A lane line data processing method and device

Abstract

This application discloses a lane line data processing method, which acquires data to be evaluated and determines the ground truth data at the second acquisition time closest to the first acquisition time corresponding to the data to be evaluated. The data to be evaluated includes first vehicle information and first lane line information. The first vehicle information includes the vehicle's first speed and first yaw rate, and the first lane line information includes a first lane line coefficient. The ground truth data includes second lane line information, which includes a second lane line coefficient. Based on the first vehicle information, the vehicle's position offset and heading angle offset are determined, and the first lane line coefficient or the second lane line coefficient is updated based on the first acquisition time, the second acquisition time, the position offset, and the heading angle offset. That is, the technical solution provided by this application can achieve alignment between the ground truth data and the data to be evaluated in the time dimension, improving the accuracy of the evaluation results.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, specifically to a lane line data processing method and apparatus. Background Technology

[0002] Lane line and road boundary detection are key technologies in the field of intelligent driving perception. The quantitative evaluation indicators of their perception capabilities are a reference for OEMs to select sensors.

[0003] Currently, laser ground truth is mainly used to evaluate lane line or road boundary detection results. However, due to the difference between the laser ground truth acquisition period and the detection period of the detection results, there is a time difference between the ground truth data and the data to be evaluated, which affects the evaluation results. Summary of the Invention

[0004] In view of this, embodiments of this application provide a lane line data processing method and apparatus, which achieves alignment of true data and data to be evaluated in the time dimension by updating lane line coefficients, thereby improving the accuracy of evaluation results.

[0005] To address the above problems, the technical solutions provided in this application are as follows:

[0006] In a first aspect of this application, a lane line data processing method is provided, the method comprising:

[0007] Acquire the data to be evaluated, which includes first vehicle information and first lane information at a first acquisition time. The first vehicle information includes the first speed and first yaw rate of the vehicle, and the first lane information includes a first lane coefficient.

[0008] Obtain the true value data closest to the first acquisition time, the true value data including the second lane line information at the second acquisition time, the second lane line information including the second lane line coefficient;

[0009] Based on the first vehicle information, determine the vehicle's position offset and heading angle offset from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time;

[0010] The first lane line coefficient or the second lane line coefficient is updated based on the first acquisition time corresponding to the data to be evaluated, the second acquisition time corresponding to the true value data, the position offset, and the heading angle offset.

[0011] In a second aspect of this application, a lane line data processing apparatus is provided, the apparatus comprising:

[0012] An acquisition unit is used to acquire data to be evaluated, the data to be evaluated including first vehicle information and first lane information at a first acquisition time, the first vehicle information including the first speed and first yaw rate of the vehicle, and the first lane information including a first lane coefficient.

[0013] The acquisition unit is further configured to acquire the true value data closest to the first acquisition time, the true value data including the second lane line information at the second acquisition time, the second lane line information including the second lane line coefficient;

[0014] The determining unit is configured to determine, based on the first vehicle information, the position offset and heading angle offset of the vehicle from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time;

[0015] The update unit is used to update the first lane line coefficient or the second lane line coefficient based on the first acquisition time corresponding to the data to be evaluated, the second acquisition time corresponding to the true value data, the position offset, and the heading angle offset.

[0016] In a third aspect of this application, an electronic device is provided, comprising: a processor and a memory;

[0017] The memory is used to store computer-readable instructions or computer programs;

[0018] The processor is configured to read the computer-readable instructions or the computer program to cause the electronic device to implement the method described in the first aspect.

[0019] In a fourth aspect of this application, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a device, cause the device to perform the method described in the first aspect.

[0020] In a fifth aspect of this application, a computer program product is provided that, when the computer program product is run on a computer, causes the computer to perform the method described in the first aspect.

[0021] Therefore, the embodiments of this application have the following beneficial effects:

[0022] In this application, data to be evaluated is acquired, and the ground truth data of the second acquisition time closest to the first acquisition time corresponding to the data to be evaluated is determined. The data to be evaluated includes first vehicle information and first lane line information. The first vehicle information includes the vehicle's first speed and first yaw rate, and the first lane line information includes a first lane line coefficient. The ground truth data includes second lane line information, which includes a second lane line coefficient. The vehicle's position offset and heading angle offset are determined based on the first vehicle information. The first lane line coefficient or the second lane line coefficient is updated based on the first acquisition time corresponding to the data to be evaluated, the second acquisition time corresponding to the ground truth data, the position offset, and the heading angle offset. If the first lane line coefficient is updated, the data to be evaluated corresponding to the second acquisition time is acquired; if the second lane line coefficient is updated, the ground truth data corresponding to the first acquisition time is acquired. That is, the technical solution provided in this application can achieve alignment between the ground truth data and the data to be evaluated in the time dimension, improving the accuracy of the evaluation results.

[0023] In addition, stability and effectiveness indicators of lane line recognition capability can be determined based on multiple sets of test data. Then, based on stability and effectiveness indicators, true value-free evaluation can be achieved. Combined with true value evaluation results (accuracy indicators), a complete lane line evaluation system can be established to achieve multi-dimensional quantitative evaluation of lane line capability. Attached Figure Description

[0024] Figure 1 A flowchart of a lane line data processing method provided in this application embodiment;

[0025] Figure 2 A schematic diagram of vehicle position estimation provided in this application embodiment;

[0026] Figure 3a A schematic diagram of GT / DT data time relationship provided in an embodiment of this application;

[0027] Figure 3b Another schematic diagram of GT / DT data time relationship provided in this application embodiment;

[0028] Figure 4 This application provides a schematic diagram comparing accuracy indicators under different scenarios.

[0029] Figure 5a A schematic diagram illustrating the comparison of lane line coefficient stability provided in this application embodiment;

[0030] Figure 5b This application provides a comparison chart of lane line coefficient stability under different scenarios.

[0031] Figure 6aThis application provides a radar-based lane capability assessment map.

[0032] Figure 6b A schematic diagram illustrating the evaluation results of different sensing schemes provided in an embodiment of this application;

[0033] Figure 6c An evaluation result diagram of different sensing schemes at night provided for an embodiment of this application;

[0034] Figure 7 This is a structural diagram of a lane line data processing device provided in an embodiment of this application. Detailed Implementation

[0035] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the embodiments of this application will be further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0036] It should be noted that the lane lines in this embodiment include not only the lane lines located in the middle lane of the road, but also the road boundary lines.

[0037] See Figure 1 This figure is a flowchart of a lane line data processing method provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes:

[0038] S101: Obtain the data to be evaluated.

[0039] The data to be evaluated refers to the data concerning lane lines, which includes, but is not limited to, vehicle information and lane line information at the first acquisition time. The first vehicle information may include the first vehicle speed and the first yaw rate; the first lane line information includes the first lane line coefficient, the start / end position of the lane line, confidence level, and validity indicator. The first lane line coefficient is used to construct the lane line equation; the confidence level refers to the confidence level of the data to be evaluated; and the validity indicator indicates that the data to be evaluated is valid.

[0040] For example, the first lane line coefficient includes four fitting coefficients, namely C D0 C D1 C D2 C D3 The lane lines corresponding to the data to be evaluated satisfy the following equation:

[0041] y(x)=C D0 +C D1 ·x+C D2 ·x 2 +C D3 ·x 3

[0042] Among them, the timestamps t corresponding to different first lane line coefficients DT different.

[0043] In this embodiment, multiple periods of data to be evaluated can be acquired, and the acquisition time for different data to be evaluated is different.

[0044] S102: Obtain the true value data closest to the first acquisition time corresponding to the data to be evaluated.

[0045] For any data to be evaluated, based on the first acquisition time corresponding to that data, the nearest true value data is determined. That is, the time interval between the second acquisition time corresponding to the determined true value data and the first acquisition time corresponding to the data to be evaluated is the smallest. It should be noted that in this embodiment, the output period of the data to be evaluated is different from the output period of the true value data. For example, the output period of the data to be evaluated is 30 milliseconds, and the output period of the true value data is 100 milliseconds.

[0046] The true data includes second vehicle information and second lane line information. The second vehicle information may include the second vehicle speed and the second yaw rate. The second lane line information includes the second lane line coefficient, the start / end position of the lane line, the confidence level, and the validity indicator. The second lane line coefficient is used to construct the lane line equation, the confidence level refers to the confidence of the true data, and the validity indicator is used to indicate that the true data is valid.

[0047] For example, the second lane line coefficient includes four fitting coefficients, namely C G0 C G1 C G2 C G3 The lane lines corresponding to the true data then satisfy the following equation:

[0048] y(x)=C G0 +C G1 ·x+C G2 ·x 2 +C G3 ·x 3

[0049] Among them, C G0 C G1 C G2 C G3 Its coefficient is t. GT The timestamp t corresponding to different second lane line coefficients. GT different.

[0050] S103: Based on the first vehicle information, determine the vehicle's position offset and heading angle offset from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time.

[0051] In calculating the vehicle position offset and heading angle offset, it is necessary to first obtain the vehicle's pose information at different times, and then determine the position offset and heading angle offset based on the pose information at different times. Specifically, the first vehicle pose information in the target coordinate system is determined based on the first vehicle speed and the first yaw rate. This first vehicle pose information includes a first coordinate position and a first heading angle. Then, the second vehicle pose information is determined based on the first vehicle pose information at the second acquisition time. This second vehicle pose information includes a second coordinate position and a second heading angle.

[0052] In this embodiment, once the first coordinate position and first heading angle of the vehicle at the first acquisition time, and the second coordinate position and second heading angle of the vehicle at the second acquisition time are obtained, the position offset can be directly calculated based on the first coordinate position and the second coordinate position; and the heading angle offset can be calculated based on the first heading angle and the second heading angle.

[0053] Specifically, determining the second vehicle pose information at the second acquisition time based on the first vehicle pose information can be achieved in the following way: forming a vehicle pose information queue for the first vehicle pose information corresponding to multiple different first acquisition times; and obtaining the second vehicle pose information corresponding to each second acquisition time by performing linear difference in the vehicle pose information queue based on the relationship between the second acquisition time and the first acquisition time.

[0054] The specific implementation of determining the vehicle's first coordinate position (x, y) and first heading angle α in the target coordinate system based on the vehicle speed v and yaw rate γ in the first vehicle information may include:

[0055] In the initial frame (initial time), x, y, and α all start from 0. The queue stores vehicle motion data for all time points. Each vehicle motion data point includes a timestamp t, vehicle speed, yaw rate, vehicle pose {x, y}, and heading angle α, i.e., (t, v, γ, x, y, α). The specific calculation method is as follows:

[0056] 1. Initialization: For the first frame, at time t=0, we have x=0, y=0, and α=0;

[0057] 2. Calculate Δx, Δy, and Δyaw at time t (not the first frame) relative to time t-1. The longitudinal velocity v at time t-1 is known. t-1 Lateral velocity (0), yawrateγ t-1 and historical heading angle α t-1 Based on time Δt, the changes in vehicle position Δx, Δy, and Δyaw at time t can be calculated, as shown in the attached figure. Figure 2 :

[0058]

[0059]

[0060]

[0061] Among them, is the longitudinal average speed of two consecutive frames (at times t-1 and t), is the lateral average speed of two consecutive frames (which is 0), is the average yaw angular velocity of two consecutive frames;

[0062] 1. Update all the vehicle historical poses x n , y n (n < t) before time t. That is,

[0063] x n = x n - Δx;

[0064] y n = y n - Δy.

[0065] The pose of the vehicle itself at the current time t is always set to 0, that is, x t = 0, y t = 0, α t = α t-1 + Δyaw, and add the information at time t to the queue.至此队列中的自车位姿信息全部更新。

[0066] After obtaining the vehicle pose information queue formed by the first vehicle pose information, according to the size relationship between the second acquisition time and each of the above first acquisition times, linearly interpolate in the above vehicle pose information queue to determine the second coordinate position (x, y) and the second heading angle α of the vehicle at the second acquisition time.

[0067] Among them, the specific value of the heading angle offset is determined by the size relationship between the first acquisition time and the second acquisition time. Specifically, if the second acquisition time is less than the first acquisition time, the difference between the first heading angle and the second heading angle is determined as the heading angle offset from the second acquisition time to the first acquisition time; if the second acquisition time is greater than the first acquisition time, the difference between the second heading angle and the first heading angle is determined as the heading angle offset from the first acquisition time to the second acquisition time.

[0068] For example, the first acquisition time is t DT 、the first coordinate position is {x DT 、y DT}、the first heading angle is α DT ; the second acquisition time is tGT The second coordinate position is {x} GT y GT The second heading angle is α. GT Calculate the position offset Δdis and the heading angle offset Δα.

[0069] If t GT <t DT , thus obtaining ΔT and Δα, where Δx=x DT -x GT ; Δy = y DT -y GT ;Δα=α DT -α GT .

[0070] If t GT >t DT , thus obtaining ΔT and Δα, where Δx=x GT -x DT ; Δy = y GT -y DT ; Δα=α GT -α DT .

[0071] S104: Update the first lane line coefficient or the second lane line coefficient based on the first acquisition time corresponding to the data to be evaluated, the second acquisition time corresponding to the true data, the position offset, and the heading angle offset.

[0072] In this embodiment, the decision to update either the first lane line coefficient or the second lane line coefficient will be determined based on the relationship between the first acquisition time and the second acquisition time. Specifically:

[0073] If the second acquisition time is less than the first acquisition time, the second lane line coefficient is updated based on the position offset and heading angle offset. That is, when the acquisition time of the ground truth data is less than the acquisition time of the data to be evaluated, the second lane line coefficient corresponding to the second acquisition time is updated, and the updated second lane line coefficient is used as the lane line coefficient of the ground truth data at the first acquisition time. For example, Figure 3a As shown, if t GT <t DT , will t GT GT lane line data predicts time to t DT time.

[0074] If the second acquisition time is greater than the first acquisition time, the first lane line coefficient is updated based on the position offset and the heading angle offset. That is, when the acquisition time of the ground truth data is greater than the acquisition time of the data to be evaluated, the first lane line coefficient corresponding to the first time is updated, and the updated first lane line coefficient is used as the lane line coefficient of the data to be evaluated at the second acquisition time. For example, Figure 3b As shown, if t GT >t DT , will t DT Time DT lane line data predicts to t GT time.

[0075] The first lane line coefficient and the second lane line coefficient each include at least four fitting coefficients, which are coefficients corresponding to different powers of the independent variable; updating the first lane line coefficient or the second lane line coefficient according to the position offset and the heading angle offset includes:

[0076] The update of the first fitting coefficient C0 includes: taking the original value of the first coefficient, the product of the position offset and the second coefficient, the product of the square of the position offset and the third coefficient, the product of the cube of the position offset and the fourth coefficient, and the sum of the negatives of the product of the first multiple of the position offset and the heading angle offset as the new value of the first coefficient. See the following calculation formula for details:

[0077] C0 = C0 + Δdis * C1 + Δdis 2 *C2+Δdis 3 *C3-0.5*Δdis*Δα;

[0078] Wherein, the first fitting coefficient C0 is the coefficient corresponding to the independent variable raised to the power of 0, the second fitting coefficient C1 is the coefficient corresponding to the independent variable raised to the power of 1, the third fitting coefficient C2 is the coefficient corresponding to the independent variable raised to the power of 2, the fourth fitting coefficient C3 is the coefficient corresponding to the independent variable raised to the power of 3, Δdis represents the position offset, and Δα represents the heading angle offset. 0.5 is the first multiple. This first multiple can be set according to actual conditions.

[0079] The update of the second fitting coefficient C1 includes: taking the original value of the second coefficient, the product of the second multiple of the position offset and the third coefficient, the product of the square of the third multiple of the position offset and the fourth coefficient, and the negative of the heading angle offset as the new value of the second coefficient. See the following calculation formula for details:

[0080] C1 = C1 + 2 * Δdis * C2 + 3 * Δdis 2 *C3-Δα;

[0081] Where 2 is the second multiple and 3 is the third multiple. The specific values ​​of the second and third multiples can be set according to the actual situation.

[0082] The update of the third fitting coefficient C2 includes: taking the sum of the original value of the third coefficient, the position offset of the fourth multiple, and the product of the fourth coefficient as the new value of the third coefficient. See the following formula for details:

[0083] C2 = C2 + 3 * Δdis * C3;

[0084] The number 3 mentioned above is the fourth multiple. The specific value of this fourth multiple can be set according to the actual situation.

[0085] For the update of the fourth fitting coefficient C3, the original value is maintained, that is:

[0086] C3 = C3.

[0087] For updates to the coefficients in the first lane line coefficient and the coefficients in the second lane line coefficient, please refer to the above content.

[0088] It should be noted that before using ground truth data to validate the data to be evaluated, the data formats of the ground truth data and the data to be evaluated can be aligned, such as using a Common-Separated Values ​​(CSV) file.

[0089] As can be seen, the above scheme can achieve alignment between the true data and the data to be evaluated in the time dimension, thereby improving the accuracy of the evaluation results.

[0090] Through the above method embodiments, alignment between ground truth data and data to be evaluated in the time dimension can be achieved. Based on the acquired multiple sets of first lane line coefficients and second lane line coefficients, the lateral error between the data to be evaluated and the ground truth data can be assessed to achieve a quantitative evaluation of accuracy indicators. The acquisition times for different sets are different, while the acquisition times for the first and second lane line coefficients within the same set are the same. The main accuracy indicators include lateral error and error stability, which will be explained below.

[0091] (I) Accuracy Calculation

[0092] For any group of first lane line coefficients and second lane line coefficients, and any independent variable among the N independent variables, calculate the difference between the dependent variable corresponding to the first lane line coefficient and the dependent variable corresponding to the second lane line coefficient under the same independent variable; determine the prediction accuracy index based on the N differences corresponding to each group.

[0093] In other words, N x values ​​are taken. For any x value, for a set of first lane line coefficients and second lane line coefficients, the difference between the y value of the first lane line coefficient at x value and the y value of the second lane line coefficient at x value is calculated, thus obtaining N difference values. That is, each set corresponds to N difference values, and the evaluation accuracy index is determined based on the N difference values ​​corresponding to multiple sets.

[0094] The range of x values ​​can be preset to determine the evaluation accuracy index within a certain range. For example, within a 30m range along the road, N points are taken at equal intervals (e.g., intervals of 10m, N=4), and the y value corresponding to the x value at each point is calculated.

[0095] Among them, the evaluation accuracy indicators may include the mean absolute error in the y-direction, the maximum absolute error, the standard deviation, the 9544th percentile error, the 9974th percentile error, and the percentage of absolute errors less than half the width of the lane line.

[0096] Specifically, it includes the following:

[0097] (1) Take the average of the absolute values ​​of a set of N differences to obtain a set of first error mean values; take the average of multiple sets of first error mean values ​​to obtain the target first error mean value.

[0098] (2) Determine the maximum mean of the first error from multiple sets of mean first errors.

[0099] (3) Take the average of a set of N differences to obtain a second set of error mean values; determine the standard deviation σ based on multiple sets of second error mean values. y .

[0100] (4) The 9544th percentile error was calculated, i.e., 2σ. y The calculation yielded a 9974th quantile error, or 3σ. y .

[0101] (5) Determine the proportion of absolute values ​​of N differences that are less than a first preset threshold to obtain a set of proportions; take the average of multiple sets of proportions to obtain the target proportion. The first preset threshold is half the lane width. For example, if the lane width is 15cm, then the first preset threshold is 7.5cm.

[0102] Typically, accuracy evaluation focuses on a set of lateral (y-direction) distance errors and their percentages within a certain longitudinal (x-direction) distance range (e.g., 30m), obtaining statistical characteristics through multiple sets of data. Key indicators for these multiple sets of statistics include:

[0103] Maximum and average lateral absolute error within 30m;

[0104] Standard deviation of lateral error within 30m (σ) y );

[0105] Lateral 9544th percentile error within 30m (2σ) y ) and 9974 quantile error (3σ) y );

[0106] The percentage of lanes with a lateral absolute error less than half the lane width within 30m (Δy) abs <7.5cm).

[0107] Among them, the accuracy evaluation based on ground truth data can more accurately quantify and evaluate the lateral accuracy of the lane line perception under test. Using the detection results of the vehicle-mounted camera as the evaluation data and the output results of the LiDAR as the ground truth data, the overall lateral accuracy indicators of the vehicle-mounted camera in multiple scenarios are evaluated as follows, allowing for a horizontal comparison of the accuracy performance of other perception solutions. Details are shown in Table 1:

[0108]

[0109]

[0110] Further analysis of accuracy performance in different daytime scenarios, such as Figure 4 As shown. (Through) Figure 4 It can be seen that the scene has a significant impact on accuracy, especially for typical scenes such as curves with large curvature and fish scale lines.

[0111] In addition to accuracy evaluation based on true values, if another set of lane line data can be obtained, the two sets of lane line data can be compared horizontally in terms of effectiveness and stability. The specific indicators are defined as follows:

[0112] (II) Validity Calculation

[0113] (1) Determine the ratio of the number of valid data in multiple sets of data to the number of sets of data to be evaluated.

[0114] Valid data refers to the first lane line information in a set of test data that includes a validity identifier, and / or the confidence level of the first lane line information is greater than a preset confidence threshold. The validity identifier indicates that the test data is valid; if the test data includes a validity identifier, it is considered valid; if the test data does not include a validity identifier, it is considered invalid.

[0115]

[0116] (2) Determine the ratio of the number of consecutive valid data to the number of valid data in multiple sets of data to be evaluated.

[0117] Continuous valid data refers to data where the difference between the ID of the currently valid data to be evaluated and the ID of the previous set of valid data to be evaluated is less than or equal to a second preset threshold. The specific value of the second preset threshold can be set according to the actual application; for example, the second preset threshold can be 1.

[0118] Typically, multiple sets of data to be evaluated will be collected, and each set of data to be evaluated will be divided into a group ID (which can be the number of collection periods, for example, data to be evaluated 1 was collected in the first period, data to be evaluated 2 was collected in the second period, etc.). If both sets of data to be evaluated are valid data and the difference between the group IDs is less than or equal to the second preset threshold, then the earlier valid data to be evaluated is considered to be continuous valid data.

[0119] Specifically, take all valid data, calculate the ID difference Δid between it and the previous set of valid data. If Δid = 1, it means that the data in this set is continuous. The proportion of all continuous valid data to all valid data is used to determine the continuity of valid data.

[0120]

[0121] (3) Based on multiple sets of data to be evaluated, determine the maximum end position of the lane line and / or the average length of the lane line.

[0122] The first lane line information includes the start and end positions of the lane lines. For a set of evaluation data, the lane line length is determined based on the start and end positions of the corresponding lane lines. Then, the average length of the lane lines is determined based on multiple sets of lane line lengths. The maximum end position is selected from the end positions corresponding to multiple sets of lane lines.

[0123] (III) Stability Calculation

[0124] (1) For the first coefficient, obtain the average value and standard deviation of multiple sets of the first coefficient, and determine the dispersion of the first coefficient based on the average value and standard deviation.

[0125] The first coefficient is the coefficient corresponding to the independent variable raised to the power of 0. For example:

[0126]

[0127] Where, μ C0 For dispersion, σ c0 For standard deviation, This is the average value.

[0128] (2) For any of the other coefficients, obtain multiple sets of standard deviations corresponding to that coefficient.

[0129] Among them, other coefficients refer to coefficients other than the first coefficient among at least four coefficients.

[0130] (3) For any of the other coefficients, determine the difference between the maximum and minimum values ​​of the coefficient in multiple groups.

[0131] Specifically, to further evaluate the magnitude of local fluctuations in the coefficients, the extreme value differences γ1, γ2, and γ3 of coefficients C1, C2, and C3 in each set of data are calculated respectively. γ1 is defined as...

[0132] γ1=C max1 -C min1

[0133] Where C max1 C is the maximum value of C1 within multiple groups. min1 It is the minimum value of C1 within multiple groups.

[0134] Typically, the multiple sets of test data can be treated as a single set. For the maximum / minimum difference corresponding to different sets, the average and / or the largest maximum / minimum difference corresponding to the multiple sets can be obtained.

[0135] (4) Based on multiple sets of data to be evaluated, determine the average and standard deviation of the lane line length; based on the standard deviation and average, determine the dispersion of the lane line length.

[0136] Specifically, based on the lane length corresponding to each group, the average and standard deviation of multiple groups of corresponding lane lengths are calculated to determine the dispersion of lane lengths. For example:

[0137]

[0138] Where, μ L For the degree of dispersion, σ L Standard deviation, This represents the average length.

[0139] Based on computational stability and effectiveness indices, the optimization performance of lane line filtering in environmental modeling can be evaluated, the optimization ratio quantified, and the lane line filtering algorithm driven in positive iteration. For example... Figure 5a As shown, this indicates the optimization ratio of lane line filtering relative to the original input in various indicators under the curve scenario. It can be clearly seen that the stability of the lane coefficient is significantly improved, especially in terms of local stability, and the overall stability of the lane line coefficient under curve is effectively improved.

[0140] Further considering lane line coefficient stability, based on common lane line types, the stability of C0 / C1 / C2 / C3 is evaluated as follows: Figure 5b As shown. (Through) Figure 5b As shown, it can be seen that the C2 / C3 coefficients of the fishbone line in the curve jump significantly, which is prone to problems. This is an aspect that the lane line filtering will focus on optimizing in the future. This is consistent with the actual performance of the intelligent driving function. The function performance at the fish scale line is unstable.

[0141] In addition, after obtaining the accuracy, stability, and effectiveness indicators through the above methods, an evaluation mechanism can be constructed based on at least one of the accuracy, stability, and effectiveness indicators. This evaluation mechanism includes a scoring standard and a weight corresponding to at least one of the accuracy, stability, and effectiveness indicators. Based on the evaluation mechanism and at least one of the accuracy, stability, and effectiveness indicators corresponding to the data to be evaluated, the capability of the device collecting the data to be evaluated can be quantitatively evaluated.

[0142] For example, the above method can be used to calculate indicators for three dimensions: accuracy, effectiveness, and stability. Based on these calculated indicators, six key evaluation indicators can be identified. A lane line capability scoring mechanism can be developed according to their importance, and presented in radar chart format, as shown in the attached figure. Figure 6a The total score is out of 5 points. See Table 2 below for details:

[0143]

[0144]

[0145] In practical applications, for evaluation data acquired through different sensors, the scores of the corresponding evaluation data for each of the aforementioned indicators can be obtained for any given sensor. The final score for that sensor is then determined by weighted summation based on the weights of each indicator in the radar chart. This horizontal comparison of scores from different sensors helps users evaluate sensor performance.

[0146] For example, Figure 6b As shown in the figure, this is an evaluation and scoring result of three lane line perception solutions based on the same test data. The figure clearly shows the differences in performance, achieving a dual quantitative evaluation of sensor cost and performance, and providing a reference for OEMs to establish a sensor selection product library.

[0147] Furthermore, by evaluating the capabilities of different sensors based on different scenarios, it can be seen that the scenario affects the lane line capability, such as... Figure 6c As shown in the figure, the left histogram represents the Mobileye sensor, and the right histogram represents the Minieye sensor. This figure describes the capability scores of different sensing solutions in different nighttime scenarios. Figure 6c It is known that congestion has the greatest impact on lane line capability, especially for Mobileye's solution, which helps in the robustness assessment of perception capability.

[0148] Based on the above method embodiments, this application provides a lane line data processing device, which will be described below with reference to the accompanying drawings.

[0149] See Figure 7 The figure shows a lane line data processing device provided in an embodiment of this application. The processing device 700 may include:

[0150] The acquisition unit 701 is used to acquire data to be evaluated, the data to be evaluated includes first vehicle information and first lane information at a first acquisition time, the first vehicle information includes the first speed and first yaw rate of the vehicle, and the first lane information includes a first lane coefficient.

[0151] The acquisition unit 701 is further configured to acquire the truth data closest to the first acquisition time, the truth data including the second lane line information at the second acquisition time, the second lane line information including the second lane line coefficient;

[0152] Determining unit 702 is used to determine the position offset and heading angle offset of the vehicle from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time based on the first vehicle information;

[0153] The updating unit 703 is also used to update the first lane line coefficient or the second lane line coefficient based on the first acquisition time corresponding to the data to be evaluated, the second acquisition time corresponding to the true value data, the position offset, and the heading angle offset.

[0154] In some embodiments, the determining unit 702 is specifically configured to: determine first vehicle pose information of the vehicle in the target coordinate system based on the first vehicle speed and the first yaw rate, wherein the first vehicle pose information includes a first coordinate position and a first heading angle; determine second vehicle pose information of the vehicle at the second acquisition time based on the first vehicle pose information, wherein the second vehicle pose information includes a second coordinate position and a second heading angle; determine a position offset from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time based on the first coordinate position and the second coordinate position; and determine a heading angle offset from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time based on the first heading angle and the second heading angle.

[0155] In some embodiments, the determining unit 702 is specifically used to perform interpolation on the vehicle pose information queue formed by the first vehicle pose information corresponding to multiple different first acquisition times, based on the relationship between the second acquisition time and the multiple different first acquisition times, to obtain the second vehicle pose information corresponding to the second acquisition time.

[0156] In some embodiments, the determining unit 702 is specifically used to determine the difference between the first heading angle and the second heading angle as the heading angle offset from the second acquisition time to the first acquisition time if the second acquisition time is less than the first acquisition time; and to determine the difference between the second heading angle and the first heading angle as the heading angle offset from the first acquisition time to the second acquisition time if the second acquisition time is greater than the first acquisition time.

[0157] In some embodiments, the updating unit 703 is specifically used to update the second lane line coefficient according to the position offset and the heading angle offset if the second acquisition time is less than the first acquisition time; and to update the first lane line coefficient according to the position offset and the heading angle offset if the second acquisition time is greater than the first acquisition time.

[0158] In some embodiments, both the first lane line coefficient and the second lane line coefficient include at least four coefficients, which are the four coefficients of the fitted cubic curve equation corresponding to the lane line. In some embodiments, the updating unit 703 is specifically used to update the first coefficient, including: taking the original value of the first coefficient, the product of the position offset and the second coefficient, the product of the square of the position offset and the third coefficient, the product of the cube of the position offset and the fourth coefficient, and the sum of the negative of the product of the first multiple of the position offset and the heading angle offset as the new value of the first coefficient; wherein, the first coefficient is the coefficient corresponding to the independent variable raised to the power of 0, the second coefficient is the coefficient corresponding to the independent variable raised to the power of 1, the third coefficient is the coefficient corresponding to the independent variable raised to the power of 2, and the fourth coefficient is the coefficient corresponding to the independent variable raised to the power of 3.

[0159] The update of the second coefficient includes: taking the original value of the second coefficient, the product of the second multiple of the position offset and the third coefficient, the product of the square of the third multiple of the position offset and the fourth coefficient, and the sum of the negative of the heading angle offset as the new value of the second coefficient;

[0160] The update of the third coefficient includes: taking the sum of the original value of the third coefficient, the position offset of the fourth multiple, and the fourth coefficient as the new value of the third coefficient;

[0161] For the fourth coefficient, maintain the original value of the fourth coefficient.

[0162] In some implementations, after acquiring multiple sets of first lane line coefficients and second lane line coefficients, the acquisition unit 701 is used to calculate, for any set of first lane line coefficients and second lane line coefficients and any of the N independent variables, the difference between the dependent variable corresponding to the first lane line coefficient and the dependent variable corresponding to the second lane line coefficient under the same independent variable; wherein, different sets correspond to different acquisition times; and the accuracy index of the evaluation is determined based on the N differences corresponding to each set.

[0163] In some embodiments, the acquisition unit 701 is specifically configured to perform one or more of the following: averaging the absolute values ​​of N differences to obtain a set of first error mean values; averaging multiple sets of first error mean values ​​to obtain a target first error mean value; determining the maximum first error mean value from the multiple sets of first error mean values; averaging the N differences to obtain a set of second error mean values; determining the standard deviation based on multiple sets of second error mean values; determining the proportion of the absolute values ​​of the N differences that are less than a first preset threshold to obtain a set of proportions; averaging multiple sets of proportions to obtain a target proportion; wherein the first preset threshold is half the width of the lane line.

[0164] In some embodiments, the acquisition unit 701 is further configured to determine the ratio of the number of valid data in multiple sets of data to be evaluated to the multiple sets of data to be evaluated, wherein valid data refers to a set of data to be evaluated in which the first lane line information includes a validity identifier, and / or the confidence level included in the first lane line information is greater than or equal to a preset confidence threshold; determine the ratio of the number of consecutive valid data in multiple sets of data to be evaluated to the number of valid data, wherein consecutive valid data refers to the difference between the frame ID of the current valid data to be evaluated and the frame ID of the previous set of valid data to be evaluated being less than or equal to a second preset threshold; and based on multiple sets of data to be evaluated, determine the maximum end position of the lane line and / or the average length of the lane line, wherein the first lane line information includes the start position / end position of the lane line.

[0165] In some embodiments, the first lane line coefficient includes at least four coefficients, which are the four coefficients of the fitted cubic curve equation corresponding to the lane line; the acquisition unit 701 is further configured to perform one or more of the following: acquire the standard deviation and average value of multiple sets of the first coefficient; determine the dispersion corresponding to the first coefficient based on the standard deviation and the average value; the first coefficient is a coefficient of the independent variable raised to the power of 0; acquire the standard deviation of multiple sets of other coefficients, which are coefficients other than the first coefficient among the at least four coefficients; for any of the other coefficients, determine the extreme value difference between the maximum and minimum values ​​of the coefficient in multiple sets; determine the standard deviation and average value of the lane line length based on multiple sets of data to be evaluated; determine the dispersion of the lane line length based on the standard deviation and the average value.

[0166] In some embodiments, the acquisition unit 701 is further configured to construct an evaluation mechanism based on at least one of accuracy indicators, stability indicators, and effectiveness indicators. The evaluation mechanism includes a scoring standard corresponding to at least one of the accuracy indicators, stability indicators, and effectiveness indicators, and a weight corresponding to that indicator. Based on the evaluation mechanism and at least one of the accuracy indicators, stability indicators, and effectiveness indicators corresponding to the data to be evaluated, the device that collects the data to be evaluated is used to quantitatively evaluate the ability of the device to identify lane lines.

[0167] It should be noted that the specific implementation of each unit in this embodiment can be found in the relevant descriptions in the above method embodiments.

[0168] This application provides a computer-readable storage medium, including instructions or a computer program, which, when run on a computer, causes the computer to perform the lane line data processing method described above.

[0169] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

[0170] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0171] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0172] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0173] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A lane line data processing method, characterized by, The method includes: Acquire the data to be evaluated, which includes first vehicle information and first lane information at a first acquisition time. The first vehicle information includes the first speed and first yaw rate of the vehicle, and the first lane information includes a first lane coefficient. Obtain the true value data closest to the first acquisition time, the true value data including the second lane line information at the second acquisition time, the second lane line information including the second lane line coefficient; Based on the first vehicle information, determine the vehicle's position offset and heading angle offset from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time; The first lane line coefficient or the second lane line coefficient is updated based on the first acquisition time corresponding to the data to be evaluated, the second acquisition time corresponding to the true data, the position offset, and the heading angle offset, including: If the second acquisition time is less than the first acquisition time, the second lane line coefficient is updated according to the position offset and the heading angle offset; If the second acquisition time is greater than the first acquisition time, the first lane line coefficient is updated according to the position offset and the heading angle offset.

2. The method of claim 1, wherein, The step of determining the position offset and heading angle offset of the vehicle from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time based on the first vehicle information includes: The vehicle pose information in the target coordinate system is determined based on the first vehicle speed and the first yaw rate. The first vehicle pose information includes the first coordinate position and the first heading angle. Based on the first vehicle pose information, the second vehicle pose information of the vehicle at the second acquisition time is determined, and the second vehicle pose information includes a second coordinate position and a second heading angle; Determine the position offset from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time based on the first coordinate position and the second coordinate position; The heading angle offset from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time is determined based on the first heading angle and the second heading angle.

3. The method of claim 2, wherein, The second vehicle pose information at the second acquisition time is determined based on the first vehicle pose information. The second vehicle pose information includes a second coordinate position and a second heading angle, including: For a vehicle pose information queue formed by multiple different first acquisition times corresponding to the first vehicle pose information, interpolation is performed on the vehicle pose information queue based on the relationship between the second acquisition time and the multiple different first acquisition times to obtain the second vehicle pose information corresponding to the second acquisition time.

4. The method according to claim 2 or 3, characterized in that, The step of determining the heading angle offset from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time based on the first heading angle and the second heading angle includes: If the second acquisition time is less than the first acquisition time, the difference between the first heading angle and the second heading angle is determined as the heading angle offset from the second acquisition time to the first acquisition time; If the second acquisition time is greater than the first acquisition time, the difference between the second heading angle and the first heading angle is determined as the heading angle offset from the first acquisition time to the second acquisition time.

5. The method of claim 1, wherein, Both the first lane line coefficient and the second lane line coefficient include at least four coefficients, which are the four coefficients of the fitted cubic curve equation corresponding to the lane line; updating the first lane line coefficient or the second lane line coefficient according to the position offset and the heading angle offset includes: The update of the first coefficient includes: The new value of the first coefficient is the sum of the original value of the first coefficient, the product of the position offset and the second coefficient, the product of the square of the position offset and the third coefficient, the product of the cube of the position offset and the fourth coefficient, and the opposite of the product of the first multiple of the position offset and the heading angle offset. Wherein, the first coefficient is the coefficient corresponding to the independent variable being raised to the power of 0, the second coefficient is the coefficient corresponding to the independent variable being raised to the power of 1, the third coefficient is the coefficient corresponding to the independent variable being raised to the power of 2, and the fourth coefficient is the coefficient corresponding to the independent variable being raised to the power of 3. The update of the second coefficient includes: The new value of the second coefficient is the sum of the original value of the second coefficient, the product of the second multiple of the position offset and the third coefficient, the product of the square of the third multiple of the position offset and the fourth coefficient, and the negative of the heading angle offset. The update of the third coefficient includes: The new value of the third coefficient is the sum of the original value of the third coefficient, the position offset of the fourth multiple, and the fourth coefficient. For the fourth coefficient, maintain the original value of the fourth coefficient.

6. The method of claim 1, wherein, After obtaining multiple sets of first lane line coefficients and second lane line coefficients, the method further includes: For any group of first lane line coefficients and second lane line coefficients, and any independent variable among N independent variables, calculate the difference between the dependent variable corresponding to the first lane line coefficient and the dependent variable corresponding to the second lane line coefficient under the same independent variable; where different groups correspond to different acquisition times. The accuracy index of the evaluation is determined based on the N differences corresponding to each group.

7. The method of claim 6, wherein, The accuracy index for evaluation, determined based on the N differences corresponding to each group, includes one or more of the following: The average of the absolute values ​​of the N differences is used to obtain a first set of error mean values. The target first error mean is obtained by averaging multiple sets of first error mean values; Determine the maximum mean first error from the plurality of mean first error values; Take the average of the N differences to obtain a second set of error mean values; The standard deviation is determined based on the mean of multiple sets of second errors; Determine the proportion of the absolute values ​​of the N differences that are less than a first preset threshold to obtain a set of proportions; The target percentage is obtained by averaging the percentages from multiple groups. The first preset threshold is half the width of the lane line.

8. The method according to claim 6 or 7, characterized in that, The method also includes one or more of the following: Determine the ratio of the number of valid data in multiple sets of data to be evaluated to the multiple sets of data to be evaluated. Valid data refers to a set of data to be evaluated in which the first lane line information includes a validity identifier, and / or the confidence level included in the first lane line information is greater than or equal to a preset confidence threshold. Determine the ratio of the number of consecutive valid data to the number of valid data in multiple sets of data to be evaluated. The consecutive valid data refers to the difference between the frame ID of the current valid data to be evaluated and the frame ID of the previous set of valid data to be evaluated, which is less than or equal to a second preset threshold. Based on multiple sets of evaluation data, the maximum end position of the lane line and / or the average length of the lane line are determined. The first lane line information includes the start position / end position of the lane line.

9. The method according to claim 8, characterized in that, The first lane line coefficient includes at least four coefficients, which are the four coefficients of the fitted cubic curve equation corresponding to the lane line; the method further includes one or more of the following: Obtain the standard deviation and mean of multiple sets of first coefficients; determine the dispersion corresponding to the first coefficient based on the standard deviation and the mean; the first coefficient is the coefficient of the independent variable raised to the power of 0; Obtain the standard deviations corresponding to multiple sets of other coefficients, wherein the other coefficients are coefficients other than the first coefficient among the at least four coefficients; For any of the other coefficients, determine the difference between the maximum and minimum values ​​of that coefficient across multiple groups; Based on multiple sets of data to be evaluated, the standard deviation and average value of the lane line length are determined; based on the standard deviation and the average value, the dispersion of the lane line length is determined.

10. The method according to any one of claims 6, 7, and 9, characterized in that, The method further includes: An evaluation mechanism is constructed based on at least one of the accuracy, stability, and effectiveness indicators. The evaluation mechanism includes a scoring standard and a weight corresponding to at least one of the accuracy, stability, and effectiveness indicators. Based on the evaluation mechanism and at least one of the accuracy, stability, and effectiveness indicators corresponding to the data to be evaluated, the ability of the device collecting the data to be evaluated to identify lane lines is quantitatively evaluated.

11. A lane line data processing device, characterized in that, The device includes: An acquisition unit is used to acquire data to be evaluated, the data to be evaluated including first vehicle information and first lane information at a first acquisition time, the first vehicle information including the first speed and first yaw rate of the vehicle, and the first lane information including a first lane coefficient. The acquisition unit is further configured to acquire the true value data closest to the first acquisition time, the true value data including the second lane line information at the second acquisition time, the second lane line information including the second lane line coefficient; The determining unit is configured to determine, based on the first vehicle information, the position offset and heading angle offset of the vehicle from the first acquisition time to the second acquisition time or from the second acquisition time to the first acquisition time; An update unit is configured to update the second lane line coefficient based on the position offset and the heading angle offset if the second acquisition time is less than the first acquisition time; and to update the first lane line coefficient based on the position offset and the heading angle offset if the second acquisition time is greater than the first acquisition time.

Images