Track traffic obstacle recognition method and device based on time sequence point cloud superposition

Abstract

The invention is suitable for the technical field of point cloud processing, and provides a time sequence point cloud superposition-based rail transit obstacle recognition method, which comprises the following steps of: obtaining current frame foreground point cloud data of a train at the current moment, the current frame of foreground point cloud data is used for representing three-dimensional coordinate information of an object in a to-be-detected area determined according to a running line of the train at the current moment, points in the current frame of foreground point cloud data and points in historical frame of foreground point cloud data of k continuous moments obtained in advance are overlapped, and current frame of overlapped point cloud data is obtained, the historical frame foreground point cloud data is used for representing three-dimensional coordinate information of an object in the to-be-detected area at a historical moment before the current moment, and performing rail transit obstacle recognition on the train according to position information corresponding to points in the current frame superposition point cloud data to obtain an obstacle recognition result of the train in the running line. According to the invention, the problem that the accuracy of detecting and identifying the obstacle by the train is low because the laser point cloud is too sparse can be solved.

Description

Technical Field

[0001] The present application belongs to the field of point cloud processing technology, and in particular relates to a rail transit obstacle identification method and device based on time-series point cloud superposition. Background Art

[0002] LiDAR is a primary sensor for high-precision obstacle detection. By emitting laser beams, LiDAR measures the distance and angle between the target object and the LiDAR, thereby constructing a point cloud map of the surrounding environment. Based on the rich spatial information contained in this point cloud data, LiDAR can identify obstacles, defective products, and other target objects. LiDAR is widely used in intelligent driving, robot navigation, drone obstacle avoidance, and product inspection. However, as the detection distance increases, the point cloud formed by the laser beam on the target object becomes increasingly sparse, and the number of measured point cloud data points gradually decreases. Consequently, the accuracy of obstacle detection based on the point cloud map and real-time measured point cloud data also decreases.

[0003] In the field of intelligent driving technology, as rail transit safety standards improve, the requirements for obstacle detection distances are also increasing. When a stationary or moving train detects obstacles at a distance, the accuracy of obstacle detection and recognition is low due to the sparse laser point cloud. Summary of the Invention

[0004] The embodiments of the present application provide a rail transit obstacle identification method, device, and storage medium based on time-series point cloud superposition, which can solve the problem of low accuracy in detecting and identifying obstacles due to the sparse laser point cloud when a stationary train or a moving train detects obstacles at a long distance.

[0005] In a first aspect, an embodiment of the present application provides a rail transit obstacle recognition method based on time-series point cloud superposition, comprising: Acquire a current frame of foreground scenic spot cloud data at a current moment during the train's operation, wherein the current frame of foreground scenic spot cloud data is used to represent three-dimensional coordinate information of an object within a to-be-detected area at the current moment, where the to-be-detected area is determined based on the train's operating route; Superimposing points in the foreground point cloud data of the current frame with points in the foreground point cloud data of k consecutive historical frames previously acquired to obtain current frame superimposed point cloud data corresponding to the foreground point cloud data of the current frame, wherein the foreground point cloud data of the historical frames is used to represent the three-dimensional coordinate information of objects in the to-be-detected area at historical moments before the current moment, and k is a constant; According to the position information corresponding to the points in the current frame superimposed point cloud data, rail transit obstacle identification is performed on the train during operation to obtain an obstacle identification result of the train in the operation line.

[0006] In a possible implementation of the first aspect, superimposing points in the foreground point cloud data of the current frame with points in foreground point cloud data of k consecutive historical frames acquired in advance to obtain current frame superimposed point cloud data corresponding to the foreground point cloud data of the current frame includes: According to a pre-constructed point cloud map, performing a first coordinate system transformation on each point in the previously acquired foreground point cloud data of k consecutive moments in the historical frames, to obtain first foreground point cloud data of k historical frames, wherein the pre-constructed point cloud map is determined based on the running route of the train using a simultaneous positioning and mapping method; According to the coordinate system corresponding to the foreground point cloud data of the current frame, a second coordinate system transformation is performed on each of the k points in the first foreground point cloud data of the historical frames to obtain k second foreground point cloud data of the historical frames; The k points in the second foreground point cloud data of the historical frames are superimposed on the foreground point cloud data of the current frame to obtain superimposed point cloud data of the current frame corresponding to the foreground point cloud data of the current frame.

[0007] In a possible implementation of the first aspect, performing rail transit obstacle identification on the running train based on position information corresponding to points in the superimposed point cloud data of the current frame to obtain an obstacle identification result of the train on the running line includes: Filtering and clustering the points in the current frame superimposed point cloud data according to position information corresponding to each point in the current frame superimposed point cloud data to obtain obstacle clusters; Selecting a static point that meets a second preset condition from the obstacle cluster, wherein the second preset condition is determined according to the motion state of each point in the obstacle cluster; Determining an obstacle bounding box according to the static points in the obstacle cluster; An obstacle recognition result of the train on the running route is determined according to the obstacle bounding box.

[0008] In a possible implementation of the first aspect, filtering and clustering points in the current frame superimposed point cloud data according to position information corresponding to each point in the current frame superimposed point cloud data to obtain obstacle clusters includes: Counting the number of first similar points of each point in the current frame superimposed point cloud data according to the position information corresponding to each point in the current frame superimposed point cloud data, wherein the first similar point of each point in the current frame superimposed point cloud data is a point cloud in the current frame superimposed point cloud data that is in the same plane as the point; Filtering, from the current frame superimposed point cloud data, current frame superimposed obstacle point cloud data that meets a first preset condition based on the number of the first similar points of each point in the current frame superimposed point cloud data, wherein the first preset condition is determined based on a preset threshold value for the number of similar points; Clustering all points in the obstacle point cloud data superimposed on the current frame to obtain multiple initial clusters; The obstacle cluster is selected from the multiple initial clusters according to the number of points in each of the initial clusters.

[0009] In a possible implementation of the first aspect, counting the number of first similar points of each point in the current frame superimposed point cloud data according to the position information corresponding to each point in the current frame superimposed point cloud data includes: Determining, based on the position information corresponding to each point in the current frame superimposed point cloud data, the distance from the train corresponding to each point in the current frame superimposed point cloud data, wherein the distance from the train corresponding to each point in the current frame superimposed point cloud data is the distance between the point and the train; For any point in the current frame overlay point cloud data: Calculate, based on the vehicle distance corresponding to each point in the current frame superimposed point cloud data, the vehicle distance difference between other points in the current frame superimposed point cloud data and the point, wherein the other points in the current frame superimposed point cloud data are points in the current frame superimposed point cloud data other than the point; Selecting, from all other points in the current frame superimposed point cloud data, a point whose distance from the point to the vehicle is less than a preset first threshold value as the first similar point to the point; Count the number of the first similar points to the point.

[0010] In a possible implementation of the first aspect, selecting a static point that meets a second preset condition from the obstacle cluster includes: Selecting, from the current frame superimposed obstacle point cloud data, current frame obstacle point cloud data that satisfies a third preset condition, wherein the third preset condition is determined based on whether a point in the current frame superimposed obstacle point cloud data is a point in the current frame foreground point cloud data; For any point in the obstacle point cloud data superimposed on the current frame: Selecting, from all points in the obstacle point cloud data of the current frame, a point whose distance from the point to the vehicle is less than a preset second threshold value as a second similar point to the point; Count the number of the second closest points to the point; Determining whether the motion state of the point is static or dynamic based on the number of the second similar points to the point; According to the motion state of each point in the obstacle point cloud data superimposed on the current frame, a static point is selected from the obstacle point cloud data superimposed on the current frame.

[0011] In a possible implementation of the first aspect, obtaining the current frame foreground scenic spot cloud data at a current moment during the operation of the train includes: Acquiring current frame point cloud data at a current moment during the operation of the train, wherein the current frame point cloud data is used to represent three-dimensional coordinate information of objects in the surrounding environment of the train; performing motion compensation on points in the current frame point cloud data according to the train's own motion information to obtain compensated current frame point cloud data, wherein the train's own motion information is estimated based on inertial measurement data of the train during operation; Determining the running route of the train based on the current position information of the train and the pre-acquired traffic track information, wherein the current position information of the train is obtained by positioning the train based on the compensated current frame point cloud data; determining the area to be inspected according to the running route of the train; The current frame foreground point cloud data is filtered out from the compensated current frame point cloud data according to the position information of the midpoint of the compensated current frame point cloud data and the area to be detected.

[0012] In a second aspect, an embodiment of the present application provides a rail transit obstacle recognition device based on time-series point cloud superposition, comprising: a point cloud data acquisition module, configured to acquire point cloud data of a current frame in front of a train at a current moment in its operation, wherein the point cloud data of the current frame in front of the train is used to represent the three-dimensional coordinate information of an object within a region to be detected at the current moment, wherein the region to be detected is determined based on the train's operating route; a point cloud data superposition module, configured to superimpose points in the foreground point cloud data of the current frame with points in the foreground point cloud data of k consecutive historical frames previously acquired, to obtain current frame superimposed point cloud data corresponding to the foreground point cloud data of the current frame, wherein the foreground point cloud data of the historical frames is used to represent the three-dimensional coordinate information of objects in the area to be detected at historical moments before the current moment, and k is a constant; The obstacle recognition module is used to identify rail transit obstacles for the train during operation based on the position information corresponding to the points in the superimposed point cloud data of the current frame, and obtain the obstacle recognition result of the train in the operating line.

[0013] In a third aspect, an embodiment of the present application provides a rail transit obstacle identification device based on time-series point cloud superposition, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the above-described methods when executing the computer program.

[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements any of the methods described above.

[0015] In a fifth aspect, an embodiment of the present application provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute any of the methods described above.

[0016] Compared with the prior art, the embodiments of the present application have the following advantages: by obtaining the current frame foreground point cloud data at the current moment during the operation of the train, the points in the current frame foreground point cloud data are superimposed with the points in the historical frame foreground point cloud data acquired in advance at k consecutive moments, so as to obtain the current frame superimposed point cloud data corresponding to the current frame foreground point cloud data, wherein the current frame foreground point cloud data is used to represent the three-dimensional coordinate information of the objects in the to-be-detected area at the current moment, the to-be-detected area is determined according to the running route of the train, and the historical frame foreground point cloud data is used to represent the three-dimensional coordinate information of the objects in the to-be-detected area at the historical moments before the current moment. , k is a constant. Since the current frame superimposed point cloud data incorporates the foreground point cloud data of k consecutive historical frames, the point cloud in the current frame superimposed point cloud data is denser. According to the position information corresponding to each point in the current frame superimposed point cloud data with denser point cloud, rail transit obstacle recognition is performed on the running train, so that the obstacle recognition result of the train in the running line is more accurate, and the distance at which obstacles can be identified is farther, thereby solving the problem of low accuracy in detecting and identifying obstacles due to the sparse laser point cloud when a stationary train or a moving train detects obstacles at a long distance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 This is a flow chart of a rail transit obstacle identification method based on time-series point cloud superposition provided by an embodiment of the present application; Figure 2 This is a flow chart of another rail transit obstacle identification method based on time-series point cloud superposition provided by another embodiment of the present application; Figure 3 1 is a flow chart of another method for filtering and clustering point clouds in superimposed point cloud data of a current frame provided by another embodiment of the present application; Figure 4 This is a schematic structural diagram of a rail transit obstacle recognition device based on time-series point cloud superposition provided by one embodiment of the present application; Figure 5 This is a structural diagram of a rail transit obstacle recognition device based on time-series point cloud superposition provided by another embodiment of the present application. DETAILED DESCRIPTION

[0019] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0020] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0021] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0022] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0023] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0024] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0025] As the primary sensor for high-precision obstacle detection, LiDAR plays a vital role in rail transit obstacle detection. When detecting obstacles at long distances, the laser point cloud obtained by LiDAR is too sparse. To prevent noise interference and a high error rate in obstacle recognition, the obstacle detection threshold cannot be set too sensitively. This can lead to difficulty in accurately detecting obstacles or insufficient detection distance.

[0026] In addition, an overly sparse point cloud can also cause problems such as excessive jitter in the detection frame and inaccurate obstacle classification.

[0027] In the field of intelligent driving technology, with the improvement of rail transit safety standards, the requirements for obstacle detection distance and detection accuracy when the train is stationary or in motion are also increasing.

[0028] Currently, the main methods to achieve long-distance obstacle detection include increasing the number of lidars or using lidars with longer detection distances and higher point cloud density to improve the accuracy of obstacle detection, but these methods are costly and difficult to implement.

[0029] It is particularly important to achieve long-distance obstacle detection without increasing additional hardware costs while further improving the stability of the detection frame and the accuracy of obstacle recognition.

[0030] The present application embodiment provides a rail transit obstacle recognition method based on time-series point cloud superposition, see Figure 1 , Figure 1 This is a flow chart of a rail transit obstacle recognition method based on time-series point cloud superposition provided by an embodiment of the present application, including: Step S11, obtaining the scenic spot cloud data of the current frame in front of the train at the current moment during the train operation, wherein the scenic spot cloud data of the current frame in front of the train is used to represent the three-dimensional coordinate information of the objects in the area to be detected at the current moment, and the area to be detected is determined according to the running route of the train; Step S12: superimposing the points in the foreground point cloud data of the current frame with the points in the foreground point cloud data of k consecutive historical frames previously acquired to obtain superimposed point cloud data of the current frame corresponding to the foreground point cloud data of the current frame, wherein the foreground point cloud data of the historical frames is used to represent the three-dimensional coordinate information of objects in the to-be-detected area at historical moments before the current moment, and k is a constant; Step S13: Based on the position information corresponding to the points in the current frame superimposed point cloud data, rail transit obstacle identification is performed on the running train to obtain the obstacle identification result of the train in the running line.

[0031] It should be noted that when the train identifies rail transit obstacles, a sensor module is installed forward on the front of the train. The sensor module includes lidar, millimeter-wave radar, and inertial measurement unit, etc., which can realize multi-dimensional perception of the surrounding environment ahead.

[0032] When a lidar scans an entire route (from a train's starting point to its destination) or an entire city, it can compensate and dedistort the resulting point cloud data. Based on the coordinate information of this dedistorted point cloud data, a high-precision point cloud map can be constructed using SLAM (Simultaneous Localization and Mapping) technology.

[0033] Since trains operate on tracks, there is a strict correspondence between the train's route and the track area. By compensating for and de-distorting the point cloud data, the precise position of the track centerline can be automatically marked in the constructed point cloud map.

[0034] Specifically, while the train is in motion, the LiDAR in front of the train acquires the current frame of point cloud data. This current frame of point cloud data is obtained by scanning the train's surroundings via the LiDAR at the current moment. This current frame of point cloud data includes the point cloud and the coordinates of each point within it. This current frame of point cloud data is then compensated and dedistorted to produce compensated current frame point cloud data. This compensated current frame of point cloud data offsets the effects of the train's movement on each point within the current frame of point cloud data.

[0035] For a moving train, obstacles within its route will affect it, while obstacles outside the route have no impact. Therefore, only obstacles within the route need to be detected. Using the compensated point cloud data of the current frame, laser SLAM (Simultaneous Localization and Mapping) technology can determine the train's position. The train's route can then be determined based on the position of the track centerline marked in the pre-built point cloud map. The train's route represents the train's trajectory or track. Based on the train's route and a pre-set detection range, the area to be inspected can be determined. For example, the detection range can be set to the width of the train.

[0036] Based on the area to be detected, the compensated point cloud data of the current frame that does not fall within the area to be detected is filtered out, and the resulting point cloud data is the foreground point cloud data of the current frame. The foreground point cloud data of the current frame is used to represent the 3D coordinate information of objects in the area to be detected at the current moment.

[0037] For any point cloud data acquired at any moment, the foreground point cloud data corresponding to that moment can be obtained by performing the same operations as described above for the current frame point cloud data. For example, for historical frame point cloud data acquired at a historical moment before the current moment, the foreground point cloud data corresponding to that historical frame can be obtained.

[0038] Overlay all points in the current frame's foreground point cloud data with all points in the previously acquired foreground point cloud data for k consecutive moments. This involves transforming the coordinate systems of all points in the k consecutive moments' foreground point cloud data into the coordinate systems corresponding to the points in the current frame's foreground point cloud data. These points are then used as the points in the current frame's overlaid point cloud data, resulting in the current frame's overlaid point cloud data corresponding to the current frame's foreground point cloud data. If the current moment is t, the k historical moments before the current moment are t-1, t-2, ..., tk.

[0039] It should be noted that the number of historical moments before the k current moment may be 5, and this embodiment does not impose any specific limitation on the value of k.

[0040] Specifically, under ideal conditions where all points in the current frame's superimposed point cloud data are noise-free and train positioning accuracy is high, rail transit obstacle identification for the running train can be performed based on the points in the current frame's superimposed point cloud data obtained after the aforementioned superposition. Obstacle bounding boxes can be determined. All obstacle bounding boxes represent the obstacle identification results for the train's route. The area enclosed by each obstacle bounding box represents an obstacle, and the obstacle bounding box represents the boundary of the obstacle. Because all points in the current frame's superimposed point cloud data obtained after the aforementioned superposition significantly enhance the point cloud density, the accuracy of obstacle identification results obtained by identifying points in the current frame's superimposed point cloud data is significantly improved.

[0041] When there is noise in the points in the current frame superimposed point cloud data, the points in the current frame superimposed point cloud data can be filtered according to the position information corresponding to each point in the current frame superimposed point cloud data to obtain the current frame superimposed obstacle point cloud data. Based on the points in the current frame superimposed obstacle point cloud data, rail transit obstacles are identified for the running train to obtain the obstacle identification result of the train in the running line.

[0042] It should be noted that the target object can be any artificially set target object. This embodiment includes but is not limited to the identification of rail transit obstacles, and the above method can be used to achieve it in the field of laser obstacle perception.

[0043] It can be understood that the technical solution provided by this embodiment obtains the current frame foreground point cloud data at the current moment during the operation of the train, and superimposes the points in the current frame foreground point cloud data with the points in the historical frame foreground point cloud data of k consecutive moments obtained in advance to obtain the current frame superimposed point cloud data corresponding to the current frame foreground point cloud data, wherein the current frame foreground point cloud data is used to represent the three-dimensional coordinate information of the object in the to-be-detected area at the current moment, the to-be-detected area is determined according to the running route of the train, and the historical frame foreground point cloud data is used to represent the three-dimensional coordinate information of the object in the to-be-detected area at the historical moment before the current moment. Information, k is a constant. Since the current frame superimposed point cloud data incorporates the foreground point cloud data of k consecutive moments of the historical frames, the points in the current frame superimposed point cloud data are denser. According to the position information corresponding to the points in the current frame superimposed point cloud data with denser point clouds, rail transit obstacles are identified for the running train, so that the obstacle identification result of the train in the running line is more accurate, and the distance at which obstacles can be identified is farther, thereby solving the problem of low accuracy in detecting and identifying obstacles caused by the laser point cloud being too sparse when the train is stationary or when the moving train detects obstacles at a long distance.

[0044] In a possible implementation, in step S12, the points in the foreground scenic spot cloud data of the current frame are superimposed with the points in the foreground scenic spot cloud data of k consecutive historical frames acquired in advance to obtain the superimposed point cloud data of the current frame corresponding to the foreground scenic spot cloud data of the current frame, including: Step S121: Based on a pre-constructed point cloud map, a first coordinate system transformation is performed on each point in the previously acquired foreground point cloud data of k consecutive historical frames, thereby obtaining first foreground point cloud data of k historical frames. The pre-constructed point cloud map is determined based on the train's route using a simultaneous positioning and mapping method. Step S122 , performing a second coordinate system transformation on each point in the first foreground point cloud data of k historical frames according to the coordinate system corresponding to the foreground point cloud data of the current frame, to obtain second foreground point cloud data of k historical frames; Step S123 , superimposing the points in the second foreground point cloud data of k historical frames onto the foreground point cloud data of the current frame to obtain superimposed point cloud data of the current frame corresponding to the foreground point cloud data of the current frame.

[0045] Specifically, when the points in the foreground scenic spot cloud data of the current frame are superimposed with the points in the foreground scenic spot cloud data of k consecutive historical frames acquired in advance, the current frame superimposed point cloud data corresponding to the foreground scenic spot cloud data of the current frame is obtained, including: By pre-scanning the entire train route, a point cloud map can be constructed based on simultaneous positioning and mapping. The point cloud map includes the point cloud data for the train route as well as the coordinate system of each point in the foreground point cloud data collected at each moment. For example, the current frame corresponds to the coordinate system of each point in the foreground point cloud data of the current frame, and the frames at k historical moments each correspond to the coordinate system of each point in the foreground point cloud data of the historical frame. Each frame corresponds to a coordinate system and a unique frame number. The foreground point cloud data includes the frame number of each point, which is used to mark the laser point cloud frame from which the point comes. The foreground point cloud data of a frame also includes the positioning pose in the coordinate system corresponding to that frame. The positioning pose includes the position coordinates and posture information of the frame.

[0046] For the historical frame foreground cloud data of k consecutive moments and the current frame foreground cloud data, they each correspond to their own coordinate system. Assume that the positioning pose corresponding to the current frame is , the positioning poses corresponding to the historical frames at k consecutive moments are , ,……, , the positioning posture is a positioning posture with 6 degrees of freedom.

[0047] For all points in the historical frame foreground cloud data corresponding to the nth frame in the historical frames of k consecutive moments, where n is any constant from 1 to k, which can also be 1 or k, obtain the positioning pose of the nth frame Perform a first coordinate system transformation on each point in the historical frame foreground point cloud data of the nth frame, converting it to coordinates in the world map coordinate system, and obtain the historical frame first foreground point cloud data of the nth frame. The coordinates of each point in the historical frame first foreground point cloud data of the nth frame are coordinates in the world map coordinate system.

[0048] The positioning pose corresponding to the current frame is Based on the coordinate system corresponding to the foreground point cloud data of the current frame, a second coordinate system transformation is performed on each point in the first foreground point cloud data of the historical frame n, converting it to the coordinate system corresponding to the foreground point cloud data of the current frame, thereby obtaining the second foreground point cloud data of the historical frame n. The coordinates of each point in the second foreground point cloud data of the historical frame n are the coordinates in the coordinate system corresponding to the current frame.

[0049] For the points in the foreground point cloud data of the historical frames at k consecutive moments, the first coordinate system conversion and the second coordinate system conversion are performed according to their respective corresponding coordinate systems to obtain the second foreground point cloud data of k historical frames.

[0050] All points in the second foreground point cloud data of k historical frames are superimposed with all points in the foreground point cloud data of the current frame to obtain the superimposed point cloud data corresponding to the foreground point cloud data of the current frame. The coordinates corresponding to each point in the superimposed point cloud data of the current frame are all coordinates in the coordinate system corresponding to the current frame. The superimposed point cloud data of the current frame also includes the frame number corresponding to the frame from which each point comes, which is used to mark the unique frame from which each point in the superimposed point cloud data of the current frame comes.

[0051] In an example, k=5, all points in the historical foreground point cloud data corresponding to 5 consecutive frames are superimposed on the current frame foreground point cloud data for superposition processing. Assume that point p is a point in the point cloud of the historical foreground point cloud data corresponding to the nth frame, its coordinates are Q=(x,y,z), and the positioning pose of the nth frame is , perform the first coordinate system transformation on point p, , then, the second coordinate system transformation is performed on point p, and the positioning pose corresponding to the current frame is , , get the coordinates of point p in the coordinate system corresponding to the current frame.

[0052] It can be understood that by superimposing the points in the foreground point cloud data of k consecutive historical frames onto the foreground point cloud data of the current frame, the point cloud density of the points in the superimposed point cloud data of the current frame obtained after superposition is improved, and is denser than that of the current frame. By identifying obstacles through the points in the superimposed point cloud data of the current frame after superposition, the detection distance can be significantly improved, and small obstacles can be effectively identified, thereby improving the accuracy of obstacle identification. In addition, only the points in the foreground point cloud data are superimposed, which improves the efficiency of superposition and makes the recognition efficiency of the entire obstacle identification process higher.

[0053] In a possible implementation, this embodiment also provides another rail transit obstacle recognition method based on time-series point cloud superposition, see the attached Figure 2 , Figure 2 This is a flow chart of another rail transit obstacle identification method based on time-series point cloud overlay provided by another embodiment of the present application. In step S13, rail transit obstacle identification is performed on the running train based on the position information corresponding to the points in the current frame overlay point cloud data, and the obstacle identification result of the train on the running line is obtained, including: Step S131 , filtering and clustering the points in the current frame superimposed point cloud data according to the position information corresponding to each point in the current frame superimposed point cloud data to obtain obstacle clusters; Step S132: selecting a static point that meets a second preset condition from the obstacle cluster, wherein the second preset condition is determined based on the motion state of each point in the obstacle cluster; Step S133, determining an obstacle bounding box based on the static points in the obstacle cluster; Step S134: Determine the obstacle recognition result of the train on the running line based on the obstacle bounding box.

[0054] Specifically, according to the position information corresponding to each point in the current frame superimposed point cloud data, rail transit obstacle recognition is performed on the train during operation, and the obstacle recognition result of the train in the running line is obtained, including: Because some points are caused by train movement, to prevent noise accumulation that could lead to low obstacle recognition accuracy, for example, identifying a bounding box composed of a point cloud that is not an obstacle as an obstacle, the points in the current frame's overlay point cloud data need to be filtered. Based on the position information corresponding to each point in the current frame's overlay point cloud data, the number of points on the plane parallel to the train is counted. Since real obstacles often have many points in a plane, the number of points on a plane parallel to the train is relatively small, and these points are noise points. The noise points in the current frame's overlay point cloud data are filtered out to obtain the current frame's overlay obstacle point cloud data.

[0055] After clustering the points in the obstacle point cloud data of the current frame, obstacle clusters are obtained. There can be one, multiple, or zero obstacle clusters.

[0056] It should be noted that the points in the obstacle point cloud data superimposed on the current frame can be clustered by using the Euclidean clustering method, that is, clustering according to the distance between the point clouds. Different thresholds can be set according to different distances. K-means clustering, density-based spatial clustering and other clustering methods can also be used for clustering. This embodiment does not specifically limit the clustering method.

[0057] Since the motion state of each point in the obstacle point cloud data overlaid in the current frame can be either static or moving, for each point in the obstacle cluster, its motion state is determined to be static or dynamic. Static motion is used as the basis for satisfying the second preset condition. Static point clouds are used to represent stationary objects, while dynamic point clouds are used to represent moving objects. Points with a static motion state are selected from the obstacle cluster as static points that meet the second preset condition.

[0058] An obstacle bounding box is determined based on the static points in the obstacle clusters obtained above, wherein all static points in each obstacle cluster constitute the obstacle bounding box, and the obstacle bounding box is the boundary of the obstacle.

[0059] Based on the obstacle bounding box, the obstacle recognition result of the train on the running route is determined. The obstacle recognition result may include whether the running route at the current moment includes an obstacle, and, if an obstacle is included, the obstacle information such as the location, shape, size and category of the obstacle.

[0060] It can be understood that by filtering and clustering the points in the current frame's superimposed point cloud data based on the position information corresponding to each point in the current frame's superimposed point cloud data, obstacle clusters are obtained. Static points that meet the second preset condition are selected from the obstacle clusters. An obstacle bounding box is determined based on the static points in the obstacle clusters. Based on the obstacle bounding box, the obstacle identification result for the train on the running route is determined. Due to the filtering, noise accumulation is prevented from affecting the accuracy of the obstacle identification results, reducing the false detection rate, ensuring the accuracy of the point cloud data and the reliability of the obstacle identification results. Furthermore, using all dynamic points in the clustering process effectively prevents the occurrence of target splitting. For example, for transparent glass, clustering only the point cloud data of the upper and lower boundaries of the glass can effectively avoid the phenomenon of each of the upper and lower boundaries of the glass being easily identified as a separate obstacle, making the identified obstacle more complete. Finally, the final obstacle bounding box is determined based on the static points, reducing the interference caused by dynamic points in the foreground point cloud data and ensuring the accuracy of obstacle identification.

[0061] In a possible implementation, this embodiment also provides a method for filtering and clustering points in the current frame superimposed point cloud data, referring to Figure 3 , Figure 3 This is a flow chart of another method for filtering and clustering point clouds in the current frame superimposed point cloud data provided by another embodiment of the present application. In step S131, based on the position information corresponding to each point in the current frame superimposed point cloud data, the points in the current frame superimposed point cloud data are filtered and clustered to obtain obstacle clusters, including: Step S1311: Counting the number of first similar points of each point in the current frame of superimposed point cloud data based on the position information corresponding to each point in the current frame of superimposed point cloud data, wherein the first similar point of each point in the current frame of superimposed point cloud data is a point cloud in the current frame of superimposed point cloud data that is in the same plane as the point; Step S1312: Filtering obstacle point cloud data of the current frame that meets a first preset condition from the current frame superimposed point cloud data based on the number of first similar points of each point in the current frame superimposed point cloud data, wherein the first preset condition is determined based on a preset threshold value of the number of similar points; Step S1313: clustering all points in the obstacle point cloud data of the current frame to obtain multiple initial clusters; Step S1314 : selecting an obstacle cluster from the multiple initial clusters according to the number of points in each initial cluster.

[0062] Specifically, for any point in the point cloud of the current frame's overlaid point cloud data, all points in the current frame's overlaid point cloud data that are in the same plane as the point are counted as the first closest points to the point, and the number of first closest points to the point is calculated. In this way, the number of first closest points for each point in the current frame's overlaid point cloud data is obtained.

[0063] For any point in the current frame's overlaid point cloud data, the number of its first similar points can be used to determine whether it is a noise point. When the number of its first similar points exceeds a preset threshold for the number of similar points, the point is determined not to be a noise point and satisfies the first preset condition. When the number of its first similar points is not greater than the preset threshold for the number of similar points, the point is determined to be a noise point and does not satisfy the first preset condition.

[0064] It should be noted that the first preset condition is determined based on a preset threshold value of the number of similar points. The preset threshold value of the number of similar points is artificially preset and can be any constant or can be set according to actual application requirements. This embodiment does not make any specific limitations on this.

[0065] Specifically, from all points in the current frame superimposed point cloud data, all noise points are filtered out, and all points that are not noise points are selected to form the points in the current frame superimposed obstacle point cloud data.

[0066] For all points in the current frame superimposed obstacle point cloud data obtained after noise removal, clustering is performed to obtain initial cluster clusters, where the initial cluster clusters can be zero, one, or multiple.

[0067] For any initial cluster, when the number of points contained in the initial cluster is too small, it may be a recognition error and not a real obstacle. The initial cluster containing too few points is filtered out, and the obstacle cluster is selected from multiple initial clusters.

[0068] In one possible implementation, in step S1311, counting the number of first similar points of each point in the current frame superimposed point cloud data according to the position information corresponding to each point in the current frame superimposed point cloud data includes: Determine the distance from the train corresponding to each point in the current frame of superimposed point cloud data based on the position information corresponding to each point in the current frame of superimposed point cloud data, wherein the distance from the train corresponding to each point in the current frame of superimposed point cloud data is the distance between the point and the train; For any point in the current frame overlay point cloud data: According to the distance from the vehicle corresponding to each point in the current frame superimposed point cloud data, calculate the difference in the distance from the vehicle between other points in the current frame superimposed point cloud data and the point, wherein the other points in the current frame superimposed point cloud data are the points in the current frame superimposed point cloud data other than the point; Selecting a point whose distance from the point to the vehicle is less than a preset first threshold from all other points in the current frame superimposed point cloud data as the first similar point to the point; Count the number of first closest points to the point.

[0069] Specifically, for each point in the point cloud of the current frame's overlaid point cloud data, the distance between each point in the current frame's overlaid point cloud data and the train is determined as the distance from the train corresponding to that point. For example, the distance from the train corresponding to point p is the x-axis coordinate value of point p in the current frame's overlaid point cloud data.

[0070] For any point in the current frame overlay point cloud data: Select all points in the current frame's overlay point cloud data except the point in question. Based on the distances to the vehicle corresponding to each point in the current frame's overlay point cloud data and the distance to the vehicle corresponding to the point in question, calculate the distance difference between the other points in the current frame's overlay point cloud data and the point in question. The distance difference is the difference between the distance between one of the two points and the vehicle and the distance between the other point and the vehicle.

[0071] For any point in the current frame's overlaid point cloud data, determine whether the distance difference between the point and the vehicle is less than a preset first threshold. If the distance difference between the point and the vehicle is less than the preset first threshold, the point is considered the first closest point to the point. The first closest point to the point is selected from all other points in the current frame's overlaid point cloud data. Count the number of first closest points to the point.

[0072] For all points in the point cloud of the current frame superimposed point cloud data, the number of first similar points of each point is counted according to the above method.

[0073] It should be noted that the first threshold is a constant and can be set as needed or according to an application scenario, and this embodiment does not specifically limit this.

[0074] In a possible implementation, in step S132, selecting a static point that meets a second preset condition from the obstacle cluster includes: Selecting, from the obstacle point cloud data superimposed on the current frame, obstacle point cloud data of the current frame that satisfies a third preset condition, wherein the third preset condition is determined based on whether a point cloud in the obstacle point cloud data superimposed on the current frame is a point cloud in the foreground point cloud data of the current frame; For any point in the obstacle point cloud data superimposed on the current frame: Selecting, from all points in the obstacle point cloud data of the current frame, a point whose distance from the point to the vehicle is less than a preset second threshold as the second closest point to the point; Count the number of the second closest points to the point; According to the number of the second closest points to the point, the motion state of the point is determined to be static or dynamic; According to the motion state of each point in the obstacle point cloud data superimposed on the current frame, a static point is selected from the obstacle point cloud data superimposed on the current frame.

[0075] Specifically, when selecting a static point that meets the second preset condition from the obstacle cluster, it includes: For each point in the obstacle point cloud data superimposed on the current frame, some points come from the scenic spot point cloud data before the current frame, and some points come from the superposition of points in the scenic spot point cloud data before the historical frame.

[0076] For each point in the obstacle point cloud data overlaid on the current frame, a determination is made as to whether the point satisfies a third preset condition based on whether the point in the record is from the scenic spot cloud data for the current frame or from the scenic spot cloud data for a previous frame. If the point in the record is from the scenic spot cloud data for the current frame, the point is determined to be in the scenic spot cloud data for the current frame and satisfies the third preset condition. The third preset condition is set in advance.

[0077] From the obstacle point cloud data superimposed on the current frame, all points that meet the third preset condition are selected, that is, all points from the foreground point cloud data of the current frame are selected as the current obstacle point cloud data.

[0078] For any point in the current frame superimposed obstacle point cloud data obtained above: after the above step S1311, the distance difference from the vehicle between the point and each point in the current frame superimposed obstacle point cloud data has been calculated, that is, the distance difference from the vehicle between the point and each point in the current obstacle point cloud data has been calculated.

[0079] For any point in the current obstacle point cloud data, the difference in distance from the vehicle between the point in the current obstacle point cloud data and the point in the current superimposed obstacle point cloud data is compared. When the difference in distance from the vehicle is less than a preset second threshold, the point in the current obstacle point cloud data is used as the second closest point to the point in the current superimposed obstacle point cloud data.

[0080] From all points in the obstacle point cloud data of the current frame, points whose distance from the point to the vehicle is less than a preset second threshold are selected as the second closest points to the point.

[0081] Count the number of second-closest points to the point. Based on the number of second-closest points to the point, determine whether the point's motion state is stationary or moving. When the number of second-closest points to the point is greater than a preset third threshold, the point's motion state is determined to be stationary, and the point is considered a static point. When the number of second-closest points to the point is not greater than the preset third threshold, the point's motion state is determined to be moving, and the point is considered a dynamic point.

[0082] According to the motion state of each point in the obstacle point cloud data superimposed on the current frame, all static points are selected from the obstacle point cloud data superimposed on the current frame to form a static point cloud.

[0083] It should be noted that the third threshold is an artificially set constant and can also be set according to actual application scenarios. This embodiment does not make any specific limitation to this.

[0084] In a possible implementation, in step S11, obtaining the current frame foreground scenic spot cloud data at the current moment during the train operation includes: Obtaining the current frame point cloud data at the current moment during the train's operation, wherein the current frame point cloud data is used to represent the three-dimensional coordinate information of objects in the train's surrounding environment; Based on the train's own motion information, motion compensation is performed on the points in the current frame point cloud data to obtain the compensated current frame point cloud data, wherein the train's own motion information is estimated based on the inertial measurement data of the train during operation; Determine the train's running route based on the train's current position information and pre-acquired traffic track information, wherein the train's current position information is obtained by positioning the train based on the compensated current frame point cloud data; Determine the area to be inspected based on the train's route; The foreground point cloud data of the current frame is filtered out from the compensated point cloud data of the current frame according to the position information of the midpoint of the compensated point cloud data of the current frame and the area to be detected.

[0085] Specifically, when obtaining the scenic spot cloud data of the current frame at the current moment during the train operation, the method includes: The laser radar is used to obtain the current frame point cloud data of the train at the current moment during its operation, wherein the current frame point cloud data is used to represent the three-dimensional coordinate information of objects in the surrounding environment of the train at the current moment.

[0086] Based on the inertial measurement data obtained during the operation of the train, the train's own motion information is estimated. Based on the train's own motion information, motion compensation is performed on each point in the current frame point cloud data to obtain the compensated current frame point cloud data.

[0087] According to the compensated current frame point cloud data, the train can be positioned and the current position information of the train can be obtained.

[0088] The pre-acquired point cloud map contains track information, such as track boundaries and centerlines. Based on the train's current position and the pre-acquired track information in the point cloud map, the train's route is determined. Based on the train's route, the area to be inspected is determined. The area to be inspected is a three-dimensional boundary area. It can also be determined based on the train's width and route, or manually set. This embodiment does not impose any specific limitations on the setting of the area to be inspected.

[0089] According to the position information of each point in the compensated current frame point cloud data and the area to be detected, the points in the current frame point cloud data located outside the area to be detected are eliminated, and all the points in the current frame point cloud data located within the area to be detected are used to form the foreground point cloud data of the current frame.

[0090] It should be noted that when positioning a train, laser SLAM technology can be used to obtain a high-precision global pose, or other SLAM methods can be used to obtain the global pose, and an odometer can be used to obtain the pose transformation relationship between adjacent frames, such as through visual SLAM or multi-sensor fusion methods. This embodiment does not specifically limit this.

[0091] It can be understood that the technical solution provided by this embodiment obtains the current frame foreground point cloud data at the current moment during the operation of the train, and superimposes the points in the current frame foreground point cloud data with the points in the historical frame foreground point cloud data of k consecutive moments obtained in advance to obtain the current frame superimposed point cloud data corresponding to the current frame foreground point cloud data, wherein the current frame foreground point cloud data is used to represent the three-dimensional coordinate information of the object in the to-be-detected area at the current moment, and the to-be-detected area is determined according to the running line of the train, and the historical frame foreground point cloud data is used to represent the three-dimensional coordinate information of the object in the to-be-detected area at the historical moment before the current moment. , k is a constant. Since the current frame superimposed point cloud data incorporates the foreground point cloud data of k consecutive historical frames, the points in the current frame superimposed point cloud data are denser. According to the position information corresponding to each point in the current frame superimposed point cloud data with denser point clouds, rail transit obstacles are identified for the running train, so that the obstacle identification result of the train in the running line is more accurate, and the distance at which obstacles can be identified is farther, thereby solving the problem of low accuracy in detecting and identifying obstacles caused by the laser point cloud being too sparse when the train is stationary or when the moving train detects obstacles at a long distance.

[0092] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0093] Corresponding to the method described in the above embodiment, Figure 4 A structural schematic diagram of a rail transit obstacle identification device based on time-series point cloud superposition provided in an embodiment of the present application is shown. For the sake of convenience, only the parts related to the embodiment of the present application are shown.

[0094] Reference Figure 4 , the device comprises: The point cloud data acquisition module 31 is used to obtain the point cloud data of the current frame in front of the train at the current moment during the train operation, wherein the point cloud data of the current frame in front of the train is used to represent the three-dimensional coordinate information of the objects in the area to be detected at the current moment. The area to be detected is determined according to the train's running route; The point cloud data superposition module 32 is configured to superimpose points in the foreground point cloud data of the current frame with points in the foreground point cloud data of k consecutive historical frames previously acquired to obtain superimposed point cloud data of the current frame corresponding to the foreground point cloud data of the current frame, wherein the foreground point cloud data of the historical frames is used to represent the three-dimensional coordinate information of objects in the to-be-detected area at historical moments before the current moment, and k is a constant; The obstacle recognition module 33 is used to identify rail transit obstacles for the train in operation based on the position information corresponding to the points in the current frame superimposed point cloud data, and obtain the obstacle recognition results of the train in the operation line.

[0095] In one possible implementation, when the points in the foreground point cloud data of the current frame are superimposed with the points in the foreground point cloud data of the historical frames obtained at k consecutive moments in advance to obtain the current frame superimposed point cloud data corresponding to the foreground point cloud data of the current frame, the cloud data superimposition module 32 is specifically used to perform a first coordinate system conversion on the points in the foreground point cloud data of the historical frames obtained at k consecutive moments in advance according to a pre-constructed point cloud map to obtain k historical frames of first foreground point cloud data, wherein the pre-constructed point cloud map is determined based on a simultaneous positioning and map construction method according to the train's running route; perform a second coordinate system conversion on the points in the first foreground point cloud data of the k historical frames according to the coordinate system corresponding to the current frame foreground point cloud data to obtain k historical frames of second foreground point cloud data; and superimpose the points in the second foreground point cloud data of the k historical frames onto the foreground point cloud data of the current frame to obtain the current frame superimposed point cloud data corresponding to the foreground point cloud data of the current frame.

[0096] In one possible implementation, when rail transit obstacle identification is performed on a running train based on the position information corresponding to the points in the current frame superimposed point cloud data to obtain the obstacle identification result of the train on the running line, the obstacle identification module 33 is specifically configured to filter and cluster the points in the current frame superimposed point cloud data based on the position information corresponding to each point in the current frame superimposed point cloud data to obtain obstacle clusters; Selecting a static point that meets a second preset condition from the obstacle cluster, wherein the second preset condition is determined based on the motion state of each point in the obstacle cluster; Determine the obstacle bounding box based on the static points in the obstacle cluster; According to the obstacle bounding box, the obstacle recognition result of the train on the running line is determined.

[0097] In one possible implementation, when the points in the current frame superimposed point cloud data are filtered and clustered according to the position information corresponding to each point in the current frame superimposed point cloud data to obtain the obstacle clusters, the obstacle identification module 33 is specifically configured to count the number of first similar points of each point in the current frame superimposed point cloud data according to the position information corresponding to each point in the current frame superimposed point cloud data, wherein the first similar points of each point in the current frame superimposed point cloud data are point clouds in the current frame superimposed point cloud data that are in the same plane as the point; Filtering, from the current frame superimposed point cloud data, obstacle point cloud data of the current frame superimposed on the point cloud data that meets a first preset condition, based on the number of first similar points of each point in the current frame superimposed point cloud data, wherein the first preset condition is determined based on a preset threshold value of the number of similar points; Cluster all points in the obstacle point cloud data of the current frame to obtain multiple initial clusters; According to the number of points in each initial cluster, an obstacle cluster is selected from the multiple initial clusters.

[0098] In one possible implementation, when counting the number of first similar points of each point in the current frame of superimposed point cloud data based on the position information corresponding to each point in the current frame of superimposed point cloud data, the obstacle identification module 33 is specifically configured to determine the distance from the train corresponding to each point in the current frame of superimposed point cloud data based on the position information corresponding to each point in the current frame of superimposed point cloud data, wherein the distance from the train corresponding to each point in the current frame of superimposed point cloud data is the distance between the point and the train; For any point in the current frame overlay point cloud data: According to the distance from the vehicle corresponding to each point in the current frame superimposed point cloud data, calculate the difference in the distance from the vehicle between other points in the current frame superimposed point cloud data and the point, wherein the other points in the current frame superimposed point cloud data are the points in the current frame superimposed point cloud data other than the point; Selecting a point whose distance from the point to the vehicle is less than a preset first threshold from all other points in the current frame superimposed point cloud data as the first similar point to the point; Count the number of first closest points to the point.

[0099] In one possible implementation, when a static point that satisfies the second preset condition is selected from the obstacle cluster, the obstacle identification module 33 is specifically configured to select obstacle point cloud data of the current frame that satisfies a third preset condition from the obstacle point cloud data superimposed on the current frame, where the third preset condition is determined based on whether a point in the obstacle point cloud data superimposed on the current frame is a point in the foreground point cloud data of the current frame; For any point in the obstacle point cloud data superimposed on the current frame: Selecting, from all points in the obstacle point cloud data of the current frame, a point whose distance from the point to the vehicle is less than a preset second threshold as the second closest point to the point; Count the number of the second closest points to the point; According to the number of the second closest points to the point, the motion state of the point is determined to be static or dynamic; According to the motion state of each point in the obstacle point cloud data superimposed on the current frame, a static point is selected from the obstacle point cloud data superimposed on the current frame.

[0100] In a possible implementation, when acquiring the point cloud data of the foreground point of the current frame at the current moment during the operation of the train, the point cloud data acquisition module 31 is specifically used to acquire the point cloud data of the current frame at the current moment during the operation of the train, wherein the point cloud data of the current frame is used to represent the three-dimensional coordinate information of objects in the surrounding environment of the train; Based on the train's own motion information, motion compensation is performed on the points in the current frame point cloud data to obtain the compensated current frame point cloud data, wherein the train's own motion information is estimated based on the inertial measurement data of the train during operation; Determine the train's running route based on the train's current position information and pre-acquired traffic track information, wherein the train's current position information is obtained by positioning the train based on the compensated current frame point cloud data; Determine the area to be inspected based on the train's route; The foreground point cloud data of the current frame is filtered out from the compensated point cloud data of the current frame according to the position information of the midpoint of the compensated point cloud data of the current frame and the area to be detected.

[0101] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of this application. Their specific functions and technical effects can be found in the method embodiment section and will not be repeated here.

[0102] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0103] The present application also provides a rail transit obstacle recognition device based on time-series point cloud superposition, referring to Figure 5 , Figure 5 This is a structural diagram of a rail transit obstacle identification device based on time-series point cloud superposition provided by another embodiment of the present application. The terminal device 6 includes: at least one processor 62, a memory 61, and a computer program stored in the memory and capable of running on the at least one processor. When the processor executes the computer program, the steps in any of the above-mentioned method embodiments are implemented.

[0104] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned various method embodiments can be implemented.

[0105] An embodiment of the present application provides a computer program product. When the computer program product is run on a mobile terminal, the mobile terminal can implement the steps in the above-mentioned various method embodiments when executing the computer program product.

[0106] If the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a camera / terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, based on legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0107] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0108] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0109] In the embodiments provided in this application, it should be understood that the disclosed devices / network equipment and methods can be implemented in other ways. For example, the device / network equipment embodiments described above are merely illustrative. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0110] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0111] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application 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 replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A rail transit obstacle identification method based on time-series point cloud superposition, characterized in that: include: Acquire the current frame foreground scenic spot cloud data at the current moment during the running process of the train, wherein the current frame foreground scenic spot cloud data is used to represent the three-dimensional coordinate information of the object in the to-be-detected area at the current moment, and the to-be-detected area is determined according to the running route of the train; Superimpose the points in the foreground point cloud data of the current frame with the points in the foreground point cloud data of k consecutive historical frames acquired in advance, to obtain the superimposed point cloud data of the current frame corresponding to the foreground point cloud data of the current frame, wherein the foreground point cloud data of the historical frame is used to represent the three-dimensional coordinate information of the object in the to-be-detected area at the historical moment before the current moment, and k is a constant; According to the position information corresponding to the points in the current frame superimposed point cloud data, rail transit obstacle identification is performed on the train during operation to obtain the obstacle identification result of the train in the operation line.

2. The rail transit obstacle identification method based on time-series point cloud superposition according to claim 1, characterized in that: The step of superimposing the points in the foreground point cloud data of the current frame with the points in the foreground point cloud data of k consecutive historical frames acquired in advance to obtain the current frame superimposed point cloud data corresponding to the foreground point cloud data of the current frame includes: According to the pre-constructed point cloud map, the points in the pre-acquired foreground point cloud data of the historical frames at k consecutive moments are respectively converted into a first coordinate system to obtain k historical frames of first foreground point cloud data, wherein the pre-constructed point cloud map is determined based on the running route of the train based on the simultaneous positioning and map construction method; According to the coordinate system corresponding to the foreground point cloud data of the current frame, respectively perform a second coordinate system transformation on the points in the first foreground point cloud data of the k historical frames to obtain the second foreground point cloud data of the k historical frames; The points in the k second foreground point cloud data of the historical frames are superimposed on the foreground point cloud data of the current frame to obtain the superimposed point cloud data of the current frame corresponding to the foreground point cloud data of the current frame.

3. The rail transit obstacle identification method based on time-series point cloud superposition according to claim 1, characterized in that: The method of performing rail transit obstacle identification on the train in operation according to the position information corresponding to the points in the current frame superimposed point cloud data to obtain the obstacle identification result of the train in the operation line includes: According to the position information corresponding to each point in the current frame superimposed point cloud data, the points in the current frame superimposed point cloud data are filtered and clustered to obtain obstacle clusters; Selecting a static point that meets a second preset condition from the obstacle cluster, wherein the second preset condition is determined according to the motion state of each point in the obstacle cluster; Determining an obstacle bounding box according to the static points in the obstacle cluster; According to the obstacle bounding box, an obstacle recognition result of the train in the running line is determined.

4. The rail transit obstacle identification method based on time-series point cloud superposition according to claim 3 is characterized in that: The filtering and clustering of points in the current frame superimposed point cloud data according to the position information corresponding to each point in the current frame superimposed point cloud data to obtain obstacle clusters includes: According to the position information corresponding to each point in the current frame superimposed point cloud data, the number of first similar points of each point in the current frame superimposed point cloud data is counted, wherein the first similar point of each point in the current frame superimposed point cloud data is a point cloud in the current frame superimposed point cloud data that is in the same plane as the point; According to the number of the first similar points of each point in the current frame superimposed point cloud data, current frame superimposed obstacle point cloud data that meets a first preset condition is screened out from the current frame superimposed point cloud data, wherein the first preset condition is determined according to a preset threshold value of the number of similar points; Clustering all points in the obstacle point cloud data superimposed on the current frame to obtain a plurality of initial clustering clusters; The obstacle cluster is selected from the multiple initial clusters according to the number of points in each of the initial clusters.

5. The rail transit obstacle identification method based on time-series point cloud superposition according to claim 4, characterized in that: The counting of the number of first similar points of each point in the current frame superimposed point cloud data according to the position information corresponding to each point in the current frame superimposed point cloud data includes: Determine the distance from the train corresponding to each point in the current frame superimposed point cloud data according to the position information corresponding to each point in the current frame superimposed point cloud data, wherein the distance from the train corresponding to each point in the current frame superimposed point cloud data is the distance between the point and the train; For any point in the current frame overlay point cloud data: According to the distance from the vehicle corresponding to each point in the current frame superimposed point cloud data, calculate the difference in the distance from the vehicle between other points in the current frame superimposed point cloud data and the point, wherein the other points in the current frame superimposed point cloud data are points in the current frame superimposed point cloud data other than the point; Selecting, from all other points in the current frame superimposed point cloud data, a point whose distance from the point to the vehicle is less than a preset first threshold value as the first similar point to the point; Count the number of the first similar points to the point.

6. The rail transit obstacle identification method based on time-series point cloud superposition according to claim 4, characterized in that: The step of selecting a static point satisfying a second preset condition from the obstacle cluster includes: Selecting the current frame obstacle point cloud data that meets a third preset condition from the current frame superimposed obstacle point cloud data, wherein the third preset condition is determined based on whether a point in the current frame superimposed obstacle point cloud data is a point in the current frame foreground point cloud data; For any point in the obstacle point cloud data superimposed on the current frame: Selecting, from all points in the obstacle point cloud data of the current frame, a point whose difference in distance from the point to the vehicle is less than a preset second threshold value as a second similar point to the point; Count the number of the second similar points to the point; According to the number of the second similar points to the point, determining whether the motion state of the point is static or dynamic; According to the motion state of each point in the obstacle point cloud data superimposed on the current frame, a static point is selected from the obstacle point cloud data superimposed on the current frame.

7. The rail transit obstacle identification method based on time-series point cloud superposition according to any one of claims 1 to 6, characterized in that: The step of obtaining the scenic spot cloud data in front of the current frame at the current moment during the operation of the train includes: Acquire current frame point cloud data at a current moment during the operation of the train, wherein the current frame point cloud data is used to represent three-dimensional coordinate information of objects in the surrounding environment of the train; According to the self-motion information of the train, motion compensation is performed on the points in the current frame point cloud data to obtain compensated current frame point cloud data, wherein the self-motion information of the train is estimated based on inertial measurement data of the train during operation; Determine the running route of the train according to the current position information of the train and the pre-acquired traffic track information, wherein the current position information of the train is obtained by positioning the train according to the compensated current frame point cloud data; Determining the area to be detected according to the running route of the train; The foreground point cloud data of the current frame is filtered out from the compensated point cloud data of the current frame according to the position information of the midpoint of the compensated point cloud data of the current frame and the area to be detected.

8. A rail transit obstacle identification device based on time-series point cloud superposition, characterized in that: include: A point cloud data acquisition module, used to acquire the point cloud data of the current frame in front of the train at the current moment during its operation, wherein the point cloud data of the current frame in front of the train is used to represent the three-dimensional coordinate information of the object in the area to be detected at the current moment, and the area to be detected is determined according to the running route of the train; A point cloud data superposition module is used to superimpose points in the foreground point cloud data of the current frame with points in the foreground point cloud data of k consecutive moments acquired in advance, so as to obtain current frame superimposed point cloud data corresponding to the foreground point cloud data of the current frame, wherein the foreground point cloud data of the historical frame is used to represent the three-dimensional coordinate information of the object in the to-be-detected area at the historical moment before the current moment, and k is a constant; The obstacle recognition module is used to identify rail transit obstacles for the train in operation according to the position information corresponding to the points in the superimposed point cloud data of the current frame, and obtain the obstacle recognition result of the train in the operation line.

9. A rail transit obstacle identification device based on time-series point cloud superposition, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

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

Images