An intelligent processing method and system for road and air integrated test data

Abstract

The application discloses a kind of road-air integrated test data intelligent processing method and system, the method includes: obtaining road-air collaborative test data;Based on the road-air collaborative test data, establish stable alignment fragment;Extract the collaborative offset feature of each synchronous test fragment;Calculate the collaborative offset of corresponding synchronous test fragment;Based on test timing, continuously update the collaborative offset of several synchronous test fragments, form test quality trajectory chain;Screen the scene reference fragment group corresponding to the to-be-checked fragment group;According to scene reference fragment group, construct adaptive cleaning threshold;The application determines the data processing level according to the latter synchronous test fragment;The application can make the test data with good quality be directly used for test evaluation, enable the data with slight offset to enter repair process, enable the data with serious offset to be excluded from test conclusion, improve road-air integrated test data test accuracy.

Description

Technical Field

[0001] This invention relates to the field of air-ground data processing, specifically to an intelligent method and system for processing integrated air-ground test data. Background Technology

[0002] With the development of intelligent connected vehicle testing, autonomous driving road testing, and vehicle-to-infrastructure (V2I) testing, the data sources during the testing process are no longer limited to single onboard devices. In actual testing, test terminals such as radar, cameras, and roadside communication units are typically deployed on the roadside, while the test vehicles are equipped with positioning modules, inertial measurement modules, onboard cameras, and onboard communication modules. Simultaneously, drones or other aerial observation equipment are used to acquire overhead images and target trajectories of the test area. The resulting road-air cooperative test data can record the test vehicle, surrounding traffic participants, and road environment conditions from roadside, vehicle-side, and aerial perspectives.

[0003] However, in actual testing, roadside data, vehicle-mounted data, and aerial observation data often suffer from problems such as different sampling frequencies, inconsistent time signatures, inconsistent coordinate references, and unstable target matching results. For example, when the same test vehicle passes through a certain detection area, the target position recorded by the roadside camera, the vehicle position recorded by the vehicle-mounted positioning module, and the vehicle position projected by the aerial observation equipment may deviate from each other; the recording time of the same lane change, following, or braking event may also differ across different data sources.

[0004] Existing processing methods typically use fixed thresholds to remove anomalies from single-source data, or simply perform time alignment on multi-source data and then directly fuse them. When the test scenario changes, such as changes in road type, vehicle speed range, lighting conditions, or aerial observation altitude, fixed thresholds can easily lead to the accidental deletion of normal data or the retention of data with large offsets, thus affecting the test conclusions of the integrated road-air test data. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent processing method and system for integrated road-air test data. By constructing a collaborative offset value that characterizes the consistency of roadside, vehicle-mounted, and aerial data, and by generating an adaptive cleaning threshold using historical stable segments from similar test scenarios, the technical problems mentioned in the background art are solved.

[0006] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention discloses an intelligent processing method and system for integrated air-ground test data, comprising the following steps: S1. Acquire road-air cooperative test data uploaded by roadside test terminals, vehicle-mounted acquisition terminals and aerial observation terminals within the test area; S2. Based on the aforementioned road-air coordination test data, establish a unified test spatiotemporal benchmark and determine stable alignment segments of the road-air data; S3. Using the stable alignment segment as the starting reference, the road-air coordination test data under the unified test spatiotemporal benchmark is segmented, and the coordination offset features of each synchronous test segment are extracted. S4. Based on the cooperative offset characteristics of each synchronous test segment, calculate the cooperative offset of the corresponding synchronous test segment; S5. Based on the test timing sequence, continuously update the collaborative offset of several synchronous test segments to form a test quality trajectory chain; S6. Determine the segment group to be checked in the test quality trajectory chain, and select the scene reference segment group corresponding to the segment group to be checked from the historical test segments; S7. Based on the group of segments to be checked and the scene reference group, construct an adaptive cleaning threshold; S8. Based on the adaptive cleaning threshold, determine the data processing level of the subsequent synchronous test segment.

[0007] S2-1. Obtain the original time and original spatial identifiers corresponding to the roadside test terminal, vehicle-mounted data acquisition terminal, and aerial observation terminal, respectively. S2-2. Map the original time identifier to a unified test time axis, and map the original spatial identifier to a unified test coordinate system; S2-3. On a unified test timeline, match the position or state of the same test object in different data sources to obtain the multi-source matching deviation. S2-4. If the multi-source matching deviation within R consecutive sampling times is less than the preset alignment tolerance, then the data segment corresponding to the R consecutive sampling times is determined as a stable alignment segment.

[0008] S3-1. According to the preset segment duration, the road-air coordination test data under the unified test spatiotemporal benchmark is divided into several synchronous test segments; S3-2. Extract at least one cooperative offset feature within each synchronous test segment; The cooperative offset features include at least one of the following: time offset features, spatial projection residual features, target matching difference features, motion state consistency features, and communication delay fluctuation features.

[0009] S4-1. Normalize each cooperative offset feature; S4-2. Based on the data missing rate and sampling jitter rate within the synchronous test segment, the basic confidence weights of each collaborative offset feature are corrected to obtain the segment adaptive weights. S4-3. Based on the adaptive weight of the segment, the normalization results of each collaborative offset feature are weighted and fused to obtain the collaborative offset of the synchronous test segment.

[0010] S5-1. Using the stable alignment segment as the starting reference for the test quality trajectory chain, record the cooperative offset of each synchronous test segment in the order of the unified test time axis. S5-2. When the number of recorded collaborative offsets reaches a preset number N, a test quality trajectory chain of length N is formed. S5-3. When a new synchronous test segment arrives, the new collaborative offset is written into the test quality trajectory chain, and the oldest collaborative offset is removed, so that the test quality trajectory chain keeps rolling and updating.

[0011] S6-1. In the test quality trajectory chain, with the latest synchronous test segment as the endpoint, select M consecutive synchronous test segments forward to form a group of segments to be checked. S6-2. Obtain the test scenario information corresponding to the segment group to be verified; S6-3. Based on the test scenario information, select candidate historical segments from the historical test segment library; S6-4. Based on the data stability of the candidate historical segments, determine the scene reference segment group from the candidate historical segments.

[0012] S7-1. Calculate the distribution drift between the group of segments to be checked and the scene reference group; S7-2. Compare the distributed drift amount with the preset drift allowance value; S7-3. If the distribution drift is less than or equal to the preset drift allowance, then a basic cleaning threshold is constructed based on the scene reference fragment group. S7-4. If the distribution drift is greater than the preset drift allowable value, a drift compensation term is introduced on the basis of the basic cleaning threshold to construct a compensated cleaning threshold. S7-5. Determine the basic cleaning threshold or the compensation cleaning threshold as the adaptive cleaning threshold.

[0013] S8-1. Obtain the collaborative offset corresponding to the subsequent synchronous test segment; S8-2, The cooperative offset of the post-synchronization test segment will be compared with the adaptive cleaning threshold; S8-3. If the cooperative offset of the post-synchronization test segment is less than or equal to the adaptive cleaning threshold, then the post-synchronization test segment is marked as reliable test data. S8-4. If the collaborative offset of the post-synchronization test segment is greater than the adaptive cleaning threshold and less than or equal to the repair upper limit, then the post-synchronization test segment is marked as test data to be repaired. S8-5. If the collaborative offset of the post-synchronization test segment is greater than the repair upper limit, then mark the post-synchronization test segment as discarded test data.

[0014] This invention provides an intelligent processing method and system for integrated air-ground test data, which has the following beneficial effects: This invention maps test data uploaded by roadside test terminals, vehicle-mounted acquisition terminals, and aerial observation terminals to a unified test spatiotemporal reference, making multi-source data with differences in sampling frequency, time stamp, and coordinate reference have a comparable basis, thereby reducing target mismatch caused by time misalignment or spatial coordinate inconsistency. Based on this, by extracting the collaborative offset features from the synchronous test segments and constructing collaborative offset values, the time deviation, spatial deviation, target matching difference, and motion state difference between roadside data, vehicle data, and aerial observation data can be uniformly quantified into data consistency evaluation results, avoiding the one-sidedness caused by relying solely on data from a single road for anomaly judgment. Furthermore, by continuously updating the collaborative offset value to form a test quality trajectory chain, the continuous change status of multi-source data quality during the test can be reflected, so that the data processing process is no longer limited to isolated judgment at a single moment. At the same time, by selecting a scene reference segment group that matches the current segment group to be checked from historical test segments, and generating an adaptive cleaning threshold based on the scene reference segment group, the cleaning standard can be adjusted according to changes in test conditions such as road type, test action, vehicle speed range, lighting conditions and aerial observation altitude, avoiding the mistaken deletion of normal data or retention of abnormal data due to fixed thresholds in different test scenarios. Finally, based on the adaptive cleaning threshold, the subsequent synchronous test segments are classified as reliable, need to be repaired, and to be removed. This allows high-quality test data to be directly used for test evaluation, data with slight deviations to enter the repair process, and data with severe deviations to be excluded from the test conclusions, thereby improving the accuracy of integrated air-ground test data.

[0015] Secondly, this invention discloses an intelligent processing method system for integrated air-ground test data, the system comprising: The road-air data acquisition module is used to acquire road-air collaborative test data uploaded by roadside test terminals, vehicle-mounted acquisition terminals and aerial observation terminals within the test area; The road-air data alignment module is used to establish a unified test spatiotemporal benchmark based on the road-air collaborative test data, and to determine stable alignment segments of the road-air data; The data segmentation module is used to segment the road-air coordination test data under the unified test spatiotemporal benchmark, with the stable alignment segment as the starting reference, and extract the coordination offset features of each synchronous test segment. The collaborative offset calculation module is used to calculate the collaborative offset of the corresponding synchronous test segment based on the collaborative offset characteristics of each synchronous test segment. The trajectory chain construction module is used to continuously update the collaborative offset of several synchronous test segments based on the test timing sequence to form a test quality trajectory chain. The segment filtering module is used to determine the segment group to be checked in the test quality trajectory chain, and to filter the scene reference segment group corresponding to the segment group to be checked from the historical test segments; The cleaning threshold construction module is used to construct an adaptive cleaning threshold based on the group of segments to be checked and the scene reference segment group; The segment level determination module is used to determine the data processing level of subsequent synchronous test segments based on the adaptive cleaning threshold.

[0016] Compared with the prior art, the beneficial effects of the intelligent processing method system for integrated road and air test data of the present invention are the same as those of the intelligent processing method for integrated road and air test data described above, so they will not be repeated here. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the processing flow of an intelligent method for processing integrated road and air test data according to the present invention. Figure 2 This is a schematic diagram illustrating the process of determining the stable aligned segment according to the present invention; Figure 3 This is a schematic diagram illustrating the process of determining the scene reference segment group described in this invention; Figure 4 This is a schematic diagram of the process for determining the adaptive cleaning threshold according to the present invention; Figure 5 This is a structural block diagram of an intelligent processing method for integrated air-ground test data according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0019] Example 1: Please refer to Figure 1 This invention provides an intelligent processing method for integrated air-ground test data, comprising the following steps: S1. Acquire road-air cooperative test data uploaded by roadside test terminals, vehicle-mounted acquisition terminals and aerial observation terminals within the test area.

[0020] Among them, the roadside test terminal is used to collect the status of roadside targets, traffic environment status and roadside communication status; the vehicle-mounted acquisition terminal is used to collect the operating status of the test vehicle, vehicle-end perception status and vehicle control status; the aerial observation terminal is used to collect the top view image of the test area, the overall trajectory of the target and the distribution status of traffic participants from an aerial perspective.

[0021] In other words, the test data processed in this embodiment consists of roadside data, vehicle-mounted data, and aerial observation data. Because these data differ in sampling frequency, time signature, spatial reference, and observation angle, direct fusion analysis can easily lead to data misalignment, target mismatch, or distorted test conclusions. Therefore, it is necessary to first perform unified processing on the road-air cooperative test data.

[0022] S2. Based on the aforementioned road-air coordination test data, establish a unified test spatiotemporal benchmark and determine stable alignment segments of the road-air data.

[0023] Specifically, data uploaded from different test terminals are mapped to a unified test timeline and a unified test coordinate system, so that roadside data, vehicle-mounted data, and aerial observation data have a common basis for comparison.

[0024] After completing the spatiotemporal mapping, consistency checks are performed on the data corresponding to the same test object from different data sources. When the target position, target state, or target motion trajectory corresponding to different data sources all meet the preset consistency conditions within multiple consecutive sampling times, the corresponding data segment is determined as a stable aligned segment.

[0025] S3. Using the stable alignment segment as the starting reference, the road-air cooperative test data under the unified test spatiotemporal benchmark is segmented, and the cooperative offset features of each synchronous test segment are extracted.

[0026] Specifically, the road-air cooperative test data is divided into several synchronous test segments according to the preset segment length. Each synchronous test segment corresponds to a continuous time interval on a unified test time axis and includes roadside data, vehicle-mounted data, and aerial observation data within that time interval.

[0027] The cooperative offset feature is used to characterize the degree of inconsistency between different data sources in the same synchronous test segment. The cooperative offset feature may include at least one of the following: time offset feature, spatial projection residual feature, target matching difference feature, motion state consistency feature, and communication delay fluctuation feature.

[0028] S4. Based on the cooperative offset characteristics of each synchronous test segment, calculate the cooperative offset of the corresponding synchronous test segment.

[0029] Specifically, various cooperative offset features are scaled uniformly, and their corresponding weights are determined based on the credibility of different cooperative offset features in the current synchronous test segment, and then fused to obtain the cooperative offset of the synchronous test segment.

[0030] The cooperative offset is used to characterize the overall degree of offset among roadside data, vehicle-mounted data, and aerial observation data within the same synchronous test segment. A larger cooperative offset indicates poorer consistency among the multi-source test data in the synchronous test segment; a smaller cooperative offset indicates better consistency among the multi-source test data in the synchronous test segment.

[0031] S5. Based on the test timing, continuously update the collaborative offset of several synchronous test segments to form a test quality trajectory chain.

[0032] Specifically, the collaborative offsets corresponding to each synchronous test segment are recorded sequentially according to the unified test timeline, and a test quality trajectory chain is formed by multiple consecutive collaborative offsets.

[0033] The test quality trajectory chain is used to reflect the continuous changes in the quality of air-road cooperative test data during the testing process. When a new synchronous test segment arrives, the new cooperative offset is added to the test quality trajectory chain, and the oldest written cooperative offset is removed according to a preset length, so that the test quality trajectory chain is kept in a rolling update state.

[0034] S6. Determine the segment group to be checked in the test quality trajectory chain, and select the scene reference segment group corresponding to the segment group to be checked from the historical test segments.

[0035] Specifically, several recent consecutive synchronous test segments are selected from the test quality trajectory chain to form a segment group to be checked. The segment group to be checked represents the recent data status that needs to be judged in the current testing phase.

[0036] Simultaneously, based on the test scenario information corresponding to the segment group to be verified, historical segments with the same or similar scenarios are selected from the historical test segment library to form a scenario reference segment group. The test scenario information may include at least one of the following: road type, test action, vehicle speed range, lighting conditions, and aerial observation altitude.

[0037] By introducing a scenario reference fragment group, this embodiment does not simply compare the current test data with a fixed threshold, but compares the current test data with historical stable data under similar test scenarios, thereby reducing misjudgments caused by changes in the test scenario.

[0038] S7. Based on the group of segments to be checked and the scene reference group, construct an adaptive cleaning threshold.

[0039] Specifically, the data quality distribution of the segment group to be checked and the scene reference segment group are compared to determine the degree of deviation of the current test data quality relative to the historical test data of the same type, and an adaptive cleaning threshold is constructed accordingly.

[0040] When the difference between the segment group to be checked and the scene reference segment group is small, a basic cleaning threshold can be constructed based on the data quality stability level of the scene reference segment group; when the difference between the segment group to be checked and the scene reference segment group is large, a compensation term can be introduced on the basis of the basic cleaning threshold to form a compensated cleaning threshold.

[0041] Therefore, the adaptive cleaning threshold can be adjusted according to the test scenario and the current data status, avoiding the data being incorrectly cleaned or missed due to the use of a fixed threshold.

[0042] S8. Based on the adaptive cleaning threshold, determine the data processing level of the subsequent synchronous test segment.

[0043] Specifically, the collaborative offset corresponding to the subsequent synchronous test segment is obtained, and this collaborative offset is compared with the adaptive cleaning threshold.

[0044] If the collaborative offset of the post-synchronization test segment is within the allowable range, the post-synchronization test segment is marked as reliable test data; if the collaborative offset of the post-synchronization test segment exceeds the adaptive cleaning threshold but is still within the repairable range, the post-synchronization test segment is marked as test data to be repaired; if the collaborative offset of the post-synchronization test segment exceeds the repairable range, the post-synchronization test segment is marked as discarded test data.

[0045] Example 2: Refer to 2 to Figure 4 Based on Example 1, this embodiment also discloses the specific implementation methods for each step.

[0046] In this embodiment, step S2 includes: S2-1. Obtain the original time and original spatial identifiers corresponding to the roadside test terminal, vehicle-mounted data acquisition terminal, and aerial observation terminal, respectively.

[0047] Among them, the original time stamp of the roadside test terminal can be the sampling time generated by the roadside edge device, the original time stamp of the vehicle-mounted acquisition terminal can be the sampling time generated by the vehicle-mounted positioning module or the vehicle-mounted controller, and the original time stamp of the aerial observation terminal can be the aerial image frame time or the flight control system time.

[0048] The original spatial identifier may include road coordinates, vehicle local coordinates, image coordinates, geographic coordinates, or intermediate coordinates obtained by transforming the above coordinates.

[0049] S2-2. Map the original time identifier to a unified test time axis and the original spatial identifier to a unified test coordinate system.

[0050] For example, the start time of the test task can be taken as the zero point of the unified test time axis, the direction of the center line of the test road can be taken as the longitudinal coordinate direction, and the lateral direction of the road can be taken as the lateral coordinate direction, thereby forming a unified test coordinate system.

[0051] For image coordinates acquired by aerial observation terminals, the coordinates can be projected into a unified test coordinate system by combining the terminal's flight altitude, attitude parameters, camera intrinsic parameters, and ground calibration parameters. For vehicle-mounted acquisition terminals, the local vehicle coordinates can be transformed into a unified test coordinate system by combining vehicle positioning and heading information.

[0052] S2-3. On a unified test timeline, match the position or state of the same test object in different data sources to obtain multi-source matching deviation.

[0053] Among them, multi-source matching deviations can include the deviation between the roadside target position and the vehicle target position, the deviation between the aerial target projection position and the roadside target position, the deviation between the aerial target projection position and the vehicle target position, and can also include the target speed difference, target heading difference, or target appearance time difference recorded by different data sources.

[0054] S2-4. If the multi-source matching deviation within R consecutive sampling times is less than the preset alignment tolerance, then the data segment corresponding to the R consecutive sampling times is determined as a stable alignment segment.

[0055] Where R is a preset positive integer. The preset alignment tolerance can be set according to the test road type, test speed, positioning accuracy, aerial observation altitude, and sensor calibration accuracy.

[0056] For example, in low-speed closed test scenarios, the preset alignment tolerance can be set relatively small; in high-speed road tests or test scenarios with high aerial observation altitudes, the preset alignment tolerance can be appropriately increased.

[0057] This embodiment determines stable aligned segments by verifying the consistency of multiple consecutive sampling moments, rather than relying solely on a single sampling moment. This avoids misjudging aligned segments due to momentary jitter, brief occlusion, or single-frame recognition errors.

[0058] In this embodiment, step S3 includes: S3-1. According to the preset segment duration, the road-air coordination test data under the unified test spatiotemporal benchmark is divided into several synchronous test segments.

[0059] The preset segment duration can be set according to the type of test task. For example, a longer segment duration can be used in low-speed following tests; a shorter segment duration can be used in emergency braking, lane changing and obstacle avoidance, or complex interaction tests to improve the responsiveness to changes in data quality.

[0060] S3-2. Extract at least one cooperative offset feature within each synchronous test segment.

[0061] The cooperative offset features include, but are not limited to, the following types: ① Time offset characteristic: This characteristic characterizes the time difference between the recording of the same test event by the roadside test terminal, the vehicle-mounted data acquisition terminal, and the aerial observation terminal. The test event can be, for example, the test vehicle passing through a preset detection line, the target vehicle entering the test area, the vehicle completing a lane change maneuver, or a change in traffic signal status.

[0062] ② Spatial projection residual characteristics: These characteristics characterize the positional differences between aerial observation data projected onto a unified test coordinate system and roadside or vehicle-mounted data. This feature reflects the spatial errors generated during the projection, calibration, or attitude compensation processes of the aerial observation data.

[0063] ③ Target matching difference feature: This feature characterizes whether the identification results of the same test object from different test data sources are consistent. This feature can be determined based on at least one of the following: difference in the number of targets, target identity matching failure rate, and target trajectory breakage.

[0064] ④ Motion state consistency characteristic: This characterizes whether the records of target velocity, acceleration, heading angle, or trajectory change trends from different test data sources are consistent. If the differences in motion state recorded by different data sources are small, it indicates that the data consistency of this synchronous test segment is high.

[0065] ⑤ Communication delay fluctuation characteristics: Communication delay fluctuation characteristics are used to characterize the delay changes of roadside communication links, vehicle-mounted communication links, or air communication links within a synchronous test segment. When communication delays fluctuate significantly, time mismatches are more likely to occur between multi-source data.

[0066] It should be noted that the aforementioned cooperative offset features can be used individually or in combination. In practical applications, the appropriate cooperative offset feature can be selected based on the test task type, sensor configuration, and data integrity.

[0067] In this embodiment, step S4 includes: S4-1. Normalize each cooperative offset feature.

[0068] For example, for the i-th type of collaborative offset feature, its normalized expression is: ; in, Let represent the normalized result of the class i cooperative offset feature in the t-th synchronous test segment. This represents the actual offset of the i-th type of cooperative offset feature in the t-th synchronous test segment. This represents the maximum allowable offset corresponding to the i-th type of cooperative offset feature.

[0069] S4-2. Based on the data missing rate and sampling jitter rate within the synchronous test segment, the basic confidence weights of each collaborative offset feature are corrected to obtain the segment adaptive weights.

[0070] For example, the segment adaptive weight can be expressed as: ; in, This represents the segment adaptive weight of the i-th class of cooperative offset features in the t-th synchronous test segment. The underlying confidence weights represent the i-th type of collaborative offset features. This represents the missing rate of the i-th type of data in the t-th synchronous test segment. This represents the sampling jitter rate of the i-th type of data in the t-th synchronous test segment.

[0071] S4-3. Based on the adaptive weight of the segment, the normalization results of each collaborative offset feature are weighted and fused to obtain the collaborative offset of the synchronous test segment.

[0072] Specifically, the normalized result of each type of collaborative offset feature is multiplied by its corresponding segment adaptive weight, and then the products are summed to obtain the collaborative offset of the synchronous test segment. The collaborative offset is used to centrally reflect the overall degree of offset between multi-source test data within the same synchronous test segment. A larger collaborative offset indicates poorer consistency among roadside data, vehicle-mounted data, and aerial observation data in the synchronous test segment; a smaller collaborative offset indicates more stable data quality in the synchronous test segment. Thus, each synchronous test segment can generate a unified data quality evaluation result, which serves as the foundational data for constructing the test quality trajectory chain.

[0073] In this embodiment, step S5 includes: S5-1. Using the stable alignment segment as the starting reference for the test quality trajectory chain, record the cooperative offset of each synchronous test segment in the order of the unified test time axis.

[0074] S5-2. When the number of recorded collaborative offsets reaches a preset number N, a test quality trajectory chain of length N is formed.

[0075] For example, a test quality trajectory chain can be represented as a set of co-offsets arranged chronologically according to the test time. Each value in this set of co-offsets corresponds to a synchronous test segment, and the order of the values ​​is consistent with the order in which the synchronous test segments were generated. Therefore, the test quality trajectory chain can reflect the continuous changes in data quality over a recent testing period.

[0076] S5-3. When a new synchronous test segment arrives, the new collaborative offset is written into the test quality trajectory chain, and the oldest collaborative offset is removed, so that the test quality trajectory chain keeps rolling and updating.

[0077] This embodiment uses a rolling update of the test quality trajectory chain, allowing the system to focus only on recent changes in test data quality without repeatedly accessing all historical data. This reduces computational load while maintaining continuous tracking of the current test status.

[0078] In this embodiment, step S6 includes: S6-1. In the test quality trajectory chain, with the latest synchronous test segment as the endpoint, select M consecutive synchronous test segments forward to form a group of segments to be checked.

[0079] Specifically, the set of segments to be checked consists of multiple adjacent synchronous test segments, used to describe the recent data quality status in the current testing phase. Compared to judging based on only a single synchronous test segment, the set of segments to be checked can mitigate the impact of fluctuations in a single sampling, making the data quality judgment results more stable.

[0080] S6-2. Obtain the test scenario information corresponding to the segment group to be verified.

[0081] The test scenario information may include at least one of the following: road type, test action, vehicle speed range, traffic density, lighting conditions, weather conditions, aerial observation altitude, and aerial observation angle.

[0082] For example, the test scenario corresponding to a certain group of segments to be verified can be represented as: a straight urban road section, following vehicle test, vehicle speed of 30km / h to 40km / h, daytime, and aerial observation altitude of 80m.

[0083] S6-3. Based on the test scenario information, select candidate historical segments from the historical test segment library.

[0084] When selecting candidate historical clips, priority can be given to historical clips that are the same or similar to the clip group to be verified in terms of road type, test action and speed range; if the number of candidate historical clips is insufficient, conditions such as lighting conditions, aerial observation altitude or traffic density can be appropriately relaxed.

[0085] S6-4. Based on the data stability of the candidate historical segments, determine the scene reference segment group from the candidate historical segments.

[0086] Specifically, candidate historical segments with low collaborative offset, low data missing rate, and small fluctuation can be selected as the scenario reference segment group. The scenario reference segment group is used to represent a relatively stable level of data quality under similar test scenarios.

[0087] The segments to be verified are not compared with arbitrary historical data, but rather with stable historical data under similar test scenarios. This reduces judgment bias caused by differences in road conditions, test actions, or aerial observation conditions.

[0088] In this embodiment, step S7 includes: S7-1. Calculate the distribution drift between the segment group to be checked and the scene reference segment group.

[0089] The distribution drift is used to represent the degree of deviation of the current test data quality distribution from the historical stable data quality distribution of the same type.

[0090] For example, the mean of the cooperative offset, the degree of dispersion, and the degree of distribution overlap of the segment group to be checked and the scene reference segment group can be calculated respectively, and the distribution drift can be constructed based on the above parameters: ; Where B represents the distribution drift, This represents the mean of the collaborative offset of the segment group to be checked. This represents the mean of the collaborative offset of the scene reference fragment group. This indicates the degree of co-offset dispersion of the segment group to be checked. H represents the degree of collaborative offset dispersion of the scene reference fragment group, H represents the degree of distribution overlap between the fragment group to be checked and the scene reference fragment group, and β1, β2, and β3 are preset adjustment coefficients.

[0091] S7-2. Compare the distribution drift amount with the preset drift allowance value.

[0092] If the distribution drift is less than or equal to the preset drift allowance, it indicates that although the quality of the current test data exhibits normal fluctuations, it is still within the acceptable range for similar test scenarios. In this case, a basic cleaning threshold is constructed based on the scenario reference segment group.

[0093] For example, the basic cleaning threshold can be expressed as: ; in, R represents the basic cleaning threshold, and R represents the set of cooperative offsets in the scene reference fragment group. This represents the median of R. R represents the median absolute deviation, and c represents the preset threshold coefficient.

[0094] The median absolute deviation can be expressed as: ; By constructing a basic cleaning threshold using the median and median absolute deviation, the impact of a small number of abnormal historical fragments on the threshold can be reduced.

[0095] S7-3. If the distribution drift is greater than the preset drift allowable value, a drift compensation term is introduced on the basis of the basic cleaning threshold to construct a compensated cleaning threshold.

[0096] For example, the compensation cleaning threshold can be expressed as: ; Where G represents the compensation cleaning threshold, η represents the drift compensation coefficient, and B represents the distribution drift amount.

[0097] S7-4. Determine the basic cleaning threshold or the compensation cleaning threshold as the adaptive cleaning threshold.

[0098] In other words, when the current test data differs little from similar historical data, a basic cleaning threshold is used; when the current test data differs significantly from similar historical data, a compensatory cleaning threshold is used. This adaptive cleaning threshold can be adjusted according to changes in the test scenario and the current data quality status.

[0099] In this embodiment, step S8 includes: S8-1. Obtain the collaborative offset corresponding to the subsequent synchronous test segment.

[0100] S8-2, compare the cooperative offset of the post-synchronization test segment with the adaptive cleaning threshold.

[0101] S8-3. If the cooperative offset of the post-synchronization test segment is less than or equal to the adaptive cleaning threshold, then the post-synchronization test segment is marked as reliable test data.

[0102] Reliable test data can be directly incorporated into test evaluation, trajectory analysis, autonomous driving behavior verification, or vehicle-road cooperative performance evaluation processes.

[0103] S8-4. If the collaborative offset of the post-synchronization test segment is greater than the adaptive cleaning threshold and less than or equal to the repair upper limit, then the post-synchronization test segment is marked as test data to be repaired.

[0104] The repair upper limit can be determined based on the adaptive cleaning threshold and the preset repair margin, or it can be preset based on the data repair capability of the test system.

[0105] The test data to be repaired can enter the compensation processing flow. The compensation processing flow may include at least one of the following: time resampling, spatial reprojection, trajectory interpolation, missing data completion, and communication delay correction.

[0106] S8-5. If the collaborative offset of the post-synchronization test segment is greater than the repair upper limit, then mark the post-synchronization test segment as discarded test data.

[0107] Test data that is excluded from test conclusion calculations or archived only as anomalies can be removed. By removing severely skewed data segments, the interference of low-quality data with test evaluation results can be avoided.

[0108] Furthermore, after determining the data processing level, trusted test data, test data to be repaired, and test data to be removed can be written into their respective datasets to facilitate test report generation, anomaly cause analysis, and test task review.

[0109] This embodiment achieves continuous quality evaluation and dynamic cleaning of integrated road-air test data through a processing flow of "cooperative offset—test quality trajectory chain—scenario reference fragment group—adaptive cleaning threshold—data processing level". This method can adapt to the data processing needs under different road conditions, different test actions, and different aerial observation conditions, improving the usability of integrated road-air test data and the reliability of test results.

[0110] This invention also provides an intelligent processing method and system for integrated air-ground test data, which is used to implement the above-described method embodiments. Details already described will not be repeated. The terms "module," "unit," and "subunit," etc., used below refer to combinations of software and / or hardware that achieve a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0111] Figure 5 This is a structural block diagram of a road-air integrated test data intelligent processing method system according to the present invention. The system includes: The road-air data acquisition module is used to acquire road-air collaborative test data uploaded by roadside test terminals, vehicle-mounted acquisition terminals and aerial observation terminals within the test area; The road-air data alignment module is used to establish a unified test spatiotemporal benchmark based on the road-air collaborative test data, and to determine stable alignment segments of the road-air data; The data segmentation module is used to segment the road-air coordination test data under the unified test spatiotemporal benchmark, with the stable alignment segment as the starting reference, and extract the coordination offset features of each synchronous test segment. The collaborative offset calculation module is used to calculate the collaborative offset of the corresponding synchronous test segment based on the collaborative offset characteristics of each synchronous test segment. The trajectory chain construction module is used to continuously update the collaborative offset of several synchronous test segments based on the test timing sequence to form a test quality trajectory chain. The segment filtering module is used to determine the segment group to be checked in the test quality trajectory chain, and to filter the scene reference segment group corresponding to the segment group to be checked from the historical test segments; The cleaning threshold construction module is used to construct an adaptive cleaning threshold based on the group of segments to be checked and the scene reference group; The segment level determination module is used to determine the data processing level of subsequent synchronous test segments based on the adaptive cleaning threshold.

[0112] This embodiment solves the problem of inaccurate test conclusions caused by changes in the test scenario through the intelligent processing method system for integrated road and air test data.

[0113] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any equivalent substitutions, simple modifications, or combinations of the above technical solutions made under the technical concept of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent processing of integrated air-ground test data, characterized in that, include: S1. Acquire road-air cooperative test data uploaded by roadside test terminals, vehicle-mounted acquisition terminals and aerial observation terminals within the test area; S2. Based on the aforementioned road-air coordination test data, establish a unified test spatiotemporal benchmark and determine stable alignment segments of the road-air data; S3. Using the stable alignment segment as the starting reference, the road-air coordination test data under the unified test spatiotemporal benchmark is segmented, and the coordination offset features of each synchronous test segment are extracted. S4. Based on the cooperative offset characteristics of each synchronous test segment, calculate the cooperative offset of the corresponding synchronous test segment; S5. Based on the test timing sequence, continuously update the collaborative offset of several synchronous test segments to form a test quality trajectory chain; S6. Determine the segment group to be checked in the test quality trajectory chain, and select the scene reference segment group corresponding to the segment group to be checked from the historical test segments; S7. Based on the group of segments to be checked and the scene reference group, construct an adaptive cleaning threshold; S8. Based on the adaptive cleaning threshold, determine the data processing level of the subsequent synchronous test segment.

2. The intelligent processing method for integrated air-ground test data according to claim 1, characterized in that, The establishment of a unified test spatiotemporal benchmark and the determination of stable alignment segments for road-space data include: S2-1. Obtain the original time and original spatial identifiers corresponding to the roadside test terminal, vehicle-mounted data acquisition terminal, and aerial observation terminal, respectively. S2-2. Map the original time identifier to a unified test time axis, and map the original spatial identifier to a unified test coordinate system; S2-3. On a unified test timeline, match the position or state of the same test object in different data sources to obtain the multi-source matching deviation. S2-4. If the multi-source matching deviation within R consecutive sampling times is less than the preset alignment tolerance, then the data segment corresponding to the R consecutive sampling times is determined as a stable alignment segment.

3. The intelligent processing method for integrated road-air test data according to claim 1, characterized in that, The extraction of cooperative offset features from each synchronous test segment includes: S3-1. According to the preset segment duration, the road-air coordination test data under the unified test spatiotemporal benchmark is divided into several synchronous test segments; S3-2. Extract at least one cooperative offset feature within each synchronous test segment; The cooperative offset features include at least one of the following: time offset features, spatial projection residual features, target matching difference features, motion state consistency features, and communication delay fluctuation features.

4. The intelligent processing method for integrated road-air test data according to claim 3, characterized in that, The calculation of the cooperative offset corresponding to the synchronous test segment includes: S4-1. Normalize each cooperative offset feature; S4-2. Based on the data missing rate and sampling jitter rate within the synchronous test segment, the basic confidence weights of each collaborative offset feature are corrected to obtain the segment adaptive weights. S4-3. Based on the adaptive weight of the segment, the normalization results of each collaborative offset feature are weighted and fused to obtain the collaborative offset of the synchronous test segment.

5. The intelligent processing method for integrated air-ground test data according to claim 1, characterized in that, The formation of the test quality trajectory chain includes: S5-1. Using the stable alignment segment as the starting reference for the test quality trajectory chain, record the cooperative offset of each synchronous test segment in the order of the unified test time axis. S5-2. When the number of recorded collaborative offsets reaches a preset number N, a test quality trajectory chain of length N is formed. S5-3. When a new synchronous test segment arrives, the new collaborative offset is written into the test quality trajectory chain, and the oldest collaborative offset is removed, so that the test quality trajectory chain keeps rolling and updating.

6. The intelligent processing method for integrated road-air test data according to claim 5, characterized in that, The process of determining the group of segments to be verified and selecting the scene reference segment group corresponding to the group of segments to be verified from historical test segments includes: S6-1. In the test quality trajectory chain, with the latest synchronous test segment as the endpoint, select M consecutive synchronous test segments forward to form a group of segments to be checked. S6-2. Obtain the test scenario information corresponding to the segment group to be verified; S6-3. Based on the test scenario information, select candidate historical segments from the historical test segment library; S6-4. Based on the data stability of the candidate historical segments, determine the scene reference segment group from the candidate historical segments.

7. The intelligent processing method for integrated road-air test data according to claim 6, characterized in that, The construction of the adaptive cleaning threshold includes: S7-1. Calculate the distribution drift between the group of segments to be checked and the scene reference group; S7-2. Compare the distributed drift amount with the preset drift allowance value; S7-3. If the distribution drift is less than or equal to the preset drift allowance, then a basic cleaning threshold is constructed based on the scene reference fragment group. S7-4. If the distribution drift is greater than the preset drift allowable value, a drift compensation term is introduced on the basis of the basic cleaning threshold to construct a compensated cleaning threshold. S7-5. Determine the basic cleaning threshold or the compensation cleaning threshold as the adaptive cleaning threshold.

8. The intelligent processing method for integrated road-air test data according to claim 1, characterized in that, The determination of the data processing level for subsequent synchronous test segments includes: S8-1. Obtain the collaborative offset corresponding to the subsequent synchronous test segment; S8-2, The cooperative offset of the post-synchronization test segment will be compared with the adaptive cleaning threshold; S8-3. If the cooperative offset of the post-synchronization test segment is less than or equal to the adaptive cleaning threshold, then the post-synchronization test segment is marked as reliable test data. S8-4. If the collaborative offset of the post-synchronization test segment is greater than the adaptive cleaning threshold and less than or equal to the repair upper limit, then the post-synchronization test segment is marked as test data to be repaired. S8-5. If the collaborative offset of the post-synchronization test segment is greater than the repair upper limit, then mark the post-synchronization test segment as discarded test data.

9. A method and system for intelligent processing of integrated air-ground test data, characterized in that, include: The road-air data acquisition module is used to acquire road-air collaborative test data uploaded by roadside test terminals, vehicle-mounted acquisition terminals and aerial observation terminals within the test area; The road-air data alignment module is used to establish a unified test spatiotemporal benchmark based on the road-air collaborative test data, and to determine stable alignment segments of the road-air data; The data segmentation module is used to segment the road-air coordination test data under the unified test spatiotemporal benchmark, with the stable alignment segment as the starting reference, and extract the coordination offset features of each synchronous test segment. The collaborative offset calculation module is used to calculate the collaborative offset of the corresponding synchronous test segment based on the collaborative offset characteristics of each synchronous test segment. The trajectory chain construction module is used to continuously update the collaborative offset of several synchronous test segments based on the test timing sequence to form a test quality trajectory chain. The segment filtering module is used to determine the segment group to be checked in the test quality trajectory chain, and to filter the scene reference segment group corresponding to the segment group to be checked from the historical test segments; The cleaning threshold construction module is used to construct an adaptive cleaning threshold based on the group of segments to be checked and the scene reference group; The segment level determination module is used to determine the data processing level of subsequent synchronous test segments based on the adaptive cleaning threshold.

Images