Initial positioning method, system and storage medium for rail train wake-up process

Abstract

The application provides an initial positioning method, system and storage medium for a rail train wake-up process, which comprises the following steps: obtaining a train wake-up instruction of a target train; obtaining an initial position of the target train based on the train wake-up instruction and through a laser radar; judging whether the target train is in a sleep wake-up track according to the initial position of the train; if the target train is in the sleep wake-up track, judging whether the target train is between two indicator boards through a camera; if the target train is between the two indicator boards, scanning the indicator boards by using the laser radar to obtain distance scanning results; and calculating an accurate position of the target train according to the distance scanning results. The application has the effect that even if an unexpected displacement occurs, the initial position of the train can be accurately obtained when the sleep train is woken up.

Description

Technical Field

[0001] This invention belongs to the field of train control and management technology, specifically relating to an initial positioning method, system, and storage medium for the wake-up process of a rail train. Background Technology

[0002] With the development of rail transit, rail transit control systems have successively undergone manual driving, semi-automatic driving, and fully automatic driving, with their safety and automation levels continuously improving. In communication-based train operation control systems, the train's safe protection distance, movement authorization, running interval, stopping accuracy, and transportation efficiency are all related to the train's real-time location. Especially for fully automatic operation systems (FAO) that achieve GoA3 and GoA4 (where GoA stands for Grade of Automation), the train can only continue the subsequent wake-up process and complete the automatic departure from the depot and entry into operation after the initial positioning is successfully activated.

[0003] For non-fully automated trains, manual power-on and daily inspections are required before departure from the depot. However, this manual operation is labor-intensive, inefficient, and may lead to energy waste and equipment wear. To address this issue, in fully automated driving, the train is controlled by a ground control center (ZC) within the depot, and an automatic power-on wake-up process is performed. Existing train wake-up and positioning methods typically record the train's position via onboard ATP before it goes into sleep mode, and then verify the wake-up position by reading information from the positioning transponder on the track after power-on. Only when the sleep position matches the position shown on the positioning transponder is the train considered successfully woken up and positioned.

[0004] However, the dormant train may experience unexpected collisions during dynamic testing, including collisions with adjacent vehicles or buildings, causing the dormant train to move unexpectedly. Since the dormant train cannot detect unexpected positional movements, the initial positioning of the train obtained when it is woken up based on the memorized position data is inaccurate, which in turn prevents the dormant train from executing the subsequent wake-up process. Summary of the Invention

[0005] This invention provides an initial positioning method, system, and storage medium for the wake-up process of a railway train, in order to solve the problem that the initial positioning of a train may be inaccurate due to unexpected displacement during the hibernation process when waking up a dormant train.

[0006] In a first aspect, the present invention provides an initial positioning method for the wake-up process of a rail train, the method comprising the following steps:

[0007] Obtain the train wake-up command of the target train. The target train is equipped with a sensing system at both the front and rear of the train. The sensing system includes a lidar and a camera device.

[0008] Based on the train wake-up command, the initial position of the target train is obtained using the lidar;

[0009] The initial position of the train determines whether the target train is in the hibernation / wake-up track. There are two signs in the hibernation / wake-up track, and the distance between the two signs is greater than the length of the target train.

[0010] If the target train is within the hibernation / wake-up track, the camera device determines whether the target train is between the two signs.

[0011] If the target train is located between the two signs, the laser radar is used to scan the signs to obtain the distance scan result;

[0012] The precise location of the target train is calculated based on the distance scan results.

[0013] Optionally, obtaining the initial position of the target train based on the train wake-up command and using the lidar includes the following steps:

[0014] The sensing system is powered on and initialized based on the train wake-up command.

[0015] The laser radar scans the surrounding track environment of the target train multiple times, and constructs a target track map based on the obtained three-dimensional point cloud information.

[0016] The target track map is compared with a preset track route map to determine the initial position of the target train.

[0017] Optionally, comparing the target track map with a preset track route map to determine the initial position of the target train includes the following steps:

[0018] The target track map is transmitted to the onboard ATP system of the target train, and the onboard ATP system has a preset track map.

[0019] The onboard ATP system compares the target track map with the track route map to determine the initial position of the target train.

[0020] Optionally, the step of repeatedly scanning the surrounding track environment of the target train using the lidar and constructing a target track map based on the obtained 3D point cloud information includes the following steps:

[0021] The laser radar is used to scan the surrounding track environment of the target train for the first time to obtain initial three-dimensional point cloud information and determine the closed-loop path;

[0022] The initial radar position of the lidar is determined based on the initial 3D point cloud information, and multiple feature points are generated.

[0023] The laser radar continues to scan the surrounding track environment of the target train, and loopback detection is performed based on the closed-loop path to obtain multiple three-dimensional point cloud information.

[0024] Multiple three-dimensional point cloud information are superimposed on the initial three-dimensional point cloud information to obtain target three-dimensional point cloud information, and the initial radar position is corrected according to the position of multiple feature points;

[0025] A target trajectory map is constructed based on the target's three-dimensional point cloud information and the corrected initial radar position.

[0026] Optionally, determining whether the target train is between the two signs using the camera device includes the following steps:

[0027] The camera device captures images of the target train from the front and rear directions.

[0028] Preprocess the images of the front and rear of the vehicle;

[0029] Based on a preset set of sign shape descriptors and using an edge detection algorithm, it is determined whether both the front view image and the rear view image contain sign feature regions.

[0030] If the sign feature area is not present in either the front view image or the rear view image, the output judgment result is that the target train is not between the two signs.

[0031] If the sign feature region exists in both the front view image and the rear view image, then the color feature of the sign feature region is extracted.

[0032] Determine whether the color feature belongs to a preset color feature range;

[0033] If the color feature belongs to the color feature range, the output judgment result is that the target train is between the two signs;

[0034] If the color feature does not belong to the color feature range, the output judgment result is that the target train is not between the two signs.

[0035] Optionally, the distance scanning results include front and rear vehicle scanning results, and the step of using the lidar to scan the sign to obtain the distance scanning results includes the following steps:

[0036] The laser radar located at the front of the target train is used to scan the sign located at the front of the target train to obtain the front scanning result;

[0037] The laser radar located at the rear of the target train is used to scan the sign located at the rear of the target train to obtain the rear scan result.

[0038] Optionally, calculating the precise position of the target train based on the distance scan results includes the following steps:

[0039] The camera device captures images of the target train from the front and rear directions.

[0040] The first sign image in the front view image and the second sign image in the rear view image are extracted using an image recognition model.

[0041] The first sign image and the second sign image are identified using OCR technology, and sign features are added to the first sign image and the second sign image respectively based on the recognition results;

[0042] The precise location of the target train is calculated by combining the features of the signpost and the distance scan results.

[0043] Optionally, the sign features include a front sign and a rear sign, and the precise train position includes the precise front position and the precise rear position. The step of calculating the precise position of the target train by combining the sign features and the distance scanning results includes the following steps:

[0044] Identify the sign features in the first sign image and the second sign image;

[0045] If the sign feature of the first sign image is the front sign, and the sign feature of the second sign image is the rear sign, then the precise position of the front of the vehicle is calculated based on the front scan result, and the precise position of the rear of the vehicle is calculated based on the rear scan result.

[0046] If the sign feature of the first sign image is the rear sign, and the sign feature of the second sign image is the front sign, then the precise position of the rear of the vehicle is calculated based on the front scan result, and the precise position of the front of the vehicle is calculated based on the rear scan result.

[0047] In a second aspect, the present invention also provides an initial positioning system for the wake-up process of a rail train, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect.

[0048] Thirdly, the present invention also provides a computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in the first aspect.

[0049] The beneficial effects of this invention are:

[0050] The initial positioning method for the train wake-up process adopted in this invention includes the following steps: obtaining a train wake-up command for the target train; obtaining the initial position of the target train based on the train wake-up command and using the lidar; determining whether the target train is within the sleep-wake track based on the initial position; if the target train is within the sleep-wake track, determining whether the target train is between two signs using the camera device; if the target train is between two signs, scanning the signs using the lidar to obtain a distance scan result; and calculating the precise position of the target train based on the distance scan result.

[0051] By setting up signs and executing the above steps, the sensing system can scan the signs to obtain the precise location of the target train in real time. Compared with the initial position acquisition method of reading the memory position before hibernation, this method can avoid the problem of inaccurate initial positioning caused by the train's unpredictable displacement during hibernation, which makes the actual position different from the memory position. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the initial positioning method for the wake-up process of a rail train in this invention.

[0053] Figure 2 This is a system structure diagram of the target train sensing system in this invention and a schematic diagram of the location of the indicator sign. Detailed Implementation

[0054] This invention discloses an initial positioning method for the wake-up process of a rail train.

[0055] Reference Figure 1 The initial positioning method for the train wake-up process specifically includes the following steps:

[0056] S101. Obtain the train wake-up command for the target train.

[0057] Among them, reference Figure 2 The target train is equipped with a perception system at both the front and rear. The perception system includes a lidar, a camera, and a perception host. The camera can be a high-definition camera. When the train wake-up command of the target train is received, the perception host will control the lidar and camera to start.

[0058] S102. Based on the train wake-up command, obtain the initial position of the target train using lidar.

[0059] S103. Determine whether the target train is in the hibernation / wake-up track based on the initial position of the train. If the target train is in the hibernation / wake-up track, proceed to step S104.

[0060] If the target train is not within the hibernation / wake-up track, then the train wake-up step should not be executed at this time, and therefore the train wake-up step is terminated. (Refer to...) Figure 2 There are two pre-set signs in the sleep / wake track, and the distance between the two signs is greater than the length of the target train.

[0061] S104. Determine whether the target train is between the two signs using the camera device. If the target train is between the two signs, proceed to step S105.

[0062] If the target train is not located between the two signs, the train wake-up process will be terminated.

[0063] S105. Use a lidar to scan the sign and obtain the distance scan results.

[0064] LiDAR (Light Detection and Ranging) is a sensor that calculates the distance to a target by emitting a laser beam and measuring the time difference between the laser beam and the target. LiDAR emits a laser beam that intersects with a target on a sign. When the laser beam intersects the target, the LiDAR receives the reflected laser beam. By measuring the time difference between the emission and reception of the laser beam, the round-trip time can be calculated. Based on the speed of light and the round-trip time, the distance between the target and the LiDAR can be calculated; this calculated distance is the distance scan result.

[0065] S106. Calculate the precise position of the target train based on the distance scan results.

[0066] The implementation principle of this method is as follows:

[0067] The process involves: obtaining the train wake-up command for the target train; acquiring the initial position of the target train using a lidar sensor based on the wake-up command; determining whether the target train is within the sleep / wake-up track based on the initial position; if the target train is within the sleep / wake-up track, determining whether the target train is between two indicator signs using a camera device; if the target train is between two indicator signs, scanning the indicator signs using a lidar sensor to obtain the distance scan result; and calculating the precise position of the target train based on the distance scan result. By setting up the indicator signs and executing the above steps, the sensing system can scan the indicator signs to achieve real-time precise positioning of the target train. Compared to the initial position acquisition method that relies on reading the memorized position before sleep, this avoids the problem of inaccurate initial positioning caused by unpredictable displacement during sleep, which results in a difference between the actual and memorized positions.

[0068] In one embodiment, step S102, which involves obtaining the initial position of the target train based on the train wake-up command and using lidar, specifically includes the following steps:

[0069] The sensing system is powered on and initialized based on the train wake-up command;

[0070] The target train's surrounding track environment is scanned multiple times using lidar, and a target track map is constructed based on the obtained 3D point cloud information.

[0071] The target track map is compared with the preset track route map to determine the initial position of the target train.

[0072] In this embodiment, refer to Figure 2 The step of comparing the target track map with the preset track route map to determine the initial position of the target train includes the following steps: transmitting the target track map to the onboard ATP system of the target train, which has a preset track route map; comparing the target track map with the track route map through the onboard ATP system, finding the target map area in the track route map that is the same as the target track map, and obtaining the area coordinates of the target map area as the initial position of the target train.

[0073] In one implementation, the process of repeatedly scanning the surrounding track environment of the target train with a lidar and constructing a target track map based on the obtained 3D point cloud information includes the following steps:

[0074] The initial three-dimensional point cloud information was obtained by scanning the surrounding track environment of the target train with lidar and the closed-loop path was determined.

[0075] The initial radar position of the lidar is determined based on the initial 3D point cloud information, and multiple feature points are generated.

[0076] The surrounding track environment of the target train is scanned by lidar, and loop detection is performed based on the closed loop path to obtain multiple three-dimensional point cloud information.

[0077] Multiple 3D point cloud information is superimposed on the initial 3D point cloud information to obtain the target 3D point cloud information, and the initial radar position is corrected according to the position of multiple feature points;

[0078] A target trajectory map is constructed based on the target's 3D point cloud information and the corrected initial radar position.

[0079] In this embodiment, the lidar emits a laser beam, scans the surrounding environment, and receives the reflected laser beam. Based on the round-trip time and speed of light, the distance between the laser beam and objects in the environment is calculated. The distance and scanning angle information are converted into three-dimensional coordinates to form initial three-dimensional point cloud information. Based on object features in the point cloud information, such as tracks and buildings, a closed-loop path is determined. Using the initial three-dimensional point cloud information, the initial lidar position is calculated using a point cloud registration algorithm. Multiple feature points, such as corner points and edge points, are extracted from the point cloud for subsequent loop detection and position correction.

[0080] The lidar continues to scan the environment around the target train, acquiring new 3D point cloud information. Using a loopback detection algorithm, the new point cloud information is matched with the initial point cloud information to determine the closed-loop path. Multiple point cloud information is superimposed on the initial point cloud information to form the target's 3D point cloud information. Using the positional information of feature points, a position correction algorithm is used to correct the initial radar position, improving positioning accuracy. Finally, using the target's 3D point cloud information, a map building algorithm is used to extract information such as the track and obstacles, and a 3D target track map is created based on the continuously accumulated point cloud data. Through these steps, compared to directly accumulating point cloud data and creating a 3D map, the problem of error accumulation can be avoided.

[0081] In one embodiment, step S104, which involves using a camera device to determine whether the target train is located between the two signs, specifically includes the following steps:

[0082] The camera device captures images of the target train from both the front and rear directions.

[0083] Preprocess the images from the front and rear of the vehicle;

[0084] Based on a pre-defined set of sign shape descriptors and using an edge detection algorithm, it is determined whether both the front and rear view images contain sign feature regions.

[0085] If either or both of the images showing the direction of the train's front or rear lack the characteristic region of a sign, the output result is that the target train is not located between the two signs.

[0086] If both the front and rear view images contain signage feature regions, then extract the color features of the signage feature regions.

[0087] Determine whether the color feature belongs to a preset color feature range;

[0088] If the color feature belongs to the color feature range, the output judgment result is that the target train is between the two signs;

[0089] If the color feature does not belong to the color feature range, the output judgment result is that the target train is not between the two signs.

[0090] In this embodiment, both signs are pre-set fixed shapes, such as circular signs, and their surface colors are also pre-set fixed colors, which are more distinctive than the surrounding colors, such as yellow. Therefore, when determining whether a target train is between the two signs using a camera device, simple image processing methods can be used to identify the signs in the image, without the need for complex methods such as image recognition models, which helps improve the computer's processing speed.

[0091] Specifically, the process begins by capturing images of the target train's front and rear directions using a camera. These images are then preprocessed, including denoising and grayscale conversion. Following this preprocessing, an edge detection algorithm extracts all edge shape features. Then, a pre-defined set of sign shape descriptors, corresponding to the fixed shape of the sign, is used to capture the sign's feature regions from these edge features. This set contains shape descriptors for the fixed shape of the sign at multiple angles. The shape descriptors can be Hu moments or Zernike moments, used to describe the shape features of the target object. The shape descriptors capture the outline of the target object, and identification is performed by comparing the similarity between the shape descriptors.

[0092] If both the front and rear images capture the sign's feature area, further confirmation is needed based on the sign's pre-set fixed color to avoid interference from obstacles of the same shape in the images. Specifically, this involves extracting the color of the sign's feature area. Since the sign's pre-set fixed color may vary slightly depending on the light intensity at different angles, a color feature range containing all variations of the fixed color under various lighting conditions needs to be generated beforehand. Then, it's determined whether the color feature belongs to the pre-set color feature range. If the color feature belongs to the color feature range, the target train is between the two signs; otherwise, it is not.

[0093] In one embodiment, the distance scanning results include front and rear scanning results. Step S105, which involves using a lidar to scan the sign to obtain the distance scanning results, specifically includes the following steps:

[0094] The laser radar located at the front of the target train is used to scan the sign located at the front of the target train to obtain the front scanning results;

[0095] The rear of the target train is scanned using a lidar sensor located at the rear of the train to obtain the rear-end scan results.

[0096] In this embodiment, step S106, which calculates the precise position of the target train based on the distance scan results, specifically includes the following steps:

[0097] The camera device captures images of the target train from both the front and rear directions.

[0098] The image recognition model is used to extract the first sign image from the image facing the front of the vehicle and the second sign image from the image facing the rear of the vehicle;

[0099] OCR technology is used to recognize the first sign image and the second sign image, and sign features are added to the first sign image and the second sign image respectively based on the recognition results;

[0100] The precise location of the target train is calculated by combining the characteristics of the signposts and the distance scan results.

[0101] In this embodiment, the specific steps for extracting the first sign image from the front view image and the second sign image from the rear view image using an image recognition model are as follows:

[0102] Collect an image dataset containing signs. Ensure the dataset includes sign images under various scenes, angles, and lighting conditions, and label each image with the sign's location and category.

[0103] Preprocessing is performed on the images from the front and rear of the vehicle, including image resizing and pixel value normalization. Ensure that the input images match the image sizes used during training.

[0104] Download or train a YOLOv4 model, using pre-trained weights as a starting point for faster convergence and improved performance.

[0105] Based on the characteristics of the sign recognition task, the configuration file of the YOLOv4 model needs to be adjusted. This includes adjusting parameters such as the number of categories, anchor boxes, and input image size.

[0106] Train the model using the prepared dataset. Ensure that the input images and labeled data are correctly passed to the model and optimize according to the loss function.

[0107] The model's performance is evaluated using a validation set, and tuning is performed based on the validation results. Hyperparameters, training epochs, and other factors are adjusted to improve the model's performance on the sign recognition task.

[0108] Use a test set to evaluate the performance of the trained model on new data. Calculate metrics such as accuracy and recall to assess the model's effectiveness.

[0109] The images of the front and rear of the vehicle are input into the trained model to obtain the first sign image and the second sign image. Non-maximum suppression (NMS) is then applied to the first and second sign images to remove overlapping detection boxes in the images.

[0110] In this embodiment, the specific steps for using OCR technology to recognize the first sign image and the second sign image, and adding sign features to the first sign image and the second sign image respectively based on the recognition results, are as follows:

[0111] Collect a dataset of images containing sign text. These images can be close-up views of the signs, ensuring the text is clearly visible. Each image needs proper text annotation for supervised learning.

[0112] The collected images undergo preprocessing, including image resizing and pixel value normalization. To improve the model's generalization ability, data augmentation may also be necessary, such as random rotation, translation, and scaling.

[0113] The CRNN (Convolutional Recurrent Neural Network) model was chosen for the text recognition task.

[0114] The OCR model is trained using the prepared dataset. The input is an image, and the label is the corresponding text. The model learns how to recognize and extract text information from images. The training process requires optimization using the OCR model's loss function.

[0115] The performance of the OCR model is evaluated using a validation set, and the model is tuned based on metrics such as recognition accuracy and character error rate. Adjustments to hyperparameters and training epochs may be necessary to improve model performance.

[0116] Use the test set to evaluate the performance of the trained OCR model on new data.

[0117] Post-processing of the text output by the OCR model may include some errors or inaccuracies, such as text correction and removal of meaningless characters, in order to improve the accuracy of the recognition results.

[0118] In the overall process of sign recognition, the first sign image and the second sign image are input into the trained OCR model for text recognition, and sign features are added to the first sign image and the second sign image respectively according to the recognition results to identify whether the sign in the first sign image is a front sign or a rear sign.

[0119] In this embodiment, the sign features include a front sign and a rear sign, and the precise train position includes the precise front position and the precise rear position. The steps for calculating the precise position of the target train by combining the sign features and distance scanning results specifically include the following steps:

[0120] Identify the sign features in the first sign image and the second sign image;

[0121] If the sign feature of the first sign image is a front sign and the sign feature of the second sign image is a rear sign, then the precise position of the front of the vehicle is calculated based on the front scan result, and the precise position of the rear of the vehicle is calculated based on the rear scan result.

[0122] If the sign feature of the first sign image is a rear sign and the sign feature of the second sign image is a front sign, then the precise position of the rear of the vehicle is calculated based on the front scan result, and the precise position of the front of the vehicle is calculated based on the rear scan result.

[0123] In this embodiment, since the precise position of the train includes both the precise position of the front and the precise position of the rear, it is necessary to determine the logical direction of the target train on the map before determining the precise position. Since the two signs are in a fixed state, image recognition and OCR technology can be combined to determine whether the front of the target train and the front sign are on the same side. If the sign features of the first sign image are those of a front sign, it indicates that the front of the target train and the front sign are on the same side. At this point, the front scanning result is converted into point cloud data, and the distance value of each measurement point is converted into coordinates in three-dimensional space. The point cloud data is filtered and denoised to remove invalid or abnormal measurement points. Based on the known position and geometry of the signs, the position and attitude information of the target train's front relative to the front sign are calculated using the valid measurement points in the point cloud data, thus finally determining the precise position of the front. The calculation method for the precise position of the rear is similar.

[0124] The present invention also discloses an initial positioning system for the wake-up process of a rail train, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the initial positioning method for the wake-up process of a rail train as described in any of the above embodiments.

[0125] The implementation principle of this method is as follows:

[0126] The following steps are executed through program invocation: 1) Obtain the train wake-up command for the target train; 2) Based on the wake-up command, obtain the initial position of the target train using lidar; 3) Determine if the target train is within the sleep / wake-up track based on the initial position; 4) If the target train is within the sleep / wake-up track, determine if the target train is between two signs using a camera device; 5) If the target train is between two signs, scan the signs using lidar to obtain distance scan results; 6) Calculate the precise position of the target train based on the distance scan results. By setting up the signs and executing the above steps, the sensing system can scan the signs to achieve real-time precise position acquisition of the target train. Compared to the initial position acquisition method that reads the memorized position before sleep, this avoids the problem of inaccurate initial positioning caused by unpredictable displacement during sleep, resulting in a difference between the actual and memorized positions.

[0127] The present invention also discloses a computer storage medium storing a computer program thereon, wherein when the computer program is executed by a processor, it implements the steps of the initial positioning method for the train wake-up process as described in any of the above embodiments.

[0128] The implementation principle of this method is as follows:

[0129] The following steps are executed through program invocation: 1) Obtain the train wake-up command for the target train; 2) Based on the wake-up command, obtain the initial position of the target train using lidar; 3) Determine if the target train is within the sleep / wake-up track based on the initial position; 4) If the target train is within the sleep / wake-up track, determine if the target train is between two signs using a camera device; 5) If the target train is between two signs, scan the signs using lidar to obtain distance scan results; 6) Calculate the precise position of the target train based on the distance scan results. By setting up the signs and executing the above steps, the sensing system can scan the signs to achieve real-time precise position acquisition of the target train. Compared to the initial position acquisition method that reads the memorized position before sleep, this avoids the problem of inaccurate initial positioning caused by unpredictable displacement during sleep, resulting in a difference between the actual and memorized positions.

[0130] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of protection of this application is limited to these examples; under the concept of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this application as described above, which are not provided in detail for the sake of brevity.

[0131] One or more embodiments in this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments in this application should be included within the protection scope of this application.

Claims

1. A method for initial positioning of a rail vehicle wake-up procedure, characterized by, The method comprises the following steps: obtaining a train wake-up instruction of a target train, the head and tail of the target train being provided with a sensing system, the sensing system comprising a laser radar and a camera device; obtaining an initial position of the target train based on the train wake-up instruction and through the laser radar; judging whether the target train is in a sleep wake-up track according to the initial position of the target train, two indicator plates being preset in the sleep wake-up track, and the distance between the two indicator plates being greater than the length of the target train; if the target train is in the sleep wake-up track, judging whether the target train is between the two indicator plates through the camera device; if the target train is between the two indicator plates, scanning the indicator plates by the laser radar to obtain distance scanning results; calculating an accurate position of the target train according to the distance scanning results; the step of obtaining the initial position of the target train based on the train wake-up instruction and through the laser radar comprises the following steps: powering on and initializing the sensing system based on the train wake-up instruction; scanning the surrounding track environment of the target train multiple times through the laser radar, and constructing a target track map according to the three-dimensional point cloud information obtained by scanning; comparing the target track map with a preset track line map to determine the initial position of the target train.

2. The method of initial positioning of a railcar wake-up process of claim 1, wherein, the step of comparing the target track map with the preset track line map to determine the initial position of the target train comprises the following steps: transmitting the target track map to an on-board ATP system of the target train, the on-board ATP system being provided with a track line map; comparing the target track map with the track line map through the on-board ATP system to determine the initial position of the target train.

3. The method of initial positioning of a railcar wake-up process of claim 1, wherein, the step of scanning the surrounding track environment of the target train multiple times through the laser radar and constructing a target track map according to the three-dimensional point cloud information obtained by scanning comprises the following steps: scanning the surrounding track environment of the target train for the first time through the laser radar to obtain initial three-dimensional point cloud information and determine a closed loop path; determining an initial radar position of the laser radar based on the initial three-dimensional point cloud information, and generating a plurality of feature points; continuing to scan the surrounding track environment of the target train through the laser radar, and performing loop detection based on the closed loop path to obtain a plurality of three-dimensional point cloud information; superimposing the plurality of three-dimensional point cloud information on the initial three-dimensional point cloud information to obtain target three-dimensional point cloud information, and correcting the initial radar position according to the positions of the plurality of feature points; constructing a target track map based on the target three-dimensional point cloud information and according to the corrected initial radar position.

4. The method of initial positioning of a railcar wake-up process of claim 1, wherein, the step of judging whether the target train is between the two indicator plates through the camera device comprises the following steps: capturing a head direction picture and a tail direction picture of the target train through the camera device; preprocessing the head direction picture and the tail direction picture; The preset sign shape descriptor set and an edge detection algorithm are used to determine whether the sign feature region exists in the front direction picture and the rear direction picture of the train; If the sign feature region does not exist in any one or both of the front direction picture and the rear direction picture, a determination result is output that the target train is not between the two signs; If the sign feature region exists in the front direction picture and the rear direction picture, the color feature of the sign feature region is extracted; It is determined whether the color feature belongs to a preset color feature interval; If the color feature belongs to the color feature interval, a determination result is output that the target train is between the two signs; If the color feature does not belong to the color feature interval, a determination result is output that the target train is not between the two signs.

5. The method for initial positioning of the railcar wake-up process of claim 1, wherein, The distance scanning result includes a front scanning result and a rear scanning result, and the laser radar is used to scan the sign to obtain the distance scanning result, including the following steps: The laser radar arranged at the front of the target train is used to scan the sign at the front of the target train to obtain the front scanning result; The laser radar arranged at the rear of the target train is used to scan the sign at the rear of the target train to obtain the rear scanning result.

6. The method of initial positioning of a railcar wake-up process of claim 5, wherein, The train accurate position of the target train is calculated according to the distance scanning result, including the following steps: The front direction picture and the rear direction picture of the target train are captured by the camera device; An image recognition model is used to extract a first sign image in the front direction picture and a second sign image in the rear direction picture; An OCR technology is used to recognize the first sign image and the second sign image, and sign features are added to the first sign image and the second sign image according to the recognition results; The train accurate position of the target train is calculated by combining the sign features and the distance scanning result.

7. The method of initial positioning of a railcar wake-up process of claim 6, wherein, The sign features include a front sign and a rear sign, the train accurate position includes a front accurate position and a rear accurate position, and the train accurate position of the target train is calculated by combining the sign features and the distance scanning result, including the following steps: The sign features of the first sign image and the second sign image are recognized; If the sign feature of the first sign image is the front sign and the sign feature of the second sign image is the rear sign, the front accurate position is calculated according to the front scanning result, and the rear accurate position is calculated according to the rear scanning result; If the sign feature of the first sign image is the rear sign and the sign feature of the second sign image is the front sign, the rear accurate position is calculated according to the front scanning result, and the front accurate position is calculated according to the rear scanning result.

8. An initial positioning system for a rail vehicle wake-up procedure, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The computer program is executed by the processor to implement the method of any one of claims 1 to 7.

9. A computer storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 7.

Images