An experimental method and system for roadside perception trajectory data quality inspection and evaluation

Abstract

The application discloses an experimental method and system for roadside perception trajectory data quality inspection and evaluation, which comprises the following steps: S1, checking the composition of the roadside equipment to be evaluated, and determining the evaluation requirements of the trajectory data quality inspection and evaluation and the evaluation system of the trajectory data quality inspection and evaluation; S2, selecting and placing a vehicle-mounted true value unit and a video true value unit, and simultaneously collecting the trajectory perception data of the vehicle-mounted true value unit and the perception data of the video true value unit; S3, comparing the perception trajectory data of the roadside equipment to be evaluated with the data true value, and calculating the accuracy evaluation and continuity evaluation indexes of the trajectory data; S4, evaluating and grading the perception trajectory data quality of the roadside equipment to be evaluated according to the grading standard, generating a data quality report and giving application suggestions. According to the application, the systematized, normalized and comprehensive roadside perception trajectory data quality inspection and evaluation requirements under the intelligent network environment can be met, and the roadside perception data system-level application and the roadside perception facility performance monitoring can be guided.

Description

Technical Field

[0001] This invention relates to the technical field of traffic data quality assessment, and in particular to an experimental method and system for assessing the quality of roadside sensing trajectory data. Background Technology

[0002] With the rapid development of smart highways, numerous edge computing devices and intelligent sensors, such as high-definition cameras, millimeter-wave radar, and lidar, will be deployed along the roadside. Through multi-source data fusion and target recognition and tracking, they can acquire real-time and all-space road traffic perception trajectory data, enabling micro-level driving behavior perception, accurate traffic condition identification, and online parallel simulation. However, the quality of road traffic perception trajectory data directly determines the effectiveness of traffic system analysis, evaluation, and control measures. Therefore, how to evaluate the quality of road traffic perception trajectory data has become a pressing problem to be solved in the context of the rapid development of smart highways and vehicle-road cooperation.

[0003] Previous research on the quality evaluation of roadside sensing trajectory data mainly focused on comparing ground truth data with roadside sensing trajectory data in the same spatiotemporal context using high-precision sensing equipment, and proposed a series of related trajectory data quality evaluation indicators. However, when applying sensing trajectory data in practical applications, it is necessary to consider both the accuracy and continuity of the trajectory data in spatial coordinates. Currently, most studies are limited by factors such as limited testing equipment and imperfect evaluation systems. The trajectory data quality assessment process lacks an evaluation system that includes both accuracy and continuity, making it difficult to obtain reliable quality evaluation indicators and results. Current roadside sensing trajectory data quality assessment methods are insufficient to meet the systematic, routine, and comprehensive requirements for roadside sensing trajectory data quality assessment in an intelligent connected vehicle environment. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide an experimental method and system for evaluating the quality of roadside sensing trajectory data. This method and system can meet the systematic, routine, and comprehensive requirements for evaluating the quality of roadside sensing trajectory data in an intelligent connected environment, guiding the system-level application of roadside sensing data and the performance monitoring of roadside sensing facilities. To achieve the above-mentioned objectives and other advantages of the present invention, an experimental system for evaluating the quality of roadside sensing trajectory data is provided, comprising:

[0005] The roadside sensing unit to be inspected, the selection and deployment of the roadside trajectory sensing system to be inspected, and the system for collecting trajectory data to be inspected;

[0006] The vehicle's true information, including vehicle true trajectory information, is collected by selecting and deploying sensors in the vehicle-mounted true information unit.

[0007] Roadside video data and vehicle information are collected by selecting and deploying video truth cells;

[0008] The data is collected from the roadside sensing unit, vehicle-mounted truth unit, and video truth unit to be evaluated through the data recording module.

[0009] The quality evaluation module calculates the quality evaluation index of the roadside sensing trajectory data to be evaluated, classifies the data quality, generates a trajectory data quality report, and provides application suggestions.

[0010] Preferably, the vehicle-mounted truth unit includes a truth vehicle, an inertial navigation subunit installed on the truth vehicle, an vehicle-mounted perception subunit, and a data acquisition subunit.

[0011] The inertial navigation integrated unit includes inertial navigation elements and GNSS positioning elements, which obtain true information about the vehicle's size, position, speed, and heading angle.

[0012] The vehicle-mounted perception subunit includes vehicle-mounted LiDAR and millimeter-wave radar;

[0013] The data acquisition subunit is used to collect data from the inertial navigation subunit and the vehicle perception subunit, and obtain the true vehicle positioning coordinates, relative positioning coordinates of surrounding vehicles, speed, and heading angle as reference true data to evaluate the quality of roadside perception data.

[0014] Preferably, the video truth unit includes a video sensing device and an offline processing server. The video sensing device collects video data and transmits it to the offline server for post-processing through the acquisition host. The correspondence between each vehicle within the range of the roadside equipment to be evaluated and the roadside sensing trajectory is determined by manual verification. This is used as the benchmark truth data to evaluate the quality of the roadside sensing data.

[0015] Furthermore, during the inspection and evaluation process, the roadside sensing unit to be inspected and the vehicle-mounted truth unit or video truth unit simultaneously collect data and transmit it to the data recording module.

[0016] An experimental method for evaluating the quality of roadside sensing trajectory data includes the following steps:

[0017] S1. Inspect the composition of the roadside equipment to be inspected and determine the inspection requirements and evaluation system for trajectory data quality inspection.

[0018] S2. Select and place the vehicle-mounted truth unit and the video truth unit, and simultaneously collect the trajectory perception data of the vehicle-mounted truth unit and the perception data of the video truth unit.

[0019] S3. Compare the perceived trajectory data of the roadside equipment to be inspected with the true data value, and calculate the accuracy evaluation and continuity evaluation indicators of the trajectory data;

[0020] S4. Evaluate and classify the quality of the perception trajectory data of the roadside to be inspected according to the grading standards, generate a data quality report and provide application suggestions.

[0021] Preferably, the inspection in step S1 includes the hardware configuration of the equipment, detection accuracy, frequency, and placement requirements;

[0022] The evaluation requirements for trajectory data quality assessment include evaluation road requirements and evaluation environment requirements;

[0023] The evaluation dimensions in the evaluation system include the accuracy assessment of trajectory data and the continuity assessment of trajectory data. The specific evaluation indicators are determined based on the actual needs of the scenario.

[0024] Preferably, step S2 specifically includes the following steps:

[0025] S21. Select appropriate ground truth vehicles based on the main vehicle types of the road to be inspected, install inertial navigation sub-units and vehicle-mounted sensing sub-units on the ground truth vehicles, calibrate the inertial navigation sub-units and vehicle-mounted sensing sub-units, and provide unified time synchronization with the roadside sensing facilities to be inspected.

[0026] S22. Within the sensing range of the sensor to be evaluated, the true value vehicle sequentially controls its speed to 20km / h, 40km / h, 60km / h, and 80km / h, respectively, driving in scenarios of low road traffic saturation (≤0.3), medium road traffic saturation (0.3~0.7), and high road traffic saturation (≥0.7). The on-board true value unit and the roadside sensor to be evaluated synchronously and repeatedly collect evaluation data, so that the true value vehicle completes at least one lane change behavior within the facility's sensing range and covers data of driving in different lanes on the road.

[0027] S23. After completing the inspection and evaluation data collection, determine the true value vehicle corresponding to each data collection from the roadside sensing trajectory data to be inspected and evaluate. Perform spatiotemporal alignment between the on-board true value unit and the roadside sensor to be inspected and evaluate. Extract the true value data of the vehicle size and position coordinates at each time based on the timestamp of the roadside trajectory data to be inspected and evaluate, and input the data into the data recording module.

[0028] Preferably, step S3 specifically includes the following steps:

[0029] S31. Select and place video truth units according to the trajectory data quality inspection requirements, so that the video truth units can acquire video data of the road within the scope of the roadside facilities to be inspected, and adjust the data acquisition frequency and accuracy according to the trajectory data quality inspection requirements. The video truth units and the roadside trajectory sensing devices to be inspected collect trajectory sensing data synchronously.

[0030] S32. The roadside equipment to be inspected and the video ground truth unit simultaneously collect inspection and evaluation data in low road traffic saturation (≤0.3), medium road traffic saturation (0.3~0.7), and high road traffic saturation (≥0.7) scenarios, with data collection time lasting 15 minutes in each traffic scenario;

[0031] S33. After completing the inspection and evaluation data collection, manually count the correspondence between the real vehicles within the scope of the roadside equipment to be inspected and the vehicle ID identifiers in the roadside perception trajectory data under each scenario, through the video data obtained by the video truth unit, count the number of tracks that are continuously tracked, added to the inspection, and missed in the roadside perception trajectory data to be inspected, calculate the length of the track that is correctly tracked, and input the data into the data recording module.

[0032] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention selects and places an onboard truth system to collect trajectory perception data from the onboard truth system for evaluating the accuracy of the roadside perception trajectory data to be inspected; it also selects and places a video truth system to collect perception data from the video truth system for evaluating the continuity of the roadside perception trajectory data to be inspected. This comprehensive evaluation of roadside perception trajectory data combines both accuracy and continuity. This invention can meet the systematic, routine, and comprehensive quality inspection requirements of roadside perception trajectory data in an intelligent connected environment, guiding the system-level application of roadside perception data and the performance monitoring of roadside perception facilities. Attached Figure Description

[0033] Figure 1 A flowchart illustrating the experimental method and system for roadside sensing trajectory data quality assessment according to the present invention;

[0034] Figure 2 A schematic diagram of the system structure of the experimental method and system for roadside sensing trajectory data quality assessment according to the present invention;

[0035] Figure 3 This is a schematic diagram illustrating the trajectory data accuracy evaluation of the experimental method and system for roadside sensing trajectory data quality assessment according to the present invention. Detailed Implementation

[0036] 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.

[0037] Reference Figure 1-3 An experimental method for evaluating the quality of roadside sensing trajectory data includes the following steps:

[0038] S1. Inspect the composition of the roadside equipment to be inspected, and determine the inspection requirements and evaluation system for trajectory data quality assessment. This includes checking the equipment's hardware configuration, detection accuracy, frequency, and placement requirements. The inspection requirements for trajectory data quality assessment include requirements for the road and the environment. The evaluation dimensions in the evaluation system include trajectory data accuracy assessment and trajectory data continuity assessment; specific evaluation indicators are determined based on actual scenario requirements. The road to be inspected should meet the following conditions:

[0039] (1) The road should be straight and flat, with good road surface conditions, and the two-way lanes should be at least two lanes;

[0040] (2) The road surface shall be made of asphalt or concrete, flat without obvious bumps, potholes, protrusions and cracks, and the road surface shall meet the relevant requirements of JTG / T F20.

[0041] (3) The lane lines shall meet the relevant requirements of GB 5768.3. The color shall be white or yellow, the line type shall be solid or dashed, and the lane lines shall be clear and complete without damage or obstruction.

[0042] The testing environment should meet the following conditions:

[0043] (1) The ambient temperature for testing is -20℃ to 60℃, and the relative humidity is 25% to 75%.

[0044] (2) The inspection environment is free from severe weather conditions such as snowfall, hail, and dust, and the horizontal visibility should be greater than 500m; the inspection environment is open and unobstructed, meeting the requirements for high-precision vehicle positioning.

[0045] (3) The evaluation environment ensures wireless signal coverage, with no signals that strongly interfere with communication, and meets the requirements for normal wireless communication and data transmission.

[0046] S2. Select and place the vehicle-mounted truth unit and the video truth unit, and simultaneously collect the trajectory perception data of the vehicle-mounted truth unit and the perception data of the video truth unit.

[0047] When selecting and placing the vehicle-mounted truth sheet: Select a suitable truth vehicle based on the main vehicle types of the road to be inspected, install an inertial navigation subunit and a vehicle-mounted sensing subunit on the vehicle body, calibrate the inertial navigation subunit and the vehicle-mounted sensing subunit, and synchronize the time with the roadside sensing facilities to be inspected.

[0048] Within the sensing range of the sensor to be evaluated, the true value vehicle sequentially controls its speed at 20km / h, 40km / h, 60km / h, and 80km / h, respectively, driving under low road traffic saturation (≤0.3), medium road traffic saturation (0.3~0.7), and high road traffic saturation (≥0.7). The on-board true value unit and the roadside sensor to be evaluated synchronously and repeatedly collect evaluation data. The above data collection process is repeated at least 3 times under each speed scenario, so that the true value vehicle completes at least one lane change behavior within the facility's sensing range and covers data of driving in different lanes on the road.

[0049] After the assessment data is collected, the ground truth vehicle roadside trajectory data corresponding to each data collection is determined from the roadside sensing trajectory data to be assessed. Further, the sensing trajectory data of the onboard ground truth unit is spatiotemporally aligned with the roadside sensing trajectory data to be assessed. For spatial alignment, the sensor coordinate system of the onboard ground truth unit and the sensing coordinate system of the roadside trajectory to be assessed are obtained. First, the sensor coordinate system of the onboard ground truth unit is transformed into the world coordinate system based on the vehicle's own parameters. Then, the sensor coordinate system of the onboard ground truth unit is aligned with the sensing coordinate system of the roadside trajectory to be assessed using the absolute position information of the sensors. For temporal alignment, the sensing system time of the onboard ground truth unit is aligned with the roadside data system time through manual parameter calibration, and the data sampling time is synchronized to ensure the temporal comparability of vehicle-road data. After completing the spatiotemporal alignment, ground truth data of the vehicle size, position coordinates, relative position coordinates, speed, and heading angle of the ground truth vehicle at each time point are extracted based on the timestamps of the roadside trajectory data to be assessed, and input into the data recording module.

[0050] When selecting and placing video truth units: synchronously collect the perceived data of the video truth system according to the data truth requirements, and simultaneously process the data. The extracted vehicle data and video data are input into the data recording module to participate in the continuous evaluation of the roadside perception trajectory data to be inspected. Specifically, this includes: selecting a suitable video truth system based on the scenario conditions and setting it near the roadside facility to be inspected, ensuring that the video truth system can acquire video data of the road within the scope of the roadside facility to be inspected, and adjusting the data acquisition frequency and accuracy according to the trajectory data quality inspection requirements, and synchronously collecting trajectory perception data with the roadside trajectory perception equipment to be inspected.

[0051] The roadside equipment and video ground truth system under inspection simultaneously collect inspection data in low road traffic saturation (≤0.3), medium road traffic saturation (0.3~0.7), and high road traffic saturation (≥0.7) scenarios, with data collection lasting 15 minutes in each traffic scenario;

[0052] After completing the data collection for the inspection and evaluation, the video data obtained by the video truth system is used to manually count the correspondence between the real vehicles within the scope of the roadside equipment to be inspected and the vehicle ID identifiers in the roadside sensing trajectory data under each scenario. The number of tracks that are continuously tracked, added to the inspection, and missed in the roadside sensing trajectory data to be inspected is counted, and the length of the track that is tracked correctly is calculated. The data is then input into the data recording module.

[0053] S3. Compare the perceived trajectory data of the roadside equipment to be inspected with the data truth value, and calculate the accuracy evaluation and continuity evaluation index of the trajectory data; use the perceived data of the vehicle-mounted truth unit to calculate the accuracy index of the perceived trajectory data of the roadside equipment to be inspected, and use the perceived data of the video truth unit to calculate the continuity index of the perceived trajectory data of the roadside equipment to be inspected, and take the average value of the evaluation results of the indexes obtained from each data collection and calculation under different inspection scenarios.

[0054] S4. After calculating the indicators, based on the quality evaluation indicators of the sensing trajectory data, evaluate and classify the quality of the sensing trajectory data of the roadside to be inspected according to the grading standards, generate a data quality report and provide application suggestions.

[0055] The specific trajectory data quality evaluation indicators in step S4 include the following aspects: trajectory data accuracy and trajectory data continuity:

[0056] In the quality evaluation of roadside sensing trajectory data, accuracy refers to the degree to which roadside sensing facilities accurately perceive the size, pose, and speed information of vehicle targets. This includes five evaluation indicators: vehicle size error, vehicle absolute positioning error, vehicle relative positioning error, vehicle speed error, and vehicle heading angle error.

[0057] (1) Vehicle size error refers to the average absolute deviation between the actual size of the vehicle and the vehicle size perceived by the roadside at the corresponding moment. The calculation formula is as follows:

[0058]

[0059] In the formula E s Vehicle dimensional error index (m); (l v (s),w v (s),h v (s) represents the true values ​​(m) of the vehicle's length, width, and height during the s-th data collection; (l r (s,i),w r (s,i),h r (s,i)) represents the length, width, and height (m) of the roadside sensing vehicle at time i during the s-th data collection.

[0060] (2) Vehicle absolute positioning error refers to the average absolute deviation between the true value of the vehicle's center point position coordinates and the corresponding time-based vehicle position coordinates perceived by the roadside. The calculation formula is as follows:

[0061]

[0062] In the formula E pa The absolute positioning error index for the vehicle is (m); (x) v (s,i),y v (s,i) represents the true coordinates of the vehicle center point at time i during the s-th data acquisition; (x) represents the true coordinates of the vehicle center point at time i during the s-th data acquisition. r (s,i),y r (s,i)) represents the vehicle position coordinates obtained by roadside sensing at time i during the s-th data collection.

[0063] (3) Vehicle relative positioning error refers to the average absolute deviation between the true value of the relative position coordinates of the vehicle and the surrounding vehicles and the relative position coordinates of the roadside-sensed vehicles at the corresponding time. The calculation formula is as follows:

[0064]

[0065] In the formula E pr The vehicle's relative positioning error index (m); The true relative position coordinates of the surrounding vehicle j and the center true value vehicle at time i during the s-th data acquisition; Let r be the relative position coordinates of the roadside sensing vehicle j and the central vehicle at time i during the s-th data acquisition; s The number of vehicles surrounding the center ground truth vehicle during the s-th data collection.

[0066] (4) Vehicle speed error refers to the average absolute deviation between the actual vehicle speed and the roadside-sensed vehicle speed at the corresponding moment. The calculation formula is as follows:

[0067]

[0068] In the formula, E is v Vehicle speed error index (m / s); v v (s,i) represents the true speed (m / s) of the vehicle at time i during the s-th data acquisition; v r (s,i) represents the vehicle speed (m / s) obtained by roadside sensing at time i during the s-th data collection.

[0069] (5) Vehicle heading angle error refers to the average absolute deviation between the vehicle's actual heading angle and the roadside-sensed vehicle heading angle at the corresponding moment. The calculation formula is as follows:

[0070]

[0071] In the formula E θ The vehicle heading angle error index (°); Let θ be the true value (°) of the vehicle's heading angle at time i during the s-th data acquisition; r (s,i) represents the vehicle heading angle (°) sensed by the roadside at time i during the s-th data collection.

[0072] In the quality evaluation of roadside sensing trajectory data, trajectory data continuity refers to the degree of continuity of the trajectory during the process of roadside sensing facilities tracking vehicle targets, including two types of evaluation indicators: multi-target tracking continuity and trajectory spatiotemporal continuity.

[0073] (1) The continuity of multi-target tracking refers to the average ratio of the number of continuously tracked vehicle trajectories to the total amount of roadside sensing data of vehicle trajectories within the time and space range corresponding to the evaluation experiment. The calculation formula is as follows:

[0074]

[0075] In the formula C mt 1 is the multi-target tracking continuity index (%); TP(s) is the number of continuously tracked vehicle trajectories in the roadside perception data at the s-th data collection time; FN(s) is the number of vehicle trajectories that were missed in the roadside perception data at the s-th data collection time; FP(s) is the number of vehicle trajectories that were added to the roadside perception data at the s-th data collection time.

[0076] (2) The spatiotemporal continuity of the trajectory refers to the average ratio of the length of the continuously tracked roadside sensing vehicle trajectory to the maximum tracking range of the roadside sensing facility within the time and space range corresponding to the evaluation experiment. The calculation formula is as follows:

[0077]

[0078] In the formula C st L is the index of spatiotemporal continuity of trajectory (%). r (s,i) represents the length of the trajectory i of the roadside sensing vehicle that is continuously tracked during the s-th data collection; L v This refers to the maximum tracking range length of the roadside sensing facility.

[0079] The trajectory data quality assessment and grading criteria in step S4 include grade classification and indicator requirements:

[0080] To enable roadside perception systems to better serve different application scenarios, the quality requirements for roadside perception data can be divided into three sensing levels. The basis for classifying the different sensing levels is as follows:

[0081] (1) Perception Level 1 (SL1): Perception data quality level for platform data applications. Vehicle trajectory roadside perception data that meets this level is complete, target tracking is basically accurate and continuous, and can support corresponding traffic data statistics, traffic incident management and related traffic control applications.

[0082] (2) Perception Level 2 (SL2): Perception data quality level for driver assistance applications. Vehicle trajectory roadside perception data that meets this level should maintain high accuracy and completeness, and can support applications that provide real-time road traffic information, risk warning information, etc., to assist drivers in making on-site decisions.

[0083] (3) Perception Level 3 (SL3): Perception data quality level for vehicle-road cooperative autonomous driving applications. Vehicle trajectory roadside perception data that meets this level should provide real-time, accurate and complete vehicle motion information for the autonomous driving system, realize the beyond-line-of-sight perception function outside the vehicle, and support the vehicle autonomous driving system to make decision-making applications.

[0084] Table 2 shows the requirements for various data quality evaluation indicators for different perception levels. The corresponding perception level can be calculated based on each indicator, and Table 3 provides corresponding application suggestions for perception data based on the perception level and the actual needs of the smart highway scenario.

[0085] Table 2: Grading Requirements for Quality Evaluation Indicators

[0086]

[0087]

[0088] Table 3: Reference for Comparison of Sensing Data Classification and Application Functions

[0089]

[0090]

[0091] To implement the above evaluation method, a quality evaluation system for roadside sensing trajectory data is proposed, including: a roadside sensing unit to be evaluated;

[0092] The vehicle's true information, including vehicle true trajectory information, is collected by selecting and deploying sensors in the vehicle-mounted true information unit.

[0093] Roadside video data and vehicle information are collected by selecting and deploying video truth cells;

[0094] The data is collected from the roadside sensing unit, vehicle-mounted truth unit, and video truth unit to be evaluated through the data recording module.

[0095] The quality evaluation module calculates the quality evaluation index of the roadside sensing trajectory data to be evaluated, classifies the data quality, generates a trajectory data quality report, and provides application suggestions.

[0096] Furthermore, the vehicle-mounted truth unit includes a truth vehicle, an inertial navigation system subunit installed on the truth vehicle, an vehicle-mounted perception subunit, and a data acquisition subunit.

[0097] The inertial navigation integrated unit includes inertial navigation elements and GNSS positioning elements, which obtain true information about the vehicle's size, position, speed, and heading angle.

[0098] The vehicle-mounted perception subunit includes vehicle-mounted lidar and millimeter-wave radar; the vehicle is equipped with no fewer than two high-precision perception sensors, which can sense the relative position coordinates of vehicle targets within a 360° range around the vehicle, with a sensing distance of ≥50m.

[0099] The data acquisition subunit is used to collect data from the inertial navigation subunit and the vehicle perception subunit, and obtain the true vehicle positioning coordinates, relative positioning coordinates of surrounding vehicles, speed, and heading angle as reference true data to evaluate the quality of roadside perception data.

[0100] The data format and corresponding data accuracy requirements exported by the vehicle-mounted truth perception system are shown in Table 4.

[0101] Table 4: Data Output Format and Accuracy Requirements of Vehicle-Mounted Truth Analysis System

[0102]

[0103] Furthermore, the video truth unit includes a video sensing device and an offline processing server. The video sensing device collects video data and transmits it to the offline server for post-processing through the acquisition host. The correspondence between each vehicle within the range of the roadside equipment to be evaluated and the roadside sensing trajectory is determined by manual verification. This is used as the benchmark truth data to evaluate the quality of the roadside sensing data.

[0104] Furthermore, during the inspection and evaluation process, the roadside sensing unit to be inspected and the vehicle-mounted truth unit or video truth unit simultaneously collect data and transmit it to the data recording module. The inspection and evaluation personnel verify the truth values ​​of the collected data and calculate various quality evaluation indicators in the quality evaluation module, and generate a data quality evaluation report.

[0105] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention, and applications, modifications and variations thereof will be apparent to those skilled in the art.

[0106] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. An experimental system for roadside perception trajectory data quality evaluation, characterized in that, The application comprises: a to-be-evaluated roadside perception unit; wherein the vehicle true value information is collected by selecting and arranging the sensors of the vehicle true value unit, and the information includes vehicle true value trajectory information; roadside video data and vehicle information are collected by selecting and arranging the video true value unit; the vehicle true value unit comprises a true value vehicle, an inertial integrated navigation subunit installed on the true value vehicle, a vehicle perception subunit and a data collection subunit; the inertial integrated navigation subunit comprises an inertial navigation element and a GNSS positioning element, and the size, position, speed and heading angle true value information of the vehicle are obtained through the inertial navigation element and the GNSS positioning element; the video true value unit comprises a video perception device and an offline processing server, the video perception device collects video data and transmits the video data to the offline server for post-processing through a collection host, the correspondence between the real vehicles within the range of the to-be-evaluated roadside device in each scene and the roadside perception trajectory data vehicle ID is manually counted, the number of trajectories that are continuously tracked, rechecked and missed in the to-be-evaluated roadside perception trajectory data is counted, and the length of the trajectory that is correctly tracked is calculated, which are used as the benchmark true value data to evaluate the quality of the roadside perception data; and in the evaluation process, the to-be-evaluated roadside perception unit, the vehicle true value unit or the video true value unit synchronously collect data and transmit the data to a data recording module; the vehicle perception subunit comprises a vehicle laser radar and a millimeter wave radar; the data collection subunit is used to collect data from the inertial integrated navigation subunit and the vehicle perception subunit, and obtain the positioning coordinates of the true value vehicle, the relative positioning coordinates of the surrounding vehicles, the speed and the heading angle as the benchmark true value data to evaluate the quality of the roadside perception data; the data collected by the to-be-evaluated roadside perception unit, the vehicle true value unit and the video true value unit are collected through the data recording module; the perception trajectory data of the vehicle true value unit and the to-be-evaluated roadside perception trajectory data are time and space aligned; for the space alignment, the sensor coordinate system of the vehicle true value unit and the measured roadside trajectory perception coordinate system are obtained, the sensor coordinate system of the vehicle true value unit is changed into a world coordinate system according to the vehicle parameters, and then the sensor coordinate system of the vehicle true value unit is aligned with the measured roadside trajectory perception coordinate system by using the absolute position information of the sensor; for the time alignment, the perception system time of the vehicle true value unit is aligned with the roadside data system time by manually calibrating the parameters, and the data sampling time is synchronized to ensure the time comparability of the vehicle and roadside data; the quality evaluation module is used to calculate the quality evaluation index of the to-be-evaluated roadside perception trajectory data, grade the data quality, generate a trajectory data quality report and give application suggestions; in order to make the roadside perception system better serve different application scenarios, the roadside perception data quality requirements can be divided into three perception levels, and the basis for dividing different perception levels is as follows: (1) perception level 1 (SL1): the perception data quality level for platform data application, the vehicle trajectory roadside perception data information of which meets the level is complete, the target tracking is basically accurate and continuous, and can support corresponding traffic data statistics, traffic event management and related traffic control applications; (2) Perception Level 2 (SL2): The perception data quality level for auxiliary driving application, the vehicle trajectory roadside perception data meeting this level should maintain high accuracy and integrity, which can support real-time road traffic information, risk warning information, and assist drivers in making on-site decision-making applications; (3) Perception Level 3 (SL3): The perception data quality level for vehicle-road cooperative automatic driving application, the vehicle trajectory roadside perception data meeting this level should provide real-time, accurate, and complete vehicle motion information for the automatic driving system, realize the function of outside-visual-range perception, and support the decision-making application of the vehicle automatic driving system. Different perception levels have different requirements for various data quality evaluation indicators. The corresponding perception level can be calculated according to each indicator, and the corresponding perception data application suggestion can be given according to the perception level and the actual demand of intelligent highway scene.

2. The experimental method for roadside perception trajectory data quality evaluation according to claim 1, wherein, The method comprises the following steps: S1, checking the composition of the to-be-inspected roadside equipment, determining the inspection requirements of the trajectory data quality inspection and evaluation system, and determining the evaluation system of the trajectory data quality inspection and evaluation; S2, selecting and placing a vehicle-mounted true value unit and a video true value unit, and simultaneously collecting trajectory perception data of the vehicle-mounted true value unit and perception data of the video true value unit; S3, comparing the perception trajectory data of the to-be-inspected roadside equipment with the data true value, and calculating the accuracy evaluation and continuity evaluation indicators of the trajectory data; S4, evaluating and grading the perception trajectory data quality of the to-be-inspected roadside according to the grading standard, generating a data quality report, and giving application suggestions.

3. The experimental method for roadside perception trajectory data quality evaluation according to claim 2, wherein, The inspection in step S1 includes the hardware composition, detection accuracy, frequency, and placement requirements of the equipment; The determination of the inspection requirements of the trajectory data quality inspection and evaluation includes the inspection road requirements and the inspection environment requirements; The evaluation dimensions in the evaluation system include trajectory data accuracy evaluation and trajectory data continuity evaluation, and the specific evaluation indicators are determined according to the actual scene demand.

4. The experimental method for roadside perception trajectory data quality evaluation according to claim 2, wherein, Step S2 specifically comprises the following steps: S21, selecting a suitable true value vehicle according to the main vehicle type of the inspection road, installing an inertial combined navigation subunit and a vehicle-mounted perception subunit on the true value vehicle, calibrating the inertial combined navigation subunit and the vehicle-mounted perception subunit, and synchronizing the time with the to-be-inspected roadside perception facility; S22, the true value vehicle controls the speed to be 20km / h, 40km / h, 60km / h, and 80km / h in turn within the perception range of the to-be-inspected sensor, and drives in low road traffic saturation, medium road traffic saturation, and high road traffic saturation scenes, respectively, the vehicle-mounted true value unit and the to-be-inspected roadside sensor synchronously and repeatedly collect inspection data, so that the true value vehicle completes at least one lane changing behavior in the facility perception range and covers the data of driving in different lanes on the road; S23, after completing the inspection data collection, determining the corresponding roadside trajectory data of the true value vehicle in each data collection from the to-be-inspected roadside perception trajectory data, time and space aligning the vehicle-mounted true value unit and the to-be-inspected roadside sensor, extracting the true value data of the vehicle size and position coordinates of the true value vehicle at each time based on the time stamp of the to-be-inspected roadside trajectory data, and inputting the data into a data recording module.

5. The experimental method for roadside perception trajectory data quality evaluation according to claim 2, wherein, Step S3 specifically comprises the following steps: S31, according to the trajectory data quality evaluation requirements selection and placement of video truth unit, so that the video truth unit acquires the video data of the road in the range of the road side facilities to be evaluated, and adjusts the data acquisition frequency and accuracy according to the trajectory data quality evaluation requirements, the video truth unit and the road side trajectory sensing device to be evaluated synchronously collect trajectory sensing data; S32, the road side device to be evaluated and the video truth unit synchronously collect evaluation data under low road traffic saturation, medium road traffic saturation and high road traffic saturation, and the data collection time lasts for 15 minutes under each traffic scenario; S33, after completing the evaluation data collection, the corresponding relationship between the real vehicles in the range of the road side device to be evaluated and the vehicle ID identification of the road side sensing trajectory data is manually counted through the video data acquired by the video truth unit, the number of trajectories of tracking continuity, additional detection and missed detection in the road side sensing trajectory data to be evaluated is counted and the length of the trajectory of correct tracking is calculated, and the data is input into the data recording module.

Images