Label data acquisition method and device, electronic equipment, medium and computer program

Abstract

The application relates to the technical field of data processing, in particular to a labeled data acquisition method and device, electronic equipment, medium and computer program, wherein the method comprises the following steps: acquiring point cloud sequence data of a target object; calculating a mathematical model of a radar according to the point cloud sequence data of the target object; and labeling public road point cloud data based on the mathematical model to obtain labeled data of the target object. Therefore, the problems that the cost is high, the time is long and the accuracy is poor when a laser radar is used to label the target object around a vehicle in the related art are solved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, electronic device, medium and computer program for acquiring labeled data. Background Technology

[0002] In autonomous driving scenarios, LiDAR is typically used to detect objects around the vehicle and identify obstacles based on the detection results, thereby enabling autonomous driving. However, LiDAR data is usually point cloud data, making it difficult to determine the boundaries of target objects. This requires a significant amount of time and effort for manual identification and suffers from low labeling accuracy. Summary of the Invention

[0003] This application provides a data acquisition device, electronic device, storage medium, and computer program for labeling data, in order to solve the problems of high cost, long time consumption, and poor accuracy in labeling target objects around vehicles using LiDAR.

[0004] The first aspect of this application provides a method for acquiring labeled data, including the following steps: acquiring point cloud sequence data of a target object; calculating a mathematical model of a radar based on the point cloud sequence data of the target object; and labeling the point cloud data of a public road based on the mathematical model to obtain labeled data of the target object.

[0005] Optionally, in one embodiment of this application, the annotation data of the target object is obtained by annotating the public road point cloud data based on the mathematical model, including: repeatedly repositioning the target object into the public road point cloud data, and removing the point cloud coordinates of the target object during each placement process to obtain a new point cloud map, while obtaining the position and orientation of the target object; generating annotation data based on the new point cloud map, the position and orientation of the target object, and the mathematical model.

[0006] Optionally, in one embodiment of this application, removing the point cloud coordinates of the target object to obtain a new point cloud map during each placement process includes: traversing public road point cloud data to determine a target area that meets preset conditions; determining the lateral constraint range of the radar scanning beam based on the target area; determining the point cloud coordinates of the target object based on the target area and the lateral constraint range, and removing the point cloud coordinates to obtain a new point cloud map.

[0007] Optionally, in one embodiment of this application, determining the point cloud coordinates of the target object based on the target region and the lateral constraint range includes: determining the lateral angle and vertical angle of each scan beam based on the lateral constraint range; constructing a straight line equation based on the lateral angle and vertical angle; and determining the intersection position of the target object in the point cloud grid based on the straight line equation to obtain the point cloud coordinates.

[0008] Optionally, in one embodiment of this application, obtaining point cloud sequence data of a target object includes: collecting raw point cloud data of the target object at its center point; converting the raw point cloud data to the center point coordinate system of the target object; and performing meshing processing on the raw point cloud data in the center point coordinate system to obtain point cloud sequence data of the target object.

[0009] Optionally, in one embodiment of this application, the point cloud data of the target object is obtained by meshing the original point cloud data, including: acquiring the surrounding rectangle of the target object acquired by the radar and calculating the height value of the target object; calculating the number of grids in the axial and auxiliary directions of the target object based on the surrounding rectangle and the height value; and generating the point cloud sequence data of the target object based on the number of grids.

[0010] A second aspect of this application provides a data annotation acquisition device, comprising: an acquisition module for acquiring point cloud sequence data of a target object; an establishment module for calculating a mathematical model of a radar based on the point cloud sequence data of the target object; and an annotation module for annotating the point cloud data of a public road based on the mathematical model to obtain annotated data of the target object.

[0011] Optionally, in one embodiment of this application, the annotation module is further configured to: repeatedly reposition the target object into the public road point cloud data, and while removing the point cloud coordinates of the target object to obtain a new point cloud map during each repositioning process, obtain the position and orientation of the target object; and generate annotation data based on the new point cloud map, the position and orientation of the target object, and the mathematical model.

[0012] Optionally, in one embodiment of this application, the annotation module is further configured to: traverse public road point cloud data to determine target areas that meet preset conditions; determine the lateral constraint range of the radar scanning beam based on the target area; determine the point cloud coordinates of the target object based on the target area and the lateral constraint range; and remove the point cloud coordinates to obtain a new point cloud map.

[0013] Optionally, in one embodiment of this application, the annotation module is further configured to: determine the point cloud coordinates of the target object based on the target region and the lateral constraint range, including: determining the lateral angle and vertical angle of each scan line bundle based on the lateral constraint range; constructing a straight line equation based on the lateral angle and vertical angle; and determining the intersection position of the target object in the point cloud grid based on the straight line equation to obtain the point cloud coordinates.

[0014] Optionally, in one embodiment of this application, the acquisition module is further configured to: collect the original point cloud data of the target object at the center point position; convert the original point cloud data to the center point coordinate system of the target object, and perform gridding processing on the original point cloud data in the center point coordinate system to obtain the point cloud sequence data of the target object.

[0015] Optionally, in one embodiment of this application, the acquisition module is further configured to: acquire the surrounding rectangle of the radar-acquired target object and calculate the height value of the target object; calculate the number of grids in the axial and auxiliary directions of the target object based on the surrounding rectangle and the height value; and generate point cloud sequence data of the target object based on the number of grids.

[0016] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to perform the annotation data acquisition method as described in the above embodiments.

[0017] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to perform the annotation data acquisition method as described in the above embodiments.

[0018] A fifth aspect of this application provides a computer program that, when executed, is used to implement the annotation data acquisition method as described in the above embodiments.

[0019] Therefore, this application has at least the following beneficial effects:

[0020] This application embodiment can establish a radar mathematical model through the point cloud sequence data of the target object, and obtain labeled data by labeling public road data based on the mathematical model, thereby improving the accuracy of labeling. It eliminates the need for manual time to determine the boundary, thereby reducing the cost of labeling. Thus, it solves the problems of high cost, long time consumption and poor accuracy in labeling target objects around vehicles using lidar in related technologies.

[0021] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0022] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0023] Figure 1 This is a flowchart of the annotation data acquisition method provided according to the embodiments of this application;

[0024] Figure 2 This is a top view of a target object provided according to an embodiment of this application;

[0025] Figure 3 This is a schematic side view of a target object according to an embodiment of this application;

[0026] Figure 4This is a flowchart illustrating the point cloud data acquisition process for a target object according to an embodiment of this application.

[0027] Figure 5 This is an example diagram of a lidar scanning model provided according to an embodiment of this application;

[0028] Figure 6 This is a schematic diagram of a laser projection model calculation according to an embodiment of this application;

[0029] Figure 7 This is a flowchart of a single-beam projection point cloud computing process according to an embodiment of this application;

[0030] Figure 8 This is a flowchart of a labeled point cloud generation process according to an embodiment of this application;

[0031] Figure 9 This is a block diagram of a label data acquisition device provided according to an embodiment of this application;

[0032] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0033] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0034] The annotation data acquisition method, apparatus, electronic device, storage medium, and computer program of this application are described below with reference to the accompanying drawings. Addressing the problems mentioned in the background section, this application provides an annotation data acquisition method. In this method, a radar mathematical model is established using point cloud sequence data of the target object. The mathematical model is then combined with publicly available road data to obtain annotation data, improving the accuracy of annotation. This eliminates the need for manual boundary determination, thereby reducing annotation costs. Thus, it solves the problems of high cost, long processing time, and poor accuracy associated with using lidar to annotate target objects around vehicles in related technologies.

[0035] Specifically, Figure 1 This is a flowchart illustrating a method for acquiring labeled data provided in an embodiment of this application.

[0036] like Figure 1 As shown, the method for obtaining labeled data includes the following steps:

[0037] In step S101, the point cloud sequence data of the target object is acquired.

[0038] In one embodiment of this application, obtaining point cloud sequence data of a target object includes: collecting raw point cloud data of the target object at its center point; converting the raw point cloud data to the center point coordinate system of the target object; and performing gridding processing on the raw point cloud data in the center point coordinate system to obtain point cloud sequence data of the target object.

[0039] To facilitate subsequent reconstruction, embodiments of this application may select, as follows: Figure 2 and Figure 3 The target object is placed as shown. This includes the following aspects:

[0040] 1) Select an open area, draw the center point of the area, and draw the horizontal and vertical axes.

[0041] 2) Place the target object on an open, level road surface.

[0042] 3) Place the center of the target object at the center point of an open road.

[0043] Furthermore, Figure 2 Example diagram showing the top-down view of the target object, such as... Figure 2 As shown, O is the center point, X is the main axis direction, Y is the auxiliary axis direction, the solid line rectangle is the horizontal flat road, and the dashed line rectangle is the box for placing the target object, which is placed at the center point O and aligned with the main axis.

[0044] Figure 3 Example diagram of a side view of the target object, such as Figure 3 As shown, O is the center point, X is the main axis direction, Y is the auxiliary axis direction, the solid line is a horizontal and flat road, and the dashed rectangle is the box for placing the target object, which is placed at the center point O and aligned with the main axis.

[0045] Furthermore, such as Figure 4 As shown in the embodiment of this application, after obtaining the target object to be calibrated, the original point cloud data of the target object can be collected first. Then, an autonomous vehicle (or other motion robot) equipped with a lidar and positioning device is configured to run around the target object 360 degrees, while recording the lidar point cloud, timestamp, and autonomous vehicle attitude information.

[0046] Furthermore, since it is an open road surface, the original point cloud data of the target object can be obtained based on the ROI region. In actual execution, the ROI region can be obtained through calibration in this embodiment of the application.

[0047] pts = pts_ori * roi

[0048] Where pts_ori is the point cloud data collected in the vehicle coordinate system, and roi is the mask of the calibrated target area. The coordinates of the point in the world coordinate system are calculated using the following formula.

[0049] world_pts = pose * pts

[0050] Where pose represents the position of the lidar in the UTM world coordinate system, which can be obtained by multiplying the pose information of the inertial navigation device with the lidar's extrinsic parameters, as shown in the formula:

[0051] pose=pose_rtk*extrincs_lidar

[0052] The pose information of the inertial navigation device can be obtained through existing measurement methods, enabling precise positioning of the device in the world UTM coordinate system. The extrinsic parameters of the lidar can be obtained through calibration.

[0053] Since all point clouds have been transformed to the world coordinate system, the combination is simply the direct addition of the sets, as shown in the following formula:

[0054] out_pt=union{worlds_pt0,{worlds_pt1,{worlds_pt2,...,{worlds_ptn}

[0055] In actual implementation, the embodiments of this application first calculate the position of the target object. Since the placement position and orientation of the target object are fixed, the target pose can be obtained through calibration and is denoted as pose_self. The coordinate system transformation is obtained by performing matrix multiplication on all target points. The calculation formula is as follows:

[0056] self_pts=pose_self*world_pts

[0057] Furthermore, embodiments of this application can obtain point cloud sequence data of a target object by performing meshing processing on all the original point cloud data, including: acquiring the surrounding rectangle of the target object acquired by radar and calculating the height value of the target object; calculating the number of grids in the axial and auxiliary directions of the target object based on the surrounding rectangle and the height value; and generating point cloud sequence data of the target object based on the number of grids.

[0058] For example, in this embodiment of the application, the mesh can be set to 1 mm. Each meshed cell stores the xyzitr value of the point located in that cell, and the calculation formula is as follows:

[0059] meshes[idx]=[pt0,pt1,pt2,...]

[0060] Here, `meshes` is an array consisting of all the meshes of the target object. The size of the target object is obtained by the following algorithm: First, determine the step size in the axial (x), auxiliary (y), and vertical (z) directions, and record it as step_x, step_y, and step_z. Calculate the rectangle around the target object along the axial direction, and denote the length of this rectangle as L and the width as W. Calculate the height of the target object, defining the height H as the difference between the maximum and minimum height values ​​of the point cloud. Based on this rectangle, calculate the number of meshes in the axial and auxiliary directions. The calculation formula is as follows:

[0061] mesh_size_x = L / step_x. If mesh_size_x is not an integer, it needs to be rounded up. Similarly, mesh_size_y and mesh_size_z can be obtained, which will not be elaborated here.

[0062] In step S102, a mathematical model of the radar is established based on the point cloud sequence data of the target object.

[0063] The point cloud sequence data is obtained by continuous acquisition or multiple placements and repeated execution of step S101. After the point cloud sequence data of the target road surface is acquired, a radar mathematical model of the target is established.

[0064] like Figure 5 As shown, the basic lidar model is the point-fire model shown in the figure below, which includes the following key parameters: horizontal scanning step size angle_h, scanning period T, and longitudinal beam angle angle_vs = [a0,a1,a2,...,an].

[0065] The lidar operates by scanning horizontally in steps of angle_h, and simultaneously emitting beams of line energy at all vertical angles within angle_vs. For example, a 64-line lidar emits 64 vertical beams. The lidar completes a 360-degree horizontal scan within a preset scan period T. This preset scan period can be set according to actual conditions, such as 100ms, without specific limitation. In a 100ms lidar scan period, the vehicle's motion can be considered as uniform, and this motion value can be derived from the lidar's pose information.

[0066] Furthermore, embodiments of this application can calculate the intensity value of the actual point of the lidar, such as... Figure 6 As shown, in this embodiment, the current position of the lidar center (xl, yl, zl) is first determined. The horizontal and vertical positions of the lidar beam emitted at the current moment are calculated to be known values; therefore, the equation of the beam's straight line is:

[0067] x / cos^2α)=y / sinαcosα=z / (tanβ cosα)

[0068] This equation is in the lidar coordinate system, with the lidar as the origin, the scanning start point as the horizontal axis, and the vertical direction as the vertical axis. Here, α is the horizontal scanning angle, and β is the vertical scanning angle; these two angles are known values ​​during the lidar's movement.

[0069] In actual implementation, the embodiments of this application can calculate the intersection point with the grid based on the linear equation. For example, points can be obtained along the linear equation with a fixed x-axis step size step_x, where the length of step_x must be less than the grid value. For the obtained points, it is calculated whether they intersect with the target grid in the corresponding x-position plane, which is divided into the following aspects:

[0070] 1) If they intersect, it means that the corresponding grid has been found;

[0071] 2) If they do not intersect, continue with the ascending steps until the target grid is found;

[0072] 3) If the maximum x value of the target grid is exceeded, the search is terminated.

[0073] Furthermore, for each line bundle, calculate the position of the first intersection point between the line bundle and the reconstructed point cloud mesh in the above figure. The intensity value of this point can be replaced by the average intensity value of all points in that mesh, such as... Figure 6 As shown in the grid, the wire bundle intersects with the grid at a point within the first point cloud data.

[0074] It should be noted that in the embodiments of this application, point cloud points are formed at the intersection locations, and the coordinates are the coordinates of that point. The intensity value of that point is given by the average of all points in the grid, the ring value of that point is given by the beam scanning angle corresponding to the lidar, and the timestamp of that point is given by the time when the lidar outputs the beam.

[0075] In summary, as Figure 7 As shown, for each scanning beam of the lidar, the horizontal scanning angle α and vertical angle β of the lidar are determined. Based on the horizontal scanning angle α and vertical angle β, a straight line equation is constructed. Through the straight line equation, the intersecting rectangular grid box is searched to determine the position of the intersection point. At the same time, the intensity value, ring value and timestamp value of the point can be calculated.

[0076] In step S103, the publicly available road point cloud data is labeled based on a mathematical model to obtain the labeled data of the target object.

[0077] In this embodiment of the application, point cloud sequence data can be understood as public road point cloud map. When collecting public road point cloud map, this embodiment of the application can simultaneously record the LiDAR attitude and LiDAR scanning time when the vehicle is moving.

[0078] In one embodiment of this application, the annotation data of the target object is obtained by annotating public road point cloud data based on a mathematical model, including: repeatedly repositioning the target object into the public road point cloud data.

[0079] During each placement process, the point cloud coordinates of the target object are removed to obtain a new point cloud map, while the position and orientation of the target object are acquired; annotation data is generated based on the new point cloud map, the position and orientation of the target object, and the mathematical model.

[0080] It is understood that, according to the radar motion recorded at the corresponding point cloud time, the present application embodiment can calculate the generated point cloud map of the target object at the current time. The calculation process is as follows: obtain the current point cloud time of the vehicle, the actual time t0, the vehicle pose information p0, the scanning period T, and the scanning step size angle angle_h.

[0081] Furthermore, in this embodiment of the application, the point cloud coordinates of the area where the generated point cloud map is located can be removed, and then placed on the road surface to form a new point cloud map. The new point cloud map and the position and orientation information of the target object are exported to complete the generation of new annotation data.

[0082] In one embodiment of this application, removing the point cloud coordinates of the target object to obtain a new point cloud map during each placement process includes: traversing public road point cloud data to determine a target area that meets preset conditions; determining the lateral constraint range of the radar scanning beam based on the target area; determining the point cloud coordinates of the target object based on the target area and the lateral constraint range, and removing the point cloud coordinates to obtain a new point cloud map.

[0083] It is understandable that, such as Figure 8 As shown, this application embodiment can traverse publicly available road point cloud data to find regions that meet the following conditions: the region contains only road surface points or approximate road surface points, and the judgment criterion is that the average z-axis height and std value of all points in the region meet the following conditions. Here, the region refers to the area enclosed by the smallest surrounding rectangle in the 3D point cloud image of the target object.

[0084] After selecting the placement point, a reasonable angle can be chosen to place the target object. The pose information p1 of the target vehicle at this time is recorded. For the found area, this embodiment of the application can calculate the generated point cloud map of the target at the current time based on the movement of the vehicle at the corresponding point cloud time recorded, and obtain the current point cloud time t0, the vehicle pose information p0, the scanning period T, and the scanning step length angle angle_h.

[0085] Furthermore, in this embodiment of the application, the scanning beam range at each of the multiple placement positions can be determined, and a lateral step can be taken within this range each time. The timestamp t1 of the step is calculated as follows:

[0086] t1 = angle / 360.0 + t0

[0087] Here, angle is the current horizontal angle. For each horizontal step, a corresponding beam is generated. For example, for a 64-line LiDAR, 64 beam rays are generated according to the configuration parameters. Based on the intersection of all scanning beams within the scanning beam range and the second point cloud data, the point cloud coordinates of the target object at each placement position are determined.

[0088] Furthermore, in this embodiment of the application, the point cloud coordinates of the area where the generated point cloud map is located can be removed, and then placed on the road surface to form a new point cloud map. After completion, the above steps can be repeated to place the target object multiple times to generate new annotation data.

[0089] According to the annotation data acquisition method proposed in the embodiments of this application, a mathematical model of radar is established through the point cloud sequence data of the target object, and the annotation data is obtained by combining the mathematical model with public road data, thereby improving the accuracy of annotation and eliminating the need for manual time to determine the boundary, thus reducing the cost of annotation. This solves the problems of high cost, long time consumption and poor accuracy in the related technology of using LiDAR to annotate target objects around vehicles.

[0090] Next, the annotation data acquisition device proposed according to the embodiments of this application is described with reference to the accompanying drawings.

[0091] Figure 9 This is a block diagram of the annotation data acquisition device according to an embodiment of this application.

[0092] like Figure 9 As shown, the annotation data acquisition device 10 includes: an acquisition module 100, an establishment module 200, and an annotation module 300.

[0093] The acquisition module 100 is used to acquire point cloud sequence data of the target object; the establishment module 200 is used to calculate the radar mathematical model based on the point cloud sequence data of the target object; and the annotation module 300 is used to annotate the public road point cloud data based on the mathematical model to obtain the annotation data of the target object.

[0094] In one embodiment of this application, the annotation module 300 is further configured to: repeatedly reposition the target object into the public road point cloud data, and while removing the point cloud coordinates of the target object to obtain a new point cloud map during each repositioning process, obtain the position and orientation of the target object; and generate annotation data based on the new point cloud map, the position and orientation of the target object, and the mathematical model.

[0095] In one embodiment of this application, the annotation module 300 is further configured to: traverse public road point cloud data to determine target areas that meet preset conditions; determine the lateral constraint range of the radar scanning beam based on the target area; determine the point cloud coordinates of the target object based on the target area and the lateral constraint range; and remove the point cloud coordinates to obtain a new point cloud map.

[0096] In one embodiment of this application, the annotation module 300 is further configured to: determine the point cloud coordinates of the target object based on the target region and the lateral constraint range, including: determining the lateral angle and vertical angle of each scan line bundle based on the lateral constraint range; constructing a straight line equation based on the lateral angle and vertical angle; and determining the intersection position of the target object in the point cloud grid based on the straight line equation to obtain the point cloud coordinates.

[0097] In one embodiment of this application, the acquisition module 100 is further configured to: acquire the original point cloud data of the target object at the center point position; convert the original point cloud data to the center point coordinate system of the target object, and perform gridding processing on the original point cloud data in the center point coordinate system to obtain the point cloud sequence data of the target object.

[0098] In one embodiment of this application, the acquisition module 100 is further configured to: acquire the surrounding rectangle of the radar-acquired target object and calculate the height value of the target object; calculate the number of grids in the axial and auxiliary directions of the target object based on the surrounding rectangle and the height value; and generate point cloud sequence data of the target object based on the number of grids.

[0099] It should be noted that the foregoing explanation of the embodiment of the annotation data acquisition method also applies to the annotation data acquisition device of this embodiment, and will not be repeated here.

[0100] According to the annotation data acquisition device proposed in the embodiments of this application, a mathematical model of radar is established through the point cloud sequence data of the target object, and the annotation data is obtained by combining the mathematical model with public road data, thereby improving the accuracy of annotation and eliminating the need for manual time to determine the boundary, thus reducing the cost of annotation. This solves the problems of high cost, long time consumption and poor accuracy in the related technology of using lidar to annotate target objects around vehicles.

[0101] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0102] The memory 1001, the processor 1002, and the computer program stored on the memory 1001 and capable of running on the processor 1002.

[0103] When the processor 1002 executes the program, it implements the annotation data acquisition method provided in the above embodiments.

[0104] Furthermore, electronic devices also include:

[0105] Communication interface 1003 is used for communication between memory 1001 and processor 1002.

[0106] The memory 1001 is used to store computer programs that can run on the processor 1002.

[0107] The memory 1001 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0108] If the memory 1001, processor 1002, and communication interface 1003 are implemented independently, then the communication interface 1003, memory 1001, and processor 1002 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0109] Optionally, in a specific implementation, if the memory 1001, processor 1002, and communication interface 1003 are integrated on a single chip, then the memory 1001, processor 1002, and communication interface 1003 can communicate with each other through an internal interface.

[0110] The processor 1002 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0111] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for acquiring labeled data.

[0112] This application also provides a computer program, which, when executed, is used to implement the annotation data acquisition method as described in the above embodiments.

[0113] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0114] Furthermore, the terms "first" and "sequence" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "sequence" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0115] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0116] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0117] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

Claims

1. A method for acquiring labeled data, characterized in that, The method comprises the following steps: obtaining point cloud sequence data of a target object; calculating a radar mathematical model according to the point cloud sequence data of the target object; annotating public road point cloud data based on the radar mathematical model to obtain annotation data of the target object; the annotation data of the target object obtained by annotating the public road point cloud data based on the radar mathematical model comprises: repeatedly placing the target object in the public road point cloud data multiple times, and obtaining the position and pose of the target object while removing the point cloud coordinates of the target object in each placement process to obtain a new point cloud image; and generating the annotation data according to the new point cloud image, the position and pose of the target object, and the radar mathematical model; the new point cloud image obtained by removing the point cloud coordinates of the target object in each placement process comprises: traversing the public road point cloud data to determine a target region that meets a preset condition; determining a transverse constraint range of a scan beam of the radar according to the target region; and determining the point cloud coordinates of the target object according to the target region and the transverse constraint range, and removing the point cloud coordinates to obtain the new point cloud image; the point cloud coordinates of the target object determined according to the target region and the transverse constraint range comprise: determining the transverse angle and the vertical angle of each scan beam according to the transverse constraint range; constructing a straight line equation according to the transverse angle and the vertical angle; and determining the intersection position of the target object in the point cloud grid to obtain the point cloud coordinates according to the straight line equation.

2. The labeled data acquisition method of claim 1, wherein, the point cloud sequence data of the target object comprises: collecting original point cloud data of the target object at a center point position; converting the original point cloud data to a center point coordinate system of the target object, and performing grid processing on the original point cloud data in the center point coordinate system to obtain the point cloud sequence data of the target object.

3. The labeled data acquisition method of claim 2, wherein, the point cloud sequence data of the target object obtained by performing grid processing on the original point cloud data comprises: obtaining a surrounding rectangle of the target object collected by the radar, and calculating a height value of the target object; calculating the number of grids in the axial direction and the auxiliary direction of the target object according to the surrounding rectangle and the height value; generating the point cloud sequence data of the target object according to the number of grids.

4. The labeled data acquisition apparatus according to claim 1, wherein comprise: an acquisition module configured to acquire point cloud sequence data of a target object; an establishment module configured to calculate a radar mathematical model according to the point cloud sequence data of the target object; an annotation module configured to annotate public road point cloud data based on the radar mathematical model to obtain annotation data of the target object; the annotation module is further configured to: repeatedly place the target object in the public road point cloud data multiple times, and obtain the position and pose of the target object while removing the point cloud coordinates of the target object in each placement process to obtain a new point cloud image; generate the annotation data according to the new point cloud image, the position and pose of the target object, and the radar mathematical model; The labeling module is further configured to traverse the public road point cloud data to determine a target region satisfying a preset condition, determine a transverse constraint range of a scan line beam of the radar according to the target region, determine a point cloud coordinate of the target object according to the target region and the transverse constraint range, and remove the point cloud coordinate to obtain the new point cloud map. The labeling module is further configured to determine a transverse angle and a vertical angle of each scan line beam according to the transverse constraint range, construct a straight line equation according to the transverse angle and the vertical angle, and determine an intersection position of the target object in a point cloud grid according to the straight line equation to obtain the point cloud coordinate.

5. An electronic device, comprising: The computer program comprises: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the labeling data acquisition method according to any one of claims 1-3.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the labeling data acquisition method according to any one of claims 1-3.

7. A computer program, characterised in that, The computer program is executed to implement the labeling data acquisition method according to any one of claims 1-3.

Images