Railway infrastructure positioning method and device assisted by transponder image

Abstract

The invention discloses a transponder image-assisted railway infrastructure positioning method and device, and the method comprises the steps: obtaining a transponder image collected by a color line-scan digital camera through a high-speed transmission interface; inputting the transponder image into a transponder identification model, and identifying the position of a transponder in the transponder image; taking the train running mileage information corresponding to the transponder image acquisition time as an index identifier of the transponder image, and generating first position information of the transponder in the railway line according to the index identifier of the transponder image and the position of the transponder in the transponder image; the standing book information of the transponder with the distance within the preset range from the first position information is obtained, wherein the standing book information comprises second position information pre-recorded by the transponder; and correcting the train running mileage information by using the second position information to obtain a positioning result of the train on the railway line. According to the invention, the robustness of the mileage positioning method based on the transponder can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of railway infrastructure, and in particular to a method and device for locating railway infrastructure assisted by a transponder image. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the invention recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.

[0003] Railway infrastructure inspection mainly involves the inspection of tracks, contact networks and communication signals. The inspection data collected by these inspection systems require the train to provide unified mileage positioning information to better guide disease control and equipment maintenance. Therefore, it is necessary to adopt a more diverse and accurate mileage positioning method to achieve the mileage positioning of the inspection train, so as to accurately and quickly detect defects in railway infrastructure.

[0004] Existing mileage positioning methods mostly use global satellite navigation system (GNSS), radio frequency tag technology (RFID), simultaneous positioning and mapping (SLAM) technology, etc. However, GNSS has high requirements on climate and geographical factors, RFID is limited by the conditions for reading tag card information, and the positioning accuracy of high-speed trains is difficult to guarantee. The calculation complexity of SLAM technology is high, and it is difficult to achieve real-time calculation on high-speed trains, and the implementation cost is high.

[0005] In addition to the above-mentioned mileage positioning methods, there is also a mileage positioning method based on a transponder: by interacting with the transponder on the railway, when the on-board equipment approaches the transponder on the railway, a signal is sent to activate the transponder, and the transponder sends the stored data to the on-board equipment by electromagnetic induction. The on-board system analyzes the transponder data to achieve train positioning. Compared with the above-mentioned mileage positioning methods, the mileage positioning method based on a transponder has a series of advantages such as high positioning accuracy, no influence from climate and location factors, easy maintenance, and low operating costs. However, once the transponder is damaged or the acquired transponder information is abnormal, it will affect the positioning accuracy, resulting in the robustness of the transponder-based mileage positioning method still needs to be improved. Summary of the invention

[0006] An embodiment of the present invention provides a transponder image-assisted railway infrastructure positioning method, which is used to improve the robustness of the transponder-based railway infrastructure positioning method. The method includes:

[0007] Through a high-speed transmission interface, the transponder image captured by the color linear array camera installed on the train is obtained, and the transponder image acquisition time is recorded; wherein the transponder is installed on the roadbed;

[0008] Inputting the transponder image into a transponder recognition model to identify the transponder position in the transponder image; the transponder recognition model is obtained by training a machine learning model using historical transponder images and transponder positions in the historical transponder images;

[0009] Using the train running mileage information corresponding to the transponder image acquisition time as the index identifier of the transponder image, and generating the first position information of the transponder in the railway line according to the index identifier of the transponder image and the transponder position in the transponder image;

[0010] Obtaining the ledger information of the transponder within a preset range from the first location information, the ledger information including the second location information pre-recorded by the transponder;

[0011] The second position information is used to correct the train running mileage information to obtain the positioning result of the train on the railway line.

[0012] The embodiment of the present invention further provides a transponder image-assisted railway infrastructure positioning device, which is used to improve the robustness of the transponder-based railway infrastructure positioning method. The device includes:

[0013] An image acquisition module is used to acquire the transponder image acquired by the color linear array camera installed on the train through a high-speed transmission interface, and record the transponder image acquisition time; wherein the transponder is installed on the roadbed;

[0014] An image processing module, for inputting the transponder image into a transponder recognition model to recognize the transponder position in the transponder image; the transponder recognition model is obtained by training a machine learning model using historical transponder images and transponder positions in the historical transponder images;

[0015] A transponder position generating module, used to use the train running mileage information corresponding to the transponder image acquisition time as the index identifier of the transponder image, and generate the first position information of the transponder in the railway line according to the transponder position in the transponder image and the index identifier of the transponder image;

[0016] A ledger data acquisition module, used to obtain ledger information of transponders within a preset range from the first location information, the ledger information including second location information pre-entered by the transponder;

[0017] The positioning module is used to correct the train running mileage information using the second position information to obtain the positioning result of the train on the railway line.

[0018] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned transponder image-assisted railway infrastructure positioning method when executing the computer program.

[0019] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the transponder image-assisted railway infrastructure positioning method is implemented.

[0020] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned transponder image-assisted railway infrastructure positioning method is implemented.

[0021] Compared with the method in the prior art in which the on-board equipment directly interacts with the transponder and analyzes the transponder data to achieve mileage positioning, the embodiment of the present invention uses image recognition technology to determine the ledger data of the transponder, and then uses the second position information pre-recorded in the ledger data to correct the train running mileage information, thereby solving the problem of reduced positioning accuracy caused by transponder damage or abnormal transponder data in the existing transponder-based mileage positioning method, thereby improving the robustness of the transponder-based mileage positioning method. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0023] Figure 1 It is a flow chart of a method for locating railway infrastructure assisted by transponder images in an embodiment of the present invention;

[0024] Figure 2 An example image showing an image recognition model correctly identifying a transponder in a grayscale image;

[0025] Figure 3 and Figure 4 These are all example images of transponders in grayscale images that were incorrectly identified by the image recognition model;

[0026] Figure 5 Schematic diagram of the imaging principle of the transponder image in an embodiment of the present invention;

[0027] Figure 6 An example diagram of the YOLOv5 network structure used in the transponder recognition model in an embodiment of the present invention;

[0028] Figure 7 Schematic diagram of the workflow of the YOLOv5 network in an embodiment of the present invention;

[0029] Figure 8This is an example diagram of a workflow of applying the transponder image-assisted railway infrastructure positioning method in an embodiment of the present invention to an actual scenario;

[0030] Fig. 9 This is an example diagram of the interaction relationship between modules in the image acquisition and processing platform in an embodiment of the present invention;

[0031] Fig.10 Schematic diagram of a transponder image-assisted railway infrastructure positioning device in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0033] In order to improve the robustness of the railway infrastructure positioning method based on transponder, a transponder image-assisted railway infrastructure positioning method is proposed in an embodiment of the present invention. Figure 1 is a flow chart of a transponder image-assisted railway infrastructure positioning method in an embodiment of the present invention. Figure 1 As shown, the method can be implemented by the following steps:

[0034] Step 101, obtaining a transponder image captured by a color linear array camera installed on a train through a high-speed transmission interface, and recording the transponder image acquisition time; wherein the transponder is installed on the roadbed;

[0035] Step 102: input the transponder image into a transponder recognition model to identify the transponder position in the transponder image; the transponder recognition model is obtained by training a machine learning model using historical transponder images and transponder positions in the historical transponder images;

[0036] Step 103: using the train running mileage information corresponding to the transponder image acquisition time as the index identifier of the transponder image, and generating first position information of the transponder in the railway line according to the index identifier of the transponder image and the transponder position in the transponder image;

[0037] Step 104: obtaining the account information of the transponders within a preset range from the first location information, the account information including the second location information pre-recorded by the transponder;

[0038] Step 105: Use the second position information to correct the train mileage information to obtain the positioning result of the train on the railway line.

[0039] The inventors have found through investigation that, in existing research on railway infrastructure image recognition, most of the image sensors used are grayscale sensors, which collect data to obtain grayscale images. Figure 2 This is an example of correctly identifying a transponder in a grayscale image. In general, image recognition models can better identify transponders in grayscale images, such as Figure 2 shown. Figure 3 and Figure 4 These are all examples of the image recognition model incorrectly identifying the transponder in the grayscale image. Due to the complex line status, there are some equipment or logo images with similar features to the transponder image, which leads to false positives when performing image analysis on the collected grayscale image, such as Figure 3 and Figure 4 As shown, the recognition accuracy of the transponder device is affected.

[0040] In order to improve the recognition accuracy of the transponder device, in an embodiment of the present invention, a transponder image captured by a color linear array camera installed on the train is obtained through a high-speed transmission interface, and the transponder image acquisition time is recorded; wherein the transponder is installed on the roadbed.

[0041] In one embodiment, the transponder image is acquired by a color linear array camera by photographing the track area illuminated by a light source; the light source is a laser light source and / or an LED light source, which is triggered by a pulse signal generated by a speed encoder installed on the train and illuminates the track area. Figure 5 FIG. 2 is an example diagram of the imaging principle of the transponder image in an embodiment of the present invention. Figure 5 As shown, in one example, an in-vehicle device used for task management and data processing can receive the pulse signal output by the speed encoder, and control the operation of under-vehicle components such as color linear array cameras, LED lights, and lasers through the pulse signal to achieve imaging of the transponder image.

[0042] Among them, the imaging scheme of the transponder image uses a high-resolution color linear array camera with a white laser light source or an LED light source. The color linear array camera and laser (or LED light) are installed on the high-speed comprehensive inspection train. After installation, quasi-static calibration is required, which mainly involves routine correction and white balance, so that the optical imaging effect meets the inspection requirements.

[0043] The specific process of transponder image acquisition is as follows: as the wheel rotates, the high-precision speed encoder installed on the wheel axle box continuously outputs equally spaced pulse signals to the device installed in the vehicle. The pulse signals processed by the device installed in the vehicle are input into the color high-definition optical imaging component installed under the vehicle, triggering the color linear array camera to acquire the transponder image and store the acquired image data to the local hard disk.

[0044] After the above process, the transponder image captured by the color linear array camera installed on the train can be directly obtained through the high-speed transmission interface, and the transponder image acquisition time can be recorded.

[0045] In an embodiment of the present invention, the transponder image is input into a transponder recognition model to identify the transponder position in the transponder image; the transponder recognition model is obtained by training a machine learning model using historical transponder images and the transponder positions in the historical transponder images.

[0046] In order to meet the needs of real-time image processing, in one embodiment, the machine learning model is a YOLOv5 network. For example, on a high-speed comprehensive inspection train with a speed of up to 400 km / h, the YOLOv5 network is used as a machine learning model to meet the needs of real-time image processing. YOLOv5 is a single-stage target detection algorithm that uses a lightweight model design. Its advantages are not only reflected in the performance of fast detection speed and high accuracy, but also in the characteristics of easy deployment.

[0047] Figure 6 This is an example diagram of the YOLOv5 network structure used in the transponder recognition model in an embodiment of the present invention.

[0048] The YOLOv5 network architecture is mainly divided into four parts: Input, Backbone, Neck, and Head. Figure 6As shown in the figure, the input end mainly includes Mosaic data enhancement, adaptive anchor box calculation, and adaptive image scaling. The backbone network adopts the New CSP-Darknet53 feature extraction network, and uses the Conv, C3 modules, and Spatial Pyramid Pooling (SPP) basic structure to extract features from the input image. The CBL module is a combination of Conv2D, BN (Batch Normalization), and activation function LeakyReLU. The CSP1_x module is a combination of CBL, Res unit (Rasidual unit), Conv, BN, and LeakyRELU. The CSP2_x module is a combination of CBL, Conv, Concat (Concatenate), BN, and LeakyRELU. The neck network uses SPP to increase the receptive field of the network, uses the New CSP-PAN feature pyramid structure for multi-scale target prediction, fuses shallow features (shallow features are conducive to detection) to the feature maps of the three scales to be detected, and fuses shallow features to the extracted feature maps, so that the feature maps have both rich semantic information and accurate location information of the object. SPP is a combination of CBL, Max pool, and Concat. The head network uses the traditional YOLOv3 Head to predict the processed feature maps, and optimizes the parameter weights according to the loss function (Classification Loss and Bounding Box Regeression Loss) and the optimizer.

[0049] like Figure 6 As shown in the figure, YOLOv5 uses the Focus module, which reduces the image scale and increases the number of channels to increase the recognition ability of the model. The Focus module is a combination of slice, Concat, and CBL; the network loss function uses BCE Loss and CIOU Loss; the activation function uses HardSwish. In addition, during model training, YOLOv5 uses a variety of data enhancement methods, including Mosaic, Copy paste, Random affine, MixUp, Albumentations, Augment HSV, and Random horizontal flip, etc., to increase the generalization and robustness of the model from the data dimension.

[0050] The training process of the transponder recognition model based on the YOLOv5 network is as follows: first, it is necessary to obtain transponder image samples in the historical inspection data, annotate each transponder in the image sample, and annotate the coordinates of the minimum circumscribed rectangle of the transponder in the image; then, according to the input features such as the number of image samples, the type of transponder, and the image resolution, adjust the YOLOv5 network parameters so that the network can converge quickly during training; finally, use a large number of annotated transponder images for training, and obtain the final transponder recognition model after the network converges. The transponder recognition model can automatically identify the location of the transponder from the image.

[0051] Figure 7 The following is a schematic diagram of the workflow of the YOLOv5 network in an embodiment of the present invention. During the railway line detection process, the real-time collected transponder image is obtained through a high-speed transmission interface. The transponder recognition model based on the YOLOv5 network is integrated into the image acquisition and processing platform, such as Figure 7 As shown in the figure, the acquired transponder image is input into the input end of the transponder recognition model, and is processed by the backbone network, the neck network and the head network in sequence to output the recognition result. The recognition result marks the location of the transponder in the image in the form of a minimum circumscribed rectangle, realizing real-time recognition of the transponder location in the captured image.

[0052] In the existing transponder-based mileage positioning method, the number of transponder deployments is limited by cost factors, and the deployment locations of transponders usually maintain a certain distance. When the transponder is damaged or the transponder information obtained is abnormal, it will affect the final positioning accuracy. The embodiment of the present invention uses image recognition technology to determine the ledger data of the transponder, and uses the ledger data to correct the transponder location information, thereby improving the robustness of the transponder-based mileage positioning method.

[0053] In an embodiment of the present invention, the train mileage information corresponding to the transponder image acquisition moment is used as the index identifier of the transponder image, and the first position information of the transponder in the railway line is generated according to the index identifier of the transponder image and the transponder position in the transponder image.

[0054] The above-mentioned embodiment describes the transponder image acquisition process. The high-precision speed encoder installed in the wheel axle box generates the time of sending the pulse signal (transponder image acquisition time) and the running distance information of the train relative to the starting point while outputting the pulse signal.

[0055] In one embodiment, before the train mileage information corresponding to the transponder image acquisition time is used as the index identifier of the transponder image, the train mileage information corresponding to the transponder image acquisition time is determined according to the correspondence between the train running time and the train running mileage.

[0056] The train mileage information can be obtained in the following way: the equipment installed in the car generates the running distance information of the train relative to the starting point according to the pulse signal sent by the high-precision speed encoder, and combines the mileage information of the train running starting point to obtain the train running mileage information. The train running mileage information can also be obtained in the following way: the equipment installed in the car receives the transponder data transmitted by electromagnetic induction, and obtains the train running mileage information by analyzing the transponder data; the transponder data contains data sending time information and location information.

[0057] In order to improve the accuracy of the transponder recognition result, the accuracy of the generated first position information can be further improved. In one embodiment of the present invention, by integrating a false alarm filtering algorithm based on the HSV color space in the image acquisition and processing platform, the output result of the transponder recognition model is filtered according to the color information to eliminate recognition false alarms, thereby improving the accuracy of the generated first position.

[0058] In one embodiment, the transponder image is an RGB image; after identifying the position of the transponder in the transponder image, the method further includes: obtaining a transponder area in the transponder image according to the position of the transponder in the transponder image; converting the transponder image from an RGB digital image to an HSV image, and calculating the H channel mean of the transponder area in the HSV image; deleting the HSV image of the transponder area whose H channel mean is greater than a first threshold or less than a second threshold; retaining the HSV image of the transponder area whose H channel mean is not greater than the first threshold and not less than the second threshold; wherein the first threshold is the maximum value of the H channel value of the HSV image of the transponder area, and the second threshold is the minimum value of the H channel value of the HSV image of the transponder area.

[0059] For example, first, we need to obtain the minimum enclosing rectangular area in the recognition result according to the recognition result output by the transponder recognition model, use the area as the transponder area sub-image, and then convert the color space of the transponder area sub-image from RGB to HSV. The conversion steps are as follows:

[0060] 1) Normalize the RGB values ​​to the range of [0, 1], that is, R'=R / 255, G'=G / 255, B'=B / 255, where R, G, B are the RGB values ​​before normalization, and R', G', B' are the RGB values ​​after normalization;

[0061] 2) Calculate the maximum value Max=max(R',G',B') and the minimum value Min=min(R',G',B') of the RGB value;

[0062] 3) Calculate the hue H: if Max = Min, then H = 0; if Max = R', then H = 60*((G'-B') / (Max-Min)+0); if Max = G', then H = 60*((B'-R') / (Max-Min)+2); if Max = B', then H = 60*((R'-G') / (Max-Min)+4); if H is less than 0, then H = H+360;

[0063] 4) Calculate the saturation S: if Max=0, then S=0; if Max≠0, then S=(Max-Min) / Max;

[0064] 5) Calculate the brightness V: V = Max.

[0065] The HSV color space of the transponder region sub-image is obtained through the above steps. A rectangular area is obtained in the region sub-image, and the H channel mean of the matrix area is calculated. The matrix area is centered at the center of the region sub-image and has a side length of 1 / 2 of the side length of the region sub-image. Since the color of the transponder is yellow, the value range of its hue H is [45, 75]. If the H channel mean of the matrix area of ​​the recognition result is greater than 75 or less than 45, the recognition result is judged as a false alarm and is filtered out. The filtered transponder image is used to generate the first position information of the transponder in the railway line.

[0066] In an embodiment of the present invention, the ledger information of the transponder within a preset range from the first location information is obtained, and the ledger information includes the second location information pre-recorded by the transponder.

[0067] In the embodiment of the present invention, the second position information is used to correct the train running mileage information to obtain the positioning result of the train on the railway line.

[0068] For example, in the train running time data, the corresponding transponder image acquisition time is marked, and then the transponder image acquisition time is associated with the train running mileage information. The train running mileage information corresponding to the transponder image acquisition time is obtained, and the second position information corresponding to the transponder image acquisition time is used to correct the train running mileage information to obtain the positioning result of the train in the railway line.

[0069] It should be noted that: although the error of the transponder data transmitted by electromagnetic induction in the prior art is very small, when the train runs at high speed for a long time, the final train positioning result may have cumulative errors. Similarly, a high-precision speed encoder in a long-term high-speed running state will also have cumulative errors. In the embodiment of the present invention, the method of correcting the train mileage information using the transponder position information pre-entered in the ledger data can also correct the cumulative errors caused by the long-term high-speed operation of the train.

[0070] Figure 8 FIG. 1 is an example of a workflow of applying the method for locating railway infrastructure assisted by transponder images in an actual scenario in an embodiment of the present invention. Figure 8 As shown, first, the on-board detection system is turned on and self-checked; then, parameters are configured on the equipment installed in the vehicle, the collection task is started, and the detection train is waiting to start running; when the detection train starts running, the components installed under the vehicle cooperate with the instructions of the equipment installed in the vehicle to collect image data in real time and store it on the local hard disk; during the image collection process, the image collection and processing platform integrated with the equipment installed in the vehicle processes the collected images in real time, analyzes the image data online, generates a transponder data packet and sends it to the positioning module. The transponder data packet contains time information and location information; finally, the train stops running and the detection is stopped.

[0071] Fig. 9 FIG. 2 is an example diagram of the interaction relationship between modules in the image acquisition and processing platform in an embodiment of the present invention. Fig. 9 As shown in the figure, the image acquisition and processing platform integrated in the in-vehicle equipment mainly includes four modules: image acquisition module, image processing module, task control and management module, and positioning module. The specific introduction is as follows:

[0072] Image acquisition module: first receives the pulse signal sent by the signal synchronization unit of the under-vehicle mounted component; controls the sensor to acquire images through the pulse signal control unit according to the control parameters transmitted by the task control and management module; receives the image acquired by the sensor, and then obtains the image source data through the image data acquisition unit, image compression unit, and image transmission unit in sequence.

[0073] Image processing module: receives the image data with index identification sent by the task control and management module, performs image recognition, color filtering and other processing on the image data to obtain the transponder recognition result; each image processing sub-unit transmits information via 1GLAN.

[0074] Positioning module: responsible for receiving the transponder identification results sent by the mission control and management module; obtaining the train positioning results by analyzing the transponder identification results; synchronizing other detection systems with the positioning information; and finally sending the positioning synchronization and control information to the mission control and management module.

[0075] Task control and management module: including identification data receiving and management unit, sensor monitoring and parameter setting unit, image acquisition task configuration unit, image index and storage unit, network sending unit, positioning synchronization information receiving unit; the above units use PCle to exchange data with the image acquisition module and image processing module, and use 1GLAN to exchange data with the positioning module.

[0076] The present invention also provides a transponder image assisted railway infrastructure positioning device in an embodiment, as described in the following embodiment. Since the principle of the device to solve the problem is similar to that of the transponder image assisted railway infrastructure positioning method, the implementation of the device can refer to the implementation of the transponder image assisted railway infrastructure positioning method, and the repeated parts will not be repeated.

[0077] Fig.10 Schematic diagram of a transponder image-assisted railway infrastructure positioning device according to an embodiment of the present invention. Fig.10 As shown, the device comprises:

[0078] The image acquisition module 1001 is used to acquire the transponder image acquired by the color linear array camera installed on the train through a high-speed transmission interface, and record the transponder image acquisition time; wherein the transponder is installed on the roadbed;

[0079] An image processing module 1002 is used to input the transponder image into a transponder recognition model to identify the transponder position in the transponder image; the transponder recognition model is obtained by training a machine learning model using historical transponder images and transponder positions in the historical transponder images;

[0080] The transponder position generating module 1003 is used to use the train running mileage information corresponding to the transponder image acquisition time as the index identifier of the transponder image, and generate the first position information of the transponder in the railway line according to the transponder position in the transponder image and the index identifier of the transponder image;

[0081] The ledger data acquisition module 1004 is used to obtain the ledger information of the transponder within a preset range from the first location information, and the ledger information includes the second location information pre-recorded by the transponder;

[0082] The positioning module 1005 is used to correct the train running mileage information using the second position information to obtain the positioning result of the train on the railway line.

[0083] In one embodiment, the transponder image is acquired by a color linear array camera by photographing the track area illuminated by a light source; the light source is a laser light source and / or an LED light source, which is triggered by a pulse signal generated by a speed encoder installed on the train and illuminates the track area.

[0084] In one embodiment, the machine learning model is a YOLOv5 network.

[0085] In one embodiment, the transponder image is an RGB image; the image processing module 1002 is further configured to, after identifying the transponder position in the transponder image:

[0086] According to the transponder position in the transponder image, a transponder area in the transponder image is obtained;

[0087] Convert the transponder image from RGB image to HSV image, and calculate the H channel mean of the transponder area in the HSV image;

[0088] Delete the HSV images of the transponder area whose H channel mean is greater than the first threshold or less than the second threshold; retain the HSV images of the transponder area whose H channel mean is not greater than the first threshold and not less than the second threshold; wherein the first threshold is the maximum value of the H channel value of the HSV image of the transponder area, and the second threshold is the minimum value of the H channel value of the HSV image of the transponder area;

[0089] The transponder position generating module is specifically used to generate the first position information of the transponder in the railway line according to the index identification of the transponder image and the HSV image of the reserved transponder area.

[0090] In one embodiment, the transponder position generating module 1003 is further configured to, before using the train running mileage information corresponding to the transponder image acquisition time as the index mark of the transponder image:

[0091] According to the correspondence between the train running time and the train running mileage, the train running mileage information corresponding to the transponder image acquisition time is determined.

[0092] It should be noted that the transponder image-assisted railway infrastructure positioning device in the embodiment of the present invention can realize real-time image data acquisition and processing when the train is running at a speed of 400 km / h. This device improves the image acquisition and processing efficiency through the following three aspects:

[0093] (1) This device uses FPGA parallel acquisition technology, edge processing technology, and PCle protocol to realize hardware system construction;

[0094] (2) This device hardware-implements the image compression algorithm to achieve low-latency, high-performance image compression functions, and transmits the compressed data to the image processing module through the PCle protocol.

[0095] (3) The image processing module uses multi-threaded parallel processing technology to realize image decoding and reasoning, and sends the processing results to the task control and management module through the PCle protocol.

[0096] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned transponder image-assisted railway infrastructure positioning method when executing the computer program.

[0097] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the transponder image-assisted railway infrastructure positioning method is implemented.

[0098] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned transponder image-assisted railway infrastructure positioning method is implemented.

[0099] In an embodiment of the present invention, a transponder image captured by a color linear array camera installed on a train is obtained through a high-speed transmission interface, and the time when the transponder image is captured is recorded; wherein the transponder is installed on the roadbed; the transponder image is input into a transponder recognition model to identify the position of the transponder in the transponder image; the transponder recognition model is obtained by training a machine learning model using historical transponder images and the transponder positions in the historical transponder images; the train mileage information corresponding to the time when the transponder image is captured is used as an index identifier of the transponder image, and the first position information of the transponder in the railway line is generated according to the index identifier of the transponder image and the transponder position in the transponder image; the ledger information of the transponders within a preset range from the first position information is obtained, and the ledger information includes the second position information pre-recorded by the transponder; the train mileage information is corrected using the second position information to obtain the positioning result of the train on the railway line. Compared with the method in the prior art in which the on-board equipment directly interacts with the transponder and analyzes the transponder data to achieve mileage positioning, the embodiment of the present invention uses image recognition technology to determine the ledger data of the transponder, and then uses the transponder position information in the ledger data to correct the train running mileage information, thereby solving the problem of reduced positioning accuracy caused by transponder damage or abnormal transponder data in the existing transponder-based mileage positioning method, thereby improving the robustness of the transponder-based mileage positioning method, and correcting the cumulative errors in the train positioning results caused by long-term train operation, thereby realizing the use of transponder images to assist in the positioning of railway infrastructure.

[0100] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0102] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0104] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for locating railway infrastructure using transponder images, characterized in that: include: Through a high-speed transmission interface, the transponder image captured by the color linear array camera installed on the train is obtained, and the transponder image acquisition time is recorded; wherein the transponder is installed on the roadbed; Inputting the transponder image into a transponder recognition model to identify the transponder position in the transponder image; the transponder recognition model is obtained by training a machine learning model using historical transponder images and transponder positions in the historical transponder images; Using the train running mileage information corresponding to the transponder image acquisition time as the index identifier of the transponder image, and generating the first position information of the transponder in the railway line according to the index identifier of the transponder image and the transponder position in the transponder image; Obtaining the ledger information of the transponder within a preset range from the first location information, the ledger information including the second location information pre-recorded by the transponder; The second position information is used to correct the train running mileage information to obtain the positioning result of the train on the railway line.

2. The method according to claim 1, characterized in that The transponder image is acquired by a color linear array camera by photographing the track area illuminated by a light source; the light source is a laser light source and / or an LED light source, which is triggered by a pulse signal generated by a speed encoder installed on the train and illuminates the track area.

3. The method according to claim 1, characterized in that The machine learning model is the YOLOv5 network.

4. The method according to claim 1, characterized in that The transponder image is an RGB image; After identifying the transponder position in the transponder image, further comprising: According to the transponder position in the transponder image, a transponder area in the transponder image is obtained; Convert the transponder image from RGB image to HSV image, and calculate the H channel mean of the transponder area in the HSV image; Delete the HSV images of the transponder area whose H channel mean is greater than the first threshold or less than the second threshold, and retain the HSV images of the transponder area whose H channel mean is not greater than the first threshold and not less than the second threshold; wherein the first threshold is the maximum value of the H channel value of the HSV image of the transponder area, and the second threshold is the minimum value of the H channel value of the HSV image of the transponder area; Generating first position information of the transponder in the railway line according to the index identifier of the transponder image and the transponder position in the transponder image, including: The first position information of the balise in the railway line is generated according to the index identifier of the balise image and the retained HSV image of the balise area.

5. The method according to claim 1, characterized in that Before using the train mileage information corresponding to the transponder image acquisition time as the index identifier of the transponder image, it also includes: determining the train mileage information corresponding to the transponder image acquisition time according to the correspondence between the train running time and the train running mileage.

6. A transponder image-assisted railway infrastructure positioning device, characterized in that: include: An image acquisition module is used to acquire the transponder image acquired by the color linear array camera installed on the train through a high-speed transmission interface, and record the transponder image acquisition time; wherein the transponder is installed on the roadbed; An image processing module, for inputting the transponder image into a transponder recognition model to recognize the transponder position in the transponder image; the transponder recognition model is obtained by training a machine learning model using historical transponder images and transponder positions in the historical transponder images; A transponder position generating module, used to use the train running mileage information corresponding to the transponder image acquisition time as the index identifier of the transponder image, and generate the first position information of the transponder in the railway line according to the transponder position in the transponder image and the index identifier of the transponder image; A ledger data acquisition module, used to obtain ledger information of transponders within a preset range from the first location information, the ledger information including second location information pre-entered by the transponder; The positioning module is used to correct the train running mileage information using the second position information to obtain the positioning result of the train on the railway line.

7. The device according to claim 6, characterized in that The transponder image is acquired by a color linear array camera by photographing the track area illuminated by a light source; the light source is a laser light source and / or an LED light source, which is triggered by a pulse signal generated by a speed encoder installed on the train and illuminates the track area.

8. The device according to claim 6, characterized in that The machine learning model is the YOLOv5 network.

9. The device according to claim 6, characterized in that The transponder image is an RGB image; The image processing module is also used after identifying the transponder position in the transponder image: According to the transponder position in the transponder image, a transponder area in the transponder image is obtained; Convert the transponder image from RGB image to HSV image, and calculate the H channel mean of the transponder area in the HSV image; Delete the HSV images of the transponder area whose H channel mean is greater than the first threshold or less than the second threshold, and retain the HSV images of the transponder area whose H channel mean is not greater than the first threshold and not less than the second threshold; wherein the first threshold is the maximum value of the H channel value of the HSV image of the transponder area, and the second threshold is the minimum value of the H channel value of the HSV image of the transponder area; The transponder position generating module is specifically used to generate the first position information of the transponder in the railway line according to the index identification of the transponder image and the HSV image of the reserved transponder area.

10. The device according to claim 6, characterized in that The transponder position generation module is also used for, before using the train running mileage information corresponding to the transponder image acquisition time as the index mark of the transponder image: According to the correspondence between the train running time and the train running mileage, the train running mileage information corresponding to the transponder image acquisition time is determined.

11. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

13. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images