Optimization methods, devices, vehicles, media, and programs for 3D target pseudo-labels

Abstract

This application relates to the field of vehicle technology, and in particular to a method, apparatus, vehicle, medium, and program for optimizing 3D target pseudo-labels. The method includes: identifying the relative positional relationship between at least one 3D target and the vehicle in its trajectory; determining an optimal observation point based on the relative positional relationship; updating the pseudo-label of each 3D target based on the optimal observation point, wherein the pseudo-label includes the size and position of the 3D target; and optimizing the updated pseudo-label of each 3D target to generate a corresponding optimal 3D target pseudo-label. This solves the problem in related technologies where the size of the same 3D target often exhibits large variance due to the observation perspective, leading to inaccurate labeling results.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and in particular to an optimization method, apparatus, vehicle, medium, and program for three-dimensional target pseudo-labels. Background Technology

[0002] In related technologies, real-world annotation results in the field of autonomous driving often require significant investment of manpower, time, and economic resources, and quality inspection after annotation is also a time-consuming and labor-intensive task. Currently, methods using lidar models for pre-annotation to generate pseudo-labels have been proposed. However, the quality of pseudo-labels is often highly correlated with the metrics of the pre-trained model, and the variance of the same 3D target size often becomes large due to the perspective of observation, leading to inaccurate annotation results. Summary of the Invention

[0003] This application provides an optimization method, apparatus, vehicle, medium, and program for three-dimensional target pseudo-labels to solve the problem that the same three-dimensional target size often exhibits large variance due to the observation perspective, leading to inaccurate labeling results.

[0004] The first aspect of this application provides a method for optimizing three-dimensional target pseudo-labels, comprising the following steps: identifying the relative positional relationship between at least one three-dimensional target and the vehicle in the vehicle's trajectory; determining an optimal observation point based on the relative positional relationship; updating the pseudo-label of each three-dimensional target based on the optimal observation point, wherein the pseudo-label includes the size and position of the three-dimensional target; and optimizing the updated pseudo-label of each three-dimensional target to generate a corresponding optimal three-dimensional target pseudo-label.

[0005] Optionally, determining the optimal observation point based on the relative positional relationship includes: identifying the relative position of the three-dimensional target at the target time and the current vehicle; calculating the angular range of the laser radar sweeping across the three-dimensional target based on the relative position; and determining the optimal observation point position based on the maximum value of the angular range.

[0006] Optionally, identifying the relative positional relationship between the three-dimensional target and the vehicle in the vehicle's running trajectory includes: acquiring the scanning data of the lidar at the target time and the vehicle position in the running trajectory; determining at least one pseudo-tag of the three-dimensional target at the target time based on the scanning data; and determining the relative positional relationship between the three-dimensional target and the vehicle based on the position of the three-dimensional target in the pseudo-tag and the vehicle position.

[0007] Optionally, determining the pseudo-label of at least one three-dimensional target at the target time based on the scan data and the vehicle's current position data includes: inputting the scan data into the target detection model, and the target detection model outputting the pseudo-label of at least one three-dimensional target.

[0008] Optionally, after determining the pseudo-label of at least one three-dimensional target at the target time based on the scanning data, the method further includes: identifying the extrinsic parameter data of the lidar; and converting the pseudo-label in the lidar coordinate system into a pseudo-label in the world coordinate system based on the extrinsic parameter data.

[0009] Optionally, after identifying the relative positional relationship between the 3D target and the vehicle in the vehicle's trajectory, the method includes: identifying whether there is occlusion in the target environment at the current moment; if there is occlusion, generating a pseudo-label for the 3D target at the current moment based on the fusion of pseudo-labels of the 3D targets in consecutive frames before and after the current moment.

[0010] A second aspect of this application provides an optimization device for three-dimensional target pseudo-labels, comprising: an identification module for identifying the relative positional relationship between at least one three-dimensional target and the vehicle in a vehicle's trajectory; a determination module for determining an optimal observation point based on the relative positional relationship; an update module for updating the pseudo-label of each three-dimensional target based on the optimal observation point, wherein the pseudo-label includes the size and position of the three-dimensional target; and an optimization module for optimizing the updated pseudo-label of each three-dimensional target to generate a corresponding optimal three-dimensional target pseudo-label.

[0011] 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 implement the optimization method for three-dimensional target pseudo-labels as described in the above embodiments.

[0012] 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 implement the optimization method for three-dimensional target pseudo-labels as described in the above embodiments.

[0013] A fifth aspect of this application provides a computer program product, which, when executed, is used to implement the optimization method for three-dimensional target pseudo-labels as described in the above embodiments.

[0014] Therefore, this application has at least the following beneficial effects:

[0015] This application identifies the relative positional relationship between at least one 3D target and the vehicle in the vehicle's trajectory; determines the optimal observation point based on the relative positional relationship; updates the pseudo-label of each 3D target based on the optimal observation point; optimizes the updated pseudo-label of each 3D target to generate the corresponding optimal 3D target pseudo-label, optimizing the true size of the 3D target. This significantly reduces target variance and improves annotation quality in both high-speed and urban scenarios; and provides more efficient training data for downstream point cloud-based and vision-based deep learning models, improving the quality of training data.

[0016] 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

[0017] 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:

[0018] Figure 1 This is a flowchart of a method for optimizing three-dimensional target pseudo-labels according to an embodiment of this application;

[0019] Figure 2 This is a flowchart illustrating the optimization method for three-dimensional target pseudo-labels provided according to an embodiment of this application.

[0020] Figure 3 This is an example diagram of an optimization device for three-dimensional target pseudo-labels provided according to an embodiment of this application;

[0021] Figure 4 This is a structural schematic diagram of a vehicle according to an embodiment of this application. Detailed Implementation

[0022] The embodiments of this application are described in detail below. Examples of the 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.

[0023] The following describes a method, apparatus, vehicle, storage medium, and program for optimizing 3D target pseudo-labels according to embodiments of this application, with reference to the accompanying drawings. Addressing the issues mentioned in the background art regarding the high labor, time, and economic costs of real annotation results, as well as the time-consuming and labor-intensive quality inspection after optimization, this application provides a method for optimizing 3D target pseudo-labels. In this method, the relative positional relationship between at least one 3D target and the vehicle in the vehicle's trajectory is identified; an optimal observation point is determined based on the relative positional relationship; the pseudo-label of each 3D target is updated based on the optimal observation point; the updated pseudo-label of each 3D target is optimized to generate a corresponding optimal 3D target pseudo-label, optimizing the true size of the 3D target. This significantly reduces target variance and improves annotation quality in both high-speed and urban scenarios; and provides more efficient training data for downstream point cloud-based and vision-based deep learning model training, improving the quality of training data. Therefore, this solves the problems of high labor, time, and economic costs of real annotation results, as well as the time-consuming and labor-intensive quality inspection after annotation in related technologies.

[0024] Specifically, Figure 1 This is a flowchart illustrating an optimization method for a three-dimensional target pseudo-label provided in an embodiment of this application.

[0025] like Figure 1 As shown, the optimization method for this 3D target pseudo-label includes the following steps:

[0026] In step S101, the relative positional relationship between at least one three-dimensional target and the vehicle in the vehicle's trajectory is identified.

[0027] It is understandable that identifying the relative positional relationship between at least one three-dimensional target and the vehicle in the vehicle's trajectory is not specifically limited, such as the three-dimensional target being located directly in front of, behind, to the left of, or to the right of the vehicle.

[0028] In this embodiment of the application, identifying the relative positional relationship between a three-dimensional target and a vehicle in a vehicle's running trajectory includes: acquiring the scanning data of the lidar at the target time and the vehicle position in the running trajectory; determining at least one pseudo-tag of a three-dimensional target at the target time based on the scanning data; and determining the relative positional relationship between the three-dimensional target and the vehicle based on the position of the three-dimensional target in the pseudo-tag and the vehicle position.

[0029] It is understood that the embodiments of this application can obtain the scanning data of the lidar and the vehicle position to determine at least one pseudo-tag of a three-dimensional target at the time of target determination, and determine the relative positional relationship between the three-dimensional target and the vehicle based on the position of the three-dimensional target and the vehicle position in the pseudo-tag, so as to facilitate the subsequent determination of the optimal observation point based on the relative positional relationship.

[0030] It should be noted that the vehicle-mounted LiDAR in this application acquires the scanning data of the LiDAR, and combines the relationship with the 3D target with RTK (Real-time kinematic, carrier phase differential technology) positioning information to determine the vehicle position.

[0031] In this embodiment of the application, determining the pseudo-label of at least one three-dimensional target at a target time based on the scanning data includes: inputting the scanning data into a target detection model, and the target detection model outputting the pseudo-label of at least one three-dimensional target.

[0032] It is understood that in this embodiment of the application, the scanned data is input into the target detection model, and the target detection model outputs at least one pseudo-label of a three-dimensional target. The three-dimensional target detection model analyzes the scanned data and the current position of the vehicle to identify and label the three-dimensional targets in the scene, such as vehicles and pedestrians, and obtains preliminary target detection results. This is so that the relative positional relationship between the three-dimensional target and the vehicle can be determined later based on the position of the three-dimensional target in the pseudo-label and the position of the vehicle.

[0033] In this embodiment of the application, after determining the pseudo-label of at least one three-dimensional target at the target time based on the scanning data, the method further includes: identifying the extrinsic parameter data of the lidar; and converting the pseudo-label in the lidar coordinate system into a pseudo-label in the world coordinate system based on the extrinsic parameter data.

[0034] It is understood that the embodiments of this application convert the pseudo-labels in the lidar coordinate system into pseudo-labels in the world coordinate system, thereby improving the accuracy of pseudo-label recognition of three-dimensional targets.

[0035] In this embodiment of the application, after identifying the relative positional relationship between the three-dimensional target and the vehicle in the vehicle's trajectory, the method includes: identifying whether there is occlusion in the target environment at the current moment; if there is occlusion, generating a pseudo-label for the three-dimensional target at the current moment based on the fusion of pseudo-labels of three-dimensional targets in consecutive frames before and after the current moment.

[0036] It is understood that the embodiments of this application identify whether there is occlusion in the target environment at the current moment; if there is occlusion, the pseudo-label of the three-dimensional target at the current moment is generated by fusing the pseudo-labels of the three-dimensional target in multiple consecutive frames before and after the current moment. By fusing multi-frame data, the ability to identify and reconstruct occluded objects can be significantly improved, making the final generated three-dimensional target point cloud more complete, reducing misjudgment or omission caused by occlusion, and improving the accuracy of three-dimensional target pseudo-label recognition.

[0037] In step S102, the optimal observation point is determined based on the relative positional relationship.

[0038] It is understood that the embodiments of this application can determine the optimal observation point based on the relative positional relationship, so as to update the pseudo-label of each three-dimensional target based on the optimal observation point.

[0039] In this embodiment of the application, determining the optimal observation point based on the relative position relationship includes: identifying the relative position of the three-dimensional target at the target time and the current vehicle; calculating the angle range of the lidar sweeping across the three-dimensional target based on the relative position; and determining the optimal observation point position based on the maximum value of the angle range.

[0040] It is understood that the embodiments of this application identify the relative position of the three-dimensional target and the current vehicle at the target time; calculate the angle range of the laser radar sweeping the three-dimensional target based on the relative position; determine the optimal observation point position according to the maximum value of the angle range, which can effectively select the most favorable observation angle, reduce the uncertainty of size estimation, make the size of the same target more consistent under different observations, and improve the accuracy of three-dimensional target pseudo-label recognition.

[0041] For example, when the target trajectory is directly in front of the vehicle, the angle of the LiDAR sweeping across the target is calculated based on the target's position. When the angle is at its maximum, which is the optimal observation point position when the target width is most accurate, the size of the same 3D target is consistent. Therefore, all widths of this target can be updated. This greatly reduces the problem of inconsistent 3D target sizes caused by the model itself or occlusion.

[0042] In step S103, the pseudo-label of each 3D target is updated based on the best observation point, wherein the pseudo-label includes the size and position of the 3D target.

[0043] It is understood that the embodiments of this application can update the pseudo-label of each 3D target based on the best observation point, optimize the true size of the 3D target, and significantly reduce the target variance and improve the annotation quality in both high-speed and urban scenes.

[0044] In step S104, the updated pseudo-labels of each 3D target are optimized to generate the corresponding optimal 3D target pseudo-labels.

[0045] It is understood that the embodiments of this application can optimize the updated pseudo-labels of each 3D target to generate corresponding optimal 3D target pseudo-labels, providing more efficient training data for downstream point cloud-based deep learning models and vision-based deep learning models. It can also be used as a data quality inspection tool to improve data quality.

[0046] The optimization method for 3D target pseudo-labels proposed in this application identifies the relative positional relationship between at least one 3D target and the vehicle in the vehicle's trajectory; determines the optimal observation point based on the relative positional relationship; updates the pseudo-label of each 3D target based on the optimal observation point; optimizes the updated pseudo-label of each 3D target to generate the corresponding optimal 3D target pseudo-label, optimizing the true size of the 3D target. This significantly reduces target variance and improves annotation quality in both high-speed and urban scenarios; and provides more efficient training data for downstream point cloud-based and vision-based deep learning model training, improving the quality of training data.

[0047] The following will combine Figure 2 The optimization method for the 3D target pseudo-label of this application is described in detail below:

[0048] Due to the different relative positions and angles between the 3D target and the LiDAR, the number of points reflected back by the LiDAR after scanning the object will vary greatly. For the LiDAR, the width of the 3D target directly in front of or behind it can be obtained very accurately, and similarly, the length of the 3D target to its left or right can be obtained very accurately. Therefore, the specific implementation of the optimization algorithm proposed in this invention consists of the following steps:

[0049] Based on the collected raw 3D point cloud and positioning data, the existing point cloud 3D target detection model is first used for inference to obtain the 3D target detection results. Then, the 3D target is transformed from the vehicle's LiDAR coordinate system to the world coordinate system by combining the positioning data and LiDAR extrinsic parameters. Motion tracking is performed by the type, position, size, speed and acceleration of the 3D target to obtain the motion trajectory of the same target in the time dimension and the probability distribution of the target's length and width.

[0050] Next, the target's trajectory is used to determine whether the object is directly in front of, behind, to the left of, or to the right of the vehicle. The maximum angle formed by the starting point of the lidar scanning of the target is calculated to determine the position of the best observation point, and the length and width of the object are updated.

[0051] For example, when the target trajectory is directly in front of the vehicle, the angle of the LiDAR sweeping across the target is calculated based on the target's position. When the angle is at its maximum, which is the optimal observation point position when the target width is most accurate, the size of the same 3D target is consistent. Therefore, all widths of this target can be updated. This greatly reduces the problem of inconsistent 3D target sizes caused by the model itself or occlusion.

[0052] It should be noted that the optimization algorithm mainly includes the following steps: the lidar model performs inference based on pseudo-labels to generate pseudo-labels; offline tracking is performed based on the target detection results of lidar; the optimal observation point position of the target is found according to the angle formed by the lidar scanning of the target; and the target size is updated.

[0053] In summary, based on statistical laws, the optimal observation point position of the target is determined by combining the relationship between the vehicle-mounted LiDAR and the 3D target with RTK (carrier phase differential technology) positioning information to optimize the 3D target. This can significantly reduce target variance and improve the data quality of pseudo-labels in both high-speed and urban scenarios, providing more efficient training data for downstream point cloud-based and vision-based deep learning models. It can also be used as a data quality inspection tool.

[0054] Next, referring to the accompanying drawings, an optimization apparatus for three-dimensional target pseudo-labels according to an embodiment of this application is described.

[0055] Figure 3 This is a block diagram of an optimization device for three-dimensional target pseudo-labels according to an embodiment of this application.

[0056] like Figure 3 As shown, the optimization device 10 for the three-dimensional target pseudo-label includes: an acquisition module 100, a determination module 200, an identification module 300, and an update module 400.

[0057] The identification module 100 is used to identify the relative positional relationship between at least one three-dimensional target and the vehicle in the vehicle's running trajectory; the determination module 200 is used to determine the optimal observation point based on the relative positional relationship; the update module 300 is used to update the pseudo-label of each three-dimensional target based on the optimal observation point, wherein the pseudo-label includes the size and position of the three-dimensional target; and the optimization module 400 is used to optimize the updated pseudo-label of each three-dimensional target to generate the corresponding optimal three-dimensional target pseudo-label.

[0058] It should be noted that the explanation of the aforementioned method for optimizing three-dimensional target pseudo-labels also applies to the optimization device for three-dimensional target pseudo-labels in this embodiment, and will not be repeated here.

[0059] The 3D target pseudo-label optimization device proposed in this application identifies the relative positional relationship between at least one 3D target and the vehicle in the vehicle's trajectory; determines the optimal observation point based on the relative positional relationship; updates the pseudo-label of each 3D target based on the optimal observation point; optimizes the updated pseudo-label of each 3D target to generate the corresponding optimal 3D target pseudo-label, optimizing the true size of the 3D target. This significantly reduces target variance and improves labeling quality in both high-speed and urban scenarios. Furthermore, it provides more efficient training data for downstream point cloud-based and vision-based deep learning model training, improving the quality of training data.

[0060] Figure 4 A schematic diagram of the structure of a vehicle provided in an embodiment of this application. The vehicle may include:

[0061] The memory 401, the processor 402, and the computer program stored on the memory 401 and capable of running on the processor 402.

[0062] When the processor 402 executes the program, it implements the optimization method for three-dimensional target pseudo-labels provided in the above embodiments.

[0063] Furthermore, the vehicle also includes:

[0064] Communication interface 403 is used for communication between memory 401 and processor 402.

[0065] The memory 401 is used to store computer programs that can run on the processor 402.

[0066] The memory 401 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0067] 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 ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) 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.

[0068] 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.

[0069] Processor 402 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of this application.

[0070] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for optimizing three-dimensional target pseudo-labels.

[0071] This application also provides a computer program product, including: a computer program or instructions, which, when executed, are used to implement the optimization method for three-dimensional target pseudo-labels as described in the above embodiments.

[0072] 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.

[0073] 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.

[0074] 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.

[0075] 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 (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0076] 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.

[0077] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. An optimization method for three-dimensional target pseudo-labels, characterized in that, The method comprises the following steps: identifying a relative position relationship between at least one three-dimensional target in a vehicle running track and the vehicle, wherein the identifying the relative position relationship between at least one three-dimensional target in a vehicle running track and the vehicle comprises: acquiring scanning data of a laser radar at a target time and a vehicle position in the running track; determining a pseudo label of at least one three-dimensional target at the target time according to the scanning data; and determining a relative position relationship between the three-dimensional target and the vehicle according to a position of the three-dimensional target in the pseudo label and the vehicle position, wherein the determining the pseudo label of at least one three-dimensional target at the target time according to the scanning data comprises: inputting the scanning data into a target detection model, and the target detection model outputs the pseudo label of at least one three-dimensional target; determining an optimal observation point based on the relative position relationship, wherein the determining the optimal observation point based on the relative position relationship comprises: identifying a relative position between the three-dimensional target at the target time and the current vehicle; calculating an included angle range of the laser radar scanning the three-dimensional target based on the relative position; and determining an optimal observation point position according to a maximum value of the included angle range; updating the pseudo label of each three-dimensional target based on the optimal observation point, wherein the pseudo label comprises a size and a position of the three-dimensional target; optimizing the updated pseudo label of each three-dimensional target to generate a corresponding optimal three-dimensional target pseudo label.

2. The method of Claim 1, wherein, After the determining the pseudo label of at least one three-dimensional target at the target time according to the scanning data, the method further comprises: identifying external parameter data of the laser radar; converting the pseudo label in the laser radar coordinate system into a pseudo label in a world coordinate system based on the external parameter data.

3. The method of Claim 1, wherein, After the identifying the relative position relationship between at least one three-dimensional target in a vehicle running track and the vehicle, the method comprises: identifying whether there is an occlusion in a target environment at a current time; if there is an occlusion, fusing and generating a pseudo label of a three-dimensional target at the current time based on pseudo labels of the three-dimensional target in multiple frames before and after the current time.

4. An apparatus for optimizing three-dimensional target pseudo-labels, comprising: The method comprises: an identifying module configured to identify a relative position relationship between at least one three-dimensional target in a vehicle running track and the vehicle, wherein the identifying the relative position relationship between at least one three-dimensional target in a vehicle running track and the vehicle comprises: acquiring scanning data of a laser radar at a target time and a vehicle position in the running track; determining a pseudo label of at least one three-dimensional target at the target time according to the scanning data; and determining a relative position relationship between the three-dimensional target and the vehicle according to a position of the three-dimensional target in the pseudo label and the vehicle position, wherein the determining the pseudo label of at least one three-dimensional target at the target time according to the scanning data comprises: inputting the scanning data into a target detection model, and the target detection model outputs the pseudo label of at least one three-dimensional target; a determining module configured to determine an optimal observation point based on the relative position relationship, wherein the determining the optimal observation point based on the relative position relationship comprises: identifying a relative position between the three-dimensional target at the target time and the current vehicle; calculating an included angle range of the laser radar scanning the three-dimensional target based on the relative position; and determining an optimal observation point position according to a maximum value of the included angle range; an updating module configured to update a pseudo label of each three-dimensional target based on the optimal observation point, wherein the pseudo label comprises a size and a position of the three-dimensional target; an optimization module configured to optimize the updated pseudo label of each three-dimensional target to generate a corresponding optimal three-dimensional target pseudo label.

5. A vehicle characterized by comprising: The method 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 method for optimizing a three-dimensional target pseudo label 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 method for optimizing a three-dimensional target pseudo label according to any one of claims 1-3.

7. A computer program product, comprising: A computer program or instructions, characterized in that the computer program or instructions are executed to implement the method for optimizing a three-dimensional target pseudo label according to any one of claims 1-3.

Images