossing and staying on the track. How to avoid such risks through effective technical means is an urgent problem to be solved.

[0003] After investigation, it is found that signals such as train approach, departure, and hump shunting can be obtained in the signal system SAM or CIPS. However, for safety reasons, the data in most track inspection yard SAM and CIPS systems does not support direct acquisition through software interfaces, which restricts the development of operation safety assistance systems. In recent years, great progress has been made in the development of artificial intelligence and Internet of Things technology. By using artificial intelligence image recognition technology to automatically identify the colors of signal lights and track lines on the SAM or CIPS interface to indirectly obtain oncoming train signals, combined with rule judgment and Internet of Things positioning and communication technology for message judgment and early warning push, it is expected to become a new solution. Summary of the Invention

[0004] In view of the above problems, the present invention proposes a method for oncoming train judgment and safety early warning during the track inspection operation on the track, which combines image recognition technology, rule judgment, and Internet of Things technology to solve the problems of difficult acquisition of oncoming train signals and untimely message early warning notifications. This method has the characteristics of high data security, clear judgment rules, and timely safety early warning. The method can be described as follows:

[0005] (1) The signal recognition module continuously collects the interface information of the SAM or CIPS system, and uses computer vision algorithms to automatically identify the status information of the tracks and signal lights included in the system interface. First step, connect the interface information of the SAM or CIPS system to a split-screen transcoding device to output a network streaming media protocol RTMP video stream. Second step, pull the RTMP video stream from the signal acquisition module in real time to obtain the video frame I. Third step, in the initialization stage, mark the track lines to be recognized on the video frame I with line segments and record the start and end point coordinates of the line segments respectively, and mark the signal lights R to be recognized with a rectangular frame light ={(xi , y i , w i , h i ) | i = 1, 2, 3…n}, Step 4, Continuously collect pictures of the track line L and signal light R in the video frame I during the initialization phase light in the area, and generate the color table C of the track line through the Kmeans algorithm L and the color table C of the signal light area R , C = {(ci, vi) | i = 1, 2, 3…n}, c represents the color category, and v = (r, g, b) represents the specific color value. Step 5, In the analysis phase, extract all the pixel points P = {pi | i = 1, 2,...M} on the track line L in the video frame I, and calculate the Euclidean distance between the color value of the pixel point pi and the color values in the color table C i respectively. The category with the closest distance is the color of the pixel point p L . Obtain the color of the entire track line through the voting method, and extract all the pixel points P = {pi | i = 1, 2,...M} included in the signal light area in the video frame I, and calculate the Euclidean distance between the color value of the pixel point pi and the color values in the color table C i respectively. The category with the closest distance is the color category of p . Obtain the color of the signal light through the voting method; R i i (2) The oncoming train judgment module judges whether there is an oncoming train according to the color information of the track line in the defined personnel operation area R

[0006] , the adjacent area R work , and the safety warning area R close , as well as the color of the signal light R between the areas safe . The following judgment rules are included: Rule 1, Train arrival warning rule, R light ←R work ←R close ←R safe . The ← direction is the train forward direction. If the track line colors in all three areas are white, it means that a train is about to arrive. When the R safe track line color turns red, trigger the train approach alarm. Rule 2, Train departure warning rule, R close →R light →R work . The → direction is the train forward direction. When the R close track line turns white and the R light is green and the R work is red, trigger the train departure alarm. Rule 3, Hump humping warning rule, R work ←R close . The ← direction is the humping direction. When the R close track line turns red, trigger the hump humping alarm;

[0007] (3) The safety warning module combines the operation plan, the positioning information of the personnel bracelets, and the oncoming train signal to accurately screen out the maintenance personnel at risk of safety and push the warning message to the bracelets to remind the operators to avoid danger. First, it judges whether there are operators entering the site according to the time of the operation plan tasks. When there are personnel working, the safety warning function is activated. Then, it judges the position information of the operators in the site according to the positioning information of the personnel bracelets. Further, according to the oncoming train signal in the specific section output by the oncoming train judgment module and the obtained personnel position information, it screens out the personnel who need to receive the warning. Finally, it pushes the warning signal to the bracelets of the operators, and the bracelet module triggers the message and vibration alarm;

[0008] Beneficial effects

[0009] The present invention discloses a method for oncoming train judgment and safety warning during the process of train inspection on the track. By integrating image recognition, rule judgment, and Internet of Things technology, this method solves the problems of difficult acquisition of oncoming train signal data and untimely pushing of safety warning messages. This method has the characteristics of high data security and timely pushing of warning messages, providing an effective technical means for the safety of train inspection personnel on the track. Brief description of the drawings

[0010] Figure 1 It is the system composition diagram of the method for oncoming train judgment and safety warning during the process of train inspection on the track of the present invention;

[0011] Figure 2 It is the interface data of the specific example SAM system of the present invention;

[0012] Figure 3 It is an example of the track and signal lamp markings of the specific example of the present invention;

[0013] Figure 4 It is an example of the drawing of the train approaching area of the specific example of the present invention;

[0014] Figure 5 It is an example of the train approaching triggering an alarm of the specific example of the present invention;

[0015] Specific implementation process

[0016] The following combines the drawings and specific examples to illustrate the implementation effect of the method for oncoming train judgment and safety warning during the process of train inspection on the track proposed by the present invention through the specific operation process. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. After reading the present invention, various equivalent forms of modification of the present invention fall within the scope defined by the appended claims of this application.

[0017] Combined with Figure 1, the present invention is based on a system for judging oncoming trains and safety warning during the process of train inspection on the track. The system includes a signal recognition module, an oncoming train judgment module, a safety warning module, and personnel bracelets. Specifically:

[0018] (1) The signal recognition module continuously collects the interface information of the SAM or CIPS system, and automatically identifies the status information of the track and signal lights included in the system interface by using machine vision algorithms;

[0019] (2) The oncoming train judgment module determines whether there is an oncoming train based on the color information of the track lines in the designated personnel operation area R work , adjacent area R close , and within the safety warning area R safe , as well as the color of the signal lights R light between the areas;

[0020] (3) The safety warning module accurately screens out maintenance personnel at risk of safety in combination with the operation plan, personnel bracelet positioning information, and oncoming train signals, and pushes warning messages to the bracelets to remind the operating personnel to avoid danger.

[0021] Select the real-scene data of a certain station to verify the effectiveness of the method for judging oncoming trains and safety warning during the process of train inspection on the track proposed by the present invention in the warning of the train arrival scenario.

[0022] (1) The signal recognition module continuously collects the RTMP video stream of the SAM system interface to extract video frames, Figure 2 showing some video stream pictures concerned by the business;

[0023] (2) In the initialization stage, the track lines to be recognized are marked with line segments on the video frames, and the signal lights to be recognized are marked with rectangular frames. The marking method is as Figure 3 shown;

[0024] (3) In the initialization stage, pictures of the track lines and signal light areas in the video frames are continuously collected, and color tables for the track lines and signal light areas are generated through the Kmeans algorithm. Table 1 shows the color table for the track lines. The format of the color table for the signal lights is the same as that for the track lines;

[0025] Table 1: Color table generated by clustering the colors of the track lines

[0026] Red 215,0,8 194,24,24 195,5,15 200,17,19 188,11,17 ...... White 219,221,208 219,220,212 230,230,228 242,245,234 224,228,214 ...... Blue 96,114,154 54,71,97 89,112,154 90,105,138 93,105,131 ......

[0027] (4) In the analysis stage, all pixel points on the track lines in the video frames are extracted, and the Euclidean distances between the pixel values of the pixel points and the color values in the color table are calculated respectively. The category with the closest distance is the color category of the pixel point. The color of the entire track line is obtained through the voting method. The color recognition method for the signal lights is the same as that for the track lines;

[0028] (5) Determine whether there is an oncoming train signal based on the color information of the middle track lines within the designated personnel operation area, adjacent area, and safety warning area. The designated areas are as shown in Figure 4 . The area marked by the yellow rectangular frame is the safety warning area, the area marked by the red rectangular frame is the adjacent area, and the area marked by the blue rectangular frame is the personnel operation area. An alarm is triggered when the track line color in the yellow area turns red, as shown in Figure 5 ;

[0029] (6) Analyze from the operation plan and personnel positioning bracelet information that Zhang San is currently working in the operation area, and send the judged warning signal to the bracelet to remind Zhang San to leave the operation area;

[0030] As can be seen from the above analysis, the method for judging oncoming trains and safety warning during the process of train inspection on the track proposed by the present invention, by integrating image recognition, rule judgment, and Internet of Things technology, solves the problems of difficult acquisition of oncoming train signal data and untimely push of safety warning messages. This method has the characteristics of high data security and timely push of warning messages, providing an effective technical means for the safety of train inspection personnel on the track.

Claims

1. A method for judging incoming vehicles and providing safety warning during train inspection on-road operation, characterized in that The method is based on a vehicle identification and safety warning system during the on-road inspection process. The system includes a signal recognition module, a vehicle identification module, a safety warning module and a personnel wristband. The specific warning steps are: (1) The signal recognition module collects SAM or CIPS system interface information in real time and uses machine vision algorithms to automatically identify the status information of tracks and signal lights contained in the system interface; (2) The vehicle identification module determines the personnel operation area R according to the designated personnel operation area R. work , adjacent area R close , R in the safety warning area safe Color information of the middle track line and the signal lights between areas R light Color, to determine whether there is an oncoming car; (3) The safety warning module combines the work plan, the positioning information of the personnel's wristband and the signal of the oncoming vehicle to accurately screen out the maintenance personnel who pose a safety risk, and pushes the warning message to the wristband to remind the operating personnel to avoid danger.

2. The method according to claim 1, characterized in that Step (1) specifically includes: (1-1) Connect the SAM or CIPS system interface information to the split-screen transcoding device to output the network streaming protocol RTMP video stream; (1-2) The signal acquisition module pulls the RTMP video stream in real time to obtain video frame I; (1-3) Initialization stage: Use line segments to mark the track lines to be identified on video frame I Record them as the starting and ending point coordinates of the line segment, and use a rectangular box to mark the signal light R that needs to be identified light ={(x i ,y i ,w i ,h i )|i=1,2,3...n}, where x i ,y i Indicates the coordinates of the upper left corner of the rectangle, w i ,h i Represents the width and height of the rectangle, and n represents the total number of marked rectangles; (1-4) Initialization phase: continuously capture the track line L and signal light R in video frame I light The image of the area, using the Kmeans algorithm to generate the color table C of the stock line L and the color table C for the signal light area R , C={(c i ,v i )|i=1,2,3...n}, c represents the color category, v=(r,g,b) represents the specific color value; (1-5) Extracting the track line L in the video frame I in the analysis stage i All pixels on P = {p i |i=1,2,...M}, calculate the pixel point p respectively i Color values ​​and color tables C L The Euclidean distance of the color value in the pixel is the closest category. i The color of the entire track line is obtained by voting; (1-6) Extract the signal light in video frame I in the analysis stage All pixels contained in the area P = {p i |i=1,2,...M}, calculate the pixel point p respectively i Color values ​​and color tables C R The Euclidean distance of the color value in the middle, the category closest to it is p i The color category of the traffic light is obtained by voting.

3. The method according to claim 1, characterized in that The judgment rules of step (2) specifically include: (2-1) Train arrival warning rules: for the "R work ←R close ←R safe " Display area, the track lines in the three areas are all white, which means that a train is about to arrive. safe The track line turns red, triggering a train approach alarm; the ← direction is the train's forward direction; (2-2) Train departure warning rules: for the "R close →R light →R work "Display area, R close The stock line turns into a white R light Green R work Red, triggering the train departure alarm; → direction is the train forward direction; (2-3) Humpback release warning rules: for the "R work ←R close "Display area, R close The track line turns red, triggering the hump release alarm; the ← direction is the release direction.

4. The method according to claim 1, characterized in that The specific process of step (3) is as follows: (3-1) Determine whether there are any workers entering the site based on the time of the work plan task, and activate the safety warning function when there are workers; (3-2) Determine the location of the operator on site based on the positioning information of the operator’s wristband; (3-3) According to the specific section vehicle signal output by the vehicle identification module and the personnel location information obtained in (3-2), select the personnel who need to receive the warning; (3-4) The warning signal is pushed to the operator’s wristband, and the wristband module triggers a message and vibration alarm.

Images