Performance evaluation method and device of road side algorithm, electronic equipment and storage medium

Abstract

The invention discloses a performance evaluation method and device of a road side algorithm, electronic equipment and a storage medium, and belongs to the technical field of vehicle-road cooperation. The method comprises the following steps: acquiring a track sequence output by a roadside algorithm, wherein each track point in the track sequence comprises position information and kinematics characteristics; performing mutation detection on the kinematic characteristics of adjacent track points in the track sequence to obtain a mutation detection result; according to the position information and the kinematics characteristics, performing anomaly detection on each track point in the track sequence to obtain an anomaly detection result; and determining a performance evaluation result of the road side algorithm according to the mutation detection result and the anomaly detection result. According to the embodiment of the invention, the performance of the roadside algorithm can be comprehensively evaluated, the limitation of single-index evaluation is avoided, and the comprehensive performance of the roadside algorithm in a complex traffic environment can be comprehensively reflected.

Description

Technical Field

[0001] This application belongs to the field of vehicle-road cooperative technology, specifically relating to a performance evaluation method, device, electronic device and storage medium for roadside algorithms. Background Technology

[0002] With the rapid development of autonomous driving and intelligent transportation systems, the need for performance evaluation of roadside algorithms is becoming increasingly urgent.

[0003] In existing technologies, performance evaluation methods for path estimation algorithms mainly focus on verifying a single metric. For example, positioning accuracy is evaluated by comparing and verifying with precision positioning equipment such as RTK; target detection accuracy is verified by manually annotating the recall and precision of the detection algorithm; and trajectory smoothness is simply evaluated based on the continuity of trajectory points.

[0004] Most existing evaluation methods focus on a single technical indicator, resulting in a single evaluation dimension that cannot fully reflect the comprehensive performance of roadside algorithms in complex traffic environments. Summary of the Invention

[0005] The purpose of this application is to provide a method, apparatus, electronic device, and storage medium for evaluating the performance of roadside algorithms, which can solve the problem that the evaluation dimension is too singular and cannot fully reflect the comprehensive performance of roadside algorithms in complex traffic environments.

[0006] To solve the above-mentioned technical problems, this application is implemented as follows: In a first aspect, embodiments of this application provide a performance evaluation method for roadside algorithms, the method comprising: Obtain the trajectory sequence output by the roadside algorithm, wherein each trajectory point in the trajectory sequence includes location information and kinematic features; Abrupt changes are detected in the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results. Based on the location information and the kinematic features, anomaly detection is performed on each trajectory point in the trajectory sequence to obtain anomaly detection results; Based on the mutation detection results and the anomaly detection results, the performance evaluation results of the roadside algorithm are determined.

[0007] Secondly, embodiments of this application provide a performance evaluation device for roadside algorithms, comprising: The trajectory acquisition module is used to acquire the trajectory sequence output by the roadside algorithm, wherein each trajectory point in the trajectory sequence includes location information and kinematic features; The mutation detection module is used to perform mutation detection on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain mutation detection results. An anomaly detection module is used to perform anomaly detection on each trajectory point in the trajectory sequence based on the location information and the kinematic features, and obtain anomaly detection results; The evaluation result determination module is used to determine the performance evaluation result of the roadside algorithm based on the mutation detection result and the anomaly detection result.

[0008] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0009] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0010] Fifthly, embodiments of this application provide a chip, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.

[0011] In this embodiment, the trajectory sequence output by the roadside algorithm is obtained. Each trajectory point in the trajectory sequence includes location information and kinematic features. Abrupt change detection is performed on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results. Based on the location information and kinematic features, anomaly detection is performed on each trajectory point in the trajectory sequence to obtain anomaly detection results. Based on the abrupt change detection results and the anomaly detection results, the performance evaluation result of the roadside algorithm is determined. Since abrupt change detection and anomaly detection of trajectory points can be performed simultaneously during the performance evaluation process, it is no longer limited to a single technical indicator. Through multi-dimensional feature analysis, the performance of the roadside algorithm can be comprehensively evaluated, avoiding the limitations of single indicator evaluation and comprehensively reflecting the overall performance of the roadside algorithm in complex traffic environments. Attached Figure Description

[0012] Figure 1 This is a flowchart of a performance evaluation method for a roadside algorithm provided in an embodiment of this application; Figure 2 This is an example diagram of the first curve in the embodiments of this application; Figure 3 This is an example diagram of the second curve in the embodiments of this application; Figure 4 This is an example diagram of the third curve in the embodiments of this application; Figure 5 This is an example diagram showing the marking results of burr points and breakpoints in the embodiments of this application; Figure 6 This is a flowchart of another performance evaluation method for roadside algorithms provided in the embodiments of this application; Figure 7 This is an example diagram showing the analysis results of the initial heading angle in an embodiment of this application; Figure 8 This is an overall framework diagram of the performance evaluation method for roadside algorithms provided in the embodiments of this application; Figure 9 This is a display example diagram of the tool GUI interface in the embodiments of this application; Figure 10 This is a schematic diagram of the structure of a performance evaluation device for a roadside algorithm provided in an embodiment of this application; Figure 11 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of this application. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0015] The performance evaluation method of the roadside algorithm provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0016] Figure 1 This is a flowchart illustrating a performance evaluation method for roadside algorithms provided in an embodiment of this application. This method can be applied to scenarios involving performance evaluation of roadside algorithms and can be executed by electronic devices such as computers. Figure 1 As shown, the performance evaluation method of this roadside algorithm may include steps 110 to 140.

[0017] Step 110: Obtain the trajectory sequence output by the roadside algorithm. Each trajectory point in the trajectory sequence includes location information and kinematic features.

[0018] Among them, roadside algorithms are algorithms used in roadside perception systems to predict vehicle position, heading, speed, etc. Kinematic features are the dynamic characteristics exhibited by a vehicle during its movement, which can include speed, acceleration, heading, etc.

[0019] A CSV (Comma Separated Values) file can be obtained through the user interface. This CSV file records the trajectory sequences output by the roadside algorithm. The user interface can include a file selection area, which includes CSV file selection controls and save path selection controls. When the user clicks the CSV file selection control, the storage directory can be displayed for the user to select the corresponding CSV file. When the user clicks the save path selection control, the user can select the save path for the performance evaluation results. CSV is a file format that stores tabular data in plain text, with fields separated by commas, and each record corresponding to a row in the table. The CSV file can include trajectory sequences from multiple vehicles, and each trajectory sequence can be distinguished by a trajectory identifier (which can be a vehicle identifier or other identifier that can differentiate between different trajectories).

[0020] After obtaining the CSV file, it can be parsed to obtain an initial trajectory sequence set. Preprocessing operations are then performed on each initial trajectory in the set, specifically determining the data volume of each initial trajectory sequence. Initial trajectory sequences with a data volume greater than or equal to the target data volume (e.g., 6 trajectory points) are used as the final trajectory sequence set. Each trajectory sequence in the final set is then evaluated according to steps 110 to 140 in the embodiments of this application. Finally, the performance evaluation results of each trajectory sequence are combined to obtain the final performance evaluation result of the roadside algorithm.

[0021] Step 120: Perform abrupt change detection on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results.

[0022] Abrupt changes can be detected in the kinematic features of adjacent trajectory points in a trajectory sequence based on a pre-obtained threshold. The result of the abrupt change detection between every two adjacent trajectory points in the trajectory sequence can be used as the result of the abrupt change detection for the trajectory sequence.

[0023] Step 130: Based on the location information and the kinematic features, perform anomaly detection on each trajectory point in the trajectory sequence to obtain anomaly detection results.

[0024] Anomaly detection can include the detection of breakpoints and burrs. A breakpoint indicates that one or more trajectory points are missing in the trajectory sequence. A burr indicates that the trajectory sequence is not smooth at that trajectory point.

[0025] Anomaly detection is performed on each trajectory point in the trajectory sequence, which is called trajectory quality detection. For roadside algorithms, trajectory quality detection mainly involves analyzing whether there are spikes or breaks in the process of each vehicle entering and exiting the intersection (i.e., each trajectory sequence). Based on this, the predictive performance of the roadside algorithm on the vehicle's driving status is expressed.

[0026] Step 140: Determine the performance evaluation result of the roadside algorithm based on the mutation detection result and the anomaly detection result.

[0027] The results of mutation detection and anomaly detection can be summarized as the final performance evaluation results of the roadside algorithm.

[0028] The performance evaluation method for roadside algorithms provided in this application involves acquiring the trajectory sequence output by the roadside algorithm. Each trajectory point in the trajectory sequence includes location information and kinematic features. Abrupt change detection is performed on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results. Based on the location information and kinematic features, anomaly detection is performed on each trajectory point in the trajectory sequence to obtain anomaly detection results. Based on the abrupt change detection results and the anomaly detection results, the performance evaluation result of the roadside algorithm is determined. Since abrupt change detection and anomaly detection of trajectory points can be performed simultaneously during the performance evaluation process, it is no longer limited to a single technical indicator. Multi-dimensional feature analysis can be used to comprehensively evaluate the performance of the roadside algorithm, avoiding the limitations of single indicator evaluation and comprehensively reflecting the overall performance of the roadside algorithm in complex traffic environments.

[0029] In some embodiments of this application, the step of performing abrupt change detection on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results may include: determining the change value of the kinematic features of the adjacent trajectory points; if the absolute value of the change value is greater than or equal to a change threshold, determining that the next trajectory point among the adjacent trajectory points has a kinematic abrupt change, and recording the abrupt change type and the change value of the next trajectory point to obtain the abrupt change detection results.

[0030] The kinematic characteristics may include at least one of velocity, acceleration, and heading angle, and the abrupt change type may include at least one of velocity abrupt change, acceleration abrupt change, and heading angle abrupt change. The change threshold may be a pre-set fixed threshold, or it may be a threshold set by the user through the operation interface during this performance evaluation.

[0031] For each kinematic feature, the change value of adjacent trajectory points with respect to that kinematic feature can be determined, that is, the change value of the kinematic feature of the next trajectory point relative to the previous trajectory point. The absolute value of the change value is compared with the change threshold. If the absolute value of the change value is greater than or equal to the change threshold, it is determined that there is a kinematic mutation in the next trajectory point. The mutation type is determined to be the type of the kinematic feature, and the mutation type and change value are recorded as the mutation detection result.

[0032] For example, for velocity, the velocity change value of adjacent trajectory points can be determined, that is, the velocity change value of the next trajectory point relative to the velocity of the previous trajectory point; the absolute value of the velocity change value is compared with the velocity change threshold. If the absolute value of the velocity change value is greater than or equal to the velocity change threshold, it is determined that there is a velocity mutation in the next trajectory point, the mutation type is determined to be a velocity mutation, and the mutation type and change value are recorded as the mutation detection result.

[0033] For example, for the heading angle, the heading angle change value of adjacent trajectory points can be determined, that is, the change value of the heading angle of the next trajectory point relative to the heading angle of the previous trajectory point. The absolute value of the heading angle change value is compared with the heading angle change threshold. If the absolute value of the heading angle change value is greater than or equal to the heading angle change threshold, it is determined that there is a heading angle abrupt change in the next trajectory point. The type of abrupt change is determined to be a heading angle abrupt change, and the type of abrupt change and the change value are recorded as the abrupt change detection result.

[0034] For example, for acceleration, the acceleration change value of adjacent trajectory points can be determined, that is, the change value of the acceleration of the next trajectory point relative to the acceleration of the previous trajectory point. The absolute value of the acceleration change value is compared with the acceleration change threshold. If the absolute value of the acceleration change value is greater than or equal to the acceleration change threshold, it is determined that there is an acceleration abrupt change in the next trajectory point. The type of abrupt change is determined to be an acceleration abrupt change, and the type of abrupt change and the change value are recorded as the abrupt change detection result.

[0035] In some embodiments, the change threshold is obtained through a settings interface. This settings interface can be a separate threshold settings interface, or it can be the threshold settings area within the aforementioned operation interface. The settings interface may include: a custom settings area for velocity, heading angle, and acceleration abrupt change thresholds; and a custom settings area for trajectory parameters (used for subsequent anomaly detection of trajectory points), including curvature threshold, velocity change threshold, angle change threshold (degrees), cluster radius, and minimum sample number.

[0036] In some embodiments, a sudden change in vehicle type can also be detected for adjacent trajectory points. The vehicle type is the type of vehicle determined by the roadside algorithm, such as a car or a truck. If the vehicle types of adjacent trajectory points are different, it is determined that there is a sudden change in vehicle type between the adjacent trajectory points.

[0037] Table 1 shows an example of mutation detection results. As shown in Table 1, the mutation detection results are for the trajectory sequence with ID 1234. The mutation detection results can include the mutation type and the difference between mutations (i.e., the change value). The mutation detection results for each trajectory sequence can also be summarized and statistically analyzed to determine the velocity mutation rate, heading angle mutation rate, and acceleration mutation rate in the trajectory sequence.

[0038] Table 1 Examples of mutation detection results

[0039] Based on the above technical solution, the method may further include at least one of the following: Generate a first graph showing the change of the speed over time; Generate a second graph showing the acceleration changing over time; A third curve showing the change of the heading angle over time is generated. Before generating the third curve, if the change in the initial heading angle of a trajectory point relative to the heading angle of the previous trajectory point is greater than 180 degrees, the initial heading angle of the trajectory point is subtracted by 360 degrees to obtain the final heading angle of the trajectory point. If the change in the initial heading angle of the trajectory point relative to the heading angle of the previous trajectory point is less than -180 degrees, the initial heading angle of the trajectory point is added by 360 degrees to obtain the final heading angle of the trajectory point.

[0040] In some embodiments, a first curve of velocity change over time in the trajectory sequence is generated based on the velocity of each trajectory point in the trajectory sequence. Figure 2 This is an example diagram of the first curve in the embodiments of this application, such as... Figure 2 As shown, different colors represent the first curve of different trajectory sequences, that is, the first curve of the trajectory of different vehicles. The trajectory sequence for which the first curve needs to be plotted can be specified by the user. The user can select one or more IDs from the ID list for speed visualization analysis. Based on the ID selected by the user, the first curve of the trajectory sequence corresponding to the corresponding ID is generated.

[0041] In some embodiments, a first curve of acceleration changing over time in the trajectory sequence is generated based on the acceleration of each trajectory point in the trajectory sequence. Figure 3 This is an example diagram of the second curve in the embodiments of this application, such as... Figure 3As shown, different colors represent different trajectory sequences in the second curve, which are the second curves of the trajectories of different vehicles. The trajectory sequence for which the second curve needs to be plotted can be specified by the user. The user can select one or more IDs from the ID list for speed visualization analysis. Based on the ID selected by the user, the second curve of the trajectory sequence corresponding to the corresponding ID is generated. Figure 3 The dashed lines in the figure represent acceleration change thresholds, which may include upper and lower thresholds.

[0042] In some embodiments, in the third curve plot of the heading angle changing over time, in order to solve the problem that the heading angle may be misinterpreted as a change of -358° due to the 0°-360° cyclic representation, and that there may be abrupt changes in the curve plot, the embodiments of this application can solve this problem in the following way: ① When the calculated change in adjacent heading angles is greater than 180°, it indicates that the 360° / 0° boundary may have been crossed. The continuity of heading angles can be restored by adding or subtracting 360°. ② If the change in adjacent heading angles is greater than 180° or less than 180°, the subsequent heading angles are calculated through intelligent processing, which effectively reduces the situation where the vehicle's large-angle turn is mistakenly judged as a sudden change. For example, the initial heading angle [358°, 1°, 4°, 7°] is processed to obtain the final heading angle [358°, 361°, 364°, 367°]. ③ Use a dual representation in the third curve of the heading angle changing over time to avoid visual jumps caused by the 0° / 360° cycle characteristic.

[0043] The above process can be represented by the following formula:

[0044] in, This represents the final heading angle of the i-th trajectory point. This represents the initial heading angle of the i-th trajectory point.

[0045] Figure 4 This is an example diagram of the third curve in the embodiments of this application, such as... Figure 4 As shown, different colors represent the third curves of different trajectory sequences, that is, the third curves of the trajectories of different vehicles. The trajectory sequences for which the third curves need to be plotted can be specified by the user. The user can select one or more IDs from the ID list for speed visualization analysis. Based on the IDs selected by the user, the third curve of the trajectory sequence corresponding to the corresponding ID is generated.

[0046] By generating the first curve, the second curve, and the third curve, a visual analysis of velocity, acceleration, and heading angle in the trajectory sequence was achieved.

[0047] In some embodiments of this application, the step of performing anomaly detection on each trajectory point in the trajectory sequence based on the position information and the kinematic features to obtain anomaly detection results includes: determining spur points in the trajectory sequence based on the position information and the kinematic features; determining breakpoints in the trajectory sequence based on the position information and the velocity in the kinematic features; determining smoothness anomalies in the trajectory sequence based on a sliding window; and determining the anomaly detection results based on the spur points, the breakpoints, and the smoothness anomalies.

[0048] The burrs can include at least one of curvature burrs, heading angle burrs, and spatial distribution anomalies. Spatial distribution anomalies refer to trajectory points that are essentially unrelated to other trajectory points.

[0049] Based on the positional information and kinematic characteristics of each trajectory point in the trajectory sequence, trajectory points with prominent curvature, heading angles, etc., relative to adjacent trajectory points are identified as spurs in the trajectory sequence. Based on the positional information and velocity in the kinematic characteristics, it is determined whether there are breakpoints among adjacent trajectory points. If a breakpoint exists, the adjacent trajectory point or the next adjacent trajectory point can be marked as a breakpoint. A sliding window is a window with a target time length that slides within the trajectory sequence to statistically analyze smoothness anomalies, i.e., to count whether there are breakpoints and spurs within the sliding window. If breakpoints and spurs exist within the sliding window, all trajectory points within the sliding window can be identified as smoothness anomalies, indicating that a segment of the trajectory within the sliding window is not smooth. Spurs, breakpoints, and smoothness anomalies can be identified as anomaly detection results, or the anomaly detection results can also include statistical data (such as percentages) of spurs, breakpoints, and smoothness anomalies.

[0050] By identifying burr points, breakpoints, and smoothness anomalies in the trajectory sequence, and thus determining the anomaly detection results, different types of anomalies in the trajectory sequence can be detected. This can comprehensively reflect the abnormal trajectory points in the trajectory sequence and improve the accuracy of anomaly detection.

[0051] In some embodiments of this application, determining the spur points in the trajectory sequence based on the location information and the kinematic features includes at least one of the following: Based on the location information, the curvature of each trajectory point is determined, and the trajectory points whose curvature satisfies the target condition are identified as the burr points. The trajectory points in the kinematic features where the change in heading angle is greater than the heading angle change threshold are identified as the spur points; The DBSCAN clustering algorithm is used to cluster the trajectory points in the trajectory sequence, and the isolated trajectory points in the clustering results are identified as the spur points.

[0052] The target condition may include: the curvature of the trajectory point is greater than or equal to a curvature threshold, and the curvature of both the preceding and following trajectory points of the trajectory point is greater than or equal to the curvature threshold in a target ratio. The preceding trajectory point is the adjacent trajectory point located before it. The following trajectory point is the adjacent trajectory point located after it.

[0053] Before identifying spurs in the trajectory sequence, multidimensional feature extraction can be performed first. These features can include the distance between adjacent trajectory points, the velocity of the trajectory points, the curvature of the corresponding trajectory points, and the change in heading angle. The distance between adjacent points can be calculated using the Haversine method. The velocity of the trajectory points can be directly obtained from the trajectory sequence. The curvature of the corresponding trajectory point can be calculated using three adjacent trajectory points; that is, when calculating the curvature of the current trajectory point, it can be based on the current trajectory point, the previous trajectory point, and the next trajectory point. The calculation formula is as follows:

[0054] in, k Let A represent the curvature of the current trajectory point, and let A represent the area of ​​the triangle formed by three adjacent trajectory points. a , b , c These represent the three side lengths of the triangle formed by three adjacent trajectory points.

[0055] The change in heading angle can be calculated using the following formula:

[0056] in, This represents the change in the heading angle of the i-th trajectory point relative to the previous trajectory point (the (i-1)-th trajectory point). This represents the position coordinates of the i-th trajectory point. This represents the position coordinates of the (i-1)th trajectory point (i.e., the trajectory point preceding the ith trajectory point). This represents the position coordinates of the (i+1)th trajectory point (i.e., the trajectory point following the i-th trajectory point).

[0057] Based on the positional information of each trajectory point in the trajectory sequence, the curvature of each trajectory point can be calculated according to the curvature calculation formula mentioned above. The trajectory points with a curvature greater than or equal to a threshold, and whose curvature of both the preceding and following trajectory points is greater than or equal to the target curvature threshold, are identified as curvature spurs. This can be expressed by the following formula:

[0058] in, This represents the set of curvature burr points. A result of 1 indicates that the i-th trajectory point is marked as a curvature burr point, and a result of 0 indicates that the i-th trajectory point is not marked as a curvature burr point. Let represent the curvature of the i-th trajectory point. Indicates the curvature threshold. This indicates the target ratio, for example, it could be 0.5.

[0059] The trajectory points where the change in heading angle exceeds the heading angle change threshold are defined as heading angle spurs, which can be expressed by the following formula:

[0060] in, This represents the set of heading angle spikes. A result of 1 indicates that the i-th trajectory point is marked as a heading angle spike, and a result of 0 indicates that the i-th trajectory point is not marked as a heading angle spike. This indicates the change in heading angle. This indicates the threshold for the change in heading angle.

[0061] Clustering spikes are points that represent spatial anomalies. Based on the DBSCAN clustering algorithm, each trajectory point in the trajectory sequence is clustered, and a cluster label is determined for each trajectory point. If a trajectory point is an isolated point that is not clustered with other trajectory points after clustering, its cluster label is -1, indicating that the trajectory point is basically unrelated to other trajectory points, representing an extreme case. Clustering spikes are trajectory points with a cluster label of -1, which can be expressed by the following formula:

[0062]

[0063] This represents the set of clustered spurs. This represents the cluster label of the i-th trajectory point. Let P represent the cluster radius, P represent the set of all trajectory points in the trajectory sequence, and minPts represent the minimum number of samples.

[0064] Finally, the union of curvature burr points, heading angle burr points, and clustered burr points is determined as the final set of burr points:

[0065] in, This represents the final set of burr points. Indicates the curvature burr point. Indicates the heading angle burr point, This represents cluster burr points.

[0066] By determining the spur points in the trajectory sequence based on curvature, heading angle, and clustering respectively, multiple detection strategies can be used in tandem to significantly improve the ability to identify complex anomaly patterns.

[0067] In some embodiments of this application, determining the breakpoint in the trajectory sequence based on the position information and the velocity in the kinematic features may include: determining the trajectory point whose velocity is greater than or equal to the velocity of the previous trajectory point and whose distance from the previous trajectory point is greater than or equal to a distance threshold as the breakpoint.

[0068] For each trajectory point, determine the ratio between the velocity of the trajectory point and the velocity of the previous trajectory point, compare the ratio with a velocity ratio threshold, and compare the distance between the trajectory point and the previous trajectory point with a distance threshold. If the ratio is greater than or equal to the velocity ratio threshold and the distance is greater than or equal to the distance threshold, then the trajectory point is determined as a breakpoint.

[0069] By determining whether the current trajectory point is a breakpoint based on the ratio of the current trajectory point's velocity to the previous trajectory point's velocity and the distance between the current trajectory point and the previous trajectory point, the accuracy of breakpoint determination can be improved.

[0070] In some embodiments of this application, determining the anomaly detection result based on the burr points, the breakpoints, and the smoothness anomalies may include: filtering out repeated trajectory points among the burr points, the breakpoints, and the smoothness anomalies, and determining the anomaly detection result based on the filtered burr points, breakpoints, and smoothness anomalies.

[0071] Since burr points, breakpoints, and smoothness anomalies are determined using different methods, duplicate anomalies may exist. Duplicate trajectory points among these anomalies can be filtered out, retaining only one. The filtered burr points, breakpoints, and smoothness anomalies are then used as the anomaly detection results. When filtering duplicate trajectory points, a minimum interval can be set. If the distance between two trajectory points is less than the minimum interval, one of the trajectory points can be retained. This filters out overly dense anomaly trajectory points.

[0072] By filtering out duplicate trajectory points among burr points, breakpoints, and smoothness anomalies, duplicate anomalies can be avoided in the final anomaly detection results.

[0073] In some embodiments of this application, determining the anomaly detection result based on the filtered burr points, breakpoints, and smoothness anomalies may include: counting the number of burr points, the number of breakpoints, the burr rate, and the breakpoint rate from the filtered burr points, breakpoints, and smoothness anomalies, wherein the burr rate is the proportion of burr points among the filtered burr points, breakpoints, and smoothness anomalies to all trajectory points, and the breakpoint rate is the proportion of breakpoints among the filtered burr points, breakpoints, and smoothness anomalies to all trajectory points; and using the number of burr points, the number of breakpoints, the burr rate, and the breakpoint rate as the anomaly detection result.

[0074] From the filtered burr points, breakpoints, and smoothness anomalies, the number of burr points and breakpoints is counted. The ratio of the number of burr points to the total number of trajectory points in the trajectory sequence is determined as the burr rate, and the ratio of the number of breakpoints to the total number of trajectory points in the trajectory sequence is determined as the breakpoint rate. The number of burr points, the number of breakpoints, the burr rate, and the breakpoint rate are used as anomaly detection results. Based on the statistical results, i.e., the number of burr points, the number of breakpoints, the burr rate, and the breakpoint rate, a trajectory quality quantification report can be generated. Table 2 is an example of the trajectory quality quantification report in this application embodiment. As shown in Table 2, for each trajectory sequence (each ID in Table 2 represents a trajectory sequence), the total number of points (total number of trajectory points), the number of burr points, the number of breakpoints, the burr rate, the breakpoint rate, and the quality score are displayed. The quality score is the integer 1 minus the burr rate and the breakpoint rate. The results of each trajectory sequence can also be summarized and statistically analyzed to determine the total number of points, the number of burr points, the number of breakpoints, the burr rate, the breakpoint rate, and the quality score for all trajectory sequences.

[0075] Table 2 Track Quality Quantification Report

[0076] By statistically analyzing the filtered burr points, breakpoints, and smoothness anomalies, the obtained statistical results can accurately reflect the anomalies of abnormal trajectory points in the trajectory sequence.

[0077] Based on the above technical solution, after determining the anomaly detection result according to the burr point, the breakpoint and the smoothness anomaly point, it may further include: generating a visualization result of the trajectory sequence according to the anomaly detection result.

[0078] The visualization results can include the trajectory quality quantification report mentioned above, and also a plot showing the marking of spurs and breakpoints. The plot showing the marking of spurs and breakpoints is created by marking spurs and breakpoints in latitude and longitude coordinates for each trajectory sequence, such as... Figure 5 As shown, burrs are marked in blue, and breakpoints are marked in red.

[0079] Visualized results of trajectory sequences are generated based on anomaly detection findings, making them easy for users to view.

[0080] This application embodiment performs trajectory quality detection, specifically trajectory point anomaly detection, by combining DBSCAN clustering, kinematic features, and trajectory smoothness. Its main advantages are as follows: ① The key to this method lies in the creative synergy and fusion of three anomaly detection algorithms based on different principles, constructing a multi-level, cross-validated trajectory quality assessment system. This is no longer a simple parallel use of multiple methods, but an organic whole; ② Multi-dimensional feature cross-validation: Traditional detection methods may rely on only one or two features (e.g., only checking for velocity abrupt changes or curvature). This application embodiment simultaneously considers geometric attributes (curvature), kinematic attributes (velocity, acceleration, heading angle), and spatial distribution attributes (DBSCAN clustering). Detection is performed from three different dimensions, greatly avoiding missed detections; ③ It solves different types of anomalies: Kinematic and geometric features (curvature spur formula and heading angle spur formula) are good at detecting local and drastic physical inconsistencies, such as sharp turns (curvature spurs), sudden turns (heading angle spurs), and rapid acceleration / deceleration (velocity breakpoints); DBSCAN clustering is an unsupervised learning method whose core advantage lies in finding "outliers". It can capture those outliers that may not be significant in kinematics, but are significantly deviated from the main trajectory in spatial distribution. These anomalies are difficult to detect by physical detection methods based on thresholds alone; Decision-level fusion, the final set of spur points is the union (∪) of the results of the three methods. This means that as long as it is judged as an anomaly by any method, it will be captured by the system. It aims to maximize the recall rate of detection, ensure that no potential trajectory quality problems are missed, and can also distinguish between spur points and breakpoints.

[0081] Figure 6 This is a flowchart of another roadside algorithm performance evaluation method provided in this application embodiment. Based on the above embodiments, this embodiment can also perform initial heading angle analysis. For example... Figure 6 As shown, the performance evaluation method of the roadside algorithm may include steps 610 to 660.

[0082] Step 610: Obtain the trajectory sequence output by the roadside algorithm, wherein each trajectory point in the trajectory sequence includes location information and kinematic features.

[0083] Step 620: Perform abrupt change detection on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results.

[0084] Step 630: Based on the location information and the kinematic features, perform anomaly detection on each trajectory point in the trajectory sequence to obtain anomaly detection results.

[0085] Step 640: Obtain the reference heading of the lane where the first trajectory point in the trajectory sequence is located.

[0086] For vehicles that have just entered the detection range of the intersection, the detection and prediction of the heading angle in the first frame (when the vehicle is in the intersection entrance lane) is crucial, as it directly affects the accuracy of subsequent tracking of changes in the vehicle's heading angle. This application provides an innovative interactive method for detecting the initial heading angle of a vehicle. The correct heading angle of a specified lane can be manually drawn, which improves adaptability and quantifies the accuracy of the heading angle of the vehicle in the first frame (i.e., the first trajectory point), thereby evaluating the initial heading angle detection of the roadside algorithm.

[0087] Users can click the initial state analysis control (Button) in the operation interface. The display interface of the intersection can receive the correct headings of the approach lanes in each direction drawn by the user, obtain the reference headings of each lane, and thus obtain the reference heading of the lane where the first trajectory point in the trajectory sequence is located.

[0088] In some embodiments, a user can be received in a visual interface to draw the correct heading for the target lane area; and the reference heading of the target lane can be determined based on the drawing input.

[0089] Step 650: Based on the reference heading, determine whether the heading angle of the first trajectory point is accurate, and obtain the initial heading angle detection result.

[0090] By comparing the heading angle of the first trajectory point with the reference heading, it can be determined whether the roadside algorithm detection is incorrect. If the heading angle of the first trajectory point is consistent with the reference heading, the heading angle of the first trajectory point is determined to be accurate; if the heading angle of the first trajectory point is inconsistent with the reference heading, the heading angle of the first trajectory point is determined to be inaccurate. When the end analysis command is received through the end analysis control in the interface, the analysis result diagram of the initial heading angle is generated, such as... Figure 7 As shown in the initial heading angle analysis results diagram, red arrows indicate incorrect initial headings, while blue arrows indicate correct initial headings. The initial heading angle analysis results also allow for statistical analysis of the initial heading angle error rate for all trajectory sequences.

[0091] Step 660: Determine the performance evaluation result of the roadside algorithm based on the mutation detection result, the anomaly detection result, and the initial heading angle detection result.

[0092] The results of mutation detection, anomaly detection, and initial heading angle detection can be summarized to obtain the performance evaluation results of the roadside algorithm.

[0093] The performance evaluation method for the roadside algorithm provided in this embodiment detects the initial heading angle based on the obtained reference heading, and combines automated detection with manual interactive verification to ensure the accuracy of the evaluation results.

[0094] This application introduces a "human-machine loop" interactive verification mechanism. Prior knowledge (correct lane direction) that is difficult to describe using fixed rules is injected into the evaluation system through manual drawing, thus solving a very specific but crucial evaluation problem—the correctness of the vehicle's initial heading angle. This addresses the blind spots in automated evaluation. The vehicle's initial heading angle is the cornerstone of all subsequent tracking and prediction algorithms; if it's wrong in the first frame, all subsequent frames will be wrong. It's difficult for algorithms alone to accurately determine whether the heading angle of a stationary or newly appearing vehicle is consistent with the lane direction due to a lack of context. This application cleverly delegates this judgment to manual input, quantifying subjective judgment. The core of this method is not simply to let people see it, but to transform the qualitative judgment of a person (the car in this lane should be going this way) into a quantifiable standard reference value (the drawn lane heading). The system calculates the accurate heading error rate by comparing the vehicle heading angle output by the roadside algorithm with this reference heading, thus making a vague concept measurable and evaluable. This improves the adaptability and credibility of the system. This method does not rely on high-precision maps or pre-set complex lane models, and has strong flexibility and adaptability. For any intersection, users can quickly draw a standard heading for evaluation. At the same time, the introduction of a manual verification process greatly enhances the credibility and authority of the final evaluation results.

[0095] Figure 8 This is an overall framework diagram of the performance evaluation method for roadside algorithms provided in the embodiments of this application, such as... Figure 8As shown, after starting the analysis tool, the GUI interface (operation interface) is initialized. Users can select files and set thresholds in the GUI, followed by data loading and preprocessing. Error handling and reloading are performed if data loading and preprocessing fail. The main data analysis process begins upon successful data loading and preprocessing. The main data analysis process includes mutation detection analysis, trajectory quality analysis, and initial heading angle analysis. Mutation detection analysis includes velocity mutation detection, heading mutation detection, acceleration mutation detection, and type (vehicle type) mutation detection, generating a mutation record table (as shown in Table 1). Trajectory quality analysis includes trajectory glitch detection, trajectory breakpoint detection, and trajectory smoothing analysis, generating a trajectory quality report and trajectory visualization results. Initial heading angle analysis displays the initial state diagram, allows for user-interactive analysis, inputs the correct heading (reference heading) via polygon annotation, performs vehicle heading matching analysis based on this heading, and generates a heading error report (initial heading angle detection result). The results of mutation detection analysis, trajectory quality analysis, and initial heading analysis are then summarized. Within the GUI interface, users can select subsequent operations. The ID filtering interface allows for ID searching and selection, enabling visualization of multiple indicators such as velocity curves, heading angle curves, and acceleration curves, with the ability to save the charts. Users can also batch generate charts, process selected IDs, generate a comprehensive report (containing all evaluation results for the selected IDs), and output all charts. Furthermore, the GUI interface allows for report export, generating an analysis report based on this command and exporting it as a CSV file, along with a statistical summary of the analysis report. After analysis, results can be displayed and saved, and the tool can be exited.

[0096] Figure 9 This is a display example diagram of the tool GUI interface in the embodiments of this application, such as... Figure 9 As shown, the overall GUI mainly consists of five areas: ① File selection area: includes a CSV file selection control and a save path control. The CSV file selection control is used to select the file of the trajectory data to be analyzed, and the save path control is used to select the save path of the analysis results; ② Threshold setting area: includes custom settings for velocity, heading angle and acceleration change thresholds, and custom settings for trajectory parameters, including curvature threshold, velocity change threshold, heading angle change threshold (degrees), cluster radius and minimum number of samples; ③ There are two controls (Buttons) for "Start Analysis" and "Initial State Analysis". After clicking "Start Analysis", data cleaning and mutation record table (including speed, acceleration and heading angle mutations) will be performed, vehicle type mutation table will be generated, overall trajectory map and trajectory quality report will be generated, and all valid IDs will be extracted and output to the ID list area, which is convenient for subsequent personalized analysis of one or more specific IDs. ④ Chart generation area: Select one or more IDs, which supports searching and selection. Click the "Start generating image" control to generate curves showing the changes in heading angle, velocity, and acceleration for the selected IDs for visualization analysis. ⑤ ID list area: Displays all valid IDs, supports one-click selection of all, deselection of all, and inversion of selection.

[0097] The advantages of this application's embodiments are reflected in the following aspects: Enhanced comprehensiveness of evaluation: Through multi-dimensional feature analysis, algorithm performance is comprehensively evaluated, avoiding the limitations of single-index evaluation; Improved anomaly detection accuracy: Multiple detection strategies are employed in synergy, significantly improving the ability to identify complex anomaly patterns; Enhanced credibility of evaluation results: Combining automated detection with manual interactive verification ensures the accuracy of evaluation results; Improved standardization: A unified performance evaluation standard and quantitative indicator system are established, facilitating comparison of results across different test scenarios; Improved visualization support: An intuitive visualization interface and detailed analysis reports are provided, facilitating rapid location and optimization of algorithm problems; Wide application scope: Applicable to multiple fields such as autonomous driving algorithm verification, traffic data quality assessment, and intelligent transportation system optimization.

[0098] It should be noted that the performance evaluation method for roadside algorithms provided in this application can be executed by a roadside algorithm performance evaluation device, or a control module within that device for executing the performance evaluation method for loading the roadside algorithm. This application uses the execution of the performance evaluation method for loading the roadside algorithm by a roadside algorithm performance evaluation device as an example to illustrate the performance evaluation method for roadside algorithms provided in this application.

[0099] Figure 10 This is a schematic diagram of the structure of a performance evaluation device for a roadside algorithm provided in an embodiment of this application, as shown below. Figure 10 The performance evaluation device for the roadside algorithm includes: The trajectory acquisition module 1010 is used to acquire the trajectory sequence output by the roadside algorithm, wherein each trajectory point in the trajectory sequence includes location information and kinematic features; The mutation detection module 1020 is used to perform mutation detection on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain mutation detection results. Anomaly detection module 1030 is used to perform anomaly detection on each trajectory point in the trajectory sequence based on the position information and the kinematic features, and obtain anomaly detection results; The evaluation result determination module 1040 is used to determine the performance evaluation result of the roadside algorithm based on the mutation detection result and the anomaly detection result.

[0100] Optionally, the mutation detection module includes: A change value determination unit is used to determine the change values ​​of the kinematic characteristics of the adjacent trajectory points; A mutation detection unit is used to determine that there is a kinematic mutation in the next trajectory point among the adjacent trajectory points if the absolute value of the change value is greater than or equal to the change threshold, and to record the mutation type and the change value of the next trajectory point to obtain the mutation detection result.

[0101] Optionally, the kinematic features include at least one of velocity, acceleration, and heading angle, and the mutation type includes at least one of velocity mutation, acceleration mutation, and heading angle mutation.

[0102] Optionally, the apparatus further includes a graph generation module, which performs at least one of the following: Generate a first graph showing the change of the speed over time; Generate a second graph showing the acceleration changing over time; A third curve showing the change of the heading angle over time is generated. Before generating the third curve, if the change in the initial heading angle of a trajectory point relative to the heading angle of the previous trajectory point is greater than 180 degrees, the initial heading angle of the trajectory point is subtracted by 360 degrees to obtain the final heading angle of the trajectory point. If the change in the initial heading angle of the trajectory point relative to the heading angle of the previous trajectory point is less than -180 degrees, the initial heading angle of the trajectory point is added by 360 degrees to obtain the final heading angle of the trajectory point.

[0103] Optionally, the change threshold is obtained through the settings interface.

[0104] Optionally, the anomaly detection module includes: A burr detection unit is used to determine burr points in the trajectory sequence based on the position information and the kinematic features. A breakpoint detection unit is used to determine breakpoints in the trajectory sequence based on the position information and the velocity in the kinematic features; A smoothing analysis unit is used to determine smoothness anomalies in the trajectory sequence based on a sliding window; An anomaly detection result determination unit is used to determine the anomaly detection result based on the burr point, the breakpoint, and the smoothness anomaly point.

[0105] Optionally, the burr detection unit is specifically used to perform at least one of the following: Based on the location information, the curvature of each trajectory point is determined, and the trajectory points whose curvature satisfies the target condition are identified as the burr points. The trajectory points in the kinematic features where the change in heading angle is greater than the heading angle change threshold are identified as the spur points; The DBSCAN clustering algorithm is used to cluster the trajectory points in the trajectory sequence, and the isolated trajectory points in the clustering results are identified as the spur points.

[0106] Optionally, the target conditions include: the curvature of the trajectory point is greater than or equal to a curvature threshold, and the curvature of the preceding and following trajectory points of the trajectory point is greater than or equal to the target proportion of the curvature threshold.

[0107] Optionally, the breakpoint detection unit is specifically used for: The breakpoint is defined as the trajectory point whose speed is greater than or equal to the speed ratio of the previous trajectory point and whose distance from the previous trajectory point is greater than or equal to the distance threshold.

[0108] Optionally, the anomaly detection result determination unit includes: The filtering subunit is used to filter out repeated trajectory points among the burr points, the breakpoints, and the smoothness anomalies. The anomaly detection result determination subunit is used to determine the anomaly detection result based on the filtered burr points, breakpoints, and smoothness anomalies.

[0109] Optionally, the anomaly detection result determination subunit is specifically used for: From the filtered burr points, breakpoints, and smoothness anomalies, the number of burr points, the number of breakpoints, the burr rate, and the breakpoint rate are counted. The burr rate is the proportion of burr points among the filtered burr points, breakpoints, and smoothness anomalies in all trajectory points, and the breakpoint rate is the proportion of breakpoints among the filtered burr points, breakpoints, and smoothness anomalies in all trajectory points. The number of burrs, the number of breakpoints, the burr rate, and the breakpoint rate are used as the anomaly detection results.

[0110] Optionally, the device further includes: The visualization result generation module is used to generate visualization results of the trajectory sequence based on the anomaly detection results.

[0111] Optionally, the device further includes: The reference heading acquisition module is used to acquire the reference heading of the lane where the first trajectory point in the trajectory sequence is located; The initial heading detection module is used to determine whether the heading angle of the first trajectory point is accurate based on the reference heading, and to obtain the initial heading angle detection result; The evaluation result determination module is specifically used for: Based on the mutation detection results, the anomaly detection results, and the initial heading angle detection results, the performance evaluation results of the roadside algorithm are determined.

[0112] The performance evaluation device for roadside algorithms provided in this application acquires the trajectory sequence output by the roadside algorithm. Each trajectory point in the trajectory sequence includes location information and kinematic features. Abrupt change detection is performed on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results. Based on the location information and kinematic features, anomaly detection is performed on each trajectory point in the trajectory sequence to obtain anomaly detection results. Based on the abrupt change detection results and the anomaly detection results, the performance evaluation result of the roadside algorithm is determined. Since abrupt change detection and anomaly detection of trajectory points can be performed simultaneously during the performance evaluation process, it is no longer limited to a single technical indicator. Through multi-dimensional feature analysis, the performance of the roadside algorithm can be comprehensively evaluated, avoiding the limitations of single indicator evaluation and comprehensively reflecting the overall performance of the roadside algorithm in complex traffic environments.

[0113] The performance evaluation device for the roadside algorithm in this application embodiment can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. For example, mobile electronic devices can be mobile phones, tablets, laptops, PDAs, in-vehicle electronic devices, wearable devices, ultra-mobile personal computers (UMPCs), netbooks, or personal digital assistants (PDAs), etc., while non-mobile electronic devices can be servers, network-attached storage (NAS), personal computers (PCs), televisions (TVs), ATMs, or self-service machines, etc. This application embodiment does not impose specific limitations.

[0114] The performance evaluation device for the roadside algorithm in this application embodiment can be a device with an operating system. This operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit it.

[0115] The performance evaluation device for roadside algorithms provided in this application embodiment can achieve... Figures 1 to 8 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0116] Optionally, embodiments of this application also provide an electronic device, including a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the various processes of the above-described roadside algorithm performance evaluation method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0117] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0118] Figure 11 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of this application. The electronic device 1100 includes, but is not limited to, components such as: a radio frequency unit 1101, a network module 1102, an audio output unit 1103, an input unit 1104, a sensor 1105, a display unit 1106, a user input unit 1107, an interface unit 1108, a memory 1109, and a processor 1110. The input unit 1104 may include an image processor 11041 and a microphone 11042; the display unit 1106 may include a display panel 11061; and the user input unit 1107 may include a touch panel 11071 and other input devices (such as a keyboard) 11072.

[0119] Those skilled in the art will understand that the electronic device 1100 may also include a power supply (such as a battery) for supplying power to various components. The power supply may be logically connected to the processor 1110 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system. Figure 11 The electronic device structure shown does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here. The memory 1109 stores a program or instruction that can run on the processor. When the program or instruction is executed by the processor 1110, it implements the various processes of the above-described roadside algorithm performance evaluation method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0120] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described roadside algorithm performance evaluation method embodiment and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0121] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0122] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described roadside algorithm performance evaluation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0123] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0124] It should be noted that, in this document, 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 that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0125] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0126] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A performance evaluation method for roadside algorithms, characterized in that, include: Obtain the trajectory sequence output by the roadside algorithm, wherein each trajectory point in the trajectory sequence includes location information and kinematic features; Abrupt changes are detected in the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results. Based on the location information and the kinematic features, anomaly detection is performed on each trajectory point in the trajectory sequence to obtain anomaly detection results; Based on the mutation detection results and the anomaly detection results, the performance evaluation results of the roadside algorithm are determined.

2. The method according to claim 1, characterized in that, The step of performing abrupt change detection on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain abrupt change detection results includes: Determine the changes in the kinematic characteristics of the adjacent trajectory points; If the absolute value of the change is greater than or equal to the change threshold, it is determined that the next trajectory point among the adjacent trajectory points has a kinematic mutation, and the mutation type and the change value of the next trajectory point are recorded to obtain the mutation detection result.

3. The method according to claim 2, characterized in that, The kinematic characteristics include at least one of velocity, acceleration, and heading angle, and the mutation type includes at least one of velocity mutation, acceleration mutation, and heading angle mutation.

4. The method according to claim 3, characterized in that, It also includes at least one of the following: Generate a first graph showing the change of the speed over time; Generate a second graph showing the acceleration changing over time; A third curve showing the change of the heading angle over time is generated. Before generating the third curve, if the change in the initial heading angle of a trajectory point relative to the heading angle of the previous trajectory point is greater than 180 degrees, the initial heading angle of the trajectory point is subtracted by 360 degrees to obtain the final heading angle of the trajectory point. If the change in the initial heading angle of the trajectory point relative to the heading angle of the previous trajectory point is less than -180 degrees, the initial heading angle of the trajectory point is added by 360 degrees to obtain the final heading angle of the trajectory point.

5. The method according to claim 1, characterized in that, The step of performing anomaly detection on each trajectory point in the trajectory sequence based on the location information and the kinematic features to obtain anomaly detection results includes: Based on the location information and the kinematic characteristics, the spur points in the trajectory sequence are determined; Based on the location information and the velocity in the kinematic features, the breakpoints in the trajectory sequence are determined; Based on a sliding window, identify smoothness anomalies in the trajectory sequence; The anomaly detection result is determined based on the burr points, the breakpoints, and the smoothness anomalies.

6. The method according to claim 5, characterized in that, Determining the spur points in the trajectory sequence based on the location information and the kinematic features includes at least one of the following: Based on the location information, the curvature of each trajectory point is determined, and the trajectory points whose curvature satisfies the target condition are identified as the burr points. The trajectory points in the kinematic features where the change in heading angle is greater than the heading angle change threshold are identified as the spur points; The DBSCAN clustering algorithm is used to cluster the trajectory points in the trajectory sequence, and the isolated trajectory points in the clustering results are identified as the spur points.

7. The method according to claim 6, characterized in that, The target conditions include: the curvature of the trajectory point is greater than or equal to a curvature threshold, and the curvature of the preceding and following trajectory points of the trajectory point is greater than or equal to the target proportion of the curvature threshold.

8. The method according to claim 5, characterized in that, Determining the breakpoints in the trajectory sequence based on the position information and the velocity in the kinematic features includes: The breakpoint is defined as the trajectory point whose speed is greater than or equal to the speed ratio of the previous trajectory point and whose distance from the previous trajectory point is greater than or equal to the distance threshold.

9. The method according to claim 5, characterized in that, The step of determining the anomaly detection result based on the burr points, the breakpoints, and the smoothness anomalies includes: Filter out repeated trajectory points among the burr points, breakpoints, and smoothness anomalies, and determine the anomaly detection result based on the filtered burr points, breakpoints, and smoothness anomalies.

10. The method according to claim 9, characterized in that, The step of determining the anomaly detection result based on the filtered burr points, breakpoints, and smoothness anomalies includes: From the filtered burr points, breakpoints, and smoothness anomalies, the number of burr points, the number of breakpoints, the burr rate, and the breakpoint rate are counted. The burr rate is the proportion of burr points among the filtered burr points, breakpoints, and smoothness anomalies in all trajectory points, and the breakpoint rate is the proportion of breakpoints among the filtered burr points, breakpoints, and smoothness anomalies in all trajectory points. The number of burrs, the number of breakpoints, the burr rate, and the breakpoint rate are used as the anomaly detection results.

11. The method according to claim 5, characterized in that, After determining the anomaly detection result based on the burr point, the breakpoint, and the smoothness anomaly point, the method further includes: Based on the anomaly detection results, a visualization of the trajectory sequence is generated.

12. The method according to any one of claims 1-11, characterized in that, Also includes: Obtain the reference heading of the lane where the first trajectory point in the trajectory sequence is located; Based on the reference heading, determine whether the heading angle of the first trajectory point is accurate, and obtain the initial heading angle detection result; The step of determining the performance evaluation result of the roadside algorithm based on the mutation detection result and the anomaly detection result includes: Based on the mutation detection results, the anomaly detection results, and the initial heading angle detection results, the performance evaluation results of the roadside algorithm are determined.

13. A performance evaluation device for a roadside algorithm, characterized in that, include: The trajectory acquisition module is used to acquire the trajectory sequence output by the roadside algorithm, wherein each trajectory point in the trajectory sequence includes location information and kinematic features; The mutation detection module is used to perform mutation detection on the kinematic features of adjacent trajectory points in the trajectory sequence to obtain mutation detection results. An anomaly detection module is used to perform anomaly detection on each trajectory point in the trajectory sequence based on the location information and the kinematic features, and obtain anomaly detection results; The evaluation result determination module is used to determine the performance evaluation result of the roadside algorithm based on the mutation detection result and the anomaly detection result.

14. An electronic device, characterized in that, The method includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the performance evaluation method for the roadside algorithm as described in any one of claims 1-12.

15. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the performance evaluation method for the roadside algorithm as described in any one of claims 1-12.

Images