Train coupler height online detection system and method

Abstract

The invention discloses a train coupler height online detection system which comprises trackside equipment, a field control center and a remote control center, and the field control center is in communication connection with the trackside equipment and the remote control center. The trackside equipment comprises an incoming line sensor, a speed measuring radar, a train number recognition device, a three-dimensional train coupler detection module, a two-dimensional train coupler detection module and an offline sensor; the car coupler three-dimensional detection module is arranged on one side close to the speed measuring radar and located between the first branch rail and the second branch rail, and images of the bottoms of the branch rails and the bottom of the car coupler which are close to each other are collected from bottom to top; the coupler two-dimensional detection module is arranged on the same side as the speed measuring radar, images of the side portion of the branch rail and the side portion of the coupler are collected in a side shooting mode, and the coupler three-dimensional detection module is within the shooting range of the coupler two-dimensional detection module. The invention further discloses a detection method using the train coupler height online detection system.

Description

Technical Field

[0001] The present invention belongs to the technical field of rail vehicle detection, and relates to a train coupler height detection system and method, and in particular to a train coupler height through-type online detection system and a detection method thereof. Background Art

[0002] In recent years, rail transit has developed rapidly with its advantages of large transportation capacity, fast speed, high safety and reliability. The construction of rail transit network has become a business card of urban development. Rail transit has gradually become the first choice for citizens' travel, effectively solving the problem of public transportation congestion in major cities.

[0003] With the rapid development of rail transit, rail transit vehicles will use different vehicle formations at different times to adapt to the changes in the number of passengers in different situations. For example, the coupling and uncoupling of trains are becoming more and more widely used. In the process of train coupling and uncoupling, the coupler is an important component used to achieve vehicle coupling in the train. It transmits traction and braking force during operation. For coupling and driving safety, the height of the coupler must be specified to be within a certain range. During the operation of the vehicle, the coupler will inevitably be worn and damaged, such as the bending of the vehicle frame beam and the bending of the hook body, which will cause the height of the coupler to drop. If the height of the coupler does not meet the requirements, it is very easy to cause the vehicle to be uncoupled during operation, which seriously threatens driving safety. Therefore, the height of the coupler needs to be measured and adjusted during vehicle maintenance to ensure that its height is within the specified range to ensure the safe operation of the vehicle.

[0004] At present, most of the railway vehicle departments in my country use homemade coupler height measuring scales for manual inspection, and a few use infrared or ultrasonic sensors to measure the coupler height. Among them, most manual inspections use homemade coupler height measuring scales to measure the coupler height. The maintenance worker holds the height measuring scale, places it on the horizontal rail surface and perpendicular to the rail plane, and then reads the height value of the coupler centerline with the naked eye. This detection method requires the vehicle to stop for inspection, which is inefficient, and due to tool and visual errors, the accuracy and stability of the coupler height measurement cannot be effectively guaranteed. Among them, the method of using ultrasonic sensors to measure the coupler height is to install a retractable cantilever beam at the center line of the coupler, and fix the ultrasonic receiving and transmitting probes on the cantilever beam, extend the cantilever beam to the top of the rail surface, and the ultrasonic transmitting probe transmits the ultrasonic signal to the rail surface and then receives it. The receiving probe receives it, and the distance from the coupler centerline to the rail surface is calculated based on the time difference between the ultrasonic transmission and reception echoes. Among them, the infrared sensor measurement method is to stick a shiny sticker on the center line of the coupler, and detect the position of the sticker through an infrared sensor installed opposite the sticker that can move up and down, and calculate the height of the coupler center line according to the height of the infrared sensor when the sticker signal is detected. The measurement methods based on ultrasonic and infrared sensors require the installation of equipment on the coupler, which is easy to fall off and affect driving safety. In addition, the above detection methods all require stopping the vehicle for detection, and the detection efficiency is low.

[0005] In addition, there are also a few methods of detecting the height of the coupler by using online detection equipment. The image of the side of the coupler is collected by installing an image acquisition device beside the track, detecting the position of the coupler in the image and further detecting the position of the center line of the coupler. At the same time, the position of the rail surface in the image is detected. The height of the coupler is calculated and analyzed by identifying the position of the center line of the coupler and the position of the rail surface. This method can achieve a fast through-type detection of the coupler height; however, this detection method is only for couplers with a clear center line drawn on the surface of the coupler and the center line is clear and not discolored, and is not blocked by dirt or light spots. However, in the actual operation of rail transit vehicles, they are exposed to wind and rain, and the surface of the coupler is often greasy, dirty, peeling, and discolored. The center line is often blocked by oil spots, discolored, and paint peeling, making it difficult to detect. Some models of couplers are sprayed with a certain center line, so the use of this detection method is limited.

[0006] Therefore, how to improve the existing coupler height detection technology, or propose a new coupler height detection system and detection method to adapt to the coupler height detection with centerline obstruction or no obvious centerline has become one of the urgent issues to be solved by technical personnel in this field. Summary of the invention

[0007] In view of the shortcomings of the current coupler height detection technology, the present invention provides a train coupler height online detection system and method, which solves the problem of low reliability of the traditional coupler height detection method by improving the image acquisition method and data processing method.

[0008] The present invention adopts the following technical solutions to solve the above technical problems:

[0009] A train coupler height online detection system, the system comprising trackside equipment, an on-site control center and a remote control center, the on-site control center is respectively connected to the trackside equipment and the remote control center in communication, the trackside equipment comprises an incoming line sensor, a speed measuring radar, a vehicle number recognition device, a three-dimensional coupler detection module, a two-dimensional coupler detection module and an offline sensor;

[0010] The three-dimensional detection module for the coupler is located between the first branch rail and the second branch rail and close to the first branch rail, and collects images of the bottom of the first branch rail and the bottom of the coupler from bottom to top;

[0011] The two-dimensional coupler detection module is arranged on the outer side of the first branch rail close to the three-dimensional coupler detection module, and images of the side of the first branch rail and the side of the coupler are captured by side shooting, and the three-dimensional coupler detection module is within the shooting range of the two-dimensional coupler detection module.

[0012] Furthermore, the two-dimensional detection module for the coupler includes an area array camera, a group of auxiliary line lasers and a group of LED white light sources, the area array camera faces the outer side of the first branch rail, the LED light source is parallel to the area array camera and faces the outer side of the first branch rail, the auxiliary line laser is placed close to the area array camera and the LED light source is located on both sides of the camera respectively, the light cross section of the auxiliary line laser is perpendicular to the ground and the outer side surface of the first branch rail at an angle less than 90°, and the laser line is within the field of view of the area array camera;

[0013] The three-dimensional coupler detection module includes an array camera and a group of line lasers, and uses laser triangulation to collect 3D images of the bottom of the coupler and the bottom of the first branch rail of the moving train.

[0014] Furthermore, the incoming line sensor is mounted on the inner side of the first branch rail through a mounting bracket as the starting position of the entire system, detects the wheels of the incoming train, obtains the train incoming line signal, and starts the control system;

[0015] The offline sensor is mounted on the inner side of the first branch rail through a mounting bracket as the end position of the entire system. It detects that the wheels of the train are offline, obtains a train departure signal, and shuts down the control system.

[0016] Furthermore, the speed measuring radar is installed on a column outside the trackside safety clearance to detect the instantaneous speed of the incoming train; the vehicle number recognition device is installed on a column outside the trackside safety clearance to recognize the vehicle number of the incoming train.

[0017] Furthermore, the three-dimensional coupler detection module and the two-dimensional coupler detection module are both provided with an air blowing and dust removal device.

[0018] Furthermore, the lenses of the incoming line sensor, speed measuring radar, vehicle number recognition device, three-dimensional coupler detection module, two-dimensional coupler detection module and offline sensor are all coated with a waterproof film.

[0019] A method for online detection of train coupler height using any one of the above-mentioned online detection systems for train coupler height comprises:

[0020] (1) When a train approaches the detection system, the detection system is started by obtaining a train approaching signal through the approaching sensor; the speed radar obtains the real-time speed of the train passing, the vehicle number recognition device obtains the vehicle number of the passing train, and the three-dimensional coupler detection module and the two-dimensional coupler detection module adjust the acquisition frame rate according to the real-time speed obtained by the speed radar and then collect images, so that the images are stretched and tend to be consistent in the direction of train movement;

[0021] (2) by processing the collected image data, a 3D image of the bottom of the train coupler and the bottom of the branch rail and a 2D image of the side of the coupler and the side of the branch rail are obtained;

[0022] (3) Through the intelligent recognition and analysis of 2D and 3D images using image processing technology and deep learning technology, the coupler position, coupler height C, distance H from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail, and branch rail height B can be detected. Further analysis and processing can be performed to obtain the height h1 from the lowest point of the lower surface of the coupler to the upper surface of the branch rail. The obtained h1 is compared with the preset h1 alarm threshold. If the preset h1 alarm threshold is exceeded, an early warning prompt is issued.

[0023] Further, the image processing technology is used to analyze the side image of the coupler and the side image of the branch rail collected by the two-dimensional detection module of the coupler, and the height B of the branch rail, the height C of the coupler and the height A from the center line of the center line coupler to the lowest point of the lower surface of the coupler are obtained by analysis;

[0024] The position of the coupler is obtained by collecting the 3D image of the coupler 3D detection module, especially the height value of the lowest point of the coupler. At the same time, the coupler 3D detection module also collects the height value of the lower edge of the branch rail. The height value of the lowest point of the coupler and the height of the lower edge of the branch rail are subtracted to obtain the distance H from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail.

[0025] The height h1 from the lowest point of the lower surface of the coupler to the upper surface of the branch rail is obtained by subtracting the distance H from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail from the height B of the branch rail.

[0026] Furthermore, firstly, the coupler side image and the branch rail side image collected by the coupler two-dimensional detection module are calibrated using a checkerboard calibration plate to obtain the relationship between the pixel points and physical dimensions of the coupler side image and the branch rail side image;

[0027] Then, the side images of the coupler and the side images of the branch rail collected by the two-dimensional coupler detection module are imported into the trained neural network model, and the output results are the coordinates of the upper left corner and the lower right corner of the target, the target category and the probability of the target category; more than 95% of the results can be screened for further analysis through the results of the probability, and the detected results can be obtained through the target category, which are the coupler, the center line of the coupler, and the laser projection line on the side of the first branch rail; the pixel height of the coupler is obtained by subtracting the coordinates of the coupler detection frame; the pixel height from the center line of the coupler to the bottom of the coupler is obtained by subtracting the coordinates of the coupler detection frame and the coordinates of the coupler center line detection frame; the pixel height of the laser projection line on the side of the first branch rail is obtained by subtracting the coordinates of the laser projection line detection frame on the side of the first branch rail;

[0028] Finally, the relationship between the pixel points and physical dimensions of the coupler side image and the branch rail side image is combined to obtain the branch rail height B, the coupler height C and the height A from the center line of the centerline coupler to the lowest point of the lower surface of the coupler.

[0029] Further, the height h1 from the lowest point of the lower surface of the coupler to the upper surface of the branch rail is added to the height A from the center line of the coupler to the lowest point of the lower surface of the coupler to obtain the height h2 from the center line of the coupler to the upper surface of the branch rail;

[0030] The height h1 from the lowest point of the lower surface of the coupler to the upper surface of the support rail is added to the height C of the coupler to obtain the height h3 from the highest point of the coupler to the upper surface of the support rail;

[0031] The obtained h2 and h3 are compared with the preset h2 alarm threshold and h3 alarm threshold, and an early warning prompt is issued if they exceed the preset alarm threshold.

[0032] The beneficial effects of the present invention are as follows:

[0033] (1) The online train coupler height detection system and detection method of the present invention can collect 3D images of the bottom of the coupler and the bottom of the track, and 2D images of the side of the coupler and the side of the track;

[0034] (2) The train coupler height online detection system and detection method of the present invention can automatically identify the coupler in 2D images and 3D images;

[0035] (3) The train coupler height online detection system and detection method of the present invention can detect the distance H from the lowest point of the coupler bottom to the bottom of the track, the coupler body height C, and the track height B;

[0036] (4) The train coupler height online detection system and detection method of the present invention can detect the distance h1 between the lowest point of the bottom of the coupler and the upper surface of the track and compare it with the set alarm threshold to automatically alarm.

[0037] (5) The online train coupler height detection system and detection method of the present invention can solve the problem of coupler center line obstruction or no obvious coupler center line during coupler height detection, expand the detection scene, and increase the reliability of the detection result. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the layout of the system of the present invention;

[0039] Figure 2 It is the working flow chart of the system of the present invention;

[0040] Figure 3 Take a track diagram for the two-dimensional coupler detection module;

[0041] Figure 4 Take a schematic diagram of the field of view for the two-dimensional detection module and the three-dimensional detection module of the coupler;

[0042] Figure 5 It is a detailed flow chart of the method for detecting the height of a coupler in the embodiment;

[0043] Figure 6 Schematic diagram of each English number in the embodiment;

[0044] Among them, 1-first rail, 2-coupler two-dimensional detection module, 3-coupler three-dimensional detection module, 4-incoming line sensor, 5-speed radar, 6-vehicle number recognition device, 7-offline sensor, 8-second rail, 9-coupler;

[0045] H-the distance from the lowest point of the lower surface of the coupler to the bottom surface of the support rail;

[0046] A-the height from the center line of the coupler to the lowest point of the lower surface of the coupler;

[0047] B-height of branch rail;

[0048] C-height of the coupler;

[0049] h1-the height from the lowest point of the lower surface of the coupler to the upper surface of the support rail;

[0050] h2-the height from the center line of the coupler to the upper surface of the branch rail;

[0051] h3-The height from the highest point of the coupler to the upper surface of the support rail. DETAILED DESCRIPTION

[0052] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0053] Embodiment 1:

[0054] Please refer to Figure 1 The train coupler height online detection system of this embodiment includes trackside equipment, a field control center and a remote control center, and the field control center is respectively connected to the trackside equipment and the remote control center. The trackside equipment includes a basic detection unit, which includes an incoming line sensor 4, a speed radar 5, a vehicle number recognition device 6, a three-dimensional coupler detection module 3, a two-dimensional coupler detection module 2 and an offline sensor 7.

[0055] Among them, the incoming line sensor 4 is mounted on the inner side of the first branch rail 1 through a mounting bracket, serving as the starting position of the entire system, detecting the wheels of the incoming train, obtaining the train incoming line signal, and starting the control system.

[0056] The speed radar 5 is installed on a column outside the trackside safety clearance to detect the instantaneous speed of the incoming train.

[0057] Among them, the vehicle number recognition device 6 is installed on a column outside the trackside safety clearance to recognize the vehicle number of the incoming train.

[0058] Among them, the offline sensor 7 is mounted on the inner side of the first branch rail 1 through a mounting bracket, serving as the end position of the entire system, detecting the wheels of the train, obtaining the train departure signal, and shutting down the control system.

[0059] The two-dimensional coupler detection module 2 is installed on a column outside the safety clearance on the same side as the speed measuring radar 5, and uses a side shot method to shoot the area where the side of the branch rail and the side of the coupler 9 are located. The two-dimensional coupler detection module 2 is used to collect images of the outer side of the area of ​​the first branch rail 1 near the installation position of the three-dimensional coupler detection module 3 and the side of the coupler 9.

[0060] The two-dimensional coupler detection module 2 in this embodiment includes an area array camera, a group of auxiliary line lasers and a group of LED white light sources. The area array camera faces the outer side of the first branch rail 1, the LED light source is parallel to the area array camera and faces the outer side of the first branch rail 1, the auxiliary line laser is placed close to the area array camera and the LED light source is located on both sides of the camera, and the light cross section of the auxiliary line laser is perpendicular to the ground and the outer side of the first branch rail 1 at an angle less than 90°. And the laser line is within the field of view of the area array camera.

[0061] There is one two-dimensional coupler detection module 2 , which is distributed on the outside of the first branch rail 1 , and the three-dimensional coupler detection module 3 is within the shooting range of the two-dimensional coupler detection module 2 .

[0062] The three-dimensional coupler detection module 3 is used to collect images of the bottom of the train coupler 9. The three-dimensional coupler detection module 3 is installed at the 1 / 3 position between the first branch rail 1 and the second branch rail 8, and close to the side of the speed radar 5. The specific placement position is in the gap between the two sleepers, and the bottom area of ​​the coupler 9 and the bottom area of ​​the branch rail are photographed from bottom to top.

[0063] The three-dimensional coupler detection module 3 in this embodiment includes an area array camera and a group of line lasers, and uses laser triangulation to collect 3D images of the bottom of the coupler 9 and the bottom of the branch rail of the moving train.

[0064] There is one three-dimensional coupler detection module 3, which is distributed on the inner side of each branch rail (the first branch rail 1 and the second branch rail 8) and close to the first branch rail 1, ensuring that the three-dimensional coupler detection module 3 can simultaneously capture the bottom of the coupler 9 and the bottom of the first branch rail 1. The line laser and camera face the lower surface of the coupler 9 in an upward shooting manner. When the coupler 9 passes through the three-dimensional coupler detection module 3, the line laser is projected onto the lower surface of the coupler 9, which facilitates the area array camera to collect the laser line projected onto the carbon slide plate, and then generate a 3D image.

[0065] The method of photographing the side of the first branch rail 1 by the two-dimensional detection module 2 of the coupler used in this embodiment is as follows: Figure 3 As shown, the two-dimensional coupler detection module 2 includes an array camera, a group of auxiliary lasers and a group of LED white light sources. Figure 3 The dotted frame in the middle represents the camera shooting field of view of the two-dimensional coupler detection module 2, the solid line in the dotted frame represents the light cross-section of the auxiliary laser line, and the thick line at the intersection of the light cross-section and the first branch rail 1 represents the projection line of the laser line on the first branch rail 1, which is used to assist in shooting the side area of ​​the first branch rail 1. The height of the first branch rail 1 is obtained by analyzing the length of the laser projection line on the first branch rail 1.

[0066] The shooting fields of the three-dimensional coupler detection module 3 and the two-dimensional coupler detection module 2 used in this embodiment are as follows: Figure 4As shown, the three-dimensional detection module 3 of the coupler can simultaneously capture the bottom surface of the coupler 9 and the bottom image of the first branch rail 1. The three-dimensional detection module 3 of the coupler can obtain a 3D image to detect the distance H from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail. The two-dimensional detection module 2 of the coupler can capture the side image of the first branch rail 1 and the side image of the coupler 9. The height of the first branch rail 1 (the height B of the branch rail) and the height of the coupler 9 (the height C of the coupler) in the shooting area are detected by the two-dimensional detection module 2 of the coupler; the distance from the coupler to the rail bottom (the distance H from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail) is combined to obtain the distance from the coupler to the rail surface (such as h1, h2, h3). Compared with the detection method in the background technology, when the center line of the coupler surface is blocked or there is no center line, the distance from the coupler to the rail surface can be obtained by analyzing 2D and 3D data.

[0067] Preferably, an air blowing and dust removal device is also provided on the three-dimensional coupler detection module 3 and the two-dimensional coupler detection module 2. The structure of the air blowing and dust removal device is detailed in Chinese patent CN202220450237.X, a hatch device for air blowing and dust removal, which will not be repeated here.

[0068] Preferably, the lenses of each detection module of the basic detection unit are coated with a nano-material waterproof film.

[0069] Among them, the on-site control center realizes the power supply, control, data and image collection, analysis, processing and storage of the basic detection unit, and communicates with the remote control center at the same time; the on-site control center includes an image processing computer, equipment control box, wheel sensor processing device, vehicle number industrial computer, switch, data server, uninterruptible power supply, and lightning protection box.

[0070] Embodiment 2:

[0071] The train coupler height online detection system in Example 1 is used to perform the coupler height online detection. The workflow is as follows: Figure 2 shown.

[0072] (1) When there is no train entering the line, the system is in a dormant state waiting for the train to arrive;

[0073] (2) When the train enters the line, the wheels pass through the line-entry sensor 4, which triggers the effective line-entry signal system to start entering the train receiving state.

[0074] (3) The three-dimensional coupler detection module 3 and the two-dimensional coupler detection module 2 simultaneously open the protective hatch and turn on the air blowing and dust removal device on the window glass. At the same time, the window glass will be heated to remove water mist.

[0075] (4) The system starts the speed radar 5 to detect the speed of the passing vehicle, and calculates the matching camera acquisition frame rate according to the vehicle speed, and sends pulses of corresponding frequency to the three-dimensional coupler detection module 3 to collect the bottom image of the coupler 9. The two-dimensional coupler detection module 2 collects the image of the side of the coupler 9 at 1 / 1000 of the pulse frequency.

[0076] (5) The system analyzes and processes the vehicle number information recognized by the vehicle number recognition device 6, the speed information of the speed radar 5, and the collected image data, and uploads them to the client.

[0077] (6) After the train passes, the train goes offline, the wheels trigger the offline sensor 7, and the number of wheels triggered by the offline sensor 7 is consistent with the incoming line sensor 4. The system completes image acquisition, the train goes offline, the three-dimensional coupler detection module 3 and the two-dimensional coupler detection module 2 are turned off, and the system enters a dormant state waiting for the train to be received.

[0078] Regarding the image acquisition and analysis processing steps in the above steps (4) and (5):

[0079] The image acquisition in step (4) is:

[0080] The three-dimensional detection module 3 for the coupler includes an array camera and a group of line lasers. There is one three-dimensional detection module for the coupler, which is distributed on the inner side of each branch rail (branch rail 1 and branch rail 8) of the track and close to the side of branch rail 1, to ensure that the three-dimensional detection module 3 for the coupler can simultaneously capture the bottom of the coupler and the bottom of branch rail 1. The two-dimensional detection module 2 for the coupler is used to capture images of the side of the coupler 9 and the outer side of the area of ​​branch rail 1 near the installation position of the three-dimensional detection module 3 for the coupler. The two-dimensional detection module 2 for the coupler includes an array camera, a group of auxiliary lasers and a group of LED white light sources. There is one two-dimensional detection module for the coupler, which is distributed on the outer side of branch rail 1 and the three-dimensional detection module 3 for the coupler is within the shooting range of the two-dimensional detection module 2 for the coupler. After receiving the signal from the incoming line sensor 4, the system starts the speed radar 5 to detect the speed of the passing vehicle, and calculates the matching camera acquisition frame rate according to the vehicle speed, and sends pulses of corresponding frequency to the three-dimensional coupler detection module 3 to collect the bottom image of the coupler, and the two-dimensional coupler detection module 2 performs image acquisition at 1 / 1000 of the pulse frequency (the 3D camera of the current three-dimensional coupler detection module 3 is 2000*2000. This setting can ensure that two 2D photos are taken in each 3D camera acquisition area, and the acquisition of 2D photos with continuous overlapping 1 / 2 areas is achieved to avoid omissions and transition overlaps). After the vehicle passes, the wheel triggers the offline sensor 7 and the number of times the offline sensor 7 triggers the wheel is consistent with the incoming line sensor 4. The system completes image acquisition, the train goes offline, the acquisition equipment is turned off, and the system enters a dormant state waiting for the train to arrive.

[0081] The coupler height detection system in step (5) analyzes and processes the collected image:

[0082] The image processing algorithm is used to analyze the side images of the coupler 9 and the first branch rail 1 collected by the two-dimensional coupler detection module 2, and the height of the first branch rail 1 (the height B of the branch rail), the height of the coupler 9 hook body itself (the height C of the coupler) and the height A from the center line of the center line coupler to the lowest point of the lower surface of the coupler are obtained.

[0083] By analyzing the 3D image collected by the three-dimensional detection module 3 of the coupler, the position of the coupler 9 in the image is detected, and the distance from the lowest point of the lower part of the coupler to the lower surface of the rail (the distance H from the lowest point of the lower surface of the coupler to the bottom surface of the support rail) is analyzed; finally, the obtained data is comprehensively analyzed and calculated to obtain the distance from the lowest point of the coupler to the upper surface of the rail (the height h1 from the lowest point of the lower surface of the coupler to the upper surface of the support rail).

[0084] Specifically, regarding the image acquisition and analysis processing in the above steps (4) and (5), as follows Figure 5 As shown:

[0085] Step 1: Image acquisition. The 2D image collected by the two-dimensional coupler detection module 2 is a color image; the 3D image collected by the three-dimensional coupler detection module 3 is 3D data in RGB-D format. The 3D data contains the depth information of each object in the image and the grayscale information of the surface color. A single-wavelength laser is used as the light source, so RGB in RGB-D is grayscale information with consistent grayscale values, and D is depth information.

[0086] Step 2: Image accumulation and preprocessing. Collect images of different types of couplers 9 on the line where the acquisition equipment is installed. Each acquisition equipment collects no less than 200 images of each type of coupler 9. At the same time, collect drawings of couplers 9 of different types of couplers that have passed through the acquisition equipment, measure the height C of the coupler 9 body and the height from the center line of the coupler to the lowest point of the bottom of the coupler (the height A from the center line of the coupler to the lowest point of the lower surface of the coupler), and set different alarm thresholds of h1, h2, and h3 according to the relevant dimensions of different types of couplers 9, such as C and A.

[0087] Step 3: Calibration of the two-dimensional coupler detection module 2. The camera is calibrated with a checkerboard calibration plate to obtain the relationship between the pixels and physical dimensions of the image captured by the camera. The checkerboard calibration plate is placed on the side of the first branch rail 1 and is close to the side of the first branch rail 1. The two-dimensional coupler detection module 2 collects the calibration plate image. The number of grids on the chessboard and the side length L of each grid are known. The number of pixels S1 on the edge of each grid on the chessboard in the image is analyzed. By calculating L / S1, the physical dimensions represented by each pixel of the side image of the first branch rail 1 captured by the two-dimensional coupler detection module 2 can be obtained.

[0088] Similarly, place the chessboard calibration plate above the center of the two rails and 850 mm from the rail surface for image acquisition. This position is the height area of ​​the coupler 9. The number of grids on the chessboard and the side length L of each grid are known. Then, the number of pixels S2 on each grid edge of the chessboard in the image at this position is analyzed. By calculating L / S2, the physical size represented by each pixel of the side image of the coupler 9 collected by the two-dimensional detection module 2 of the coupler can be obtained.

[0089] Step 4: Image analysis of the two-dimensional coupler detection module 2. The 200 images collected by the two-dimensional coupler detection module 2 in the first step are annotated as a data set. The images are annotated using the labelImg image annotation software. The side area of ​​the coupler 9, the center line of the coupler 9, and the laser projection line on the side of the first branch rail 1 are finely annotated respectively. The annotated data are input into the PP-YOLOE neural network training target detection model.

[0090] After the training is completed, the 2D image to be detected is input into the PP-YOLOE neural network, and the output results are the coordinates of the upper left corner and the lower right corner of the target, the target category and the probability of the target category; the results above 0.95 are further analyzed by screening the probability results, and the detected results are the coupler 9, the coupler centerline, and the laser projection line on the side of the first branch rail 1. The pixel height of the coupler can be obtained by difference calculation of the coupler detection frame coordinates; the pixel height from the coupler centerline to the bottom of the coupler can be obtained by difference calculation of the coupler detection frame coordinates and the coupler centerline detection frame coordinates; the pixel length of the laser projection line on the side of the branch rail 1, that is, the pixel height of the branch rail 1, can be obtained by difference calculation of the detection frame coordinates of the coupler detection frame and the coupler centerline. Combined with the calibration information in the third step, the relationship between the pixel point at the coupler and the physical size, the relationship between the pixel point at the first branch rail 1 and the physical size are obtained, and finally the actual height C of the coupler, the actual height from the coupler centerline to the bottom of the coupler (the height A from the coupler centerline to the lowest point of the lower surface of the coupler), and the length of the laser projection line on the side of the branch rail 1, that is, the height B of the branch rail 1, are obtained.

[0091] Step 5: Image analysis of the three-dimensional detection module 3 of the coupler. In the first step, the 200 3D images collected by the three-dimensional detection module 3 of the coupler are annotated as a data set. The 3D image is 3D data in RGB-D format. The 3D data contains the depth information of each object in the image and the grayscale information of the surface color. A single-wavelength laser is used as the light source, so RGB in RGB-D is grayscale information with consistent grayscale values, and D is depth information. The grayscale information and depth information in the RGB-D data are separated and saved as two png format image data of the same size. The grayscale image data is 8 bits, and the image data of the depth information is 16 bits. The grayscale image is annotated by the labelImg image annotation software, and the bottom area of ​​the side of the coupler is finely annotated (it means that the annotation box fits the edge of the coupler), and the annotated data is input into the PP-YOLOE neural network training target detection model.

[0092] After the training is completed, the grayscale image obtained by RGB-D data separation is input into the PP-YOLOE neural network, and the output results are the coordinates of the upper left corner and the lower right corner of the target, the target category and the probability of the target category; the results with probability results above 0.95 are further analyzed, and the detected result can be obtained by category as the coupler 9, and the coordinate frame of the coupler 9 is obtained at the same time. The maximum depth value is retrieved in the same frame coordinate area of ​​the depth information image obtained by RGB-D data separation, and the mean depth information of each point in the rectangle with a length of 9 and a width of 9 with this point as the center is extracted. The mean is used as the coupler 9 The height value of the lowest point, at the same time, the three-dimensional detection module 3 of the coupler can collect the lower edge of the first branch rail 1. Since the three-dimensional detection module 3 of the coupler adopts the shooting method of the laser triangulation method, the lower edge of the first branch rail 1 collected is always fixed. The coordinate point of the lower edge of the first branch rail 1 is selected by observing the grayscale image and then the depth information of the corresponding point in the depth information image is extracted, which is the height of the lower edge of the first branch rail 1; the difference between the height value of the lowest point of the coupler and the height of the lower edge of the first branch rail 1 is the height from the coupler 9 to the lower edge of the first branch rail 1 (the distance from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail, i.e. H).

[0093] The height H is subtracted from the height of the first branch rail 1 obtained in the fourth step to obtain the height of the lowest point of the coupler from the upper surface of the branch rail 1 (the height from the lowest point of the lower surface of the coupler to the upper surface of the branch rail, i.e. h1).

[0094] The height from the center line of the coupler to the upper surface of the first branch rail 1 (the height from the center line of the coupler to the upper surface of the branch rail, i.e., h2) can be obtained by summing h1 with the height from the center line of the coupler to the bottom of the coupler obtained in the fourth step (the height A from the center line of the coupler to the lowest point of the lower surface of the coupler).

[0095] The height from the highest point of the coupler to the upper surface of the first branch rail 1 can be obtained by summing h1 with the coupler height C obtained in the fourth step (the height from the highest point of the coupler to the upper surface of the branch rail, i.e., h3).

[0096] Compare the obtained h1, h2, and h3 with the manually set alarm thresholds. If they exceed the alarm thresholds, an early warning prompt will be issued.

[0097] The present invention has been disclosed as above in terms of preferred embodiments, but they are not intended to limit the present invention. Anyone familiar with this technology can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection defined by the claims of this application.

Claims

1. A train coupler height online detection system, characterized in that: The system includes trackside equipment, an on-site control center and a remote control center. The on-site control center is respectively connected to the trackside equipment and the remote control center for communication. The trackside equipment includes an incoming line sensor, a speed measuring radar, a vehicle number recognition device, a three-dimensional coupler detection module, a two-dimensional coupler detection module and an offline sensor. The three-dimensional detection module for the coupler is located between the first branch rail and the second branch rail and close to the first branch rail, and collects images of the bottom of the first branch rail and the bottom of the coupler from bottom to top; The two-dimensional coupler detection module is arranged on the outer side of the first branch rail close to the three-dimensional coupler detection module, and images of the side of the first branch rail and the side of the coupler are captured by side shooting, and the three-dimensional coupler detection module is within the shooting range of the two-dimensional coupler detection module.

2. A train coupler height online detection system as claimed in claim 1, characterized in that: The two-dimensional detection module for the coupler includes an area array camera, a group of auxiliary line lasers and a group of LED white light sources. The area array camera faces the outer side of the first branch rail. The LED light source is parallel to the area array camera and faces the outer side of the first branch rail. The auxiliary line laser is placed close to the area array camera and is located on both sides of the camera with the LED light source. The light cross section of the auxiliary line laser is perpendicular to the ground and forms an angle of less than 90° with the outer side surface of the first branch rail, and the laser line is within the field of view of the area array camera. The three-dimensional coupler detection module includes an array camera and a group of line lasers, and uses laser triangulation to collect 3D images of the bottom of the coupler and the bottom of the first branch rail of the moving train.

3. A train coupler height online detection system as claimed in claim 1 or 2, characterized in that: The incoming line sensor is mounted on the inner side of the first branch rail through a mounting bracket as the starting position of the entire system, detects the wheels of the incoming train, obtains the train incoming line signal, and starts the control system; The offline sensor is mounted on the inner side of the first branch rail through a mounting bracket as the end position of the entire system. It detects that the wheels of the train are offline, obtains a train departure signal, and shuts down the control system.

4. A train coupler height online detection system as claimed in claim 1 or 2, characterized in that: The speed measuring radar is installed on a column outside the trackside safety clearance to detect the instantaneous speed of the incoming train; the vehicle number recognition device is installed on a column outside the trackside safety clearance to recognize the vehicle number of the incoming train.

5. A train coupler height online detection system as claimed in claim 1 or 2, characterized in that: The three-dimensional coupler detection module and the two-dimensional coupler detection module are both provided with an air blowing and dust removal device.

6. A train coupler height online detection system as claimed in claim 1 or 2, characterized in that: The lenses of the incoming line sensor, speed measuring radar, vehicle number recognition device, three-dimensional coupler detection module, two-dimensional coupler detection module and offline sensor are all coated with a waterproof film.

7. A method for online detection of train coupler height using a train coupler height online detection system according to any one of claims 1 to 6, characterized in that: The method comprises: (1) When a train approaches the detection system, the detection system is started by obtaining a train approaching signal through the approaching sensor; the speed radar obtains the real-time speed of the train passing, the vehicle number recognition device obtains the vehicle number of the passing train, and the three-dimensional coupler detection module and the two-dimensional coupler detection module adjust the acquisition frame rate according to the real-time speed obtained by the speed radar and then collect images, so that the images are stretched and tend to be consistent in the direction of train movement; (2) by processing the collected image data, a 3D image of the bottom of the train coupler and the bottom of the branch rail and a 2D image of the side of the coupler and the side of the branch rail are obtained; (3) Through the intelligent recognition and analysis of 2D and 3D images using image processing technology and deep learning technology, the coupler position, coupler height C, distance H from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail, and branch rail height B can be detected. Further analysis and processing can be performed to obtain the height h1 from the lowest point of the lower surface of the coupler to the upper surface of the branch rail. The obtained h1 is compared with the preset h1 alarm threshold. If the preset h1 alarm threshold is exceeded, an early warning prompt is issued.

8. The method according to claim 7, characterized in that (3): using image processing technology to analyze the side image of the coupler and the side image of the branch rail collected by the two-dimensional coupler detection module, and analyzing to obtain the height B of the branch rail, the height C of the coupler, and the height A from the center line of the centerline coupler to the lowest point of the lower surface of the coupler; The position of the coupler is obtained by collecting the 3D image of the coupler 3D detection module, especially the height value of the lowest point of the coupler. At the same time, the coupler 3D detection module also collects the height value of the lower edge of the branch rail. The height value of the lowest point of the coupler and the height of the lower edge of the branch rail are subtracted to obtain the distance H from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail. The height h1 from the lowest point of the lower surface of the coupler to the upper surface of the branch rail is obtained by subtracting the distance H from the lowest point of the lower surface of the coupler to the bottom surface of the branch rail from the height B of the branch rail.

9. The method according to claim 8, characterized in that Firstly, the coupler side image and the branch rail side image collected by the coupler two-dimensional detection module are calibrated using a checkerboard calibration plate to obtain the relationship between the pixel points and physical dimensions of the coupler side image and the branch rail side image; Then, the side images of the coupler and the side images of the branch rail collected by the two-dimensional coupler detection module are imported into the trained neural network model, and the output results are the coordinates of the upper left corner and the lower right corner of the target, the target category and the probability of the target category; more than 95% of the results can be screened for further analysis through the results of the probability, and the detected results can be obtained through the target category, which are the coupler, the center line of the coupler, and the laser projection line on the side of the first branch rail; the pixel height of the coupler is obtained by subtracting the coordinates of the coupler detection frame; the pixel height from the center line of the coupler to the bottom of the coupler is obtained by subtracting the coordinates of the coupler detection frame and the coordinates of the coupler center line detection frame; the pixel height of the laser projection line on the side of the first branch rail is obtained by subtracting the coordinates of the laser projection line detection frame on the side of the first branch rail; Finally, the relationship between the pixel points and physical dimensions of the coupler side image and the branch rail side image is combined to obtain the branch rail height B, the coupler height C and the height A from the center line of the centerline coupler to the lowest point of the lower surface of the coupler.

10. The method according to claim 8 or 9, characterized in that The height h1 from the lowest point of the lower surface of the coupler to the upper surface of the branch rail is added to the height A from the center line of the coupler to the lowest point of the lower surface of the coupler to obtain the height h2 from the center line of the coupler to the upper surface of the branch rail; The height h1 from the lowest point of the lower surface of the coupler to the upper surface of the support rail is added to the height C of the coupler to obtain the height h3 from the highest point of the coupler to the upper surface of the support rail; The obtained h2 and h3 are compared with the preset h2 alarm threshold and h3 alarm threshold, and an early warning prompt is issued if they exceed the preset alarm threshold.

Images