Vehicle visual truth value acquisition method and device, vehicle, medium and program

Abstract

The application relates to the technical field of vehicles, in particular to a vehicle visual truth value acquisition method and device, a vehicle, a medium and a program, wherein the method comprises the following steps: acquiring image data of a target vehicle at a target moment and a plurality of frames of image data corresponding to the same timestamp of a sampling vehicle and the target vehicle before and after the target moment; identifying annotation data of the same three-dimensional target in the image data at the target moment and the plurality of frames of image data; converting the annotation data of the three-dimensional target of the sampling vehicle to the pose of the corresponding three-dimensional target in the coordinate system of the target vehicle; and generating the visual truth value in the coordinate system of the target vehicle according to the image data at the target moment and the converted pose of the three-dimensional target. Therefore, the problems in the prior art that different deep learning models are trained for different vehicle types to identify image data, which leads to high cost, repeated collection of data, waste of manpower and material resources, and the same deep learning model cannot be applied to multiple vehicles, which leads to low applicability, are solved.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and in particular to a method, device, vehicle, medium and program for obtaining vehicle visual truth. Background Technology

[0002] Currently, the generation of ground truth data in the field of autonomous driving often relies on data collection devices installed on the collection vehicle, such as LiDAR, cameras, and millimeter-wave radar. Data is collected through these sensors, labeled manually, and then used for model training. However, deep learning models trained on data collected on the collection vehicle can often only be applied to the same vehicle type. If other vehicle types are used, the model often performs very poorly. Essentially, this is because the datasets used in the training and testing processes of the model do not come from the same sample distribution space.

[0003] In related technologies, training different deep learning models for different vehicle models requires installing a set of data acquisition equipment on each vehicle model, especially since LiDAR is very expensive. Data needs to be collected and labeled repeatedly, wasting time and manpower. The cost increases exponentially with each new vehicle model. Therefore, training different deep learning models for different vehicle models to recognize image data results in high costs, requires repeated data collection which wastes manpower and resources, and the same deep learning model cannot be applied to multiple vehicles, resulting in low applicability. Summary of the Invention

[0004] This application provides a method, apparatus, vehicle, storage medium, and program product for obtaining vehicle visual ground truth, in order to solve the problems in related technologies such as high cost due to training different deep learning models to recognize image data for different vehicle models, waste of manpower and resources due to repeated collection, and low applicability due to the inability of the same deep learning model to be applicable to multiple vehicles.

[0005] The first aspect of this application provides a method for obtaining vehicle visual ground truth, comprising the following steps: acquiring image data of a target vehicle at a target time and multiple frame image data of a sampled vehicle corresponding to the same timestamp as the target vehicle; identifying the annotation data of the same three-dimensional target in the image data at the target time and the multiple frame image data; converting the annotation data of the three-dimensional target of the sampled vehicle to the pose of the corresponding three-dimensional target in the target vehicle coordinate system; and generating visual ground truth in the target vehicle coordinate system based on the image data at the target time and the converted pose of the three-dimensional target.

[0006] Optionally, the step of converting the labeled data of the three-dimensional target of the sampling vehicle to the pose of the corresponding three-dimensional target in the target vehicle coordinate system includes: identifying the extrinsic parameters and positioning information of the sensors and cameras of the sampling vehicle; converting the labeled data of the three-dimensional target of the sampling vehicle from the sampling vehicle coordinate system to the labeled data of the three-dimensional target in the world coordinate system according to the extrinsic parameters and the positioning information, and determining the actual pose of the three-dimensional target in the world coordinate system according to the labeled data; and converting the actual pose of the three-dimensional target in the world coordinate system into the pose of the corresponding three-dimensional target in the target vehicle coordinate system.

[0007] Optionally, determining the actual pose of the 3D target in the world coordinate system based on the labeled data further includes: calculating the rotation angle of the 3D target pose in the labeled data for multiple consecutive frames using quaternion spherical interpolation; calculating the translation vector of the 3D target pose in the labeled data for multiple consecutive frames using linear interpolation; and determining the actual pose of the 3D target in the world coordinate system corresponding to the same timestamp of the sampled vehicle and the target vehicle based on the rotation angle and translation vector.

[0008] Optionally, converting the actual pose of the three-dimensional target in the world coordinate system into the pose of the corresponding three-dimensional target in the target vehicle coordinate system includes: identifying the relative position of the sampling vehicle and the target vehicle, as well as the sensor and camera extrinsic parameters of the target vehicle; and converting the actual pose of the three-dimensional target in the world coordinate system into the pose of the corresponding three-dimensional target in the target vehicle coordinate system based on the relative position and the sensor and camera extrinsic parameters of the target vehicle.

[0009] Optionally, before identifying the annotation data of the same three-dimensional target in the image data and multi-frame image data at the target time, the method includes: preprocessing all three-dimensional targets in the image data and multi-frame image data at the target time; inputting the preprocessed image data into the annotation model, and the annotation model outputting the annotation data of the corresponding three-dimensional target, wherein the annotation data includes the pose, size, yaw angle and target name of the three-dimensional target.

[0010] Optionally, it also includes controlling the distance between the target vehicle and the sampling vehicle within a preset range.

[0011] A second aspect of this application provides a vehicle visual truth acquisition device, comprising: an acquisition module for acquiring image data of a target vehicle at a target time and multiple frame image data of a sampled vehicle corresponding to the same timestamp as the target vehicle; an identification module for identifying the annotation data of the same three-dimensional target in the image data at the target time and the multiple frame image data; and a conversion module for converting the annotation data of the three-dimensional target of the sampled vehicle to the pose of the corresponding three-dimensional target in the target vehicle coordinate system, and generating a visual truth value in the target vehicle coordinate system based on the image data at the target time and the converted pose of the three-dimensional target.

[0012] A third aspect of this application provides a vehicle, 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 vehicle visual truth acquisition method as described in the above embodiments.

[0013] 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 vehicle visual truth acquisition method as described in the above embodiments.

[0014] A fifth aspect of this application provides a computer program product, including a computer program or instructions, characterized in that, when the computer program or instructions are executed, they implement the vehicle visual truth acquisition method as described in the above embodiments.

[0015] Therefore, this application has at least the following beneficial effects:

[0016] This application embodiment can acquire image data of the target vehicle at the target time and multiple frame image data of the sampling vehicle and the target vehicle corresponding to the same timestamp; identify the annotation data of the same three-dimensional target in the image data at the target time and the multiple frame image data; convert the annotation data of the three-dimensional target of the sampling vehicle to the pose of the corresponding three-dimensional target in the target vehicle coordinate system; and generate the visual ground truth in the target vehicle coordinate system based on the image data at the target time and the pose of the converted three-dimensional target. This can save a lot of time, manpower, material resources and overall cost, and has high acquisition efficiency and more accurate data.

[0017] 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

[0018] The above and 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, in which:

[0019] Figure 1This is a flowchart of a method for obtaining vehicle visual truth according to an embodiment of this application;

[0020] Figure 2 This is a schematic diagram of a vehicle visual truth acquisition method provided according to an embodiment of this application;

[0021] Figure 3 This is a block diagram of a vehicle visual truth acquisition device provided according to an embodiment of this application;

[0022] Figure 4 This is a structural schematic diagram of a vehicle provided according to an embodiment of this application. Detailed Implementation

[0023] 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.

[0024] The following description, with reference to the accompanying drawings, outlines a method, apparatus, vehicle, storage medium, and program for acquiring vehicle visual ground truth according to embodiments of this application. Addressing the issues raised in the background section of related technologies, such as the high cost and repeated data acquisition required for different vehicle models, leading to waste of manpower and resources, and the limited applicability of a single deep learning model for multiple vehicles, this application provides a method for acquiring vehicle visual ground truth. This method acquires image data of the target vehicle at a target time and multiple frames of image data corresponding to the same timestamp of the sampled vehicle and the target vehicle; identifies the annotation data of the same 3D target in the image data at the target time and the multiple frames of image data; converts the annotation data of the 3D target of the sampled vehicle to the pose of the corresponding 3D target in the target vehicle coordinate system; and generates the visual ground truth in the target vehicle coordinate system based on the image data at the target time and the converted 3D target pose. This method saves significant time, manpower, resources, and overall cost, while also achieving high acquisition efficiency and more accurate data. Therefore, it solves the problems in related technologies, such as the high cost and repeated data acquisition required for different vehicle models, and the limited applicability of a single deep learning model for multiple vehicles.

[0025] Specifically, Figure 1 This is a flowchart illustrating a method for obtaining vehicle visual truth values ​​provided in an embodiment of this application.

[0026] like Figure 1 As shown, the method for obtaining the vehicle's visual truth value includes the following steps:

[0027] In step S101, image data of the target vehicle at the target time and image data of multiple frames before and after the target vehicle corresponding to the same timestamp of the sampled vehicle are acquired.

[0028] It is understood that the embodiments of this application can acquire image data of the target vehicle at the target time and multiple frame image data of the sampled vehicle and the target vehicle corresponding to the same timestamp, so as to facilitate subsequent identification of the image data at the target time and the annotation data of the same three-dimensional target in the multiple frame image data.

[0029] It should be noted that the sampling vehicle equipped with a complete set of data acquisition equipment and the target vehicle equipped with a camera collect data simultaneously.

[0030] In step S102, the image data at the target time and the annotation data of the same three-dimensional target in the multi-frame image data are identified.

[0031] It is understood that the embodiments of this application can identify the labeled data of the same three-dimensional target in the image data at the target time and the image data of multiple frames, so as to generate the visual ground truth in the target vehicle coordinate system for the data collected at the same time stamp, thereby improving the accuracy.

[0032] In this embodiment of the application, before identifying the annotation data of the same three-dimensional target in the image data and multi-frame image data at the target time, the method includes: preprocessing all three-dimensional targets in the image data and multi-frame image data at the target time; inputting the preprocessed image data into the annotation model, and the annotation model outputting the annotation data of the corresponding three-dimensional target, wherein the annotation data includes the pose, size, yaw angle and target name of the three-dimensional target.

[0033] The preprocessing includes extracting the image data corresponding to the closest features from the image data collected from the sampling vehicle and the image data collected from the target vehicle through information such as appearance features, position, and speed, and then performing distortion removal, noise reduction, and edge enhancement on the data, without making specific limitations.

[0034] It is understood that the embodiments of this application can preprocess all three-dimensional targets in the image data at the target time and the multi-frame image data; input the preprocessed image data into the annotation model, and the annotation model outputs the corresponding annotation data of the three-dimensional targets, thereby making the annotation data more accurate, so as to facilitate the subsequent generation of the visual ground truth in the target vehicle coordinate system.

[0035] In step S103, the labeled data of the three-dimensional target of the sampled vehicle is converted to the pose of the corresponding three-dimensional target in the target vehicle coordinate system, and the visual ground truth in the target vehicle coordinate system is generated based on the image data of the target time and the pose of the converted three-dimensional target.

[0036] It is understood that the embodiments of this application convert the labeled data of the three-dimensional target of the sampled vehicle into the pose of the corresponding three-dimensional target in the target vehicle coordinate system, and generate the visual ground truth in the target vehicle coordinate system based on the image data of the target time and the pose of the converted three-dimensional target. This can save a lot of time, manpower and material resources and overall cost, and the acquisition efficiency is high and the data is more accurate.

[0037] In this embodiment, converting the labeled data of the three-dimensional target of the sampling vehicle into the pose of the corresponding three-dimensional target in the target vehicle coordinate system includes: identifying the extrinsic parameters and positioning information of the sensors and cameras of the sampling vehicle; converting the labeled data of the three-dimensional target of the sampling vehicle from the sampling vehicle coordinate system into the labeled data of the three-dimensional target in the world coordinate system according to the extrinsic parameters and positioning information, and determining the actual pose of the three-dimensional target in the world coordinate system according to the labeled data; and converting the actual pose of the three-dimensional target in the world coordinate system into the pose of the corresponding three-dimensional target in the target vehicle coordinate system.

[0038] It is understood that the embodiments of this application can identify the sensor and camera extrinsic parameters and positioning information of the sampling vehicle; based on the extrinsic parameters and positioning information, the annotation data of the three-dimensional target of the sampling vehicle is converted from the sampling vehicle coordinate system to the annotation data of the three-dimensional target in the world coordinate system, and the actual pose of the three-dimensional target in the world coordinate system is determined based on the annotation data; converting the actual pose of the three-dimensional target in the world coordinate system into the pose of the corresponding three-dimensional target in the target vehicle coordinate system can save a lot of time, manpower and material resources and overall costs, and the acquisition efficiency is high and the data is more accurate.

[0039] In this embodiment of the application, determining the actual pose of a 3D target in the world coordinate system based on the labeled data further includes: calculating the rotation angle of the 3D target pose in the labeled data for multiple frames before and after using quaternion spherical interpolation; calculating the translation vector of the 3D target pose in the labeled data for multiple frames before and after using linear interpolation; and determining the actual pose of the 3D target pose in the world coordinate system corresponding to the same timestamp of the sampled vehicle and the target vehicle based on the rotation angle and translation vector.

[0040] It is understood that the embodiments of this application can use quaternion spherical interpolation to calculate the rotation angle of the three-dimensional target pose in the annotation data for multiple frames before and after; use linear interpolation to calculate the translation vector of the three-dimensional target pose in the annotation data for multiple frames before and after; and determine the actual pose of the three-dimensional target pose in the world coordinate system corresponding to the same timestamp of the sampling vehicle and the target vehicle based on the rotation angle and translation vector, so as to improve the accuracy of the three-dimensional target pose data.

[0041] Specifically, using quaternion spherical interpolation, suppose there are two rotation states R1 and R2, corresponding to two different time points, and we want to calculate the rotation state Rt at any given time t between these two time points; where the new quaternion qt is calculated using the spherical linear interpolation formula:

[0042]

[0043] Where t represents the interpolation ratio from q1 to q2, and 0 ≤ t ≤ 1.

[0044] Calculate the new translation vector Pt using the linear interpolation formula:

[0045] Pt=P1+t·ΔP,

[0046] Where t represents the interpolation ratio from P1 to P2, 0≤t≤1.

[0047] In this embodiment of the application, the actual pose of a three-dimensional target in the world coordinate system is converted into the pose of the corresponding three-dimensional target in the target vehicle coordinate system, including: identifying the relative position of the sampling vehicle and the target vehicle and the extrinsic parameters of the target vehicle's sensors and cameras; and converting the actual pose of the three-dimensional target in the world coordinate system into the pose of the corresponding three-dimensional target in the target vehicle coordinate system based on the relative position and the extrinsic parameters of the target vehicle's sensors and cameras.

[0048] It is understood that the embodiments of this application can identify the relative position of the sampling vehicle and the target vehicle, as well as the sensor and camera extrinsic parameters of the target vehicle; based on the relative position and the sensor and camera extrinsic parameters of the target vehicle, the actual pose of the three-dimensional target in the world coordinate system is converted into the pose of the corresponding three-dimensional target in the target vehicle coordinate system, which can save a lot of time, manpower and material resources and overall cost, and has high acquisition efficiency and more accurate data.

[0049] In this embodiment of the application, it also includes: controlling the distance between the target vehicle and the sampling vehicle within a preset range.

[0050] The preset range can be 20m or 10m, and can be set according to actual needs without specific limitations.

[0051] It is understood that the embodiments of this application control the distance between the target vehicle and the sampling vehicle within a preset range, thereby improving the quality and real-time performance of data collection, ensuring the accuracy of data collection, and reducing the complexity of subsequent calculations.

[0052] According to the vehicle visual ground truth acquisition method proposed in the embodiments of this application, image data of the target vehicle at the target time and multiple frame image data of the sampled vehicle and the target vehicle corresponding to the same timestamp are acquired; the annotation data of the same three-dimensional target in the image data at the target time and the multiple frame image data are identified; the annotation data of the three-dimensional target of the sampled vehicle is converted to the pose of the corresponding three-dimensional target in the target vehicle coordinate system; and the visual ground truth in the target vehicle coordinate system is generated based on the image data at the target time and the pose of the converted three-dimensional target. This method can save a lot of time, manpower and material resources and overall cost, and has high acquisition efficiency and more accurate data.

[0053] The following will combine Figure 2 The method for obtaining the vehicle visual truth value in this application is described in detail below:

[0054] First, the sampling vehicle and the target vehicle collect data simultaneously, with the target vehicle in front. The sampling vehicle needs to follow closely behind the target vehicle, maintaining a distance of within 20 meters. Do not be too far apart, and do not insert other vehicles in between. The data collected by the sampling vehicle includes point cloud, image, and positioning data, while the data collected by the target vehicle generally only includes image data and positioning data.

[0055] Secondly, after the data collection is completed, the data collected by the sampling vehicle needs to be labeled. The labeling is generally done on the point cloud, and can be done manually or by pre-labeling. After the labeling is completed, each 3D target has its corresponding position, size, yaw angle and target ID.

[0056] Finally, for the image data collected from the target vehicle, it is necessary to find the two frames of data before and after the timestamp corresponding to the sampling vehicle. The 3D targets labeled in these two frames are unified into the world coordinate system through extrinsic parameters and positioning information to obtain the pose of the same target in the two frames before and after in the world coordinate system. Quaternion spherical interpolation is used for rotation and linear interpolation is used for translation to obtain the pose of the target at the timestamp collected by the target vehicle. Then, the target can be transformed into the target vehicle coordinate system through the positioning information and extrinsic parameters of the target vehicle to generate the visual ground truth value corresponding to the target vehicle.

[0057] In summary, this application only requires the installation of one data acquisition device on the data collection vehicle, and other vehicle models can generate the corresponding true values ​​using this method, which can save a lot of time, manpower, and economic costs.

[0058] Next, the vehicle visual truth acquisition device according to the embodiments of this application is described with reference to the accompanying drawings.

[0059] Figure 3 This is a block diagram of a vehicle visual truth acquisition device according to an embodiment of this application.

[0060] like Figure 3As shown, the vehicle visual truth acquisition device 10 includes: an acquisition module 100, a recognition module 200, and a conversion module 300.

[0061] The acquisition module 100 is used to acquire image data of the target vehicle at the target time and multiple frames of image data of the sampled vehicle and the target vehicle corresponding to the same timestamp; the recognition module 200 is used to recognize the annotation data of the same three-dimensional target in the image data at the target time and the multiple frames of image data; the conversion module 300 is used to convert the annotation data of the three-dimensional target of the sampled vehicle to the pose of the corresponding three-dimensional target in the target vehicle coordinate system, and generate the visual ground truth in the target vehicle coordinate system based on the image data at the target time and the pose of the converted three-dimensional target.

[0062] It should be noted that the foregoing explanation of the vehicle visual truth acquisition method embodiment also applies to the vehicle visual truth acquisition device of this embodiment, and will not be repeated here.

[0063] The vehicle visual truth acquisition device proposed in the embodiments of this application

[0064] This method acquires image data of the target vehicle at the target time and multiple frames of image data corresponding to the same timestamp of the sampled vehicle and the target vehicle; identifies the annotation data of the same 3D target in the image data at the target time and the multiple frames of image data; transforms the annotation data of the 3D target of the sampled vehicle into the pose of the corresponding 3D target in the target vehicle coordinate system; and generates the visual ground truth in the target vehicle coordinate system based on the image data at the target time and the transformed 3D target pose. This method can save a lot of time, manpower, material resources and overall cost, and has high acquisition efficiency and more accurate data.

[0065] Figure 4 A schematic diagram of the structure of a vehicle provided in an embodiment of this application. The vehicle may include:

[0066] The memory 401, the processor 402, and the computer program stored on the memory 401 and capable of running on the processor 402.

[0067] When the processor 402 executes the program, it implements the vehicle visual truth acquisition method provided in the above embodiments.

[0068] Furthermore, the vehicle also includes:

[0069] Communication interface 403 is used for communication between memory 401 and processor 402.

[0070] The memory 401 is used to store computer programs that can run on the processor 402.

[0071] The memory 401 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0072] If the memory 401, processor 402, and communication interface 403 are implemented independently, then the communication interface 403, memory 401, and processor 402 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. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 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.

[0073] Optionally, in a specific implementation, if the memory 401, processor 402, and communication interface 403 are integrated on a single chip, then the memory 401, processor 402, and communication interface 403 can communicate with each other through an internal interface.

[0074] Processor 402 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.

[0075] This application also provides a computer-readable storage medium storing a computer program or instructions thereon, which, when executed by a processor, implements the above-described method for obtaining vehicle visual truth values.

[0076] This application also provides a computer program product, including a computer program or instructions, characterized in that, when the computer program or instructions are executed, they implement the above-described method for obtaining vehicle visual truth values.

[0077] 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.

[0078] Furthermore, the terms "first" and "second" 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 "second" 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.

[0079] 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.

[0080] 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.

[0081] 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 vehicle visual ground truth acquisition method, characterized by, The method comprises the following steps: obtaining image data of a target vehicle at a target time and a plurality of frames of image data corresponding to the same timestamp of the target vehicle and a sampling vehicle; identifying annotation data of the same three-dimensional target in the image data at the target time and the plurality of frames of image data; converting the annotation data of the three-dimensional target of the sampling vehicle to the pose of the corresponding three-dimensional target in the coordinate system of the target vehicle, and generating a visual ground truth in the coordinate system of the target vehicle according to the image data at the target time and the converted pose of the three-dimensional target; wherein the conversion of the annotation data of the three-dimensional target of the sampling vehicle to the pose of the corresponding three-dimensional target in the coordinate system of the target vehicle comprises: identifying the sensor and camera extrinsic parameters and positioning information of the sampling vehicle; converting the annotation data of the three-dimensional target of the sampling vehicle from the coordinate system of the sampling vehicle to the annotation data of the three-dimensional target in the world coordinate system according to the extrinsic parameters and the positioning information, and determining the actual pose of the three-dimensional target in the world coordinate system according to the annotation data; and converting the actual pose of the three-dimensional target in the world coordinate system to the pose of the corresponding three-dimensional target in the coordinate system of the target vehicle. The determination of the actual pose of the three-dimensional target in the world coordinate system according to the annotation data further comprises: calculating the rotation angle of the three-dimensional target pose in the annotation data using spherical interpolation of quaternions; calculating the translation vector of the three-dimensional target pose in the annotation data using linear interpolation; and determining the actual pose of the three-dimensional target pose in the world coordinate system corresponding to the same timestamp of the sampling vehicle and the target vehicle according to the rotation angle and the translation vector.

2. The vehicle visual ground truth acquisition method of claim 1, wherein, The conversion of the actual pose of the three-dimensional target in the world coordinate system to the pose of the corresponding three-dimensional target in the coordinate system of the target vehicle comprises: identifying the relative position of the sampling vehicle and the target vehicle and the sensor and camera extrinsic parameters of the target vehicle; converting the actual pose of the three-dimensional target in the world coordinate system to the pose of the corresponding three-dimensional target in the coordinate system of the target vehicle according to the relative position and the sensor and camera extrinsic parameters of the target vehicle.

3. The vehicle visual ground truth acquisition method of claim 1, wherein, Before the identification of the annotation data of the same three-dimensional target in the image data at the target time and the plurality of frames of image data, the method comprises: preprocessing all three-dimensional targets in the image data at the target time and the plurality of frames of image data; inputting the preprocessed image data into a labeling model, wherein the labeling model outputs corresponding annotation data of the three-dimensional target, and the annotation data comprises the pose, size, yaw angle and target name of the three-dimensional target.

4. The vehicle vision ground truth acquisition method of claim 1, wherein, The method further comprises: controlling the distance between the target vehicle and the sampling vehicle to be within a preset range.

5. A vehicle visual truth acquisition apparatus characterized by comprising: The method comprises: an acquisition module, configured to obtain image data of a target vehicle at a target time and a plurality of frames of image data corresponding to the same timestamp of the target vehicle and a sampling vehicle; an identification module, configured to identify annotation data of the same three-dimensional target in the image data at the target time and the plurality of frames of image data; The conversion module is configured to convert the annotation data of the three-dimensional target of the sampling vehicle to the pose of the corresponding three-dimensional target in the target vehicle coordinate system, and generate the visual ground truth in the target vehicle coordinate system according to the image data of the target moment and the converted pose of the three-dimensional target; wherein the conversion of the annotation data of the three-dimensional target of the sampling vehicle to the pose of the corresponding three-dimensional target in the target vehicle coordinate system comprises: identifying the sensor and camera extrinsic parameters and positioning information of the sampling vehicle; converting the annotation data of the three-dimensional target of the sampling vehicle from the sampling vehicle coordinate system to the annotation data of the three-dimensional target in the world coordinate system according to the extrinsic parameters and the positioning information, and determining the actual pose of the three-dimensional target in the world coordinate system according to the annotation data; and converting the actual pose of the three-dimensional target in the world coordinate system into the pose of the corresponding three-dimensional target in the target vehicle coordinate system. The determination of the actual pose of the three-dimensional target in the world coordinate system according to the annotation data further comprises: calculating the rotation angle of the three-dimensional target pose in the front and rear multiple frames of annotation data by using quaternion spherical interpolation; calculating the translation vector of the three-dimensional target pose in the front and rear multiple frames of annotation data by using linear interpolation; and determining the actual pose of the three-dimensional target pose in the world coordinate system corresponding to the same time stamp of the sampling vehicle and the target vehicle according to the rotation angle and the translation vector.

6. A vehicle characterized by comprising: The computer program or instructions are executed by the processor to implement the vehicle visual ground truth acquisition method according to any one of claims 1-4. The computer program or instructions are executed by the processor to implement the vehicle visual ground truth acquisition method according to any one of claims 1-4.

7. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions are executed by the processor to implement the vehicle visual ground truth acquisition method according to any one of claims 1-4.

8. A computer program product comprising computer programs or instructions, characterized in that, ​

Images