An automatic driving data collection multi-dimension segmentation method and system

Abstract

This invention discloses a multi-dimensional segmentation method for autonomous driving data collection, comprising the following steps: obtaining a segmentation configuration, including a start maximum time, an end maximum time, a start effective distance, an end effective distance, and a segment time interval; reading the data collection packet and extracting multi-sensor data and vehicle speed information; calculating the cumulative driving distance using timestamps and vehicle speed information; performing a first validity check on the starting segment of the data collection packet using the start maximum time and the start effective distance, and performing a second validity check on the ending segment of the data collection packet using the end maximum time and the end effective distance; if either check fails, the segmentation is abandoned; if both checks pass, a first and a second valid time point are determined, and the interval between the two is taken as the valid time interval; within this interval, the multi-sensor data is segmented according to the segment time interval to obtain one or more data segments.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving data processing technology, specifically to a multi-dimensional segmentation method and system for autonomous driving data collection. Background Technology

[0002] The development of autonomous driving technology relies heavily on large-scale, high-quality real-world road data collection. This data is typically collected continuously by vehicles equipped with various sensors (such as LiDAR, cameras, and millimeter-wave radar) while driving on actual roads, and then packaged and stored as data packets according to a specific strategy.

[0003] In the training and evaluation of autonomous driving perception, prediction, and planning algorithms, it is often necessary to divide long periods of continuously collected data into several shorter data segments to facilitate data management, annotation, distribution, and model training. Currently, traditional segmentation methods are mainly based on a single time dimension, such as cutting data segments at fixed intervals (e.g., every 20 seconds). However, this segmentation strategy has significant technical drawbacks: when the vehicle briefly stops during operation (e.g., waiting at a traffic light), the vehicle body does not actually move, but time-based segmentation still generates corresponding data segments. These segments record essentially unchanged road scenes, resulting in highly repetitive content.

[0004] In such scenarios, repetitive data fragments not only fail to provide effective information increments for subsequent downstream tasks such as mapping or algorithm training, but also lead to the following problems: the total amount of data that needs to be processed in mapping operations increases unnecessarily, while the mapping quality does not improve accordingly; a large amount of invalid data flow and computational resource consumption are generated; at the same time, this redundant data will continue to occupy storage space and increase the corresponding storage costs. Summary of the Invention

[0005] In order to avoid generating redundant data fragments that repeat in the same scenario from the source and to reduce the technical defects of data processing pressure and storage burden in the production process, this invention provides a multi-dimensional segmentation method and system for autonomous driving data collection.

[0006] To solve the above problems, the present invention is implemented according to the following technical solution:

[0007] In a first aspect, the present invention provides a multi-dimensional segmentation method for autonomous driving data collection, comprising the following steps: obtaining segmentation configuration information, the segmentation configuration information including a start maximum time, an end maximum time, a start effective distance, an end effective distance, and a segment time interval; reading a data collection packet, extracting multi-sensor data from the data collection packet, and obtaining vehicle driving speed information; calculating the cumulative driving distance corresponding to the data collection packet based on the timestamps of the multi-sensor data and the vehicle driving speed information; performing a first validity check on the starting segment of the data collection packet using the start maximum time and the start effective distance, and performing a second validity check on the ending segment of the data collection packet using the end maximum time and the end effective distance; when the first validity check or the second validity check fails, the data collection packet is determined as an invalid data packet, and segmentation is abandoned; when both the first validity check and the second validity check pass, a first effective time point and a second effective time point are determined, and the time interval between the first effective time point and the second effective time point is taken as the effective time interval; within the effective time interval, the multi-sensor data is segmented according to the segment time interval to obtain one or more data segments.

[0008] In conjunction with the first aspect, the present invention provides a first specific implementation of the first aspect, wherein the multi-sensor data includes point cloud data; the vehicle speed information is obtained by parsing the CAN bus data in the acquisition data packet, and the CAN bus data includes the correspondence between timestamps and vehicle speeds.

[0009] In conjunction with the first aspect, the present invention provides a second specific implementation of the first aspect. Specifically, the step of calculating the cumulative driving distance corresponding to the collected data packet based on the timestamps of the multi-sensor data and the vehicle driving speed information includes: aligning the timestamps of the point cloud data with the timestamps of the CAN bus data, and assigning a corresponding vehicle speed to each frame of point cloud data; calculating the time difference between two adjacent frames of point cloud data, and using the product of the time difference and the corresponding vehicle speed as the inter-frame driving distance; and accumulating all the inter-frame driving distances in chronological order to obtain the cumulative driving distance.

[0010] In conjunction with the first aspect, the present invention provides a third specific implementation of the first aspect. Specifically, determining the first valid time point and the second valid time point includes: starting from the start time of the data collection, searching along the time sequence for the moment when the cumulative driving distance first reaches the starting valid distance, and taking this as the first valid time point; starting from the end time of the data collection, searching backward along the time axis, and taking the moment when the remaining driving distance first reaches the end valid distance as the second valid time point; the remaining driving distance is the driving distance from the current time to the end time.

[0011] In conjunction with the first aspect, the present invention provides a fourth specific implementation of the first aspect. Specifically, the step of segmenting the multi-sensor data according to the segment time interval within the effective time interval to obtain one or more data segments includes: obtaining the duration of the effective time interval, and determining the number of segments to be segmented and the time margin based on the duration and the segment time interval; for each segment to be segmented, calculating the segment start time and segment end time based on the time margin, the segment time interval, and the segment number; and extracting data located between the segment start time and the segment end time from the multi-sensor data to form a data segment.

[0012] In conjunction with the first aspect, the present invention provides a fifth specific implementation of the first aspect. Specifically, the calculation of the segment start time and segment end time includes: shifting the segment start time forward by a first extended duration, the first extended duration being the time length from the start time of the acquired data packet to the first valid time point; and shifting the segment end time backward by a second extended duration, the second extended duration being the time length from the second valid time point to the end time of the acquired data packet.

[0013] In conjunction with the first aspect, the present invention provides a sixth specific implementation of the first aspect. Specifically, if the vehicle is detected to be stationary for more than a preset third time threshold within the maximum start time or the maximum end time, the data corresponding to the stationary period is discarded, and the cumulative driving distance is recalculated based on the remaining data, and the first validity check or the second validity check is performed again.

[0014] Secondly, the present invention also provides a multi-dimensional segmentation system for autonomous driving data collection, comprising: a configuration acquisition module for acquiring segmentation configuration information, the segmentation configuration information including a start maximum time, an end maximum time, a start effective distance, an end effective distance, and a segment time interval; a data extraction module for reading the data collection data packet, extracting multi-sensor data from the data collection data packet, and acquiring vehicle driving speed information; a distance calculation module for calculating the cumulative driving distance corresponding to the data collection data packet based on the timestamps of the multi-sensor data and the vehicle driving speed information; and a validity verification module for performing a first validity check on the starting segment of the data collection data packet using the start maximum time and the start effective distance. The system includes a validity check module, which performs a second validity check on the end segment of the acquired data packet using the maximum end time and the effective end distance; an invalidation processing module, which determines the acquired data packet as invalid and abandons the segmentation when either the first or second validity check fails; an interval determination module, which determines a first effective time point and a second effective time point when both the first and second validity checks pass, and defines the time interval between the first and second effective time points as the effective time interval; and a segmentation module, which segments the multi-sensor data within the effective time interval according to the segment time interval to obtain one or more data segments.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] Introducing driving distance as a segmentation criterion complements the approach of simply segmenting by time. When a vehicle remains stationary due to waiting at a red light or temporary stop, the cumulative driving distance during this stationary period does not increase. During the validity verification phase, this data is deemed invalid and discarded because it fails to reach the start or end valid distance, directly preventing the generation of redundant segments with no change in the scene. The advantage of this approach is a significant reduction in the amount of data transmitted to downstream stages such as annotation, mapping, and model training. Furthermore, the data eliminated consists of repetitive data that does not add contextual information, thus not affecting the quality of the work. Simultaneously, the storage space usage and associated storage costs are also reduced.

[0017] With independently configurable start and end effective distances, it can flexibly adapt to different data acquisition tasks and data quality requirements. For example, it can ensure that the beginning and end of a data packet have sufficient travel distance, avoiding the cutting out of useless segments when the scene is just starting or about to stop, so that each cut-out data segment corresponds to a real driving process, making the scenarios diverse and effective.

[0018] Ultimately, the flow of invalid data in the entire data processing chain is reduced, the workload of downstream operations decreases, and the processing speed and overall efficiency are improved. Attached Figure Description

[0019] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0020] Figure 1 This is a flowchart of a multi-dimensional segmentation method for autonomous driving data collection according to the present invention.

[0021] Figure 2 This is a block diagram of an electronic device used to implement embodiments of the present invention;

[0022] In the picture:

[0023] 100 - Electronic device, 101 - Computing unit, 102 - ROM, 103 - RAM, 104 - Bus, 105 - I / O interface, 106 - Input unit, 107 - Output unit, 108 - Storage unit, 109 - Communication unit. Detailed Implementation

[0024] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0025] like Figures 1-2 As shown, this invention provides a multi-dimensional segmentation method and system for autonomous driving data collection.

[0026] Example 1

[0027] A multi-dimensional segmentation method for autonomous driving data collection includes the following steps: obtaining segmentation configuration information, which includes a start maximum time, an end maximum time, a start effective distance, an end effective distance, and a segment time interval; reading a data collection packet, extracting multi-sensor data from the data collection packet, and obtaining vehicle speed information; calculating the cumulative driving distance corresponding to the data collection packet based on the timestamps of the multi-sensor data and the vehicle speed information; performing a first validity check on the starting segment of the data collection packet using the start maximum time and the start effective distance, and performing a second validity check on the ending segment of the data collection packet using the end maximum time and the end effective distance; when either the first validity check or the second validity check fails, the data collection packet is determined as an invalid data packet, and segmentation is abandoned; when both the first validity check and the second validity check pass, a first effective time point and a second effective time point are determined, and the time interval between the first effective time point and the second effective time point is taken as the effective time interval; within the effective time interval, the multi-sensor data is segmented according to the segment time interval to obtain one or more data segments.

[0028] Specifically, the multi-dimensional segmentation method for autonomous driving data acquisition described in this embodiment is applied to an autonomous driving data preprocessing system to segment the raw sensor data packets (rosbag) acquired by the acquisition vehicle. This method extracts effective data clips from the acquisition data packets that meet both duration and distance requirements through joint constraints of time and space dimensions, thus solving the problem of large amounts of repetitive data generated by traditional fixed-time segmentation methods in scenarios where the vehicle is stationary or moving slowly.

[0029] S100: Obtain segmentation configuration information, which includes start maximum time, end maximum time, start effective distance, end effective distance, and segment time interval.

[0030] Specifically, when a task is initiated, the system first obtains the preset splitting configuration information. This splitting configuration information includes: the maximum start time. End maximum time Start effective distance End of effective distance and the time interval between segments Among them, the maximum start time Used to limit the maximum time window during the initial phase of a data packet's transmission to reach the initial effective distance; maximum end time. Used to limit the maximum time window allowed to accumulate until the effective end distance is reached during the packet termination phase; fragment time interval This parameter is used to limit the minimum duration of each clip. The above configuration parameters can be flexibly set according to the actual application scenario, including the type of road being collected, vehicle driving conditions, and downstream task requirements.

[0031] S200: Read the data acquisition data packet, extract multi-sensor data from the data acquisition data packet, and obtain vehicle driving speed information.

[0032] Specifically, the data acquisition data packet is a rosbag format data log file containing time-series data recorded by multiple sensors during vehicle operation. By parsing this data acquisition data packet, point cloud radar data, image data, and other multi-sensor data are extracted. Simultaneously, CAN bus information is read from the data acquisition data packet, and the timestamps and corresponding vehicle speed values ​​contained in the CAN bus messages are extracted. Each frame of data from the multi-sensor sensors carries a corresponding timestamp, serving as the basis for time-dimension processing.

[0033] As an optional embodiment, the multi-sensor data includes point cloud data; the vehicle speed information is obtained by parsing the CAN bus data in the data acquisition packet, and the CAN bus data contains the correspondence between timestamps and vehicle speeds.

[0034] In a more specific implementation, the multi-sensor data specifically includes point cloud data acquired by a lidar sensor. The point cloud data consists of several frames, each frame corresponding to a timestamp indicating the time the lidar acquired that frame. When acquiring vehicle speed information, CAN bus data is extracted from the data acquisition packet. The CAN bus data is a sequence of vehicle speed messages recorded by the CAN bus during the acquisition process; each message in this sequence contains a... A timestamp and a The vehicle speed value and the time difference form a one-to-one correspondence. When calculating the cumulative travel distance, for two adjacent point cloud frames, the timestamps of each frame are obtained, and the time difference is calculated. Simultaneously, based on the time interval to which this time difference belongs, the corresponding `can_speed` vehicle speed value is retrieved from the CAN bus data. This vehicle speed value is used as the speed parameter for the vehicle's travel between the two adjacent point cloud frames. Multiplying the time difference by the vehicle speed value yields the displacement distance traveled by the vehicle between the two adjacent point cloud frames.

[0035] S300: Based on the timestamps of the multi-sensor data and the vehicle speed information, calculate the cumulative driving distance corresponding to the collected data packet.

[0036] Specifically, for two adjacent frames of data from multiple sensors, the time difference between their timestamps is calculated. This time difference is multiplied by the vehicle's speed at the corresponding moment to obtain the displacement distance between the two adjacent frames. This process is repeated for each pair of adjacent data frames arranged along the time axis in the data acquisition packet, calculating the displacement distance for each segment and summing all the displacement distances to obtain the cumulative distance traveled by the data acquisition packet throughout the entire acquisition period. This cumulative distance reflects the actual spatial movement range of the vehicle during data acquisition and is used for validity assessment based on the distance dimension.

[0037] As an optional embodiment, the step of calculating the cumulative driving distance corresponding to the collected data packet based on the timestamps of the multi-sensor data and the vehicle driving speed information includes: aligning the timestamps of the point cloud data with the timestamps of the CAN bus data, and assigning a corresponding vehicle speed to each frame of point cloud data;

[0038] Specifically, the point cloud data consists of several point cloud frames, each with its own acquisition timestamp. The CAN bus data contains a series of message records consisting of timestamps and vehicle speed values. The timestamp sequence of the point cloud data is aligned and matched with the timestamp sequence of the CAN bus data on the time axis, so that each point cloud frame is assigned a corresponding vehicle speed value based on its timestamp. For a given point cloud frame, if its timestamp is exactly equal to the timestamp of a CAN bus message, the vehicle speed value of that message is directly assigned to that frame; if its timestamp falls between the timestamps of two adjacent CAN bus messages, the vehicle speed value corresponding to that frame can be determined using nearest neighbor matching or linear interpolation. Thus, the vehicle speed information corresponding to each frame of point cloud data is constructed, obtaining a one-to-one correspondence between point cloud frames and vehicle speeds.

[0039] Calculate the time difference between two adjacent frames of point cloud data, and use the product of the time difference and the corresponding vehicle speed as the inter-frame travel distance.

[0040] Specifically, after aligning and matching the point cloud data with the vehicle speed information, each pair of adjacent frames in the point cloud data sequence is traversed along the time axis. For the nth frame and the (n+1)th frame point cloud, their respective timestamps are obtained, and the absolute value of the difference between the two timestamps is calculated to obtain the time interval between the two adjacent frame point clouds. The corresponding vehicle speed value within this time interval is determined, using the vehicle speed assigned to the nth frame point cloud or the average of the vehicle speeds assigned to the nth and (n+1)th frame point clouds as the speed parameter for calculation. The time interval is multiplied by the speed parameter for calculation, and the product is the inter-frame travel distance of the vehicle between the nth and (n+1)th frame point clouds.

[0041] The cumulative travel distance is obtained by summing up the travel distances between all the frames in chronological order.

[0042] Specifically, starting from the first point cloud frame with the earliest timestamp in the data acquisition packet, the inter-frame travel distances calculated for each pair of adjacent frames are accumulated sequentially until the last point cloud frame with the latest timestamp is reached. The accumulated result of all inter-frame travel distances is taken as the cumulative travel distance corresponding to the data acquisition packet. This cumulative travel distance reflects the total amount of actual spatial movement of the acquisition vehicle along the travel path during the entire data acquisition process.

[0043] S400: Perform a first validity check on the starting segment of the collected data packet using the maximum start time and the effective start distance, and perform a second validity check on the ending segment of the collected data packet using the maximum end time and the effective end distance.

[0044] Specifically, the processing logic for the first validity check is as follows: Starting from the initial moment of the data packet collection, a time interval with a duration equal to the maximum initial time is extracted along the time axis. The cumulative distance traveled by the vehicle within this initial time interval is calculated. This cumulative distance is then compared with the initial valid distance. If the cumulative distance is less than the initial valid distance, the first validity check is deemed to have failed, indicating that the vehicle was stationary or moving slowly at the beginning, and the collected data scene showed no substantial spatial change.

[0045] The second validity check is processed as follows: Starting from the end time of the data packet acquisition, a time interval equal to the maximum end time is extracted along the time axis, and the cumulative distance traveled by the vehicle within this time interval is calculated. This cumulative distance is then compared with the valid end distance. If the cumulative distance is less than the valid end distance, the second validity check fails, indicating that the vehicle was also stationary or moving slowly during the end phase.

[0046] S500: When the first validity check or the second validity check fails, the collected data packet is determined to be an invalid data packet, and the segmentation is abandoned.

[0047] Specifically, when the first validity check fails, it indicates that the starting segment of the collected data packet does not meet the distance dimension requirement, and the data packet is directly classified as an invalid data packet. When the second validity check fails, it indicates that the ending segment of the collected data packet does not meet the distance dimension requirement, and the data packet is also classified as an invalid data packet. For data packets determined to be invalid, the system will not execute the subsequent segmentation process, thereby filtering out low-quality data with repetitive scenes and insufficient spatial coverage at the source, preventing such data from entering the downstream mapping or annotation production chain.

[0048] S600: When both the first validity check and the second validity check pass, determine the first valid time point and the second valid time point, and take the time interval between the first valid time point and the second valid time point as the valid time interval.

[0049] Specifically, locating forward along the time axis, the moment when the cumulative travel distance in the collected data packet first reaches the starting effective distance is determined and marked as the first effective time point. Locating backward along the time axis, the moment when the cumulative travel distance in the collected data packet first reaches the ending effective distance is determined and marked as the second effective time point. The continuous time interval between the first and second effective time points is defined as the effective time interval. This effective time interval eliminates any stationary or slowly moving portions that may exist in the initial and final segments, retaining the effective data segments with sufficient spatial movement and variation in the middle, which serve as the input range for the segmentation operation.

[0050] As an optional embodiment, determining the first valid time point and the second valid time point includes: starting from the start time of the data collection, searching chronologically for the moment when the cumulative driving distance first reaches the starting valid distance, and using this moment as the first valid time point;

[0051] Specifically, the inter-frame travel distance is calculated frame by frame along the timeline, starting with the earliest timestamp in the data packet. During the accumulation process, the current cumulative travel distance is compared with the initial effective distance in real time. When the cumulative travel distance is greater than or equal to the initial effective distance for the first time, the corresponding timestamp is recorded, which is the first effective time point. The first effective time point indicates that the data acquisition vehicle has completed sufficient spatial displacement in the initial stage, and the data collected from this moment onwards contains valid scene change information.

[0052] Starting from the end time of the data packet collection, a reverse search is performed along the time axis, and the moment when the remaining driving distance first reaches the end effective distance is taken as the second effective time point; the remaining driving distance is the driving distance from the current time to the end time.

[0053] Specifically, starting with the latest timestamped data frame in the data acquisition packet, the inter-frame travel distance is accumulated frame by frame in reverse along the timeline. The direction of accumulation is reversed from the chronological order, calculating the distance frame by frame from the last frame to earlier times. During this reverse accumulation process, the currently accumulated travel distance is compared in real-time with the final valid distance. When the accumulated travel distance first exceeds or equals the final valid distance, the corresponding timestamp is recorded; this timestamp is the second valid time point. This accumulated travel distance is the remaining travel distance, representing the total distance traveled by the vehicle from this moment until the end of the acquisition process. The second valid time point signifies that the vehicle has completed sufficient spatial displacement in the subsequent final phase; the data segment before this time point contains the acquisition content that includes valid scene changes.

[0054] By combining the above forward and reverse search methods, the first and second effective time points are determined respectively. The resulting time interval [first effective time point, second effective time point] is the effective time interval that simultaneously satisfies the distance requirements of the starting segment and the ending segment.

[0055] It should be noted that this embodiment does not limit the specific implementation of forward search and reverse search. For example, frame-by-frame traversal, binary search, or fast positioning based on a pre-calculated distance accumulation array can be used. Those skilled in the art can make adaptive choices according to the data scale and processing efficiency requirements.

[0056] S700: Within the effective time interval, the multi-sensor data is segmented according to the segment time interval to obtain one or more data segments.

[0057] Specifically, the total duration corresponding to the effective time interval is calculated, and this total duration is divided by the segment time interval to obtain the integer division result and the remainder result. The integer division result is the number of data segments that can be divided, and the remainder result is the remaining duration after division. For each data segment to be generated, its start time is calculated as follows: based on the first effective time point, add the remaining duration, add the product of the segment number minus one and the segment time interval, and then back a certain duration from this time point. The backdated duration is the time occupied by the corresponding initial effective travel segment before the first effective time point, thus obtaining the segment division start time of this data segment. Its end time is calculated as follows: based on the first effective time point, add the remaining duration, add the product of the segment number and the segment time interval, and then extend a certain duration from this time point. The extended duration is the time occupied by the corresponding final effective travel segment after the second effective time point, thus obtaining the segment division end time of this data segment. According to the above start and end times, all data within the corresponding time period is extracted from the multi-sensor data and output as a data segment. Repeat the above process until all data fragments are generated.

[0058] As an optional embodiment, the step of segmenting the multi-sensor data according to the segment time interval within the effective time interval to obtain one or more data segments includes: obtaining the duration of the effective time interval, and determining the number of segments to be segmented and the time margin according to the duration and the segment time interval;

[0059] Specifically, based on the first and second valid time points, the time difference between them is calculated, and this time difference represents the duration of the valid time interval. The duration is then divided by the segment time interval to obtain a divisibility result and a remainder result. The divisibility result represents the number of segments to be divided, indicating the number of data segments that can be completely divided within the valid time interval. The remainder result represents the time margin, indicating the remaining time length after dividing into an integer number of segments. For example, if the duration of the valid time interval is 75 seconds and the segment time interval is 30 seconds, then 75 divided by 30 results in a divisibility result of 2, indicating that 2 complete data segments can be divided, and the remainder result is 15, indicating a 15-second time margin.

[0060] For each segment to be segmented, the segment start time and segment end time are calculated based on the time margin, the segment time interval, and the segment number.

[0061] Specifically, to ensure a balanced distribution of the segmented data along the time axis, time margins are preferentially allocated before the first data segment or between data segments. Let the segment number to be segmented be i, where i is a positive integer starting from 1. The start time of the i-th segment is calculated as follows: based on the first valid time point, add the time margin, add the product of segment number i - 1 and the segment time interval, and then extend backward by a starting extension duration from this time point. This starting extension duration is the time occupied by the corresponding initial valid travel segment before the first valid time point; its purpose is to ensure that the segmented data includes spatial movement information of the initial stage. The end time of the i-th segment is calculated as follows: based on the first valid time point, add the time margin, add the product of segment number i and the segment time interval, and then extend backward by a finishing extension duration from this time point. This finishing extension duration is the time occupied by the corresponding final valid travel segment after the second valid time point; its purpose is to ensure that the segmented data includes spatial movement information of the final stage. The above calculation method ensures that each data segment not only meets the duration requirement of the segment time interval, but also includes the spatial movement changes at the beginning and end, avoiding the problem of incomplete scenes caused by fixed time segmentation.

[0062] Data located between the start time and end time of the segment is extracted from the multi-sensor data to form a data segment.

[0063] Specifically, after determining the start and end times of the i-th segment, all data frames in the multi-sensor data are traversed, and all data frames with timestamps between the segment start and end times are selected. This selection covers all types of multi-sensor data, including point cloud data and image data extracted from the acquisition data packets. All selected data frames are then organized according to the original data structure to form a data segment containing complete temporal and spatial information, which is used for downstream operations such as mapping.

[0064] It should be noted that this embodiment does not limit the specific allocation method of the time margin. For example, the time margin can be evenly distributed to the beginning and end of all data segments, or the time margin can be concentrated on the first or last data segment. Those skilled in the art can make adaptive choices according to the actual scenario requirements and the continuity requirements of the data segments.

[0065] As an optional embodiment, calculating the segment start time and segment end time includes: shifting the segment start time forward by a first extended duration, where the first extended duration is the time length from the start time of the acquired data packet to the first valid time point;

[0066] Specifically, after determining the start time of a data segment, the time interval between the start time of the data packet acquisition and the first valid time point is calculated; this time interval is the first extended duration. The segment start time is shifted forward along the time axis by the first extended duration to obtain the extended segment start time. The first extended duration corresponds to the time elapsed when the data acquisition vehicle reaches the initial valid distance during the initial phase. Although the data within this time period is determined to be in a stationary or slow-moving phase by validity checks, the scene information it contains is still valuable for downstream tasks. By shifting forward and incorporating this segment of data, the segmented data retains the complete transition process from a stationary state to a valid moving state in the time dimension, which is beneficial for mapping or annotation tasks to continuously understand the scene.

[0067] The end time of the segment is shifted backward by a second extended duration, which is the length of time from the second valid time point to the end time of the acquired data packet.

[0068] Specifically, after determining the end time of a data segment, the time interval between the second valid time point and the end time of the acquired data packet is calculated; this time interval is the second extended duration. The segment end time is shifted backward along the time axis by the second extended duration to obtain the extended segment end time. The second extended duration corresponds to the time elapsed when the acquired vehicle reaches the final valid distance during the final stage of travel. The data within this time period includes the transition scenario of the vehicle gradually entering a stopped or slow-moving state from an effective driving state. By shifting backward and incorporating this data segment, the segmented data covers the complete change process from the effective driving state to the final stationary state.

[0069] By extending the start time forward and the end time backward as described above, the actual time span of the final segmented data is [segment start time - first extension duration, segment end time + second extension duration]. This not only meets the basic duration requirement of the segment time interval, but also ensures that each data segment has complete scene transition semantics when entering the downstream annotation or mapping process.

[0070] As an optional embodiment, if the vehicle is detected to be stationary for more than a preset third time threshold within the maximum start time or the maximum end time, the data corresponding to the stationary period is discarded, and the cumulative driving distance is recalculated based on the remaining data, and the first validity check or the second validity check is performed again.

[0071] Specifically, during the first validity check, starting from the beginning of the data packet acquisition, the vehicle speed is detected frame by frame along the time axis. If, within the time window defined by the maximum start time, the vehicle speed is continuously zero or below a preset stationary speed threshold for a duration exceeding a preset third time threshold, it is determined that a continuous stationary interval exists within the initial segment. The scene content corresponding to the data frames within this continuous stationary interval has no substantial spatial change and is considered redundant data. In this case, all data frames within this continuous stationary interval are discarded from the initial segment, retaining only the remaining data before and after this stationary interval. Based on the remaining initial segment data after the discarding process, the cumulative driving distance is recalculated, and the recalculated cumulative driving distance is compared with the initial valid distance again to perform the first validity check.

[0072] During the second validity check, starting from the end of the data packet acquisition, the vehicle speed is detected frame by frame in reverse along the timeline. If, within the time window defined by the maximum end time, the vehicle speed is continuously zero or below a preset stationary speed threshold for a duration exceeding a preset third time threshold, it is determined that a continuous stationary interval exists within the end segment. All data frames within this continuous stationary interval are discarded from the end segment, retaining only the remaining data before and after this stationary interval. Based on the remaining end segment data after the discarding process, the cumulative travel distance is recalculated, and the recalculated cumulative travel distance is compared with the valid end distance again to perform the second validity check.

[0073] By introducing a third time threshold and a static segment discarding mechanism, this embodiment can re-evaluate distance validity after removing invalid static data in the initial or final segment where there is prolonged static activity. This process avoids the situation where valid moving data that meets the distance criteria is mistakenly judged as invalid packets due to a continuous static segment occupying a time window, thus improving the retention rate of valid data segments while filtering redundant data.

[0074] Example 2

[0075] A multi-dimensional segmentation system for autonomous driving data collection includes: a configuration acquisition module for acquiring segmentation configuration information, including a start maximum time, an end maximum time, a start effective distance, an end effective distance, and a segment time interval; a data extraction module for reading acquisition data packets, extracting multi-sensor data from the acquisition data packets, and acquiring vehicle speed information; a distance calculation module for calculating the cumulative driving distance corresponding to the acquisition data packets based on the timestamps of the multi-sensor data and the vehicle speed information; and a validity verification module for performing a first validity verification on the starting segment of the acquisition data packets using the start maximum time and the start effective distance, and using... The maximum end time and the effective end distance are used to perform a second validity check on the end segment of the acquired data packet; the invalidation processing module is used to determine the acquired data packet as invalid data packet and abandon the segmentation when the first validity check or the second validity check fails; the interval determination module is used to determine a first effective time point and a second effective time point when both the first validity check and the second effective time point pass, and determine the time interval between the first effective time point and the second effective time point as the effective time interval; the segmentation module is used to segment the multi-sensor data according to the segment time interval within the effective time interval to obtain one or more data segments.

[0076] Example 3

[0077] According to embodiments of the present invention, the present invention also provides an electronic device, a readable storage medium, and a computer program product.

[0078] Figure 2 A schematic block diagram of an example electronic device 100 that can be used to implement embodiments of the present invention is shown. Electronic device 100 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 100 may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their links and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0079] like Figure 2As shown, the electronic device 100 includes a computing unit 101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 102 or a computer program loaded from a storage unit 108 into a random access memory (RAM) 103. The RAM 103 may also store various programs and data required for the operation of the electronic device 100. The computing unit 101, ROM 102, and RAM 103 are interconnected via a bus 104. An input / output (I / O) interface 105 is also linked to the bus 104.

[0080] Multiple components in electronic device 100 are linked to I / O interface 105, including: input unit 106, such as keyboard, mouse, etc.; output unit 107, such as various types of displays, speakers, etc.; storage unit 108, such as disk, optical disk, etc.; and communication unit 109, such as network card, modem, wireless transceiver, etc. Communication unit 109 allows electronic device 100 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0081] The computing unit 101 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 101 performs the various methods and processes described above, such as a multi-dimensional segmentation method for autonomous driving data acquisition. For example, in some embodiments, a multi-dimensional segmentation method for autonomous driving data acquisition can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 108. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 100 via ROM 102 and / or communication unit 109. When the computer program is loaded into RAM 103 and executed by the computing unit 101, one or more steps of the multi-dimensional segmentation method for autonomous driving data acquisition described above can be performed. Alternatively, in other embodiments, the computing unit 101 may be configured, by any other suitable means (e.g., by means of firmware), to perform a multi-dimensional segmentation method for autonomous driving data acquisition.

[0082] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0083] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0084] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical links based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0085] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0086] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0087] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0088] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A multi-dimensional segmentation method for autonomous driving data collection, characterized in that, Includes the following steps: Obtain segmentation configuration information, which includes start maximum time, end maximum time, start effective distance, end effective distance, and segment time interval; Read the data acquisition data packet, extract multi-sensor data from the data acquisition data packet, and obtain vehicle speed information; Based on the timestamps of the multi-sensor data and the vehicle speed information, the cumulative driving distance corresponding to the collected data packet is calculated. The starting segment of the collected data packet is subjected to a first validity check based on the maximum start time and the effective start distance, and the ending segment of the collected data packet is subjected to a second validity check based on the maximum end time and the effective end distance. If either the first validity check or the second validity check fails, the collected data packet is determined to be an invalid data packet, and the segmentation is abandoned. When both the first validity check and the second validity check pass, the first valid time point and the second valid time point are determined, and the time interval between the first valid time point and the second valid time point is taken as the valid time interval. Within the effective time interval, the multi-sensor data is segmented according to the segment time interval to obtain one or more data segments.

2. The method for multi-dimensional segmentation of autonomous driving data collection according to claim 1, characterized in that, The multi-sensor data includes point cloud data; the vehicle speed information is obtained by parsing the CAN bus data in the data acquisition packet, and the CAN bus data contains the correspondence between timestamps and vehicle speeds.

3. The method for multi-dimensional segmentation of autonomous driving data collection according to claim 2, characterized in that, The calculation of the cumulative travel distance corresponding to the collected data packet based on the timestamps of the multi-sensor data and the vehicle speed information includes: Align the timestamp of the point cloud data with the timestamp of the CAN bus data, and assign a corresponding vehicle speed to each frame of point cloud data. Calculate the time difference between two adjacent frames of point cloud data, and use the product of the time difference and the corresponding vehicle speed as the inter-frame travel distance; The cumulative travel distance is obtained by summing up the travel distances between all the frames in chronological order.

4. The method for multi-dimensional segmentation of autonomous driving data collection according to claim 1, characterized in that, Determining the first valid time point and the second valid time point includes: Starting from the start time of the collected data packet, the moment when the cumulative driving distance first reaches the starting effective distance is found in chronological order and is taken as the first effective time point; Starting from the end time of the data packet collection, search backwards along the time axis, and take the moment when the remaining driving distance first reaches the end effective distance as the second effective time point; The remaining driving distance is the driving distance from the current time to the end time.

5. The method for multi-dimensional segmentation of autonomous driving data collection according to claim 1, characterized in that, Within the effective time interval, the multi-sensor data is segmented according to the segment time interval to obtain one or more data segments, including: Obtain the duration of the effective time interval, and determine the number of segments to be segmented and the time margin based on the duration and the segment time interval; For each segment to be segmented, the segment start time and segment end time are calculated based on the time margin, the segment time interval, and the segment number. Data located between the start time and end time of the segment is extracted from the multi-sensor data to form a data segment.

6. The method for multi-dimensional segmentation of autonomous driving data collection according to claim 5, characterized in that, The calculation of the segment's start time and end time includes: The start time of the segment is shifted forward by a first extended duration, where the first extended duration is the length of time from the start time of the acquired data packet to the first valid time point; The end time of the segment is shifted backward by a second extended duration, which is the length of time from the second valid time point to the end time of the acquired data packet.

7. The method for multi-dimensional segmentation of autonomous driving data collection according to claim 1, characterized in that, If the vehicle remains stationary for more than a preset third time threshold within the maximum start time or the maximum end time, the data corresponding to that period of stationary motion is discarded, and the cumulative driving distance is recalculated based on the remaining data. The first validity check or the second validity check is then performed again.

8. A multi-dimensional segmentation system for autonomous driving data collection, characterized in that, include: The configuration acquisition module is used to acquire segmentation configuration information, which includes start maximum time, end maximum time, start effective distance, end effective distance, and segment time interval. The data extraction module is used to read the data acquisition data packet, extract multi-sensor data from the data acquisition data packet, and obtain vehicle driving speed information; The distance calculation module is used to calculate the cumulative driving distance corresponding to the collected data packet based on the timestamp of the multi-sensor data and the vehicle driving speed information; The validity verification module is used to perform a first validity verification on the starting segment of the collected data packet using the maximum start time and the effective start distance, and to perform a second validity verification on the ending segment of the collected data packet using the maximum end time and the effective end distance. An invalid processing module is used to determine the collected data packet as an invalid data packet and abandon the segmentation when the first validity check or the second validity check fails. The interval determination module is used to determine a first valid time point and a second valid time point when both the first validity check and the second validity check pass, and to determine the time interval between the first valid time point and the second valid time point as the valid time interval. The segmentation module is used to segment the multi-sensor data according to the segment time interval within the effective time interval to obtain one or more data segments.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the multi-dimensional segmentation method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that enables a processor to execute the multi-dimensional segmentation method according to any one of claims 1-7.

Images