Data processing method and device, and vehicle

Abstract

The application discloses a data processing method and device and a vehicle. The method comprises the following steps: in the process of collecting real-time data of the vehicle from a starting collection time point, determining the accumulated time between the starting time point and the current time point; in response to receiving a target label corresponding to the real-time data, setting the receiving time point of the received target label as the starting time point, and continuing to perform the step of obtaining the accumulated time between the starting time point and the current time point; in response to receiving a stop collection instruction, obtaining target data, wherein the target data is the real-time data collected from the starting collection time point to a stop collection time point when the stop collection instruction is received and the target label corresponding to the real-time data; and training an initial algorithm based on the target data to obtain a target algorithm. The application solves the technical problem of low data training efficiency in the related art.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a data processing method, apparatus, and vehicle. Background Technology

[0002] Data acquisition is a crucial part of developing advanced driver assistance perception algorithms. The acquired data includes camera video data, LiDAR data, millimeter-wave radar data, ultrasonic radar data, and bus data, which are used for algorithm training and validation during the development process. Among this data, camera data comes in various formats and is quite large. The acquisition process can lead to format differences. When a camera captures an image, it generates Raw (RAW Image Format) data after passing through an image sensor. This Raw data is then processed by an Image Signal Processor (ISP) to generate YUV ("Y" represents luminance, "U" and "V" represent chrominance) and RGB (red, green, and blue color channels) formats. The data used for algorithm training is typically YUV or RGB format image data after ISP processing.

[0003] The current mainstream solution is to integrate the camera ISP into the advanced driver assistance domain controller. This means that the camera sends raw data to the domain controller, which cannot be directly used for algorithm training.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a data processing method, apparatus, and vehicle to at least address the technical problem of low data training efficiency in related technologies.

[0006] According to one aspect of the present invention, a data processing method is provided, comprising: determining the cumulative time between a starting time point and a current time point during the process of collecting real-time data of a vehicle from a start collection time point, wherein the starting time point is the timing start point for collecting real-time data, and the current time point is the current time point for collecting real-time data; in response to receiving a target tag corresponding to the real-time data, setting the receiving time point of the target tag as the starting time point, and continuing to execute the step of obtaining the cumulative time between the starting time point and the current time point; in response to receiving a stop collection command, acquiring target data, wherein the target data consists of real-time data collected from the start collection time point to the stop collection time point at which the stop collection command is received and the target tag corresponding to the real-time data; and training an initial algorithm based on the target data to obtain a target algorithm.

[0007] Furthermore, the method also includes: comparing the cumulative time with a preset invalid time to determine whether to delete the real-time data collected within the cumulative time; and deleting the real-time data collected within the cumulative time in response to the cumulative time being greater than the preset invalid time.

[0008] Further, the initial algorithm is trained based on the target data to obtain the target algorithm, including: splitting the target data to generate first data and second data, wherein the first data and second data have different data formats; training the initial algorithm using the first data to obtain the trained initial algorithm; verifying the trained initial algorithm using the second data to obtain a verification result, wherein the verification result is used to characterize whether the trained initial algorithm is accurate; and updating the trained initial algorithm based on the verification result to obtain the target algorithm.

[0009] Further, the target data is split to obtain first data and second data, including: using a splitter board to deserialize the real-time data to obtain deserialized data; copying the deserialized data to obtain copied data; and serializing the deserialized data and the copied data respectively to generate the first data and the second data.

[0010] Furthermore, the method also includes: when the vehicle travels to the preset data collection route, activating the vehicle's data collection function and determining the activation time as the start time of data collection.

[0011] Furthermore, the method also includes: determining the stop collection time point based on the start collection time point and the preset collection duration; and generating a stop collection command when the current time point is the stop collection time point.

[0012] Furthermore, the method also includes: the stop acquisition command is a command issued by the client to stop acquiring real-time data.

[0013] Furthermore, the method also includes: displaying at least one preset label and the collected real-time data; receiving a target label corresponding to the real-time data, wherein the target label is a label determined from at least one preset label that corresponds to the real-time data.

[0014] According to another aspect of the present invention, a data processing apparatus is also provided, comprising: a cumulative time determination module, configured to determine the cumulative time between a start time point and a current time point during the process of collecting real-time data of a vehicle from a start collection time point, wherein the start time point is the timing start point for collecting real-time data, and the current time point is the current time point for collecting real-time data; an instruction execution module, configured to, in response to receiving a target tag corresponding to real-time data, set the receiving time point of the target tag as the start time point, and continue to execute the step of obtaining the cumulative time between the start time point and the current time point; a target data acquisition module, configured to, in response to receiving a stop collection instruction, acquire target data, wherein the target data consists of real-time data collected from the start collection time point to the stop collection time point at which the stop collection instruction is received and the target tag corresponding to the real-time data; and an algorithm training module, configured to train an initial algorithm based on the target data to obtain a target algorithm.

[0015] According to another aspect of the present invention, a vehicle is also provided, including one or more processors and a storage device, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to perform the above-described data processing method.

[0016] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to perform the above-described data processing method.

[0017] According to another aspect of the present invention, a processor is also provided, which is used to run a program, wherein the program executes the above-described data processing method during runtime.

[0018] In this embodiment of the invention, during the process of collecting real-time data of the vehicle from the start time point, the cumulative time between the start time point and the current time point is determined, where the start time point is the timing start point for collecting real-time data, and the current time point is the current time point for collecting real-time data; in response to receiving the target tag corresponding to the real-time data, the receiving time point of the target tag is set as the start time point, and the step of obtaining the cumulative time between the start time point and the current time point continues; in response to receiving a stop collection command, target data is obtained, where the target data is the real-time data collected from the start time point to the stop collection time point when the stop collection command is received and the target tag corresponding to the real-time data; the initial algorithm is trained based on the target data to obtain the target algorithm. It is conceivable that, in the process of determining the target data, the original video data of the vehicle can be split, and the split data can be processed by a domain controller for one channel to meet the data requirements for algorithm training, and the bypass output of the other channel to meet the data requirements for algorithm verification. This technology ensures that the data collection scenarios for algorithm verification can cover those for algorithm training. By collecting verification data and training data in parallel, it improves data collection efficiency and solves the technical problem of low data training efficiency in related technologies. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0020] Figure 1 This is a flowchart of a data processing method according to an embodiment of the present invention;

[0021] Figure 2 This is a flowchart illustrating the data acquisition process of a data processing method according to an embodiment of the present invention;

[0022] Figure 3 This is a flowchart of a data pre-slicing method according to an embodiment of the present invention;

[0023] Figure 4 This is a flowchart illustrating the acquisition of sensor data and vehicle data according to a data processing method based on an embodiment of the present invention.

[0024] Figure 5 This is a schematic diagram of a data processing apparatus according to an embodiment of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] Example 1

[0028] According to an embodiment of the present invention, an embodiment of a data processing method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0029] Figure 1 This is a flowchart of a data processing method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0030] Step S102: During the process of collecting real-time data of the vehicle from the start collection time point, determine the cumulative time between the start time point and the current time point, where the start time point is the time point at which the real-time data is collected, and the current time point is the time point at which the real-time data is collected.

[0031] Specifically, the aforementioned start time point for data collection can be understood as the time point when data collection begins when the vehicle travels to the planned data collection section.

[0032] The aforementioned real-time data includes, at a minimum, vehicle sensor data and vehicle bus data for the vehicle on the planned road section.

[0033] The vehicle sensor data includes at least the tilt angle, throttle position, steering angle, and brake or accelerator pedal position collected by the position sensor; the distance between the vehicle and obstacles and point cloud data collected by the radar sensor and ultrasonic sensor; the wheel speed and crankshaft speed data collected by the speed sensor; and video data collected by the camera sensor.

[0034] Vehicle bus data reflects the communication network data of the underlying vehicle devices or vehicle instruments interconnected in the vehicle network.

[0035] The aforementioned starting time point is the time point at which the real-time data is collected, and this starting time point is later than the aforementioned start time point for data collection.

[0036] The current time point mentioned above is the time point at which the above real-time data is currently collected, and this current time point is later than the above starting time point.

[0037] The cumulative time mentioned above is the cumulative time used to mark data tags between the aforementioned starting time point and the aforementioned current time point. The cumulative time must be consistent with the time format of the aforementioned starting time point and current time point, that is, all of them are recorded in the same time unit.

[0038] In one optional embodiment, during data collection, the planned data collection route for the vehicle is first determined according to a scenario library. When the vehicle reaches the starting point of the planned data collection segment, real-time vehicle data is collected during the vehicle's journey from the starting point, resulting in the aforementioned real-time data. The scenario library is a database used in the driving system to manage vehicle data collection scenarios.

[0039] Step S104: In response to receiving the target tag corresponding to the real-time data, set the receiving time point of the target tag as the starting time point, and continue to execute the step of obtaining the cumulative time between the starting time point and the current time point;

[0040] Specifically, the aforementioned target labels can be understood as labels used to characterize the vehicle's driving status or labels for the data collection scenario.

[0041] For example, the positions of tilt angle, throttle position, steering angle, and brake or accelerator pedal in real-time data can be corresponding to vehicle position labels; the distance between the vehicle and obstacles can be corresponding to vehicle distance labels; wheel speed, crankshaft speed, and other data can be corresponding to speed labels; tire pressure value, fuel tank pressure value, and other data can be corresponding to pressure labels; the scene label for the vehicle in front decelerating suddenly; the scene label for the vehicle in front cutting in; and the scene label for the vehicle in front cutting out.

[0042] In one optional embodiment, time a is taken as the starting time point to start collecting real-time data of the vehicle. At the same time, during the data collection process, the target tag corresponding to the real-time data is received at time b. The system can re-determine time b as the starting time point and continue to execute the above steps of obtaining the cumulative time between the starting time point and the current time point.

[0043] Step S106: In response to receiving a stop acquisition command, acquire target data, wherein the target data is the real-time data acquired from the start acquisition time to the stop acquisition time when the stop acquisition command is received and the target label corresponding to the real-time data;

[0044] Specifically, the aforementioned stop data acquisition command is used to halt the real-time data acquisition process. This stop command can be manually triggered or system-triggered.

[0045] The target data mentioned above can be understood as the real-time data collected from the start time to the stop time and its corresponding target label.

[0046] In one optional embodiment, at the start time point, real-time data during the vehicle's driving process is collected. The collected real-time data includes at least a speed tag corresponding to the vehicle's driving speed and a fuel level tag corresponding to the fuel tank pressure value. As described above, the real-time data and its tags are continuously collected until a stop collection command is received, at which point the collection of real-time data can be stopped. Then, the real-time data and its corresponding tags collected during the time period from the start time point to the stop time point are statistically analyzed.

[0047] Step S108: Train the initial algorithm based on the target data to obtain the target algorithm.

[0048] Specifically, the initial algorithm described above can be understood as converting the Raw format data captured by the vehicle's camera into YUV or RGB formats via an ISP for algorithm training. In other words, the advanced driver assistance perception algorithm for the vehicle is trained using YUV or RGB format data to verify the algorithm's stability and accuracy.

[0049] The target algorithm described above is the algorithm obtained by training the initial algorithm using the target data collected above, thereby ensuring the authenticity of the algorithm through compressed data verification and the consistency of the algorithm training data.

[0050] Specifically, during data collection, the planned collection route is first determined according to the scenario library. Data tags are then extracted based on the specific scenario and object being collected, or vehicle bus signals are selected as tags. The data tags are then set on the host computer. Once the vehicle reaches the planned collection route, data collection begins. A preprocessing invalidity period is set. During this process, preset tags can be manually triggered, or tags can be triggered by the set bus signals. During continuous data collection, if no valid tags are found within the continuous invalidity period, the system will pre-slice and automatically delete this invalid data.

[0051] Once a data acquisition route is completed, data acquisition is stopped. After all acquisition routes for the day are completed, the pre-sliced ​​data is formally sliced. Based on valid data labels, different slice durations are set for data slicing. After data slicing, multiple valid data segments are obtained. If these data segments are in YUV or RGB format, they can be used for algorithm training. If they are in Raw format (the original video format from the camera), they can be used for algorithm verification and integration testing, comprehensively verifying the hardware and software performance of the visual perception module.

[0052] This data acquisition method simultaneously completes image data acquisition before and after ISP during scene capture, satisfying the needs of visual perception algorithm verification and training, respectively. By balancing algorithm training and verification scenarios during acquisition, it avoids repetitive scene capture and improves data acquisition efficiency.

[0053] To illustrate the above data collection process, Figure 2 This is a flowchart of data acquisition in a data processing method according to an embodiment of the present invention.

[0054] like Figure 2 As shown, the domain controller first triggers the camera to expose the image. The raw data from the camera is output to the video splitter board via a serializer. After receiving the raw data, the video splitter board first deserializes the data, then copies the data, re-serializes it, and outputs it again. This achieves bypass output of the raw data, with one path going to the domain controller and the other to the data acquisition host. The data entering the domain controller is deserialized by the deserializer and then processed by the ISP to generate YUV or RGB data. The domain controller compresses this data and outputs it to the data acquisition host via the Ethernet port.

[0055] The data acquisition host simultaneously acquires raw video data and compressed YUV or RGB data, meeting the needs of algorithm training and verification during algorithm development. Algorithm training can directly utilize the acquired YUV or RGB data, while algorithm verification uses the acquired raw video data injected into the domain controller without needing to mask the domain controller's ISP, thus eliminating the need for developing special test software. This method utilizes the domain controller's own ISP during verification, maximizing the reproduction of the domain controller's visual perception module's hardware and software performance, closely mirroring the domain controller's real-world operating state, and facilitating more accurate algorithm verification and hardware / software integration testing.

[0056] Optionally, the method further includes: comparing the cumulative time with a preset invalid time to determine whether to delete the real-time data collected within the cumulative time; and deleting the real-time data collected within the cumulative time in response to the cumulative time being greater than the preset invalid time.

[0057] Specifically, the aforementioned preset invalid time can be understood as the time period covered by the preset period during which no valid data was collected. The invalid time can be preset through the system or manually.

[0058] In one optional embodiment, if no valid data is collected within the cumulative time period from the aforementioned start time point to the current time point, it can be determined whether the cumulative time is greater than a preset invalid time. If it is greater, the real-time data collected within the cumulative time period is deleted. Generally, if the cumulative time is greater than the preset invalid time, there is a problem that too much invalid data collected occupies the data storage space, resulting in a significant reduction in data write-to-disk efficiency. Therefore, it is necessary to delete the real-time data collected within the cumulative time period that is greater than the preset invalid time. Conversely, if the cumulative time is less than the preset invalid time, there is no need to delete the data.

[0059] The process of real-time data deletion can also be understood as pre-slicing the collected data, that is, removing data that is irrelevant to the algorithm training in advance. To illustrate the pre-slicing process, Figure 3 This is a flowchart of a data pre-slicing method according to an embodiment of the present invention. Figure 3 As shown:

[0060] S11, After data acquisition is started, the system records the timing start time Ts as the current time T1;

[0061] S12, start calculating the cumulative time Tt from the start time Ts;

[0062] S13, then determine whether a trigger signal has been received;

[0063] S14, When the system receives the data tag corresponding to the trigger signal, it will reset the timing start time Ts to the time Tg when the data tag was received;

[0064] S15, if the system does not receive the data tag corresponding to the trigger signal, then jump to S17;

[0065] S16, reset the cumulative time Tt to zero and start calculating the cumulative time Tt again;

[0066] S17, The system continuously checks whether the accumulated time Tt exceeds the set invalid time Tc;

[0067] S18, if Tc is exceeded, the system will cut off and automatically delete the data from the start time Ts to the current time, and the pre-slicing process will end.

[0068] S19, if not exceeded, return to S12 and recalculate the cumulative time.

[0069] In summary, when performing data pre-slicing, an invalid time Tc is set in the system. If there are no data tags for data written to disk within a continuous Tc time period, this data is considered invalid and will be automatically sliced ​​and deleted by the system. In this way, continuous invalid data is cleared in real time, saving data storage space and completing the data pre-slicing.

[0070] Optionally, training the initial algorithm based on the target data to obtain the target algorithm includes: splitting the target data to obtain first data and second data, wherein the first data and second data have different data formats; training the initial algorithm using the first data to obtain the trained initial algorithm; verifying the trained initial algorithm using the second data to obtain a verification result, wherein the verification result is used to characterize whether the trained initial algorithm is accurate; and updating the trained initial algorithm based on the verification result to obtain the target algorithm.

[0071] Specifically, the first data mentioned above is used for algorithm training and is data in YUV or RGB format obtained after ISP processing.

[0072] The aforementioned second data is used for algorithm verification and is in Raw format without image signal processing. In an optional embodiment, the original data is split using a video splitter board. Image signal processing can be performed on the first stream of split data to obtain YUV or RGB format data. Based on this data, the initial algorithm for advanced driver assistance perception is trained to obtain the trained algorithm.

[0073] Meanwhile, the trained algorithm is verified using the second stream of data after splitting. This provides a verification result that characterizes the accuracy of the initial trained algorithm. If the verification result shows that the initial trained algorithm is accurate, it is updated to obtain the updated target algorithm. Based on this target algorithm, the vehicle's driving process can be assisted in perception.

[0074] To illustrate the above data processing procedure, Figure 4 This is a flowchart illustrating the acquisition of sensor data and vehicle data according to an embodiment of the present invention.

[0075] like Figure 4 As shown, an example can be taken of a Level 3 autonomous driving vehicle for illustration.

[0076] During data acquisition, the collected data includes vehicle sensor data and vehicle bus data. The data types can be divided into three categories: video data, Ethernet data, and CAN (Controller Area Network) data. Data is split into three streams—video splitter, Ethernet switch, and CAN isolation gateway—with one stream going to the domain controller and the other to the data acquisition host, achieving bypass data acquisition. Simultaneously, to acquire the data required for visual algorithm training and development, the video data from the ISP is output and acquired by the data acquisition host.

[0077] Meanwhile, NVMe (Non-Volatile Memory Express, Logical Device Interface) solid-state drives are used for data storage. The data acquisition host is also equipped with a portable host computer, which performs operations such as configuring the data acquisition host, setting acquisition tags, starting and stopping acquisition actions, setting data pre-slicing, and data slicing.

[0078] Another common problem encountered in data acquisition is the amount of data collected. L3 autonomous driving systems currently typically use 8-megapixel front-facing cameras. Taking an 8-megapixel camera outputting Raw12 format at 30 FPS as an example, the raw data volume generated per second is 343 MB / s. If three 8-megapixel front-facing cameras are used, the camera data volume generated per second exceeds 1 GB / s. Adding data from panoramic cameras, LiDAR, millimeter-wave radar, and other sensors, the total data volume reaches 2 GB / s. This places very high demands on data storage and disk storage.

[0079] To meet the high-bandwidth disk persistence requirements, data is persisted to the NVMe solid-state drive via a PCIe (Peripheral Component Interconnect Express) interface. The typical data acquisition process involves continuous data collection, during which different data tags are applied based on the acquisition scenario. After collection, the large continuous data segments are then segmented into numerous data fragments based on these tags, including valid and invalid data fragments. This process is called data slicing. If continuous collection is performed for 6 hours per day, 43.2TB of data will be generated daily. After data slicing, the effective data volume is approximately 2 hours (14.4TB).

[0080] When performing formal data slicing, the system reads the pre-sliced ​​data, further segments the data according to the scene tags taken during data acquisition and the slice duration set by the system, and further removes invalid data. Due to the existence of data pre-slicing, the workload of formal data slicing is reduced, less invalid data is generated, and data acquisition efficiency is improved.

[0081] By monitoring data tags, real-time pre-slicing of data written to disk reduces the storage of invalid data and lowers storage costs. Simultaneously, after reading pre-sliced ​​data, the amount of data decreases, and the amount of invalid data also decreases, reducing the workload of formal data slicing and improving the efficiency of data slicing.

[0082] Optionally, the target data is split to obtain first data and second data, including: using a splitter board to deserialize the real-time data to obtain deserialized data; copying the deserialized data to obtain copied data; and serializing the deserialized data and the copied data respectively to obtain first data and second data.

[0083] Specifically, the data obtained after deserialization is the data obtained by deserializing the original data from the camera.

[0084] The copied data mentioned above is the data obtained by copying the data obtained from deserializing the serial sequence.

[0085] The first and second data are not generated directly from the split stream; the video is split into two data streams.

[0086] The first data mentioned above, namely the first data transmitted serially to the domain controller, is deserialized by the deserializer and then input to the ISP. After processing, the ISP compresses the data and outputs it, and the first data is obtained after acquisition.

[0087] The second data mentioned above, namely the second channel data, is transmitted serially to the data acquisition host, and after passing through the deserializer, it is stored to obtain the second data.

[0088] Generally, the first set of data needs to be decompressed before it can be used for algorithm training, while the second set of data can be used directly for algorithm testing and verification.

[0089] In one optional embodiment, the raw camera data is output to a video splitter board via a serializer. Upon receiving the raw data, the video splitter board first deserializes the data, then copies it, re-serializes it, and outputs the copies. This achieves bypass output of the raw data, with one path going to the domain controller and the other to the data acquisition host. This allows the camera data to simultaneously meet the needs of algorithm training and algorithm development.

[0090] Optionally, the initial algorithm is trained using the first data to obtain the trained initial algorithm, including: compressing the first data to obtain compressed data; and training the initial algorithm using the compressed data to obtain the trained initial algorithm.

[0091] Specifically, the compressed data mentioned above refers to the data obtained by compressing the YUV or RGB format data generated by the ISP processing.

[0092] In one alternative embodiment, such as Figure 2 As shown, the data entering the domain controller passes through a deserializer and is then processed by the ISP to generate YUV or RGB data. The domain controller compresses this data and outputs it to the data acquisition host via the Ethernet port. The data acquisition host simultaneously acquires both the raw video data and the compressed YUV or RGB data, meeting the needs of algorithm training and verification during algorithm development.

[0093] Optionally, the method further includes: when the vehicle travels to the preset data collection route, activating the vehicle's data collection function and determining the activation time as the start time of data collection.

[0094] Specifically, the aforementioned preset data collection route is a planned data collection route determined from multiple data collection routes in the scene library, used to collect real-time vehicle data on that route.

[0095] In one optional embodiment, the vehicle's data acquisition function can be activated via the driving system or manually. Specifically, the vehicle's camera can capture images of the view ahead, and the system can then determine whether the vehicle has reached the starting point of the preset data acquisition route. If so, the system triggers a data acquisition command to activate the vehicle's data acquisition function. Alternatively, the user can manually determine whether they have reached the starting point of the preset data acquisition route based on their own view, thus deciding whether to activate the vehicle's data acquisition function. After activating the vehicle's data acquisition function, the vehicle's start time is determined as the start time for data acquisition, enabling effective data collection from the vehicle.

[0096] Optionally, the method further includes: determining a stop collection time point based on the start collection time point and a preset collection duration; and generating a stop collection command when the current time point is the stop collection time point, wherein the stop collection command is a command issued by the client to stop collecting real-time data.

[0097] Specifically, the aforementioned preset data collection duration is the system's preset time for collecting vehicle data.

[0098] Once the start time and preset collection duration are known, the stop time can be determined so that continuous collection of vehicle data stops when the vehicle data collection time is the stop time.

[0099] The aforementioned stop data collection command is a command issued by the client to stop collecting real-time data.

[0100] In one optional embodiment, the entire data collection process of the vehicle can be counted down by stopping the collection time point. When the current time point gradually approaches and reaches the stop collection time point, the user can issue a stop collection command through the client to stop the continuous collection of vehicle data.

[0101] Example 2

[0102] According to an embodiment of the present invention, a data processing apparatus is also provided, which can execute a data processing method provided in Embodiment 1 above. The specific implementation method and preferred application scenario are the same as those in Embodiment 1 above, and will not be described again here.

[0103] Figure 5 This is a schematic diagram of a data processing apparatus according to an embodiment of the present invention, such as... Figure 5 As shown, the device includes:

[0104] The cumulative time determination module 502 is used to determine the cumulative time between the starting time point and the current time point during the process of collecting real-time data of the vehicle from the start time point. The starting time point is the time point at which the real-time data is collected, and the current time point is the time point at which the real-time data is collected.

[0105] The instruction execution module 504 is used to respond to the target tag corresponding to the received real-time data, set the receiving time point of the target tag as the starting time point, and continue to execute the step of obtaining the cumulative time between the starting time point and the current time point;

[0106] The target data acquisition module 506 is used to acquire target data in response to receiving a stop acquisition command. The target data consists of real-time data collected from the start acquisition time to the stop acquisition time when the stop acquisition command is received, and the target tags corresponding to the real-time data.

[0107] The algorithm training module 508 is used to train the initial algorithm based on the target data to obtain the target algorithm.

[0108] Optionally, the device further includes: a time comparison module for comparing the accumulated time with a preset invalid time to determine whether to delete the real-time data collected within the accumulated time; and a data deletion module for deleting the real-time data collected within the accumulated time in response to the accumulated time being greater than the preset invalid time.

[0109] Optionally, the algorithm training module 508 includes: a data splitting module for splitting the target data to generate first data and second data, wherein the first data and second data have different data formats; an algorithm training module for training an initial algorithm using the first data to obtain a trained initial algorithm; an algorithm verification module for verifying the trained initial algorithm using the second data to obtain a verification result, wherein the verification result is used to characterize whether the trained initial algorithm is accurate; and an algorithm update module for updating the trained initial algorithm based on the verification result to obtain the target algorithm.

[0110] Optionally, the data splitting module includes: a sequence deserialization unit, used to deserialize the serial sequence of real-time data using a splitting board to obtain deserialized data; a data copying unit, used to copy the deserialized data to obtain copied data; and a data output unit, used to serialize and output the deserialized data and the copied data respectively to generate first data and second data.

[0111] Optionally, the device further includes a function activation module, used to activate the vehicle's data acquisition function when the vehicle travels to the preset data acquisition route, and to determine the activation time as the start time of data acquisition.

[0112] Optionally, the device further includes: a time point determination module, used to determine the stop acquisition time point based on the start acquisition time point and the preset acquisition duration; and an instruction generation module, used to generate a stop acquisition instruction when the current time point is the stop acquisition time point.

[0113] Optionally, the device further includes: a data display module for displaying at least one preset tag and the collected real-time data; and a tag receiving module for receiving a target tag corresponding to the real-time data, wherein the target tag is a tag determined from at least one preset tag that corresponds to the real-time data.

[0114] Example 3

[0115] According to an embodiment of the present invention, a vehicle is also provided, including one or more processors and a storage device, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to perform the above-described data processing method.

[0116] Example 4

[0117] According to an embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein, when the program is running, it controls the device where the computer-readable storage medium is located to perform the above-described data processing method.

[0118] Example 5

[0119] According to an embodiment of the present invention, a processor is also provided, which is used to run a program, wherein the program executes the above-described data processing method during runtime.

[0120] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0121] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0122] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0123] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0124] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0125] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0126] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A data processing method, characterized in that, include: During the process of collecting real-time data of the vehicle from the start time point, the cumulative time between the start time point and the current time point is determined. The start time point is the time point at which the real-time data is collected, and the current time point is the current time point at which the real-time data is collected. The real-time data includes at least: vehicle sensor data and vehicle bus data of the vehicle in the planned collection section. In response to receiving the target tag corresponding to the real-time data, the receiving time point of the target tag is set as the starting time point, and the step of obtaining the cumulative time between the starting time point and the current time point continues to be executed; In response to receiving a stop collection command, target data is acquired, wherein the target data is the real-time data collected from the start collection time to the stop collection time when the stop collection command is received and the target label corresponding to the real-time data; The initial algorithm is trained based on the target data to obtain the target algorithm; Training an initial algorithm based on the target data to obtain a target algorithm includes: splitting the target data to obtain first data and second data, wherein the first data and second data have different data formats, the first data is YUV or RGB format data obtained after ISP processing, and the second data is raw image encoded data in the original image encoding format without image signal processing; training the initial algorithm using the first data to obtain a trained initial algorithm; verifying the trained initial algorithm using the second data to obtain a verification result, wherein the verification result is used to characterize whether the trained initial algorithm is accurate; and updating the trained initial algorithm based on the verification result to obtain the target algorithm. The method further includes: comparing the cumulative time with a preset invalid time to determine whether to delete the real-time data collected within the cumulative time.

2. The data processing method according to claim 1, characterized in that, The method further includes: In response to the cumulative time exceeding the preset invalid time, the real-time data collected within the cumulative time is deleted.

3. The data processing method according to claim 1, characterized in that, The target data is split to obtain first data and second data, including: The serial sequence of the real-time data is deserialized using a splitter board to obtain the deserialized data; The data after the release is copied to obtain copied data; The deactivated data and the copied data are serialized and output separately to generate the first data and the second data.

4. The data processing method according to claim 1, characterized in that, The method further includes: When the vehicle travels to the preset data collection route, the data collection function of the vehicle is activated, and the activation time is determined as the start time of data collection.

5. The method according to claim 1, characterized in that, The method further includes: The stop collection time point is determined based on the start collection time point and the preset collection duration; If the current time point is the time point at which data acquisition is stopped, the stop data acquisition command is generated.

6. The method according to claim 1, characterized in that, The stop data collection command is a command issued by the client to stop collecting the real-time data.

7. The method according to claim 1, characterized in that, The method further includes: Display at least one preset label and the collected real-time data; Receive the target tag corresponding to the real-time data, wherein the target tag is a tag corresponding to the real-time data determined from the at least one preset tag.

8. A data processing apparatus, characterized in that, The device includes: The cumulative time determination module is used to determine the cumulative time between the starting time point and the current time point during the process of collecting real-time data of the vehicle from the start time point. The starting time point is the time point at which the real-time data is collected, and the current time point is the time point at which the real-time data is collected. The real-time data includes at least: vehicle sensor data and vehicle bus data of the vehicle in the planned collection section. The instruction execution module is used to, in response to receiving the target tag corresponding to the real-time data, set the receiving time point of the target tag as the starting time point, and continue to execute the step of obtaining the cumulative time between the starting time point and the current time point; The target data acquisition module is used to acquire target data in response to receiving a stop acquisition command, wherein the target data is the real-time data collected from the start acquisition time to the stop acquisition time when the stop acquisition command is received and the target label corresponding to the real-time data; The algorithm training module is used to train the initial algorithm based on the target data to obtain the target algorithm; The algorithm training module is further configured to split the target data into first data and second data, wherein the first data and second data have different data formats: the first data is YUV or RGB format data obtained after ISP processing, and the second data is raw image encoded data in the original image encoding format without image signal processing; the first data is used to train the initial algorithm to obtain the trained initial algorithm; the second data is used to verify the trained initial algorithm to obtain a verification result, wherein the verification result is used to characterize whether the trained initial algorithm is accurate; and the trained initial algorithm is updated based on the verification result to obtain the target algorithm. The instruction execution module is also used to compare the cumulative time with a preset invalid time to determine whether to delete the real-time data collected within the cumulative time.

9. A vehicle, characterized in that, include: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the data processing method according to any one of claims 1 to 7.

Images