Turnout state detection method, storage medium and controller

Abstract

The present disclosure relates to a turnout state detection method, a storage medium and a controller. The method comprises: obtaining, based on a sensor of a first train, measured point cloud data in a running direction of the first train; determining, from a pre-established point cloud semantic map, target point cloud data matched with the measured point cloud data, the point cloud semantic map comprising point cloud data of different states of each turnout on a full line of the first train and turnout state information corresponding to the point cloud data; and determining a turnout state in the running direction of the first train according to the turnout state information corresponding to the target point cloud data in the point cloud semantic map. The present scheme can realize autonomous detection of the turnout state by the train without relying on a ground subsystem, thus having stronger universality and reducing signal overhead between the train and a trackside subsystem.

Description

Technical Field

[0001] This disclosure relates to the field of rail transit technology, specifically to a turnout status detection method, storage medium, and controller. Background Technology

[0002] Detecting the direction of turnouts is crucial for train safety. However, the methods and technologies used for turnout direction detection often rely on a ground subsystem to obtain information about the ground conditions. If the ground subsystem malfunctions, the turnout direction detection cannot be completed. Summary of the Invention

[0003] The purpose of this disclosure is to provide a turnout condition detection method, storage medium, and controller to solve the above-mentioned technical problems.

[0004] To achieve the above objectives, in a first aspect, this disclosure provides a turnout condition detection method, comprising:

[0005] Based on the sensors of the first train itself, the measured point cloud data in the direction of travel of the first train is obtained;

[0006] From the pre-established point cloud semantic map, target point cloud data that matches the measured point cloud data is determined. The point cloud semantic map includes point cloud data of different states of each turnout on the entire line of operation of the first train and the turnout state information corresponding to the point cloud data.

[0007] The turnout status in the direction of travel of the first train is determined based on the turnout status information corresponding to the target point cloud data in the point cloud semantic map.

[0008] Optionally, the establishment of the point cloud semantic map includes:

[0009] Obtain point cloud data of different states of each turnout on the entire operating line;

[0010] The point cloud data is converted into two-dimensional data, and the geometric features of the turnout are determined based on the two-dimensional data.

[0011] Based on the geometric features, the opening state and type of each turnout are determined, and the correspondence between the geometric features, the opening state, and the turnout type is established to obtain the point cloud semantic map.

[0012] Optionally, before acquiring the measured point cloud data along the direction of travel of the first train based on its own sensors, the method further includes:

[0013] Obtain the point cloud semantic map established by the second train;

[0014] The point cloud semantic map is established by the second train by performing the following steps: acquiring point cloud data of different states of each turnout on the entire operating line; converting the point cloud data into two-dimensional data and determining the geometric features of the turnout based on the two-dimensional data; determining the opening state and turnout type of each turnout based on the geometric features, and establishing the correspondence between the geometric features, the opening state, and the turnout type to obtain the point cloud semantic map.

[0015] Optionally, acquiring point cloud data of different states of each turnout along the entire operating line includes:

[0016] The point cloud data collected by the second train's own sensors during each of the multiple runs of the entire line is obtained. The opening state of the switches on the entire line is different for each of the multiple runs. The number of multiple runs is the product of the number of opening states of each switch on the entire line.

[0017] For each collection of point cloud data of the entire running line, the turnout point cloud data of the entire running line is extracted and segmented from the point cloud data of the entire running line;

[0018] For each turnout point cloud data, the absolute position of the turnout point cloud data is determined based on the relative position of the turnout point cloud data with respect to the sensor of the second train itself, and the absolute position of the second train at the time of collection of the turnout point cloud data.

[0019] Point cloud data of turnouts with the same absolute position are treated as point cloud data of different states of the same turnout.

[0020] Optionally, converting the point cloud data into two-dimensional data includes:

[0021] From the point cloud data, a set of collinear point clouds is determined, which includes point cloud data that are adjacent and whose normal vectors have an angle less than a preset threshold.

[0022] The point cloud data is projected onto a horizontal plane to obtain multiple data points on the horizontal plane;

[0023] The two-dimensional data is obtained by connecting the data points formed by each point cloud data in the collinear point cloud set on the horizontal plane.

[0024] Optionally, the geometric features include the number of intersection angles formed by the turnouts presented in the two-dimensional data, and the positional relationship between the intersection angles;

[0025] The turnout type includes at least one of single turnout, double turnout, triple turnout, and multiple turnout;

[0026] The opening states of the turnout include the first position and the reverse position.

[0027] Optionally, acquiring measured point cloud data along the direction of travel of the first train based on its own sensors includes:

[0028] Acquire point cloud data collected by the sensors of the first train itself;

[0029] The point cloud data shall be preprocessed by at least one of the following: removing outliers from the point cloud data, and downsampling the point cloud data;

[0030] Determine whether there are point cloud data that match the characteristics of a turnout in the preprocessed point cloud data;

[0031] If point cloud data that conforms to the characteristics of a turnout exists, the point cloud data that conforms to the characteristics of a turnout is extracted and segmented from the preprocessed point cloud data to obtain the measured point cloud data.

[0032] Optionally, the point cloud semantic map also includes the absolute location of the point cloud data, and determining the target point cloud data that matches the measured point cloud data from the pre-established point cloud semantic map includes:

[0033] Determine the absolute position of the first train at the time of data acquisition of the measured point cloud data, and the relative position of the measured point cloud data with respect to the sensors of the first train itself.

[0034] The absolute position of the measured point cloud data is determined based on the absolute position and the relative position.

[0035] From the point cloud semantic map, determine the target point cloud data whose absolute position is consistent with the absolute position of the measured point cloud data.

[0036] In a second aspect, this disclosure provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described in the first aspect.

[0037] Thirdly, this disclosure provides a controller, including:

[0038] A memory on which computer programs are stored;

[0039] A processor for executing the computer program in the memory to implement the steps of the method described in the first aspect.

[0040] The above technical solution, based on a pre-constructed point cloud semantic map, acquires measured point cloud data along the train's direction of travel using the train's own sensors. Based on this measured point cloud data, it determines target point cloud data from the point cloud semantic map and directly uses the turnout status information corresponding to the target point cloud data as the detection result for the train's turnout status. Since the train can acquire this measured point cloud data using its own sensors (e.g., lidar), this solution can achieve autonomous turnout status detection without relying on a ground subsystem, offering greater versatility and reducing signal overhead between the train and the trackside subsystem.

[0041] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0042] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings:

[0043] Figure 1 A schematic diagram of a single turnout provided in an exemplary embodiment is shown;

[0044] Figure 2 A schematic diagram of a double turnout provided in an exemplary embodiment is shown;

[0045] Figure 3 A schematic diagram of a three-way turnout provided in an exemplary embodiment is shown;

[0046] Figure 4 A schematic diagram of a multi-turnout provided in an exemplary embodiment is shown;

[0047] Figure 5 A flowchart of a turnout status detection method provided in an exemplary embodiment is shown;

[0048] Figure 6 A flowchart illustrating the creation of a point cloud semantic map in an exemplary embodiment is shown.

[0049] Figure 7 A flowchart illustrating a specific implementation of S210 in an exemplary embodiment is shown;

[0050] Figure 8 A flowchart illustrating a specific implementation of S220 in an exemplary embodiment is shown;

[0051] Figure 9 A flowchart illustrating a specific implementation of S110 in an exemplary embodiment is shown;

[0052] Figure 10A flowchart illustrating a specific implementation of S120 in an exemplary embodiment is shown;

[0053] Figure 11 A block diagram of a controller provided in an exemplary embodiment is shown. Detailed Implementation

[0054] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0055] To enable those skilled in the art to quickly understand the technical solutions provided in the embodiments of this disclosure, the definitions of the technical terms involved in the embodiments of this disclosure are explained below.

[0056] First, the technical solutions provided in the embodiments of this disclosure involve detecting the state of a turnout, which may include, but is not limited to, detecting the type of the turnout and / or detecting the opening state of the turnout.

[0057] The two opening states of a turnout can be either the normal position or the reverse position. Specifically, the normal and reverse positions are relative to the train's travel route. The normal position of the turnout refers to the position where the turnout is normally open, while the reverse position is the position that is temporarily changed when arranging the route. Depending on the route arrangement, the opening direction of the turnout needs to be adjusted so that the train can travel according to the arranged route.

[0058] There are four types of turnouts: single turnout, double turnout, triple turnout, and multi-turnout. For example, such as... Figure 1 The single turnout shown has the following geometric characteristics: the turnout has two angles in all directions, and the vertices of the two angles do not coincide. Angle 1 indicates the turnout is in the correct position and the train travels from track A to track C, while angle 2 indicates the turnout is in the reverse position and the train travels from track A to track B.

[0059] For example Figure 2 The double turnout shown has the following geometric characteristics: the turnout has two angles in all directions, and the vertices of the two angles do not coincide, but the angle bisectors of the two angles are collinear with the two tracks. Angle 3 indicates that the turnout is in the correct position and the train is traveling from track A to track C, while angle 4 indicates that the turnout is in the reverse position and the train is traveling from track A to track B.

[0060] For example Figure 3The three-way turnout shown has the following geometric characteristics: the turnout has four angles in all directions, and two sets of angles with a common vertex are axially symmetric about the collinear sides. Specifically, angles 5 and 6 indicate that the turnout is in the correct position and the train is traveling from track A to track C; angle 7 indicates that the turnout is in the reverse position and the train is traveling from track A to track D; and angle 8 indicates that the turnout is in the reverse position and the train is traveling from track A to track B.

[0061] For example Figure 4 The multi-turnout shown has the following geometric features: the turnout has four angles in all directions, and two sets of angles with a common vertex are symmetrical about the center. Angles 9 and 10 indicate that the turnout is in the correct position and the train travels from track A to track C, while angles 11 and 12 indicate that the turnout is in the reverse position and the train travels from track B to track D.

[0062] The technical solution provided in this disclosure can use the number of angles between turnouts in a two-dimensional plane and the positional relationship of each angle as the geometric features of the turnout. A point cloud semantic map, including the correspondence between the geometric features and turnout types, is established based on these geometric features. This pre-established point cloud semantic map is then used to detect the state of the turnouts during actual train operation. Furthermore, the semantic information in the point cloud semantic map can be encoded into information recognizable by the train signaling system, which can be, but is not limited to, strings or numerical codes.

[0063] Specifically, Figure 5 A flowchart of a turnout condition detection method provided in an exemplary embodiment is shown. (Refer to...) Figure 1 The process includes:

[0064] S110 acquires measured point cloud data along the direction of travel of the first train based on the train's own sensors.

[0065] The measured point cloud data represents the point cloud data monitored in real time before the first train passes the switch. This point cloud data may include the coordinates of data points formed by objects sensed by the train's own sensors.

[0066] For example, the objects sensed by the sensors of the first train include the track area scanned at the corresponding time, fixed objects outside the track area, and moving objects outside the track area.

[0067] S120, from the pre-established point cloud semantic map, determine the target point cloud data that matches the measured point cloud data. The point cloud semantic map includes point cloud data of different states of each turnout on the entire line of the first train's operation, as well as the turnout state information corresponding to the point cloud data.

[0068] The target point cloud data refers to the point cloud data in the point cloud semantic map that corresponds to the measured point cloud data, such as the absolute position coordinates of the measured point cloud data in the point cloud semantic map, and point cloud data with the same geometric features.

[0069] S130, determine the turnout status in the direction of travel of the first train based on the turnout status information corresponding to the target point cloud data in the point cloud semantic map.

[0070] In one possible implementation, the turnout status includes the turnout type and the turnout's orientation. For definitions of turnout type and orientation, please refer to the above. Figures 1-4 The explanation will not be repeated here.

[0071] For example, with Figure 4 Taking the multi-turnout example, assuming the train is on track A and its direction of travel is from track A to track C, the measured point cloud data of track C and track D ahead of track A is obtained. Through the point cloud semantic map, the measured point cloud data of track C and track D are matched with the point cloud data in the point cloud semantic map to find the corresponding point cloud data. Thus, it can be known that the turnout type is a multi-turnout, and the geometric feature angles formed by track A to track C are angle 9 and angle 10. Based on this feature and the direction of travel of the train, it can be concluded that the turnout is currently in a position.

[0072] It is worth noting that the train's onboard sensors include one or more of the following: lidar, Global Navigation Satellite System (GNSS), and Inertial Measurement Unit (IMU). The lidar can be solid-state, semi-solid-state, or mechanical.

[0073] In a specific implementation, before step S110 above, the following is also included:

[0074] Obtain a point cloud semantic map established by the second train; wherein, the point cloud semantic map is established by the second train by performing the following steps: obtaining point cloud data of different states of each turnout on the entire line; converting the point cloud data into two-dimensional data, and determining the geometric features of the turnouts based on the two-dimensional data; determining the opening state and turnout type of each turnout based on the geometric features, and establishing the correspondence between the geometric features, the opening state and the turnout type to obtain the point cloud semantic map.

[0075] As can be seen from the above process, based on a pre-constructed point cloud semantic map, the train's own sensors acquire measured point cloud data along the train's direction of travel. Based on this measured point cloud data, target point cloud data is determined from the point cloud semantic map, and the turnout status information corresponding to the target point cloud data is directly used as the detection result of the train's turnout status. Since the train can acquire this measured point cloud data using its own sensors, this solution can achieve autonomous turnout status detection by the train without relying on a ground subsystem, thus having greater versatility and reducing signal overhead between the train and the trackside subsystem.

[0076] Figure 6 A flowchart illustrating the creation of a point cloud semantic map in an exemplary embodiment is shown. (Refer to...) Figure 6 The process includes:

[0077] S210, acquire point cloud data of different states of each turnout on the entire line.

[0078] The point cloud data for different states of each turnout can be obtained from raw point cloud data collected by the train's own sensors. For example, the raw point cloud data can undergo at least one of the following preprocessing steps: removing outliers and downsampling the point cloud data. Outlier removal, while preserving the point cloud characteristics, can be achieved using statistical filtering or radius filtering; downsampling can be performed using voxel grid filtering or sample consistency filtering to reduce data volume and improve real-time system efficiency. Furthermore, the preprocessed point cloud data is then analyzed to determine if point cloud data matching the turnout characteristics exists. If such data exists, it is extracted from the preprocessed point cloud data to obtain the point cloud data representing the different states of the turnout.

[0079] Furthermore, to obtain all possible opening states of the turnouts, point cloud data can be collected multiple times, with each collection showing a different opening state. The point cloud data from each collection of different turnout opening states is stored in a turnout point cloud data set, increasing the completeness and richness of the set. Simultaneously, because point cloud data is collected multiple times from the same turnout area along the same line, interference from moving objects or weather conditions can be reduced. Subsequent processing based on this point cloud data set also enhances the system's robustness.

[0080] S220 converts point cloud data into two-dimensional data and determines the geometric features of the turnout based on the two-dimensional data.

[0081] S230: Determine the opening state and turnout type of each turnout based on geometric features, and establish the correspondence between geometric features, opening state and turnout type to obtain a point cloud semantic map.

[0082] The correspondence between geometric features, orientation states, and turnout types in the point cloud semantic map can include, for example, the correspondence between geometric features and orientation states, and the correspondence between geometric features and turnout types. The orientation state in the geometric feature-orientation state correspondence can, for example, represent information about the permitted train travel direction of a turnout with that geometric feature. Thus, step S130 can specifically determine the turnout state based on the correspondence between the geometric features and orientation states of the target point cloud data. If the permitted train travel direction represented by the orientation state is inconsistent with the train travel direction when the turnout is normally open, the turnout state is determined to be reversed; if the permitted train travel direction represented by the orientation state is consistent with the train travel direction when the turnout is normally open, the turnout state is determined to be in the correct position.

[0083] For example, taking a multi-turnout as an example, the point cloud data of the turnout for all its opening directions is obtained. This point cloud data is then converted into two-dimensional data to obtain the geometric features of the turnout. The geometric features of the turnout include the number of angular features and the geometric relationships between these angular features, such as... Figure 4 As shown, angles 9 and 10, and angles 11 and 12 represent the geometric features of the point cloud of the turnout area when the turnout is in different directions. There are four angle features for each direction, and there are two sets of angles with a common vertex. Each set of angles is symmetrical about the common vertex. Based on the above geometric features, a point cloud semantic map of the correspondence between geometric features and direction states and geometric features and turnout types can be constructed.

[0084] It is worth noting that the point cloud semantic map can be established based on the point cloud data of different states of each turnout on the entire line obtained by the first train, or it can be established based on the point cloud data of different states of each turnout on the entire line obtained by other trains. This disclosure does not limit it.

[0085] As can be seen from the above process, the technical solution provided in this disclosure, after acquiring point cloud data of all opening states of turnouts along the entire train line, can convert these turnout point cloud data into two-dimensional data, thereby obtaining accurate geometric features of each turnout on the horizontal plane. Based on the acquired geometric features, a correspondence between geometric features and opening states, as well as a correspondence between geometric features and turnout types, can be established, so that the point cloud data collected by the train in real time can be accurately matched with the pre-constructed point cloud semantic map, thereby accurately identifying the opening state and type of the turnout.

[0086] Figure 7 A flowchart illustrating a specific implementation of S210 in an exemplary embodiment is shown. (Refer to...) Figure 7 The process includes:

[0087] S211, acquire point cloud data collected by the second train's own sensors each time the second train runs multiple times on the entire line. The opening state of the switches on the entire line is different for each run in the multiple runs. The number of runs is the product of the number of opening states of each switch on the entire line.

[0088] The second train can be any train different from the first train. The point cloud data for the entire line includes point cloud data of the track area, point cloud data of fixed objects outside the track area, and point cloud data of moving objects outside the track area.

[0089] S212, for each time the point cloud data of the entire running line is collected, extract and segment the turnout point cloud data of the entire running line from the point cloud data of the entire running line.

[0090] For example, the point cloud data of the turnout can be segmented from the point cloud data of the entire train line, and then the point cloud data of the turnout can be segmented from the point cloud data of the track area.

[0091] The segmentation of point cloud data in the track area allows for the filtering out of point clouds irrelevant to turnout direction detection, thereby reducing the amount of data required for further point cloud data processing and improving processing efficiency. The segmentation of point cloud data in the track area primarily involves extracting and segmenting track features, which mainly refer to the track's geometric characteristics. The point cloud data segmentation method can be, but is not limited to, any of the following: linear model, hyperbola model, and multi-feature fusion method.

[0092] S213, for each turnout point cloud data, determine the absolute position of the turnout point cloud data based on the relative position of the turnout point cloud data with respect to the sensor of the second train and the absolute position of the second train at the time of data collection.

[0093] Specifically, the relative position L of the turnout point cloud data with respect to the sensors of the second train itself. i The data was collected by the lidar sensor in the second train's own sensors and recorded as follows:

[0094] L i (x i y i , z i )

[0095] Where i = 1, 2, ..., N, i represents the number of points contained in the point cloud in the current frame.

[0096] The absolute position W of the second train at the time of data acquisition of the point cloud at the turnout is obtained by GNSS data collected from the second train's own sensors and is denoted as:

[0097] W(x, y, z)

[0098] Among them, the absolute position of the second train at the time of data collection of the point cloud data of the turnout can be, but is not limited to, the absolute position information in the geocentric coordinate system.

[0099] In addition, it also includes acquiring the attitude information P of the second train at the moment of data acquisition of the point cloud data at the turnout, which is acquired by the IMU in the second train's own sensor and denoted as:

[0100] P(roll, pitch, yaw)

[0101] Where roll represents the roll angle around the Y-axis, pitch represents the pitch angle around the X-axis, and yaw represents the yaw angle around the Z-axis.

[0102] Based on the relative position of the turnout point cloud data with respect to the sensors of the second train, the absolute position of the second train at the time the turnout point cloud data was collected, and the attitude information of the second train at the time the turnout point cloud data was collected, the formula for calculating the absolute position of the turnout point cloud data is as follows:

[0103] L′ i =ψ(L i (W, P)

[0104] Among them, L′ i The absolute position of the turnout point cloud data is represented by ψ, which represents the coordinate transformation, and L is the absolute position of the turnout point cloud data. i W represents the relative position of the turnout point cloud data with respect to the sensor of the second train itself, W represents the absolute position of the second train at the time of data acquisition of the turnout point cloud data, and P represents the attitude information of the second train at the time of data acquisition of the turnout point cloud data.

[0105] S214, treat the point cloud data of turnouts with the same absolute position as point cloud data of different states of the same turnout.

[0106] The absolute position L′ i Using the same turnout point cloud data, the position of the turnout (either its original or reversed position) can be determined. For example, such as... Figure 4 The absolute positions L′ of the multiple turnouts shown are for the turnouts from track A to track C and from track A to track D. i Similarly, the turnout position for the turnout direction from track A to track C is set to the fixed position, and the turnout position for the turnout direction from track A to track D is set to the reverse position.

[0107] As can be seen from the above process, the technical solution provided in this disclosure acquires point cloud data of different opening states of the turnouts along the entire railway line during each train operation, then segments the turnout point cloud data, and calculates the absolute position of the turnout point cloud based on the relative position of the acquired turnout point cloud data with respect to the sensor, the absolute position of the train at the time of turnout point cloud data acquisition, and the train's attitude information at the time of turnout point cloud data acquisition. Based on point cloud data with the same absolute position, point cloud data representing different states of the turnout are determined. This solution achieves precise positioning of the turnout area and precise positioning of different turnout states.

[0108] Figure 8 A flowchart illustrating a specific implementation of S220 in an exemplary embodiment is shown. (Refer to...) Figure 8 The process includes:

[0109] S221, From the point cloud data, determine the set of collinear point clouds, which includes adjacent point cloud data whose angle between normal vectors is less than a preset threshold.

[0110] Among them, neighboring points whose normal vectors are equal or approximately equal can be considered to be on the same plane. Therefore, in point cloud data, point cloud data with normal vectors between point cloud data that are less than a preset threshold are found. All point cloud data that meet this condition constitute a set of collinear point cloud data.

[0111] Specifically, find the absolute position L′ of each turnout point cloud data in the point cloud data. i The method for finding neighboring points can be, but is not limited to, the K-nearest neighbor search method.

[0112] The method searches for the set of three-dimensional coordinates {L′} of K neighboring points. i1 , L′ i2 , ..., L′ i3}, where K is a positive integer. The relative coordinates (L′) of the neighboring points are calculated using the three-dimensional coordinate set of the neighboring points. iK -L′ i ), and the Euclidean distance ||L′ of the neighboring points. iK -L′ i ||.

[0113] Among them, the three-dimensional coordinate set of the neighboring points, the relative coordinates of the neighboring points, and the Euclidean distance of the neighboring points represent the absolute position L′ of the turnout point cloud data. i The point cloud features are used to obtain the normal vector of each point cloud data.

[0114] For example, in point cloud point L′ iWithin K neighborhood points, select 3 non-collinear point cloud points. Using the relative coordinates of these 3 points, obtain two vectors a and b that share a common point. Find vector c that satisfies a·c = 0 and b·c = 0. Then vector c is the point cloud point L′. i The characteristics of the normal vector.

[0115] S222 projects the point cloud data onto a horizontal plane to obtain multiple data points on the horizontal plane.

[0116] The found collinear point cloud dataset is projected onto a horizontal plane, i.e., z is set to 0, resulting in a two-dimensional plane dataset. Projecting the point cloud data onto the horizontal plane allows for the extraction of geometric features representing the opening states of different turnouts.

[0117] For example, the point cloud points L′ in the same plane i (x i y i , z i Projected onto z i =0 plane, obtain the coordinates L″ of the two-dimensional plane point. i (x i y i ).

[0118] S223 connects the data points formed on the horizontal plane by each point cloud data in the collinear point cloud set to obtain two-dimensional data.

[0119] Connecting the data points in the set of data points in a two-dimensional plane yields two-dimensional data, which can be used to determine the geometric characteristics of the turnout.

[0120] Then, collinear points are found within the same plane. If the point cloud points belong to different straight lines, then the point is the intersection of the different straight lines, forming an angle at that point. The angle represents the geometric characteristics of the turnout.

[0121] As can be seen from the above process, in the technical solution provided by this disclosure, by acquiring a set of collinear point cloud data and projecting these point cloud data onto a horizontal plane, two-dimensional data points can be obtained. By connecting the non-collinear two-dimensional data points, the geometric features of the turnout can be obtained. This solution can obtain the geometric features of the turnout by setting z to 0 to obtain data points in a two-dimensional plane. This solution is simple, easy to obtain, and highly operable.

[0122] Figure 9 A flowchart illustrating a specific implementation of S110 in an exemplary embodiment is shown. (Refer to...) Figure 9 The process includes:

[0123] S111: Obtain point cloud data collected by the sensors on the first train.

[0124] S112, perform at least one of the following preprocessing steps on the point cloud data: remove outliers from the point cloud data, and perform downsampling on the point cloud data.

[0125] Among these methods, outliers can be removed while preserving the point cloud features. This can be achieved by, but is not limited to, statistical filtering or radius filtering. Downsampling can be performed to reduce the amount of data and improve the real-time operating efficiency of the system. This can be achieved by, but is not limited to, voxel grid filtering or sample consistency filtering.

[0126] S113, determine whether there is point cloud data that conforms to the characteristics of a turnout in the preprocessed point cloud data.

[0127] S114. If point cloud data that conforms to the characteristics of a turnout exists, extract and segment point cloud data that conforms to the characteristics of a turnout from the preprocessed point cloud data to obtain the measured point cloud data.

[0128] After noise reduction and downsampling processing of all point cloud data collected multiple times along the entire train line, the point cloud data with turnout characteristics is segmented from all point cloud data, which is the measured point cloud data at the time of collection.

[0129] As can be seen from the above process, in the technical solution provided by the embodiments of this disclosure, by preprocessing the point cloud data of the entire train operation line to obtain data that meets the requirements, and then segmenting the turnout point cloud data from the preprocessed data, the geometric features of the points are preserved and the amount of data is reduced, thereby improving the computational efficiency.

[0130] Figure 10 A flowchart illustrating a specific implementation of S120 in an exemplary embodiment is shown. The point cloud semantic map also includes the absolute position of the point cloud data, as referenced... Figure 10 The process includes:

[0131] S121, determine the absolute position of the first train at the time of data acquisition of the measured point cloud data, and the relative position of the measured point cloud data with respect to the sensors of the first train itself.

[0132] The method for determining the absolute position of the first train at the time of data acquisition of the measured point cloud data is the same as the method for determining the absolute position of the second train at the time of data acquisition of the turnout point cloud data. The method for determining the relative position of the measured point cloud data with respect to the sensors of the first train is the same as the method for determining the relative position of the turnout point cloud data with respect to the sensors of the second train.

[0133] Specifically, the relative position L of the measured point cloud data with respect to the sensors of the first train itself. i1 The data collected by the lidar sensor in the first train itself is recorded as follows:

[0134] Li1 (x i1 y i1 , z i1 )

[0135] Where i1 = 1, 2, ..., N, i1 represents the number of points contained in the point cloud in the current frame.

[0136] The absolute position W1 of the first train at the moment of data acquisition of the measured point cloud is obtained by the GNSS sensor of the first train itself and is denoted as:

[0137] W1(x1, y1, z1)

[0138] Among them, the absolute position of the first train at the time of data collection of the measured point cloud data includes, but is not limited to, the absolute position information in the geocentric coordinate system.

[0139] S122, determine the absolute position of the measured point cloud data based on the absolute position and the relative position.

[0140] In addition to determining the absolute and relative positions according to step S121, the attitude information P1 of the first train at the moment of data acquisition in the measured point cloud is also determined. This information is acquired by the IMU in the first train's own sensor and denoted as:

[0141] P1(roll1, pitch1, yaw1)

[0142] Where roll1 represents the roll angle around the Y-axis, pitch1 represents the pitch angle around the X-axis, and yaw1 represents the yaw angle around the Z-axis.

[0143] Based on the relative position of the measured point cloud data with respect to the sensors of the first train, the absolute position of the first train at the time of data acquisition, and the attitude information of the first train at the time of data acquisition, the formula for calculating the absolute position of the measured point cloud data is as follows:

[0144] L′ i1 =ψ(L i1 (W1, P1)

[0145] Among them, L′ i1 The absolute position of the measured point cloud data is represented by ψ, which represents the coordinate transformation, and W... i1 W1 represents the relative position of the measured point cloud data with respect to the sensor of the first train itself, W1 represents the absolute position of the first train at the time of data acquisition, and P1 represents the attitude information of the first train at the time of data acquisition.

[0146] S123, from the point cloud semantic map, determine the target point cloud data whose absolute position is consistent with the absolute position of the measured point cloud data.

[0147] Because different opening states of the same turnout have different point cloud data (i.e., point cloud data of different states of the same turnout as mentioned above), in practical implementation, multiple point cloud data points may be determined in the point cloud semantic map based on the absolute position of the measured point cloud data. These multiple point cloud data points correspond to different opening states of the same turnout. In this case, the target point cloud data can be further determined from these multiple point cloud data points based on geometric features. For details on how to obtain the geometric features of the measured point cloud data, please refer to [link / reference needed]. Figure 8 The steps and methods shown will not be repeated here.

[0148] As can be seen from the above process, in this technical solution, by determining the absolute position of the train and the relative position of the measured point cloud data, a coordinate transformation is performed to obtain the absolute position of the measured point cloud data. This absolute position is then matched with the absolute position of the point cloud data in the point cloud semantic map to obtain consistent point cloud data. Finally, the target point cloud data can be obtained through geometric feature matching. This solution is simple, readily available, and highly operable.

[0149] In one possible implementation of this disclosure, before step S120, it can be determined whether the ground subsystem has failed. If it is determined that the ground subsystem has not failed, the ground subsystem can still be used to obtain ground condition information. If the ground subsystem fails, then step S120 is executed.

[0150] Figure 11 A block diagram of a controller provided in an exemplary embodiment is shown. For example, controller 1100 may be an object controller. Figure 11 As shown, the controller 1100 includes a processor 1101, which may be one or more, and a memory 1102 for storing computer programs executable by the processor 1101. The computer programs stored in the memory 1102 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor 1101 may be configured to execute the computer program to perform the aforementioned turnout status detection method.

[0151] Additionally, the controller 1100 may also include a power supply component 1103 and a communication component 1104. The power supply component 1103 can be configured to perform power management of the controller 1100, and the communication component 1104 can be configured to enable communication of the controller 1100, such as wired or wireless communication, through which communication with the vehicle and with the ATS is achieved. Furthermore, the controller 1100 may also include an input / output interface 1105. The controller 1100 can operate on an operating system, such as Windows Server, stored in the memory 1102. TM Mac OS X TM Unix TM Linux TM etc.

[0152] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the turnout status detection method described above. For example, the computer-readable storage medium may be the memory 1102 including program instructions, which may be executed by the processor 1101 of the controller 1100 to complete the turnout status detection method described above.

[0153] In another exemplary embodiment, a computer program product is also provided, which includes a computer program executable by a programmable device, the computer program having a code portion for performing the above-described turnout status detection method when executed by the programmable device.

[0154] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0155] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0156] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A method for detecting the condition of a turnout, characterized in that, include: Based on the sensors of the first train itself, the measured point cloud data in the direction of travel of the first train is obtained; From the pre-established point cloud semantic map, target point cloud data that matches the measured point cloud data is determined. The point cloud semantic map includes point cloud data of different states of each turnout on the entire line of operation of the first train, the turnout state information and absolute position corresponding to the point cloud data. The turnout status in the direction of travel of the first train is determined based on the turnout status information corresponding to the target point cloud data in the point cloud semantic map. The step of determining the target point cloud data that matches the measured point cloud data from the pre-established point cloud semantic map includes: Determine the absolute position of the first train at the time of acquisition of the measured point cloud data, the relative position of the measured point cloud data with respect to the sensors of the first train itself, and the attitude information of the first train at the time of acquisition of the measured point cloud data. The absolute position of the measured point cloud data is determined based on the absolute position, the relative position, and the attitude information. From the point cloud semantic map, determine the target point cloud data whose absolute position is consistent with the absolute position of the measured point cloud data.

2. The method according to claim 1, characterized in that, The establishment of the point cloud semantic map includes: Obtain point cloud data of different states of each turnout on the entire operating line; The point cloud data is converted into two-dimensional data, and the geometric features of the turnout are determined based on the two-dimensional data. Based on the geometric features, the opening state and type of each turnout are determined, and the correspondence between the geometric features, the opening state, and the turnout type is established to obtain the point cloud semantic map.

3. The method according to claim 1, characterized in that, Before acquiring measured point cloud data along the direction of travel of the first train based on its own sensors, the method further includes: Obtain the point cloud semantic map established by the second train; The point cloud semantic map is established by the second train by performing the following steps: acquiring point cloud data of different states of each turnout on the entire operating line; converting the point cloud data into two-dimensional data and determining the geometric features of the turnout based on the two-dimensional data; determining the opening state and turnout type of each turnout based on the geometric features, and establishing the correspondence between the geometric features, the opening state, and the turnout type to obtain the point cloud semantic map.

4. The method according to claim 3, characterized in that, The acquisition of point cloud data of different states of each turnout along the entire operating line includes: The point cloud data collected by the second train's own sensors during each of the multiple runs of the entire line is obtained. The opening state of the switches on the entire line is different for each of the multiple runs. The number of multiple runs is the product of the number of opening states of each switch on the entire line. For each collection of point cloud data of the entire running line, the turnout point cloud data of the entire running line is extracted and segmented from the point cloud data of the entire running line; For each turnout point cloud data, the absolute position of the turnout point cloud data is determined based on the relative position of the turnout point cloud data with respect to the sensor of the second train itself, and the absolute position of the second train at the time of collection of the turnout point cloud data. Point cloud data of turnouts with the same absolute position are treated as point cloud data of different states of the same turnout.

5. The method according to claim 2 or 3, characterized in that, The step of converting the point cloud data into two-dimensional data includes: From the point cloud data, a set of collinear point clouds is determined, which includes point cloud data that are adjacent and whose normal vectors have an angle less than a preset threshold. The point cloud data is projected onto a horizontal plane to obtain multiple data points on the horizontal plane; The two-dimensional data is obtained by connecting the data points formed by each point cloud data in the collinear point cloud set on the horizontal plane.

6. The method according to claim 2 or 3, characterized in that, The geometric features include the number of intersection angles formed by the turnouts presented in the two-dimensional data, and the positional relationship between each intersection angle; The turnout type includes at least one of single turnout, double turnout, triple turnout, and multiple turnout; The opening states of the turnout include the first position and the reverse position.

7. The method according to claim 1, characterized in that, The acquisition of measured point cloud data along the direction of travel of the first train, based on the first train's own sensors, includes: Acquire point cloud data collected by the sensors of the first train itself; The point cloud data shall be preprocessed by at least one of the following: removing outliers from the point cloud data, and downsampling the point cloud data; Determine whether there are point cloud data that match the characteristics of a turnout in the preprocessed point cloud data; If point cloud data that conforms to the characteristics of a turnout exists, the point cloud data that conforms to the characteristics of a turnout is extracted and segmented from the preprocessed point cloud data to obtain the measured point cloud data.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-7.

9. A controller, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-7.

Images