Automatic driving data acquisition system and acquisition method thereof

By introducing a data acquisition host into the autonomous driving data acquisition system to receive and distribute raw image data, the problems of data lack of verification and loss of details in the existing technology are solved, and the perception algorithm is trained with high accuracy and robustness, meeting the reliability requirements of the autonomous driving system.

CN119094869AActive Publication Date: 2024-12-06ZHEJIANG GEELY HLDG GRP CO LTD +2

Patent Information

Application Number
CN202411213387.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-06
Estimated Expiration
2044-08-30

AI Technical Summary

Technical Problem

Existing autonomous driving data acquisition equipment directly collects processed video data, resulting in a lack of verification data and significant loss of details. This fails to meet the training requirements of perception algorithms, leading to inaccurate and unrobust perception models, which in turn affects the reliability and safety of autonomous driving systems.

Method used

By introducing a data acquisition host between the data acquisition source device and the domain controller, raw image data is received and transmitted, and data splitting is performed to generate raw image data for perception training and embedded row data for vehicle underlying software development, ensuring data integrity and verification.

Benefits of technology

It achieves complete acquisition of raw image data, meets the high requirements of autonomous driving for perception accuracy and reliability, ensures smooth and stable data transmission channels, and supports high accuracy and robust training of perception algorithms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119094869A_ABST
    Figure CN119094869A_ABST
Patent Text Reader

Abstract

The invention provides an automatic driving data acquisition system and an acquisition method thereof, which realize automatic driving data acquisition based on the automatic driving data acquisition system comprising data acquisition source equipment, a data acquisition host and a domain controller with a sensing module. And the data acquisition host is electrically connected between the data acquisition source equipment and the domain controller and is used for receiving the original image data and re-outputting the original image data to the domain controller, so that the acquisition of the original raw format data is realized, and the high requirements of automatic driving on perception accuracy and reliability are met.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to an autonomous driving data acquisition system and a data acquisition method thereof. Background Art

[0002] In autonomous driving technology, the perception link is the most core part of the technology, which is crucial for accurately identifying and understanding the surrounding environment, and has a large demand for video data. Data acquisition equipment, as a key data closed-loop medium in the development of autonomous driving, can adapt to smart driving models for full data collection and software algorithm development for data collection to meet the perception link's demand for video data.

[0003] At present, the equipment that collects video data directly collects processed video data. Since the processed video data does not contain verification data and the details of the video data are seriously lost, when the data is used to feed back to the domain control and conduct joint bottom soft module training and verification of perception algorithms, a large number of errors will be generated, the overall link cannot be run through, and the results of perception training are poor. This is because the processed video data lacks verification data, which means that the integrity and accuracy of the data cannot be effectively verified and guaranteed. In complex autonomous driving systems, any errors or omissions in the data may lead to serious consequences, such as incorrect perception judgments and decisions. At the same time, the perception algorithm requires rich and accurate information to learn and optimize. Video data with lost details cannot provide enough effective features for the algorithm, resulting in inaccurate and non-robust trained perception models.

[0004] Due to these issues, the current links cannot operate normally and cannot meet the high requirements of autonomous driving for perception accuracy and reliability. Summary of the invention

[0005] In order to overcome the problems existing in the related art, this specification provides an autonomous driving data collection system and a collection method thereof.

[0006] According to a first aspect of an embodiment of this specification, there is provided an autonomous driving data acquisition system, the system comprising: a data acquisition source device, a data acquisition host, and a domain controller having a perception module;

[0007] The data acquisition source device is used to acquire original image data of the vehicle;

[0008] The data acquisition host is electrically connected between the data acquisition source device and the domain controller, and is used to receive the original image data and re-output the original image data to the domain controller so that the domain controller performs perception analysis based on the original image data for vehicle driving.

[0009] According to an autonomous driving data acquisition system provided by the present application, the data acquisition host utilizes a serial link of a bidirectional data channel to transmit data with the data acquisition source device and the domain controller respectively; so that the data acquisition host receives the original image data sent by the data acquisition source device and receives the verification data sent by the domain controller.

[0010] According to an autonomous driving data acquisition system provided by the present application, the data acquisition host performs data diversion processing on the original image data to obtain a first data packet that is re-output to the domain controller and a second data packet that is stored on the disk in the data acquisition host, and the second data packet is used to be fed back into the domain controller for perception training.

[0011] According to an autonomous driving data acquisition system provided by the present application, the data acquisition host is also used to strip the diverted original image data through a preset data structure to obtain raw image data for perception training and embedded row data for vehicle bottom software development and verification.

[0012] According to a second aspect of an embodiment of this specification, a method for collecting autonomous driving data is provided, the method comprising:

[0013] Collecting original image data of the vehicle through the data acquisition source device;

[0014] receiving the original image data through the data acquisition host, and re-outputting the original image data to the domain controller;

[0015] The domain controller performs perception analysis based on the original image data for vehicle driving.

[0016] According to an autonomous driving data acquisition method provided by the present application, before the data acquisition host re-outputs the original image data to the domain controller, the method further includes:

[0017] Performing data splitting processing on the original image data by the data acquisition host to obtain a first data packet and a second data packet;

[0018] Re-outputting the first data packet to the domain controller;

[0019] Processing the second data packet to a disk, and storing it in the data acquisition host, so as to be fed back to the domain controller for perception training;

[0020] The data in the first data packet is consistent with the data in the second data packet.

[0021] According to an autonomous driving data acquisition method provided by the present application, before the data acquisition host performs data splitting processing on the original image data to obtain the first data packet and the second data packet, the method further includes:

[0022] In the case where the data acquisition source device performs string addition processing on the collected original image data, the data acquisition host performs deserialization processing on the original image data;

[0023] The data acquisition host performs data splitting processing on the original image data to obtain a first data packet and a second data packet, including:

[0024] The data acquisition host performs data splitting processing on the original image data after deserialization processing to obtain a first data packet and a second data packet.

[0025] According to an autonomous driving data collection method provided by the present application, the re-outputting the first data packet to the domain controller includes:

[0026] The first data packet is subjected to string adding processing by the data acquisition host, and the processed first data packet is re-output to the domain controller.

[0027] According to an automatic driving data collection method provided by the present application, before collecting the original image data of the vehicle by the data collection source device, the method further includes:

[0028] Verify the data link of the autonomous driving data acquisition system.

[0029] According to an autonomous driving data collection method provided by the present application, the verification of the data link of the autonomous driving data collection system includes:

[0030] After accessing the data acquisition source device, verifying the data flow status of a series of data transmitted by the data acquisition source device through the data acquisition host to obtain a first verification result;

[0031] Verifying the verification data sent by the domain controller through the data acquisition host to obtain a second verification result;

[0032] When both the first verification result and the second verification result are verified to be passed, a step of collecting original image data of the vehicle by using the data collection source device is performed.

[0033] The present application also provides an autonomous driving data collection device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the autonomous driving data collection method described above is implemented.

[0034] The present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for collecting autonomous driving data.

[0035] The autonomous driving data acquisition system and its acquisition method in the embodiments of this specification implement autonomous driving data acquisition based on an autonomous driving data acquisition system including a data acquisition source device, a data acquisition host, and a domain controller with a perception module. In the autonomous driving data acquisition system, the data acquisition host is electrically connected between the data acquisition source device and the domain controller, and is used to receive the original image data and re-output the original image data to the domain controller, thereby realizing the acquisition of original raw format data and meeting the high requirements of autonomous driving for perception accuracy and reliability.

[0036] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the specification and, together with the description, serve to explain the principles of the specification.

[0038] Figure 1 is a schematic diagram of an automatic driving data collection system according to an exemplary embodiment of the present specification;

[0039] Figure 2 is a schematic diagram of another autonomous driving data collection system according to an exemplary embodiment of the present specification;

[0040] Figure 3 is a schematic diagram of data offloading processing according to an exemplary embodiment of the present specification;

[0041] Figure 4 is a flow chart of an automatic driving data collection method according to an exemplary embodiment of the present specification;

[0042] Figure 5 is a schematic diagram of an automatic driving data collection device according to an exemplary embodiment of the present specification;

[0043] Figure 6 It is a schematic block diagram of an autonomous driving data collection device shown in this specification according to an exemplary embodiment. DETAILED DESCRIPTION

[0044] Here, the technical solutions in the embodiments (or "implementations") of the present application will be described clearly and completely in conjunction with the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements.

[0045] If there are terms involving directional indications or positional relationships in the embodiments of the present application (such as up, down, left, right, front, back, inside, outside, top, bottom, center, vertical, horizontal, longitudinal, transverse, length, width, counterclockwise, clockwise, axial, radial, circumferential, etc.), such terms are only used to explain the relative positional relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings); if the specific posture changes, the directional indication or positional relationship will also change accordingly. In addition, the terms "first", "second", etc. involved in the embodiments of the present application are only used for the purpose of convenience of description and cannot be understood as indicating or implying relative importance.

[0046] The present application provides an autonomous driving data collection system and a collection method thereof. The present application is described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the features of the following embodiments and implementations can be combined with each other.

[0047] Data collection equipment has been widely used in autonomous driving development, such as Udacity data collector, Waymo data collection system, comma.ai data collector, etc. As a key data closed-loop medium in autonomous driving development, data collection equipment can adapt to smart driving models for full data collection and software algorithm development for data collection.

[0048] Perception is the core part of autonomous driving technology and has a huge demand for video data. In current use, the coupling degree between the domain controller and the camera is relatively high. The domain controller is embedded with the corresponding processing algorithm, and directly processes the video data after obtaining it from the camera. Therefore, the current data acquisition device solution is to directly collect processed video data from the domain controller. When the data is fed back to the domain controller for joint bottom soft module training and verification of perception algorithms, a large number of errors are generated, because the processed video data does not contain verification data and the details of the video data are lost seriously. The overall link cannot be run through and the results of perception training are poor.

[0049] like Figure 1As shown in the figure, in the current data collection process, the data collection host directly collects the data output by the domain controller. By configuring in the data_collect module (responsible for filtering and sorting the data to be output) of the AUTOSAR AP in the domain controller, the data is sent out through the zmq framework (the zmq framework is an efficient message passing framework that can quickly transmit data between different components). On the data collection host side, the data is received based on the existing zmq framework, and the data can be stored on the disk after deserialization. This solution can only collect data output by the domain controller. For image / video data, only YUV image / video data processed by the domain controller ISP (image signal processing) can be collected. These compressed and processed data lose some image details and cannot be used for data re-injection to simulate the real vehicle environment and combine the bottom software to verify the perception algorithm.

[0050] For example, if the subtle features of an important traffic sign in the real vehicle environment are lost in the compression process, the perception algorithm may not be able to correctly identify the traffic sign in the re-injected data simulation, resulting in incorrect decisions and potential safety risks.

[0051] In view of the above problems and the demand for autonomous driving, the development of raw video data acquisition equipment has emerged.

[0052] In order to solve the above-mentioned technical problems, this specification provides an autonomous driving data collection method.

[0053] By connecting the data acquisition host in series between the data acquisition source device and the domain controller, the original image data transmitted by the data acquisition source device in GMSL is received, and the original image data is transmitted to the domain controller for analysis and processing. Since the original image data contains verification data and the video data details are complete, when the data is fed back to the domain controller for perception algorithm training, the perception algorithm has high accuracy and robustness, thereby meeting the high requirements of autonomous driving for perception accuracy and reliability.

[0054] like Figure 2 As shown, the autonomous driving data acquisition system described in this article includes: a data acquisition source device, a data acquisition host, and a domain controller with a perception module.

[0055] The data acquisition source device generally refers to a source that can generate raw data, such as a camera, various sensors, etc., which are used to collect raw image data of a vehicle. The following description takes a camera as an example.

[0056] The data acquisition host focuses on receiving, processing, storing and distributing data from different data acquisition source devices, and is used to receive original image data from the data acquisition source devices.

[0057] As an important component in the vehicle electronic system, the domain controller is responsible for receiving and processing data from various vehicle subsystems.

[0058] In the above-mentioned autonomous driving data acquisition system, the data acquisition host is electrically connected to a key position between the data acquisition source device and the domain controller, playing a series connection and important data processing role. For the camera, the data acquisition host simulates the function of the domain controller to receive the raw image data from the camera. This means that when the camera transmits data, it sends the data to the data acquisition host, just like sending it to a real domain controller.

[0059] For the domain controller, the data acquisition host simulates the camera to send data to complete the data verification on the domain controller side, so that the entire data transmission channel can be opened, that is, to ensure the smooth and stable operation of the data transmission channel from the camera to the domain controller. Therefore, after receiving the data, the data acquisition host re-outputs the original image data to the domain controller, so that the domain controller can perform further analysis and decision-making based on the algorithm and processing module within it and the original image data, including perception analysis, for vehicle driving.

[0060] In this embodiment, the data acquisition host collects the original image / video data without any image processing, retains as much image information as possible, provides a larger space for image processing and application, and expands the resources for image processing algorithm development and verification.

[0061] In some embodiments, after the data acquisition host receives the data, the data acquisition host will immediately start the data diversion mechanism. During the data diversion process, a portion of the original video data will be collected and stored in the hard disk of the device. This part of the stored data provides the original material for subsequent analysis, backtracking and research. At the same time, another part of the data will be reorganized and output, and continue to be transmitted in the direction of the intelligent driving domain as input data for the domain control for subsequent processing and decision-making. As an example, the data acquisition host includes a data processing module, which performs data diversion processing on the original image data to obtain a first data packet that is re-output to the domain controller and a second data packet that is stored in the data acquisition host, and the second data packet is used to be fed back to the domain controller for perception training.

[0062] For example, when the camera collects real-time video data on the road, the data collection host is like an intelligent "transfer station". On the one hand, it saves part of the data for subsequent possible problem tracing and in-depth research, and on the other hand, it promptly transmits key data to the intelligent driving domain controller. At the same time, it also simulates the interaction between the camera and the domain controller to ensure that the domain controller can correctly receive and verify the data, so that the entire data transmission channel from the camera to the domain controller can be smoothly opened, providing strong support for the stable operation and continuous optimization of the autonomous driving system.

[0063] Furthermore, the data verification on the domain controller side includes CRC verification.

[0064] In this embodiment, the data acquisition host can bypass the original image / video data, one way is used for data acquisition and retention, and the other way is re-output to the domain controller as the input of the intelligent driving domain. This diversion method helps to ensure that data is not lost during the acquisition process, and also ensures the operation of the intelligent driving domain controller.

[0065] In some embodiments, the data acquisition host also includes a communication module, and the communication module uses a serial link of a bidirectional data channel (such as a GMSL2 link) to transmit data with the data acquisition source device and the domain controller respectively; so that the data acquisition host receives the original image data sent by the data acquisition source device and receives the verification data sent by the domain controller. As an example, the data output by the camera end is transmitted in D-PHY via the CSI-2 protocol and output to the data acquisition host via GMSL2, and the data acquisition host transmits the data to the domain controller via GMSL2.

[0066] In this embodiment, the same GMSL2 link will also be used to send and receive bidirectional control channel data and accept verification data returned by the domain controller. The verification ensures that the data received by the domain controller from the camera is reliable, accurate and available, thereby ensuring the stability and effectiveness of the entire data transmission channel from the camera to the domain controller.

[0067] In some embodiments, the data acquisition host also includes a data stripping module, which is connected to the data processing module and is used to strip the diverted raw image data through a preset data structure to obtain raw image data for perception algorithm training and embedded row data containing real-time dynamic register information that can be used for bottom software development and verification. As an example, after completing the acquisition of the raw image data, a data processing script is used in combination with a data structure to realize the stripping of raw image data and embedded row data, wherein the data processing script reads the collected raw image data. These data usually have a specific data structure, including raw image data and row data that may be embedded therein. By analyzing the data structure, the script can identify the characteristics and positions of the raw image data and embedded row data. Then, according to the preset rules and algorithms, the script automatically performs the separation operation. After the separation is completed, the obtained raw image data and embedded row data can be processed and applied separately.

[0068] For example, if the embedded row data contains metadata about the image, such as shooting time, geographic location, etc., it can be extracted for data management and analysis; and the separated raw image data can be used for further image processing operations such as image enhancement, compression, and recognition.

[0069] In other embodiments, the data acquisition host further includes a data stripping module, which is connected to the data processing module and is used to strip the original image data before diversion through a preset data structure to obtain raw image data for perception training and embedded line data for vehicle bottom soft development and verification. Thereafter, the raw image data for perception training is diverted and the diverted data is transmitted to the domain controller for the domain controller to analyze and make decisions based on the raw image data with higher correlation.

[0070] As an example, an example of building an autonomous driving data collection system based on the above structure is as follows:

[0071] First, taking the V4L2 architecture camera adapted in this manual as an example, the raw image data generated at the sensor end is transmitted on the D-PHY physical layer through the CSI-2 protocol. The serializer on the camera side will perform string operations on these raw image data, convert them into a format suitable for transmission on the GMSL2 serial link, and output them on the GMSL2 serial link transceiver. On the data acquisition host side, imitate the domain controller, implement data deserialization through the deserializer, and perform PHY-COPY, mirror bypass the raw image / video data, and transmit the same video data to two D-PHYs, one of which completes the data storage on the disk in the device, and the other is re-output to the device-side serializer to add strings to the domain controller as input.

[0072] In this autonomous driving data acquisition system, the GMSL2 serial link has its specific transmission protocol and signal requirements. The data format generated by the camera may not meet the direct transmission standard of the link. The serializer can convert and encode the raw data to adapt it to the electrical characteristics and communication protocol of the GMSL2 link, thereby ensuring that the data can be transmitted accurately and stably on the link.

[0073] It should be noted that the same GMSL2 link is not only used to transmit raw image data, but also acts as a bidirectional control channel for sending and receiving. It receives verification data returned from the domain controller to ensure the accuracy and integrity of data transmission.

[0074] Next, we continue to take the adapted V4L2 architecture camera as an example. In order for the camera to adapt and work properly, the camera still needs to be adapted on the domain controller. Providing a complete camera framework based on NVIDIA orin chip and driveos plays a key role. In this framework, the nvsipl framework on Linux works with nvmedia, and nvmedia is responsible for obtaining the camera data. nvsipl not only cooperates with nvmedia, but also interacts by calling the kernel layer to realize communication and data transmission between different layers.

[0075] In order for the camera to work properly and output images, it is also necessary to flash the correct NITO (Image Tuning Object File) in the domain controller, which contains the ISP settings and related characterization parameters of the camera module, including automatic exposure, white balance and other parameter configurations. Among them, the automatic exposure parameters determine the camera's capture and adjustment of light under different lighting conditions to ensure that the image is not overexposed or too dark. The white balance parameters are used to correct the color deviation in the image so that white objects can still appear true white under different light sources, thereby ensuring the accuracy and naturalness of the overall image color. For example, if the automatic exposure is not set properly, an image taken in a strong light environment may have a white overexposed area and lose details; and inaccurate white balance may cause images taken under indoor lighting to be yellowish or bluish. Correctly configuring these parameters enables the camera to output high-quality images that meet actual needs, providing a good foundation for subsequent image processing and applications.

[0076] Finally, after completing the device debugging, connect the camera, start the camera_serive module in the soc (system-on-chip) (this module is used to monitor and manage camera-related services), and verify the data stream status. By verifying the data stream status, you can obtain key information about data transmission, such as data continuity, integrity, rate, etc., and you can determine the camera working status, such as whether it is working properly. If the data stream status is stable and meets expectations, it usually means that the camera is in good working condition; otherwise, there may be a problem.

[0077] On the domain controller, through the nvsipl module, create and configure an instance of CNvSIPLCamera to complete camera initialization and prepare for subsequent data acquisition and processing. Then call NvMedia to capture the output yuv image, frame rate and other data, and perform a simple verification of the quality and stability of the output image in the memory, such as checking whether the image is clear, distorted, accurate in color, and stable in frame rate. Then indirectly infer whether the entire link from collecting data from the camera, through transmission and processing, to the final output is correct and reliable, thereby verifying the correctness of the entire link.

[0078] When the autonomous driving data collection link in the autonomous driving data collection system is normal, the system is used to collect and process raw image / video data to expand the resources for image processing algorithm development and verification.

[0079] Based on the embodiment of the autonomous driving data collection system provided above, this specification proposes an autonomous driving data collection method.

[0080] Figure 4 This is a flow chart of an automatic driving data collection method provided in an embodiment of this specification, which includes the following steps:

[0081] S100: Collect original image / video data of the vehicle through the data acquisition source device.

[0082] The original image / video data of the vehicle is collected on the data collection element device, and the original image / video data will be transmitted to the data collection host through GMSL for processing.

[0083] In some embodiments, before the step of collecting raw image data of the vehicle by the data collection source device in step S100, the method further includes:

[0084] S110, verifying the data link of the autonomous driving data acquisition system.

[0085] Only by ensuring that the entire data transmission channel is open and verifying the correctness of the entire link can autonomous driving data be collected, and the collection of original raw format data can be achieved.

[0086] As an example, the step S110 of verifying the data link of the autonomous driving data acquisition system includes:

[0087] After accessing the data acquisition source device, the data acquisition host verifies the data flow status of a series of data transmitted by the data acquisition source device to obtain a first verification result; taking the above-mentioned V4L2 architecture camera as an example, after completing the device debugging, the camera is connected, and the camera working status can be judged by starting the camera_serive module in the soc and verifying the data flow status to obtain the first verification result.

[0088] The verification data sent by the domain controller is verified by the data acquisition host to obtain a second verification result; for example, on the domain control side, the nvsipl module is used to create an instance of the configured CNvSIPLCamera to complete the camera initialization, and NvMedia is called to capture the output yuv image, frame rate and other data, and the quality and stability of the output image are simply verified in the memory to obtain the second verification result.

[0089] When both the first verification result and the second verification result are verified to be passed, a step of collecting original image data of the vehicle by using the data collection source device is performed.

[0090] S200, receiving the original image data through the data acquisition host, and re-outputting the original image data to the domain controller.

[0091] When the original image data reaches the data acquisition host, it will imitate the function of the domain controller and re-output the original image data to the domain controller for further processing and decision-making by the domain controller.

[0092] In some embodiments, before the data acquisition host re-outputs the original image data to the domain controller in S200, the method further includes:

[0093] In the case where the data acquisition source device performs string addition processing on the collected original image data, the data acquisition host performs deserialization processing on the original image data.

[0094] This means that when the original image data reaches the data acquisition host, it will imitate the function of the domain controller. The deserializer will deserialize the original image data and restore it to its original format. The original image data will be re-output to the domain controller for further processing and decision-making.

[0095] Among them, the data conversion logic and method of the deserializer on the data acquisition host side and the serializer on the camera side are consistent.

[0096] In this embodiment, in the data transmission link, the serializer at the transmitting end serializes the data for efficient transmission on the transmission medium. At the receiving end, the deserializer restores the serial data to the original parallel format for subsequent processing and use. The configuration of the deserializer and the serializer ensures that the data is still accurate when it reaches the receiving end, ensuring that the data is accurately and stably transmitted on the link.

[0097] In the data acquisition host, after the original image data is deserialized, PHY-COPY (physical layer copy) is performed to mirror the original image / video data. This means that the same video data is copied into two copies. Figure 3 As shown, the same video data is transmitted to two D-PHYs, one of which completes data storage on the disk within the device, and the other is re-output to the device-side serializer and added to the string to the domain controller as input.

[0098] In some embodiments, before the data acquisition host re-outputs the original image data to the domain controller, the method further includes:

[0099] S210, performing data splitting processing on the original image data by the data acquisition host to obtain a first data packet and a second data packet, wherein the data in the first data packet is consistent with the data in the second data packet.

[0100] In other examples, the data acquisition host performs data splitting processing on the original image data after deserialization processing to obtain a first data packet and a second data packet.

[0101] S220: re-output the first data packet to the domain controller for analysis and decision-making.

[0102] In the process of re-outputting the first data packet to the domain controller, in order to ensure the stability and accuracy of data transmission, the first data packet is serialized by the serializer on the data acquisition host side, and the processed first data packet is re-output to the domain controller.

[0103] S230, performing data write processing on the second data packet and storing it in the data acquisition host for feeding it back to the domain controller for perception training.

[0104] This diversion method helps to ensure that the original image data is not lost during the acquisition process, while also ensuring the operation of the intelligent driving domain controller.

[0105] S300, performing perception analysis based on the original image data by the domain controller for vehicle driving.

[0106] When the domain controller receives the raw image data, it can extract valuable information from the raw image data, understand the current scene conditions, and make accurate decisions and controls to meet the needs of specific applications, such as safe driving of autonomous driving and anomaly detection of intelligent monitoring.

[0107] The present application provides an autonomous driving data acquisition system and a collection method thereof, and realizes autonomous driving data acquisition based on an autonomous driving data acquisition system including a data acquisition source device, a data acquisition host and a domain controller with a perception module. In the autonomous driving data acquisition system, the data acquisition host is electrically connected between the data acquisition source device and the domain controller, and is used to receive the original image data and re-output the original image data to the domain controller, thereby realizing the collection of original raw format data and meeting the high requirements of autonomous driving for perception accuracy and reliability.

[0108] Based on the same application concept as the above method, the present application embodiment also proposes an automatic driving data collection device, such as Figure 5 shown.

[0109] The device comprises:

[0110] The data acquisition module 602 is used to acquire original image data of the vehicle through the data acquisition source device;

[0111] A data processing module 604 is used to receive the original image data through the data acquisition host and re-output the original image data to the domain controller;

[0112] The data analysis module 606 is used to perform perception analysis based on the original image data through the domain controller for vehicle driving.

[0113] The implementation process of the functions and effects of each module / submodule / unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and the same technical effect can be achieved, so it will not be repeated here.

[0114] Figure 6 The following is a schematic diagram of the physical structure of an autonomous driving data collection device. Figure 6 As shown, the autonomous driving data collection device may include: a processor (processor) 810, a communication interface (CommunicationsInterface) 820, a memory (memory) 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the autonomous driving data collection method.

[0115] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0116] On the other hand, the present application also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the autonomous driving data collection method provided by the above-mentioned methods.

[0117] On the other hand, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the autonomous driving data collection method provided by the above-mentioned methods.

[0118] It should be noted that the technical solutions or technical features described in the above embodiments can be combined or supplemented with each other without causing conflicts. The scope of protection of this application is not limited to the precise structures described in the above embodiments and shown in the drawings; all modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application should be included in the scope of protection of this application.

Claims

1. An automatic driving data collection system, characterized in that: The system comprises: a data acquisition source device, a data acquisition host and a domain controller having a perception module; The data acquisition source device is used to acquire original image data of the vehicle; The data acquisition host is electrically connected between the data acquisition source device and the domain controller, and is used to receive the original image data and re-output the original image data to the domain controller so that the domain controller performs perception analysis based on the original image data for vehicle driving.

2. The automatic driving data collection system according to claim 1, characterized in that: The data acquisition host utilizes a serial link of a bidirectional data channel to transmit data with the data acquisition source device and the domain controller respectively, so that the data acquisition host receives the original image data sent by the data acquisition source device and receives the verification data sent by the domain controller.

3. The automatic driving data collection system according to claim 1, characterized in that: The data acquisition host is used to perform data diversion processing on the original image data to obtain a first data packet that is re-output to the domain controller and a second data packet that is stored in the data acquisition host. The second data packet is used to be fed back into the domain controller for perception training.

4. The automatic driving data collection system according to claim 2, characterized in that: The data acquisition host is also used to strip the diverted original image data through a preset data structure to obtain raw image data for perception training and embedded line data for vehicle bottom soft development and verification.

5. A method for collecting data for autonomous driving, characterized in that: The autonomous driving data collection method is applied to the autonomous driving data collection system according to any one of claims 1 to 4, and the method comprises: Collecting original image data of the vehicle through the data acquisition source device; receiving the original image data through the data acquisition host, and re-outputting the original image data to the domain controller; The domain controller performs perception analysis based on the original image data for vehicle driving.

6. The method for collecting data for autonomous driving according to claim 5, wherein: Before the data acquisition host re-outputs the original image data to the domain controller, the method further includes: Performing data splitting processing on the original image data by the data acquisition host to obtain a first data packet and a second data packet; Re-outputting the first data packet to the domain controller; Processing the second data packet to a disk, and storing it in the data acquisition host, so as to be fed back to the domain controller for perception training; The data in the first data packet is consistent with the data in the second data packet.

7. The method for collecting data for autonomous driving according to claim 6, wherein: Before the data acquisition host performs data splitting processing on the original image data to obtain the first data packet and the second data packet, the method further includes: In the case where the data acquisition source device performs string addition processing on the collected original image data, the data acquisition host performs deserialization processing on the original image data; The data acquisition host performs data splitting processing on the original image data to obtain a first data packet and a second data packet, including: The data acquisition host performs data splitting processing on the original image data after deserialization processing to obtain a first data packet and a second data packet.

8. The method for collecting data for autonomous driving according to claim 7, wherein: The step of re-outputting the first data packet to the domain controller comprises: The first data packet is subjected to string adding processing by the data acquisition host, and the processed first data packet is re-output to the domain controller.

9. The method for collecting data for autonomous driving according to claim 5, wherein: Before collecting the original image data of the vehicle by the data collection source device, the method further includes: Verify the data link of the autonomous driving data acquisition system.

10. The method for collecting data for autonomous driving according to claim 9, wherein: The verifying of the data link of the autonomous driving data acquisition system includes: After accessing the data acquisition source device, verifying the data flow status of a series of data transmitted by the data acquisition source device through the data acquisition host to obtain a first verification result; Verifying the verification data sent by the domain controller through the data acquisition host to obtain a second verification result; When both the first verification result and the second verification result are verified to be passed, a step of collecting original image data of the vehicle by using the data collection source device is performed.

Citation Information

Patent Citations

  • Evaluation method and device of perception fusion algorithm, equipment and storage medium

    CN116580271A

  • Ethernet offline detection method for intelligent driving domain controller

    CN116582464A

  • Camera data processing method and device based on domain controller ISP

    CN117857933A

  • Model training method and device based on intelligent driving data acquisition

    CN117991642A

  • Access verification method and system of domain controller

    CN118041628A

Cited By

  • Data acquisition method and device applied to intelligent driving vehicle and medium

    CN121300322A