Method, system, device and medium for identifying obstacles in the track area ahead of a train

Abstract

The present invention belongs to the field of train control technology, and in particular, relates to a method, system, device and medium for identifying obstacles in a track area ahead of a train. The method comprises: collecting first visual data of a track area ahead of a train in a train operation line, and determining obstacle-free visual data of the track area ahead of the train; collecting second visual data of the track area ahead of the train, and using the obstacle-free visual data as a reference standard to identify obstacle data in the second visual data; using a machine learning algorithm to train sample data consisting of obstacle-free visual data and obstacle data to obtain an obstacle recognition model ahead of the train; inputting the collected real-time visual data of the track area ahead of the train into the obstacle recognition model for recognition, and completing train braking according to the recognition result. The technical solution of the present invention can assist the driver in identifying obstacles in the track area and improve the reliability and safety of rail transit operations.

Description

Technical Field

[0001] The present invention belongs to the technical field of train operation control, and in particular relates to a method, system, equipment and medium for identifying obstacles in a track area ahead of a train. Background Art

[0002] The onboard ATP (Automatic Train Protection) equipment in the high-speed railway CTCS-2 / 3 train control system is the core control equipment and plays a huge role in the running safety of trains. The onboard equipment generates a continuous speed control curve and monitors the safe operation of trains by receiving the movement authorization information from the ground signal control equipment.

[0003] The ground signal control system mainly calculates the movement authorization based on the idle state of the block section, the interlocking route and the switch state. The mainstream method for detecting the idle state of the section is to use the track circuit for detection. However, in actual use, the track circuit is prone to poor branching due to rust on the track surface or dust pollution. At the same time, it does not have the ability to detect some large non-metallic objects such as fallen rocks. If a foreign object intrudes in front of the section when the train runs to a block section, the idle state detection method of the ground signal control system can no longer distinguish, and in extreme cases, it is easy to cause driving risks. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a method, system, device and medium for identifying obstacles in the track area ahead of a train.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a method for identifying obstacles in a track area ahead of a train, which is characterized by comprising:

[0007] Collecting the first visual data of the track area ahead of the train on the train running line;

[0008] Determining obstacle-free visual data of a track area ahead of the train based on the first visual data;

[0009] Collecting second visual data of a track area ahead of the train in the train running line, using the obstacle-free visual data as a reference standard, and identifying obstacle data in the second visual data;

[0010] The obstacle-free visual data and obstacle data of the track area ahead of the train are used as sample data, and the sample data are trained using a machine learning algorithm to obtain an obstacle recognition model ahead of the train.

[0011] The collected real-time visual data of the track area ahead of the train is input into the obstacle recognition model ahead of the train for identification, and the train braking is completed according to the identification results.

[0012] Further, based on the first visual data, determining the obstacle-free visual data of the track area ahead of the train includes:

[0013] Based on the point cloud data collected by the laser radar in the first vision data, traverse each point cloud in the point cloud data and calculate the vertical angle of the scanning ray where the point cloud is located;

[0014] Taking the front of the train as the axis, the horizontal angle of the point cloud is calculated in a clockwise direction.

[0015] Based on the horizontal angle of the point cloud, traverse each point cloud in the point cloud data in turn, obtain the index of each point cloud on the laser radar scanning circle, and organize the point cloud data in order according to the index of each point cloud;

[0016] Based on the orderly organized point cloud data combined with the installation position height of the laser radar and the vertical angle of the scanning ray where the point cloud is located, the expected distance from the point cloud on the laser radar scanning circle to the central axis of the laser radar is calculated;

[0017] The horizontal distance and the expected distance difference of the point cloud between adjacent scanning rays are calculated, and the visual data including the horizontal distance and the expected distance difference are determined as the obstacle-free visual data of the track area ahead of the train.

[0018] Furthermore, the calculation formula for calculating the vertical angle of the scanning ray where the point cloud is located is:

[0019]

[0020] In the formula, For a point cloud The coordinates of is the vertical angle of the i-th scanning ray.

[0021] Furthermore, the calculation formula for calculating the horizontal angle of the point cloud is:

[0022]

[0023] In the formula, Point Cloud Projection coordinates on the XOY plane, Point Cloud The horizontal angle, is an integer.

[0024] Furthermore, the calculation formula for obtaining the index of each point cloud on the laser radar scanning circle is:

[0025]

[0026] In the formula, Point Cloud The index on the scan circle, Point Cloud The horizontal angle, is the floor function.

[0027] Furthermore, the expected distance from the point cloud on the laser radar scanning circle to the central axis of the laser radar is calculated, including:

[0028]

[0029] In the formula, is the expected distance from the point cloud on the i-th scanning ray to the center of the laser radar, H is the height of the sensor installation position, is the vertical angle of the i-th scanning ray.

[0030] Furthermore, the horizontal distance of the point cloud between adjacent scanning rays and the expected distance difference are calculated, where the calculation formula of the horizontal distance is:

[0031]

[0032] In the formula, They are , The horizontal and vertical coordinates of the point cloud;

[0033] The expected distance difference is calculated as:

[0034]

[0035] In the formula, They are , The vertical angle of the ray, H is the height of the sensor installation location.

[0036] Furthermore, the objectives of the obstacle recognition model ahead of the train are:

[0037] Obstacles and obstacle classifications, as well as obstacle location information, are identified from visual data of a track area ahead of a train on a train running line.

[0038] Further, train braking is completed according to the recognition result, including:

[0039] According to the classification information and location information of the obstacles in the obstacle identification results, the train braking command is determined, and the train braking command is executed to complete the train braking.

[0040] In a second aspect, the present invention further provides a system for identifying obstacles in a track area ahead of a train, which is characterized by comprising:

[0041] A collection module, used for collecting first visual data of a track area ahead of a train in a train running line;

[0042] A determination module, configured to determine obstacle-free visual data of a track area ahead of the train based on the first visual data;

[0043] The recognition module is used to collect the second visual data of the track area ahead of the train in the train running line, and use the obstacle-free visual data as a reference standard to identify the obstacle data in the second visual data;

[0044] A training module is used to use the obstacle-free visual data and obstacle data of the track area ahead of the train as sample data, and use a machine learning algorithm to train the sample data to obtain an obstacle recognition model ahead of the train;

[0045] The braking module is used to input the collected real-time visual data of the track area ahead of the train into the obstacle recognition model ahead of the train for identification, and complete the train braking according to the identification results.

[0046] Furthermore, the module is determined to be further used for:

[0047] Based on the point cloud data collected by the laser radar in the first vision data, traverse each point cloud in the point cloud data and calculate the vertical angle of the scanning ray where the point cloud is located;

[0048] Taking the front of the train as the axis, the horizontal angle of the point cloud is calculated in a clockwise direction.

[0049] Based on the horizontal angle of the point cloud, traverse each point cloud in the point cloud data in turn, obtain the index of each point cloud on the laser radar scanning circle, and organize the point cloud data in order according to the index of each point cloud;

[0050] Based on the orderly organized point cloud data combined with the installation position height of the laser radar and the vertical angle of the scanning ray where the point cloud is located, the expected distance from the point cloud on the laser radar scanning circle to the central axis of the laser radar is calculated;

[0051] The horizontal distance and the expected distance difference of the point cloud between adjacent scanning rays are calculated, and the visual data including the horizontal distance and the expected distance difference are determined as the obstacle-free visual data of the track area ahead of the train.

[0052] Furthermore, the calculation formula for calculating the vertical angle of the scanning ray where the point cloud is located is:

[0053]

[0054] In the formula, For a point cloud The coordinates of is the vertical angle of the i-th scanning ray.

[0055] Furthermore, the calculation formula for calculating the horizontal angle of the point cloud is:

[0056]

[0057] In the formula, Point Cloud Projection coordinates on the XOY plane, Point Cloud The horizontal angle, is an integer.

[0058] Furthermore, the calculation formula for obtaining the index of each point cloud on the laser radar scanning circle is:

[0059]

[0060] In the formula, Point Cloud The index on the scan circle, Point Cloud The horizontal angle, is the floor function.

[0061] Furthermore, the expected distance from the point cloud on the laser radar scanning circle to the central axis of the laser radar is calculated, including:

[0062]

[0063] In the formula, is the expected distance from the point cloud on the i-th scanning ray to the center of the laser radar, H is the height of the sensor installation position, is the vertical angle of the i-th scanning ray.

[0064] Furthermore, the horizontal distance of the point cloud between adjacent scanning rays and the expected distance difference are calculated, where the calculation formula of the horizontal distance is:

[0065]

[0066] In the formula, They are , The horizontal and vertical coordinates of the point cloud;

[0067] The expected distance difference is calculated as:

[0068]

[0069] In the formula, They are , The vertical angle of the ray, H is the height of the sensor installation location.

[0070] Furthermore, the objectives of the obstacle recognition model ahead of the train are:

[0071] Obstacles and obstacle classifications, as well as obstacle location information, are identified from visual data of a track area ahead of a train on a train running line.

[0072] Further, train braking is completed according to the recognition result, including:

[0073] According to the classification information and location information of the obstacles in the obstacle identification results, the train braking command is determined, and the train braking command is executed to complete the train braking.

[0074] In a third aspect, the present invention further provides an electronic device, comprising: a processor and a memory;

[0075] The processor is coupled to the memory;

[0076] The processor is used to read and execute the program or instructions stored in the memory, so that the device executes the method of the first aspect.

[0077] In a fourth aspect, the present invention further provides a computer-readable storage medium storing a computer program, which implements the method of the first aspect when executed by a processor.

[0078] In summary, the technical solution provided by the present invention has at least the following technical effects or advantages:

[0079] The technical solution of the present invention utilizes computer vision technology combined with ATP on-board equipment to detect active obstacles in rail transit, which can assist the driver in observing obstacles in the track area. When foreign objects are detected, the driver is alarmed in real time, and the type and distance of the identified foreign objects are detected, and a braking request is output to the ATP system, so as to improve the reliability and safety of rail transit line operations and reduce operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0081] Figure 1A schematic diagram of a method for identifying obstacles in a track area ahead of a train provided by an embodiment of the present invention;

[0082] Figure 2 A schematic diagram of an obstacle recognition system for a track area ahead of a train provided by an embodiment of the present invention;

[0083] Figure 3 It is a structural schematic diagram of installing a visual detection device on an accumulator in an embodiment of the present invention;

[0084] Figure 4 A schematic diagram of the connection relationship between the visual detection device and the ATP vehicle-mounted device in an embodiment of the present invention;

[0085] Figure 5 It is an information flow diagram among the vehicle-mounted ATP, the visual detection device and the vehicle in an embodiment of the present invention;

[0086] Figure 6 A schematic structural diagram of an electronic device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0088] In recent years, urban rail transit construction has been in full swing, my country's high-speed railway network has been fully laid, and intercity railways and urban express rails have developed rapidly. With the vigorous development of new-generation information technologies such as artificial intelligence and cloud computing, computer vision, lidar detection and other intelligent perception technologies have been introduced into the field of rail transit operation control, which has had a profound impact on the national transportation experience. Therefore, the intelligence of rail transit has become the main focus of the transportation field.

[0089] Active obstacle detection in rail transit, which combines computer vision technology with ATP on-board equipment, can provide warning information about obstacles in the track area ahead of the train that may affect driving safety. It can also help train drivers to be reminded of the situation ahead when they make misjudgments in visual observation due to weather or driving fatigue, helping to improve the reliability and safety of railway operations.

[0090] The technical solution of the present invention is achieved by installing a visual detection device on the train and maintaining two-way communication between the visual detection device and the station ATP device. The visual detection device includes: a visual detection host and a sensor acquisition device. The on-board ATP device and the visual detection host are connected through a network (wired network or wireless network) and communicate based on the Ethernet (transport layer) protocol. The on-board visual detection host is connected to the sensor acquisition device through a network and data communication, and the external information obtained by the sensor acquisition device is processed to detect obstacles such as vehicles, pedestrians, large animals, etc. that intrude into the dynamic limits of the train. The visual detection host outputs a braking command to the ATP based on the detection results, and the ATP outputs a braking command to the train to complete the braking of the train.

[0091] Figure 1 A schematic diagram of a method for identifying obstacles in a track area ahead of a train provided by an embodiment of the present invention, as shown in the figure, the method includes:

[0092] S101, collecting first visual data of a track area ahead of a train on a train running line;

[0093] Exemplarily, the first visual data of the track area ahead of the train in the train running line is collected by a sensor collection device, the sensor collection device includes a laser radar, and the laser radar collects point cloud data. The collected point cloud data is the first visual data ahead of the train in the train running line, and the point cloud data includes multiple point clouds.

[0094] S102, determining obstacle-free visual data of the track area ahead of the train based on the point cloud data collected by the laser radar in the first visual data;

[0095] The specific steps include:

[0096] (1) Based on the point cloud data collected by the laser radar in the first vision data, traverse each point cloud in the point cloud data and calculate the vertical angle of the scanning ray where each point cloud is located. The calculation formula is:

[0097] (1)

[0098] In the formula, For a point cloud The coordinates of is the vertical angle of the i-th scanning ray;

[0099] (2) Taking the front of the train as the axis, the horizontal angle of the axis is set to 0°, and the horizontal angle of each point cloud in the point cloud data is calculated in a clockwise direction, and the point cloud is stored according to the horizontal angle resolution. The calculation formula of the horizontal angle is:

[0100] (2)

[0101] In the formula, Point Cloud Projection coordinates on the XOY plane, Point Cloud The horizontal angle, is an integer.

[0102] Horizontal angle resolution refers to the angle difference between two adjacent point clouds in the point cloud data on a horizontal plane. When storing point clouds, the 3D coordinates of each point cloud are recorded, as well as other information contained, such as color, intensity, and other general information of the point cloud data collected by the LiDAR. The stored point clouds can be easily arranged and processed in an orderly manner later.

[0103] (3) Traverse each point cloud in the point cloud data in turn, obtain the index of each point cloud on the laser radar scanning circle, organize the point cloud data in order according to the index of each point cloud, and calculate the index formula as follows:

[0104] (3)

[0105] In the formula, Point Cloud The index on the scan circle, Point Cloud The horizontal angle, is the floor function.

[0106] Ordered organization means storing each point cloud in the point cloud data in a double-layer storage container according to the horizontal angle and vertical angle, and organizing the point clouds in the point cloud data into an ordered structure according to the index of each point cloud, making data access and query more efficient, especially when performing spatial query and neighborhood search.

[0107] (4) Based on the orderly organized point cloud data combined with the installation height of the lidar and the vertical angle of the scanning ray where the point cloud is located, the expected distance from the point cloud on the lidar scanning circle to the central axis of the lidar is calculated. The calculation formula is:

[0108] (4)

[0109] In the formula, is the expected distance from the point cloud on the i-th scanning ray to the center of the laser radar, H is the height of the sensor installation position, is the vertical angle of the i-th scanning ray.

[0110] The centerline axis of the LiDAR refers to the line perpendicular to the ground where the LiDAR sensor is installed. It is used as a reference for position to facilitate the calculation of the position relationship of the point cloud in the point cloud data under this reference. The expected distance refers to the distance from the ground point cloud to the centerline axis of the LiDAR.

[0111] (5) Calculate the horizontal distance of the point cloud between adjacent scanning rays and the expected distance difference. The calculation formula for the horizontal distance is:

[0112] (5)

[0113] In the formula, They are , The horizontal and vertical coordinates of the point cloud.

[0114] The expected distance difference is calculated as:

[0115] (6)

[0116] In the formula, They are , The vertical angle of the ray, H is the height of the sensor installation location.

[0117] Adjacent scanning rays refer to the scanning rays formed by two adjacent transmissions and receptions of laser signals during the continuous scanning process of the laser radar.

[0118] Through the above calculation, the horizontal distance and expected distance difference corresponding to all point cloud data in the first visual data can be obtained, and then the obstacle-free visual data corresponding to the ground information of the track area in front of the train (the area in front of the train) when there are no obstacles can be obtained, that is, the visual data of the track area in front of the train when there are no obstacles, including the horizontal distance and the expected distance difference. The values ​​of the horizontal distance and the expected distance difference corresponding to all point clouds in the obstacle-free visual data are fixed.

[0119] S103, collecting second visual data of a track area ahead of the train in the train running line, using the obstacle-free visual data as a reference standard, and identifying obstacle data in the second visual data;

[0120] The laser radar in the sensor acquisition device collects point cloud data of the track area in front of the train in the train running line as the second visual data;

[0121] Taking the obstacle-free visual data as the reference standard, if the horizontal distance and expected distance difference of the point cloud in the point cloud data in the second visual data of the track area in front of the train are inconsistent with the horizontal distance and expected distance difference of the point cloud in the obstacle-free visual data, that is, if the standard of obstacle-free visual data is not met, then the point cloud is considered to be an obstacle point cloud, and the distribution of all obstacle point clouds in front of the train is obtained by traversing and checking in sequence, and then the location information of the obstacle can be determined. The visual data corresponding to the obstacle point cloud, including the horizontal distance and the expected distance difference, is the obstacle data.

[0122] S104, using the obstacle-free visual data and obstacle data of the track area ahead of the train as sample data, and using a machine learning algorithm to train the sample data to obtain an obstacle recognition model ahead of the train;

[0123] The goal of the obstacle recognition model ahead of the train is to identify obstacles and obstacle classification, as well as obstacle location information, from the visual data of the track area ahead of the train on the train running line.

[0124] In order to more accurately detect obstacles ahead of the train and filter out interference from some non-obstacles, when constructing sample data based on the obstacle-free visual data and obstacle data of the track area ahead of the train, it is also necessary to consider interference factors such as natural dimensions, external dimensions, and target dimensions. That is, when collecting visual data of the track area ahead of the train, the data is collected under the conditions of multiple interference factors. Natural dimensions include: weather, lighting, visibility, etc. External dimensions: traffic environment such as platform edge features, traffic signs such as signal styles, etc. Target dimensions: size, shape, style of the tail of the train ahead, pedestrians, etc. The obstacle-free visual data and obstacle data of the track area ahead of the train taking into account the above-mentioned interference factors are constructed as sample data. The obstacle recognition model ahead of the train is trained with the sample data to improve the recognition accuracy and speed of the model.

[0125] S105: input the collected real-time visual data of the track area ahead of the train into an obstacle recognition model ahead of the train for recognition, and complete train braking according to the recognition result.

[0126] During the train's travel, the visual detection equipment collects real-time visual data ahead of the train, and inputs the real-time visual data into the above-mentioned obstacle recognition model for detection and recognition. If an obstacle is detected ahead of the train, an obstacle warning is issued to the driver based on the obstacle classification information and location information. If the driver determines to take braking measures after confirmation, a braking request message is sent to the visual detection equipment, and the visual detection equipment outputs a braking command to the onboard ATP, which outputs the braking to the train to complete the train braking. In some cases, if an obstacle is detected ahead of the train, the visual detection equipment can also directly output a braking command to the onboard ATP, which outputs the braking to the train to complete the train braking.

[0127] Figure 2 A schematic diagram of an obstacle recognition system for a track area ahead of a train provided by an embodiment of the present invention. As shown in the figure, the system includes:

[0128] A collection module, used for collecting first visual data of a track area ahead of a train in a train running line;

[0129] A determination module, configured to determine obstacle-free visual data of a track area ahead of the train based on the first visual data;

[0130] The recognition module is used to collect the second visual data of the track area ahead of the train in the train running line, and use the obstacle-free visual data as a reference standard to identify the obstacle data in the second visual data;

[0131] A training module is used to use the obstacle-free visual data and obstacle data of the track area ahead of the train as sample data, and use a machine learning algorithm to train the sample data to obtain an obstacle recognition model ahead of the train;

[0132] The braking module is used to input the collected real-time visual data of the track area ahead of the train into the obstacle recognition model ahead of the train for identification, and complete the train braking according to the identification results.

[0133] It should be noted that, for the sake of convenience, Figure 2 For example, only the main modules of the obstacle identification system structure in the track area ahead of the train are shown. In actual applications, the system may also include modules or components not shown in the figure; the system is not limited to the above module structure, and may also be other module structures that implement the above method embodiments.

[0134] For example, Figure 3 This is a schematic diagram of the structure of installing visual detection equipment on a train, as shown in the figure.

[0135] The structure of the visual detection equipment includes:

[0136] 1) Both ends of the train share a visual detection host.

[0137] 2) The acquisition sensors include visual sensors and laser radars, which are installed on the inside of the driver's cab windshield to detect the line ahead and collect visual data of the track area ahead. It has a built-in camera and a laser radar. The visual sensor assists the laser radar to complete the visual data acquisition.

[0138] 3) The display screen is used to display system interface information, camera images, visual data collected by the lidar, and warning information of obstacles, and provide interactive operations for confirming warnings.

[0139] The interactive communication scheme between the visual detection host of the visual detection equipment and the on-board ATP is as follows:

[0140] 1. Communication mechanism description

[0141] According to the connection relationship Figure 4 , a two-way communication is established between the visual detection host and the on-board ATP, point cloud to point cloud communication, which are redundant channels for each other.

[0142] 2. Communication status management

[0143] The visual detection host / vehicle ATP should perform a legitimacy check on the application message received from the other party. If the check fails, it is considered that the application message of the peer party has not been received in this cycle or the packet is discarded. The specific inspection method is as follows:

[0144] (1) Message content consistency check: including information field legitimacy check (all values ​​except the specified field values ​​are illegal), field combination legitimacy check, and information integrity check.

[0145] (2) Check the integrity of general information.

[0146] (3) Outlier protection logic check.

[0147] (4) The visual detection host and the on-board ATP should be able to judge the communication connection status respectively:

[0148] 1) The timeout period during which the visual detection host considers that the communication with the vehicle ATP is interrupted is defined as T-Visual Detection Host Timeout.

[0149] 2) The timeout period during which the vehicle-mounted ATP considers that the communication with the visual detection host is interrupted is defined as T-ATPTimeout.

[0150] 3) If the visual detection host does not receive any message from the vehicle-mounted ATP dual-channel within the T-visual detection host Timeout period, the visual detection host should consider that the communication with the vehicle-mounted ATP is interrupted.

[0151] 4) If the on-board ATP does not receive any message from the dual-channel of the visual detection host within the T-ATPTimeout time, the on-board ATP should consider that the communication with the visual detection host is interrupted.

[0152] 5) If the visual detection host determines that the delay in receiving the application information of the on-board ATP has reached T-visual detection host Timeout, the visual detection host should discard this information packet or consider that the communication with the on-board ATP is interrupted.

[0153] 6) If the on-board ATP determines that the delay in receiving the application information from the visual detection host has reached T-ATPTimeout, the on-board ATP should discard this information packet or consider that the communication with the visual detection host is interrupted.

[0154] 3. Communication fault handling

[0155] The communication fault handling at the application layer of the vehicle-ground wireless communication (visual detection host-vehicle ATP) is divided into the following situations:

[0156] (1) The visual detection host and the on-board ATP directly discard the received duplicate, reversed, or damaged application information.

[0157] (2) Handling in the event of communication delay: If the visual detection host / on-board ATP determines that the received application information is delayed, safety-side processing should be adopted to manually ensure the identification of obstacles ahead.

[0158] (3) When the communication between the on-board ATP and the visual detection host is interrupted, the on-board ATP should display an alarm message on the display screen to alert the driver.

[0159] 4. Communication interaction process

[0160] as follows Figure 5 As shown, the information flow diagram between the on-board ATP, visual detection equipment and the vehicle.

[0161] The display screen is mainly used for the visual detection device to provide display information to the driver. The on-board ATP device and the visual detection device must first complete the power-on self-test to ensure that their own equipment is working properly. The visual detection device initiates a connection request. After receiving the connection request, the on-board ATP device establishes Ethernet communication with the visual detection device and sends the current working mode of the on-board ATP to the visual detection device, including but not limited to the train running direction, handle direction, current operating mode, driver-side activation signal, etc. The visual detection device combines the information sent by the on-board ATP and its own status to display the corresponding content to the driver through the display screen. During the operation of the equipment, if an obstacle is detected in front, an obstacle warning is issued to the driver based on the identified obstacle classification information. If the driver judges and takes braking measures, a braking request information is sent to the visual detection device, and the visual detection device outputs a braking command to the on-board ATP, and the on-board ATP outputs a braking command to the train to complete the train braking.

[0162] Based on the above disclosed content, the present invention also provides an electronic device. Figure 6 As shown, the electronic device of an embodiment of the present invention includes at least one electrically connected processor and at least one storage medium, wherein the storage medium is electrically connected to the processor, wherein the storage medium stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method as described above.

[0163] Based on the same inventive concept, the present invention also provides a storage medium, which stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method as described above.

[0164] It should be noted that the electrical connection between the above-mentioned units does not necessarily mean the connection between lines. An indirect connection method can be applied to the embodiments of the present invention as long as the purpose of the present invention is achieved.

[0165] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying obstacles in the track area ahead of a train. Its characteristics include: Collecting the first visual data of the track area ahead of the train on the train running line; Determine obstacle-free visual data of the track area ahead of the train based on the first visual data, wherein the horizontal distances corresponding to all point clouds in the obstacle-free visual data and the values ​​of the expected distance differences are fixed; Collecting the second visual data of the track area ahead of the train in the train running line, taking the obstacle-free visual data as a reference standard, and determining the obstacle data in the second visual data by judging the consistency between the point cloud data in the second visual data and the point cloud data in the obstacle-free visual data; The obstacle-free visual data and obstacle data of the track area ahead of the train are used as sample data, and the sample data are trained using a machine learning algorithm to obtain an obstacle recognition model ahead of the train. The collected real-time visual data of the track area ahead of the train is input into the obstacle recognition model ahead of the train for recognition, and the train braking command is determined based on the classification information and location information of the obstacles in the recognition results, and the train braking command is executed to complete the train braking; The first visual data, the second visual data, and the real-time visual data are acquired through the visual detection equipment installed on the train, and the visual detection equipment includes: a visual detection host and a sensor acquisition device; the visual detection host and the on-board ATP device are connected through a network and communicate based on the Ethernet protocol; the visual detection host and the sensor acquisition device are connected to the network and communicate data; the communication between the visual detection host and the on-board ATP device is two-way communication, and the communication connection status between the visual detection host and the on-board ATP device is judged during communication, including: defining the timeout time for the communication interruption between the visual detection host and the on-board ATP device as T-visual detection host Timeout; defining the timeout time for the communication interruption between the on-board ATP device and the visual detection host as T-ATPTimeout; if the If the visual detection host does not receive any message from the two-way channel of the on-board ATP device within the T-visual detection host Timeout time, it is determined that the communication between the on-board ATP device and the on-board ATP device is interrupted; if the on-board ATP device does not receive any message from the two-way channel of the visual detection host within the T-ATPTimeout time, it is determined that the communication between the on-board ATP device and the visual detection host is interrupted; if the delay of the on-board ATP device receiving the message from the visual detection host reaches the T-ATPTimeout time, the on-board ATP device discards the message of the visual detection host and determines that the communication between the on-board ATP device and the visual detection host is interrupted; when the communication between the on-board ATP device and the visual detection host is interrupted, the on-board ATP device prompts an alarm message; The determining, based on the first visual data, the obstacle-free visual data of the track area ahead of the train includes: Based on the point cloud data collected by the laser radar in the first vision data, traverse each point cloud in the point cloud data and calculate the vertical angle of the scanning ray where the point cloud is located; Taking the front of the train as the axis, the horizontal angle of the point cloud is calculated in a clockwise direction. Based on the horizontal angle of the point cloud, traverse each point cloud in the point cloud data in turn, obtain the index of each point cloud on the laser radar scanning circle, and organize the point cloud data in order according to the index of each point cloud; Based on the orderly organized point cloud data combined with the installation height of the laser radar and the vertical angle of the scanning ray where the point cloud is located, the expected distance from the point cloud on the laser radar scanning circle to the central axis of the laser radar is calculated; The horizontal distance and the expected distance difference of the point cloud between adjacent scanning rays are calculated, and the visual data including the horizontal distance and the expected distance difference are determined as the obstacle-free visual data of the track area ahead of the train.

2. The method for identifying obstacles in the track area ahead of a train according to claim 1, characterized in that: The calculation formula for the vertical angle of the scanning ray where the calculation point cloud is located is: In the formula, For a point cloud The coordinates of is the vertical angle of the i-th scanning ray.

3. The method for identifying obstacles in the track area ahead of a train according to claim 1, characterized in that: The calculation formula for the horizontal angle of the calculation point cloud is: In the formula, Point Cloud Projection coordinates on the XOY plane, Point Cloud The horizontal angle, is an integer.

4. The method for identifying obstacles in the track area ahead of a train according to claim 1, characterized in that: The calculation formula for obtaining the index of each point cloud on the laser radar scanning circle is: In the formula, Point Cloud The index on the scan circle, Point Cloud The horizontal angle, is the floor function.

5. The method for identifying obstacles in the track area ahead of a train according to claim 1, characterized in that: The step of calculating the expected distance from the point cloud on the laser radar scanning circle to the central axis of the laser radar includes: In the formula, is the expected distance from the point cloud on the i-th scanning ray to the center of the laser radar, H is the height of the sensor installation position, is the vertical angle of the i-th scanning ray.

6. The method for identifying obstacles in the track area ahead of a train according to claim 1, characterized in that: The horizontal distance and expected distance difference of the point cloud between adjacent scanning rays are calculated, wherein the calculation formula of the horizontal distance is: In the formula, They are , The horizontal and vertical coordinates of the point cloud; The expected distance difference is calculated as: In the formula, They are , The vertical angle of the ray, H The height of the sensor installation location.

7. The method for identifying obstacles in the track area ahead of a train according to claim 1, characterized in that: The objectives of the obstacle recognition model ahead of the train are: Obstacles and obstacle classifications, as well as obstacle location information, are identified from visual data of a track area ahead of a train on a train running line.

8. A system for identifying obstacles in the track area ahead of a train. Its characteristics are that it includes: visual detection equipment and on-board ATP equipment installed on a train, the visual detection equipment includes: a visual detection host and a sensor acquisition device; the visual detection host and the on-board ATP equipment are connected through a network and communicate based on the Ethernet protocol; the visual detection host and the sensor acquisition device are connected to the network and communicate data; the communication between the visual detection host and the on-board ATP equipment is two-way communication, and the communication connection status between the visual detection host and the on-board ATP equipment is judged during communication, including: defining the timeout time of the communication interruption between the visual detection host and the on-board ATP equipment as T-visual detection host Timeout; defining the timeout time of the communication interruption between the on-board ATP equipment and the visual detection host as T-ATPTimeout; if the visual detection host is in T - If the visual detection host does not receive any message from the two-way channel of the vehicle-mounted ATP device within the timeout period of the visual detection host Timeout, it is determined that the communication between the vehicle-mounted ATP device and the vehicle-mounted ATP device is interrupted; if the vehicle-mounted ATP device does not receive any message from the two-way channel of the visual detection host within the timeout period of T-ATPTimeout, it is determined that the communication between the vehicle-mounted ATP device and the visual detection host is interrupted; if the delay of the vehicle-mounted ATP device receiving the message from the visual detection host reaches the timeout period of T-ATPTimeout, the vehicle-mounted ATP device discards the message from the visual detection host and determines that the communication between the vehicle-mounted ATP device and the visual detection host is interrupted; when the communication between the vehicle-mounted ATP device and the visual detection host is interrupted, the vehicle-mounted ATP device prompts an alarm message; The sensor acquisition device is used to collect first visual data of the track area ahead of the train in the train running line; The visual detection host is used to determine obstacle-free visual data of the track area ahead of the train based on the first visual data, wherein the horizontal distances corresponding to all point clouds in the obstacle-free visual data and the values ​​of the expected distance differences are fixed; The sensor acquisition device is also used to collect second visual data of a track area ahead of a train in a train running line, and use the obstacle-free visual data as a reference standard to determine obstacle data in the second visual data by judging the consistency between the point cloud data in the second visual data and the point cloud data in the obstacle-free visual data; The visual detection host is also used to use the obstacle-free visual data and obstacle data of the track area ahead of the train as sample data, and use a machine learning algorithm to train the sample data to obtain an obstacle recognition model ahead of the train; The visual detection host is also used to input the collected real-time visual data of the track area ahead of the train into the obstacle recognition model ahead of the train for recognition, and determine the train braking command according to the classification information and position information of the obstacle in the recognition result, and output the braking command to the on-board ATP to complete the train braking; The visual detection host is also used for: Based on the point cloud data collected by the laser radar in the first vision data, traverse each point cloud in the point cloud data and calculate the vertical angle of the scanning ray where the point cloud is located; Taking the front of the train as the axis, the horizontal angle of the point cloud is calculated in a clockwise direction. Based on the horizontal angle of the point cloud, traverse each point cloud in the point cloud data in turn, obtain the index of each point cloud on the laser radar scanning circle, and organize the point cloud data in order according to the index of each point cloud; Based on the orderly organized point cloud data combined with the installation height of the laser radar and the vertical angle of the scanning ray where the point cloud is located, the expected distance from the point cloud on the laser radar scanning circle to the central axis of the laser radar is calculated; The horizontal distance and the expected distance difference of the point cloud between adjacent scanning rays are calculated, and the visual data including the horizontal distance and the expected distance difference are determined as the obstacle-free visual data of the track area ahead of the train.

9. The system for identifying obstacles in the track area ahead of a train according to claim 8, characterized in that: The calculation formula for the vertical angle of the scanning ray where the calculation point cloud is located is: In the formula, For a point cloud The coordinates of is the vertical angle of the i-th scanning ray.

10. The obstacle recognition system for the track area ahead of the train according to claim 8, characterized in that: The calculation formula for the horizontal angle of the calculation point cloud is: In the formula, Point Cloud Projection coordinates on the XOY plane, Point Cloud The horizontal angle, is an integer.

11. The system for identifying obstacles in the track area ahead of a train according to claim 8, characterized in that: The calculation formula for obtaining the index of each point cloud on the laser radar scanning circle is: In the formula, Point Cloud The index on the scan circle, Point Cloud The horizontal angle, is the floor function.

12. The system for identifying obstacles in the track area ahead of a train according to claim 8, characterized in that: The step of calculating the expected distance from the point cloud on the laser radar scanning circle to the central axis of the laser radar includes: In the formula, is the expected distance from the point cloud on the i-th scanning ray to the center of the laser radar, H is the height of the sensor installation position, is the vertical angle of the i-th scanning ray.

13. The system for identifying obstacles in the track area ahead of a train according to claim 8, characterized in that: The horizontal distance and expected distance difference of the point cloud between adjacent scanning rays are calculated, wherein the calculation formula of the horizontal distance is: In the formula, They are , The horizontal and vertical coordinates of the point cloud; The expected distance difference is calculated as: In the formula, They are , The vertical angle of the ray, H is the height of the sensor installation location.

14. The system for identifying obstacles in the track area ahead of a train according to claim 8, characterized in that: The objectives of the obstacle recognition model ahead of the train are: Obstacles and obstacle classifications, as well as obstacle location information, are identified from visual data of a track area ahead of a train on a train running line.

15. An electronic device, characterized in that: include: Processor and memory; The processor is coupled to the memory; The processor is used to read and execute the program or instruction stored in the memory, so that the electronic device executes the method according to any one of claims 1 to 7.

16. A computer-readable storage medium, characterized in that: A computer program is stored, and when the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images