Typical dangerous scene autonomous identification method and device for automatic driving test, and readable storage medium

Abstract

The invention relates to an automatic driving test-oriented typical dangerous scene autonomous identification method and device, and a readable storage medium. The typical dangerous scene autonomous identification method comprises the following steps: S1, acquiring an accident scene data set; s2, coordinate transformation; s3, determining an accident type and marking the accident type; s4, calculating longitudinal behavior characteristic parameters based on the longitudinal driving behavior characteristics of the accident scene data; s5, calculating transverse behavior characteristic parameters based on the transverse driving behavior characteristics of the accident scene data; s6, constructing an annotation data set based on the scene key feature parameters and the annotation result in the step S3; s7, based on the annotation data set, constructing a typical dangerous scene autonomous identification model; and S8, automatically identifying typical dangerous scenes. The typical dangerous scene autonomous identification model of the support vector machine is constructed based on the driving behavior characteristics of the vehicle, and the test level of automatic driving is improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving testing technology, and in particular to a method, device and readable storage medium for autonomously identifying typical dangerous scenarios for autonomous driving testing. Background Art

[0002] With the rapid development of emerging technologies such as artificial intelligence, big data, and cloud computing, autonomous driving technology is constantly being updated and upgraded, and the level of autonomous driving is also constantly improving. In the future, autonomous driving will be an inevitable development trend in the road transportation industry, and will bring about tremendous changes in people's travel. Based on this, in order to promote the implementation of autonomous driving technology and ensure the safety of autonomous driving functions, autonomous driving testing is an indispensable and important link. Compared with traditional mileage-based real vehicle testing, scenario-based virtual testing technology has been favored by many scholars and engineering technicians due to its low testing cost, flexible scenario configuration, high testing efficiency, and strong test repeatability. Therefore, scenario-based virtual testing has become an indispensable and important part of autonomous driving test evaluation, which has important practical significance for accelerating the implementation of autonomous driving technology and improving autonomous driving test specifications and standards.

[0003] However, the current scenario-based virtual testing technology still has the following obvious defects:

[0004] 1) There is a lack of overall connection between existing scene recognition methods and autonomous driving testing technologies, ignoring strategic and tactical considerations that serve testing needs.

[0005] 2) The degree of automation of scenario testing is low, and the scenarios used in the test still need to be manually identified and screened according to the test requirements, resulting in low testing efficiency.

[0006] 3) The analysis error of driving behavior in scene data is large. The longitudinal and lateral behaviors of the vehicle in the scene are coupled with the road curvature, resulting in the longitudinal and lateral driving behavior parameters of the vehicle being unable to accurately reflect the driving behavior characteristics of the vehicle in the scene. Summary of the invention

[0007] In response to the above-mentioned problems in the prior art, the present invention proposes a method, device and readable storage medium for autonomous identification of typical dangerous scenarios for autonomous driving testing, and constructs a typical dangerous scenario autonomous identification model of a support vector machine based on the driving behavior characteristics of the vehicle, thereby improving the testing level of automated driving.

[0008] Specifically, the present invention proposes a typical dangerous scene autonomous identification method for autonomous driving testing, comprising the steps of:

[0009] S1, obtain the original accident scene dataset;

[0010] S2, converting the vehicle coordinate information in the accident scene data set from a Cartesian coordinate system to a Frenet coordinate system through coordinate transformation;

[0011] S3, determining the accident type of each piece of accident scene data in the accident scene data set based on the accident analysis tool, and marking the accident scene data set;

[0012] S4, calculating longitudinal behavior characteristic parameters of the vehicle in the accident scene based on the longitudinal driving behavior characteristics of the accident scene data;

[0013] S5, calculating a lateral behavior characteristic parameter of the vehicle in the corresponding accident scene based on the lateral driving behavior characteristics of the accident scene data, and combining the longitudinal behavior characteristic parameter and the corresponding lateral behavior characteristic parameter into a scene key characteristic parameter;

[0014] S6, constructing a labeled data set based on the key feature parameters of the scene and the labeling result of step S3;

[0015] S7, based on the labeled data set, constructing a typical dangerous scene autonomous recognition model based on a support vector machine;

[0016] S8, automatically identifying typical dangerous scenes through the typical dangerous scene autonomous identification model.

[0017] According to an embodiment of the present invention, in step S1, the acquired accident scene data set is preprocessed, and the preprocessing includes removing redundant data and missing data.

[0018] According to one embodiment of the present invention, in step S2, the vehicle coordinate information in the accident scene is transformed from the Cartesian coordinate system (x t ,y t , v x , a x ,θ t , k t ) is converted to the Frenet coordinate system The following steps are involved:

[0019] S21, using the center line of the road as the reference line of the Frenet coordinate system, based on the normal vector of the reference line and the tangent vector Establish a coordinate system, and record (l, s) as the coordinate position in the Frenet coordinate system;

[0020] S22, calculating the relative difference between the coordinates of the vehicle's driving trajectory line and the coordinates of the reference line in the Cartesian coordinate system:

[0021] d x =x t -xr ;

[0022] d y =y t -y r ;

[0023] d θ =θ t -θ r ;

[0024] Among them, (x t ,y t ) is the vehicle’s driving trajectory point in the Cartesian coordinate system, x t is the horizontal coordinate, y t is the vertical coordinate; θ t is the heading angle of the vehicle in the Cartesian coordinate system;

[0025] (x r ,y r ) is the trajectory point (x t ,y t ) is the projection point on the reference line, θ r is the tangent vector The angle with the horizontal axis;

[0026] S23, calculate the lateral coordinate l of the vehicle in the Frenet coordinate system:

[0027]

[0028] Among them, I(·) is the indicator function, which means:

[0029]

[0030] ||(d x , d y )||2 is represented as a vector (d x , d y )’s second norm;

[0031] S24, calculate the derivative l′ of the horizontal coordinate l with respect to the vertical coordinate in the Frenet coordinate system, and the calculation formula is:

[0032]

[0033] Among them, k r is the reference line curvature;

[0034] S25, calculate the derivative of l′ with respect to the ordinate in the Frenet coordinate system:

[0035]

[0036] Among them, k tis the curvature of the vehicle trajectory;

[0037] S26, calculate the longitudinal coordinate s of the vehicle in the Frenet coordinate system:

[0038] S=S r ;

[0039] Among them, s r From the starting point of the reference line to the projection point (x r ,y r ) between the reference lines;

[0040] S27, calculate the rate of change of the longitudinal coordinates of the vehicle in the Frenet coordinate system The calculation formula is:

[0041]

[0042] Among them, v x is the vehicle's speed in the Cartesian coordinate system;

[0043] S28, calculate the second-order derivative of the vehicle's ordinate with respect to time in the Frenet coordinate system The calculation formula is:

[0044]

[0045] Among them, a x is the vehicle's acceleration in the Cartesian coordinate system.

[0046] According to one embodiment of the present invention, in step S3, based on the PC-crash traffic accident analysis tool, the accident scene corresponding to each piece of the accident scene data is replayed, the accident type is determined, and the accident scene data set is labeled.

[0047] According to one embodiment of the present invention, in step S4, the longitudinal behavior characteristic parameters of the vehicle in the accident scene are calculated based on the longitudinal driving behavior characteristics, and the longitudinal behavior characteristic parameters include the collision time TTC, and the calculation thereof includes the following steps:

[0048] S41, calculating the headway distance between the front and rear vehicles, the headway distance H is calculated as follows:

[0049] H = s1-s2;

[0050] Wherein, s1 is the longitudinal coordinate position of the center of mass of the front vehicle in the Frenet coordinate system, and s2 is the longitudinal coordinate position of the center of mass of the rear vehicle in the Frenet coordinate system;

[0051] S42, calculate the speed difference between the front and rear vehicles, the speed difference The calculation formula is as follows:

[0052]

[0053] in, is the speed of the front vehicle in the Frenet coordinate system, is the speed of the following vehicle in the Frenet coordinate system;

[0054] S43, calculating the collision time TTC of each frame of trajectory data in each group of dangerous accident scenarios, and the calculation formula is as follows:

[0055]

[0056] S44, setting a TTC threshold TTCm, extracting segments with TTC≤TTCm in the scene as longitudinal behavior feature parameters of the scene data.

[0057] According to one embodiment of the present invention, in step S4, based on the lateral driving behavior characteristics, calculating the lateral behavior characteristic parameters of the vehicle in the accident scene, the lateral behavior characteristic parameters including lateral spacing and trajectory slope parameters, comprises the following steps:

[0058] S41', calculate the lateral distance L between vehicles in the scene. The calculation formula of the lateral distance L is as follows:

[0059] L = l1-l2;

[0060] l1 and l2 represent the lateral coordinate positions of the centroids of two adjacent vehicles in the Frenet coordinate system;

[0061] S42', calculate the heading angle of each vehicle in the scene in the Frenet coordinate system;

[0062] S43', calculating the lateral spacing L and the trajectory slope parameter l' of each frame of trajectory data in each group of dangerous accident scenarios as lateral behavior characteristic parameters;

[0063] S44', based on the extracted longitudinal behavior feature parameters, extract the transverse behavior feature parameters corresponding to the segment, and combine the longitudinal behavior feature parameters and the transverse behavior feature parameters of the segment to form a complete scene key feature parameter.

[0064] According to one embodiment of the present invention, step S7 specifically includes the following steps:

[0065] S71, dividing the labeled data set into a model training set and a model test set;

[0066] S72, gridding the hyperparameters in the support vector machine to obtain multiple groups of parameters, training and testing the model with each group of parameters, and outputting parameters corresponding to the best test accuracy as the selected parameters of the typical dangerous scene autonomous recognition model based on the support vector machine;

[0067] S73, inputting a model training set to train the typical danger scene autonomous recognition model parameters of the support vector machine;

[0068] S74, input the test set data to verify the accuracy of the typical dangerous scene autonomous recognition model of the support vector machine. The verification indicators include precision rate P, recall rate R, accuracy rate A, and F1 score. The corresponding indicator calculation formula is as follows:

[0069]

[0070] Among them, TP is the number of positive classes predicted as positive classes; TN is the number of negative classes predicted as negative classes; FP is the number of negative classes predicted as positive classes; FN is the number of positive classes predicted as negative classes.

[0071] According to one embodiment of the present invention, step S8 specifically includes the following steps:

[0072] S81, extracting longitudinal key feature parameters and transverse key feature parameters of an accident scene data set of a typical dangerous scene to form scene key feature parameters;

[0073] S82, placing the scene key feature parameters into the typical dangerous scene autonomous identification model for classification calculation;

[0074] S83, determine the accident types of typical dangerous scenarios.

[0075] The present invention also provides an automatic identification device for typical dangerous scenarios for autonomous driving tests, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any of the aforementioned methods for autonomously identifying typical dangerous scenarios are implemented.

[0076] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the aforementioned methods for autonomously identifying typical dangerous scenarios are implemented.

[0077] The present invention provides a method, device and readable storage medium for autonomous identification of typical dangerous scenarios for autonomous driving testing. Based on Frenet coordinate transformation, feature parameter identification and support vector machine method, a typical dangerous scenario autonomous identification model of a support vector machine is constructed, and the accident classification of typical dangerous scenarios is automatically identified through the model, thereby improving the test level of automated driving.

[0078] It is to be understood that both the foregoing general description and the following detailed description of the present invention are exemplary and explanatory and are intended to provide further explanation of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] The accompanying drawings are included to provide further explanation of the present invention, and are incorporated into and constitute a part of this application. The accompanying drawings illustrate embodiments of the present invention and together with the description serve to explain the principle of the present invention.

[0080] In the attached figure:

[0081] Figure 1 A flowchart of a typical dangerous scenario autonomous identification method for autonomous driving testing according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0082] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application may be combined with each other.

[0083] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present application and its application or use. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0084] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0085] Unless otherwise specifically stated, the relative arrangement, numerical expressions and numerical values ​​of the parts and steps set forth in these embodiments do not limit the scope of the application. Meanwhile, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to actual proportional relationships. The technology, methods and equipment known to those of ordinary skill in the relevant art may not be discussed in detail, but in appropriate cases, the technology, methods and equipment should be considered as a part of the specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, and therefore, once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0086] In the description of the present application, it should be understood that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "lateral, vertical, perpendicular, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present application and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the device or element referred to must have a specific direction or be constructed and operated in a specific direction, and therefore cannot be understood as limiting the scope of protection of the present application; the directional words "inside and outside" refer to the inside and outside relative to the outline of each component itself.

[0087] In addition, it should be noted that the use of words such as "first" and "second" to define components is only for the convenience of distinguishing the corresponding components. If not otherwise stated, the above words have no special meaning and cannot be understood as limiting the scope of protection of this application. In addition, although the terms used in this application are selected from well-known and commonly used terms, some terms mentioned in the specification of this application may be selected by the applicant at his or her discretion, and their detailed meanings are explained in the relevant parts of the description of this article. In addition, it is required to understand this application not only by the actual terms used, but also by the meaning implied by each term.

[0088] Figure 1 A flowchart of a typical dangerous scene autonomous identification method for autonomous driving test according to an embodiment of the present invention is shown. As shown in the figure, a typical dangerous scene autonomous identification method for autonomous driving test comprises the following steps:

[0089] S1, obtaining an original accident scene data set, which is formed by collecting original accident scene data, for example, by capturing the real scene when the accident occurs in all directions and from multiple angles based on a roadside camera.

[0090] S2, through coordinate transformation, converts the vehicle coordinate information in the accident scene dataset from the Cartesian coordinate system to the Frenet coordinate system. The purpose of this step is to eliminate the influence of road curvature and provide a more accurate data basis for subsequent accident analysis.

[0091] S3, determining the accident type of each piece of accident scene data in the accident scene data set based on the accident analysis tool, and labeling the accident scene data set.

[0092] S4, based on the longitudinal driving behavior characteristics of the accident scene data, calculating longitudinal behavior characteristic parameters of the vehicle in the accident scene. The longitudinal behavior characteristic parameters are used to show the behavior characteristics of the vehicle in the front and rear directions.

[0093] S5, based on the lateral driving behavior characteristics of the accident scene data, calculate the lateral behavior characteristic parameters of the vehicle in the corresponding accident scene. The lateral behavior characteristic parameters are used to show the behavior characteristics of the vehicle in the left and right directions. The longitudinal behavior characteristic parameters and the corresponding lateral behavior characteristic parameters are integrated and merged into the scene key characteristic parameters. The scene key characteristic parameters comprehensively describe the behavior performance of the vehicle in the accident scene.

[0094] S6, constructing a labeled data set based on the key feature parameters of the scene and the labeling results of step S3.

[0095] S7, based on the labeled dataset, builds a typical dangerous scene autonomous recognition model based on support vector machine;

[0096] S8, automatically identifies typical dangerous scenes through the typical dangerous scene autonomous recognition model. By extracting and analyzing the features in the labeled data set and taking advantage of the support vector machine algorithm, the autonomous recognition model can accurately identify typical dangerous scenes, thereby improving the test level of automated driving.

[0097] In some examples, in step S1, the acquired accident scene data set is preprocessed, and the preprocessing includes removing redundant data and missing data.

[0098] In some examples, in step S2, the vehicle coordinate information in the accident scene is transformed from the Cartesian coordinate system (x t ,y t , v x , a x ,θ t , k t ) is converted to the Frenet coordinate system The coordinate transformation includes the following steps:

[0099] S21, using the center line of the road as the reference line of the Frenet coordinate system, based on the normal vector of the reference line and the tangent vector Establish a coordinate system, and record (l, s) as the coordinate position in the Frenet coordinate system;

[0100] S22, calculating the relative difference between the coordinates of the vehicle's driving trajectory line and the coordinates of the reference line in the Cartesian coordinate system:

[0101] d x =x t -x r ;

[0102] d y =y t -y r ;

[0103] d θ =θ t -θ r ;

[0104] Among them, (x t ,y t ) is the vehicle’s driving trajectory point in the Cartesian coordinate system, x t is the horizontal coordinate, y t is the vertical coordinate; θ t is the heading angle of the vehicle in the Cartesian coordinate system, that is, the angle between the vehicle's velocity direction and the horizontal coordinate;

[0105] (x r ,y r ) is the trajectory point (x t ,y t ) on the reference line, that is, the distance from the trajectory point (x t ,y t ) nearest point; θ r is the tangent vector The angle with the horizontal axis;

[0106] S23, calculate the lateral coordinate l of the vehicle in the Frenet coordinate system:

[0107]

[0108] Among them, I(·) is the indicator function, which means:

[0109]

[0110] ||(d x , d y )||2 is represented as a vector (d x , d y ), that is, (d x , d y )’s length;

[0111] S24, calculate the derivative of the horizontal coordinate l with respect to the vertical coordinate in the Frenet coordinate system, that is, calculate the tangent slope l′ of the vehicle trajectory in the Frenet coordinate system:

[0112]

[0113] Among them, k r is the reference line curvature;

[0114] S25, calculate the derivative of l′ with respect to the ordinate in the Frenet coordinate system:

[0115]

[0116] Among them, k t is the curvature of the vehicle trajectory;

[0117] S26, calculate the longitudinal coordinate s of the vehicle in the Frenet coordinate system:

[0118] s=S r ;

[0119] Among them, s r From the starting point of the reference line to the projection point (x r ,y r ) between the reference lines;

[0120] S27, calculate the rate of change of the longitudinal coordinate of the vehicle in the Frenet coordinate system, that is, the speed of the vehicle in the Frenet coordinate system:

[0121]

[0122] Among them, v x is the vehicle's speed in the Cartesian coordinate system;

[0123] S28, calculate the second-order derivative of the vehicle's longitudinal coordinate with respect to time in the Frenet coordinate system, that is, the acceleration of the vehicle in the Frenet coordinate system:

[0124]

[0125] Among them, a x is the vehicle's acceleration in the Cartesian coordinate system.

[0126] In some examples, in step S3, based on the PC-crash traffic accident analysis tool, the accident scenes corresponding to the accident scene data are replayed to determine the accident type. Accident types include rear-end collision, cut-in / cut-out and collision in intersection. The accident scene data set is annotated, which can be digitally annotated. For example, rear-end collision, cut-in / cut-out and collision in intersection are respectively corresponded to integers "0", "1", "2", ... for category labeling. As an example and not limitation, accident types may also include collision, scratch, rollover, etc.

[0127] In some examples, in step S4, the longitudinal behavior characteristic parameters of the vehicle in the accident scene are calculated based on the longitudinal driving behavior characteristics, and the longitudinal behavior characteristic parameters include the collision time TTC, and the calculation thereof includes the following steps:

[0128] S41, calculating the headway distance between the front and rear vehicles, that is, the distance from the head of the front vehicle to the head of the rear vehicle. The calculation formula of the headway distance H is as follows:

[0129] H = s1-s2;

[0130] Wherein, s1 is the longitudinal coordinate position of the center of mass of the front vehicle in the Frenet coordinate system, and s2 is the longitudinal coordinate position of the center of mass of the rear vehicle in the Frenet coordinate system;

[0131] S42, calculating the speed difference between the front and rear vehicles, that is, the difference between the speed of the front vehicle and the speed of the rear vehicle, the speed difference The calculation formula is as follows:

[0132]

[0133] in, is the speed of the front vehicle in the Frenet coordinate system, is the speed of the following vehicle in the Frenet coordinate system;

[0134] S43, calculating the collision time TTC of each frame of trajectory data in each group of dangerous accident scenarios, and the calculation formula is as follows:

[0135]

[0136] S44, setting a TTC threshold TTCm, extracting segments with TTC≤TTCm in the scene as longitudinal behavior feature parameters of the scene data.

[0137] In some examples, in step S4, calculating lateral behavior characteristic parameters of the vehicle in the accident scene based on the lateral driving behavior characteristics, the lateral behavior characteristic parameters including lateral spacing and trajectory slope parameters, includes the following steps:

[0138] S41', calculating the lateral distance L between vehicles in the scene, the calculation formula of the lateral distance L is as follows:

[0139] L = l1-l2;

[0140] l1 and l2 represent the lateral coordinate positions of the centroids of two adjacent vehicles in the Frenet coordinate system;

[0141] S42', calculate the heading angle of each vehicle in the scene in the Frenet coordinate system;

[0142] S43', calculate the lateral spacing L and the trajectory slope parameter l of each frame of trajectory data in each group of dangerous accident scenarios ′ As a parameter of horizontal behavior characteristics;

[0143] S44', based on the extracted longitudinal behavior feature parameters, extract the transverse behavior feature parameters in the corresponding segment, and combine the longitudinal behavior feature parameters and the transverse behavior feature parameters of the segment to form a complete scene key feature parameter.

[0144] In some examples, step S7 specifically includes the following steps:

[0145] S71, dividing the labeled data set into a model training set and a model test set;

[0146] S72, gridding the hyperparameters in the support vector machine to obtain multiple groups of parameters, training and testing the model with each group of parameters, outputting the parameters corresponding to the best test accuracy as the selected parameters of the typical dangerous scene autonomous identification model based on the support vector machine. In other words, the hyperparameter gridding process refers to enumerating the combinations of hyperparameters, and then training and testing the model under each group of hyperparameter combinations to select the hyperparameter combination that can optimize the model accuracy.

[0147] S73, inputting a model training set to train the typical danger scene autonomous recognition model parameters of the support vector machine;

[0148] S74, input the test set data to verify the accuracy of the typical dangerous scene autonomous recognition model of the support vector machine. The verification indicators include precision rate P, recall rate R, accuracy rate A, and F1 score. The corresponding indicator calculation formula is as follows:

[0149]

[0150] Among them, TP is the number of positive classes predicted as positive classes; TN is the number of negative classes predicted as negative classes; FP is the number of negative classes predicted as positive classes; FN is the number of positive classes predicted as negative classes.

[0151] In some examples, step S8 specifically includes the following steps:

[0152] S81, extracting longitudinal key feature parameters and transverse key feature parameters of an accident scene data set of a typical dangerous scene to form scene key feature parameters;

[0153] S82, placing the key characteristic parameters of the scene into a typical dangerous scene autonomous identification model for classification calculation;

[0154] S83, determine the accident types of typical dangerous scenarios.

[0155] The present invention also provides an automatic identification device for typical dangerous scenarios for autonomous driving tests, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the steps of the aforementioned method for autonomous identification of typical dangerous scenarios for autonomous driving tests.

[0156] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the aforementioned methods for autonomously identifying typical dangerous scenarios for autonomous driving testing.

[0157] Among them, the specific implementation methods and technical effects of the automatic identification equipment of typical dangerous scenes for autonomous driving tests and computer-readable storage media can be found in the above-mentioned embodiment of an autonomous identification method of typical dangerous scenes for autonomous driving tests provided by the present invention, which will not be repeated here.

[0158] The present invention provides a typical dangerous scene autonomous identification method for autonomous driving test, a typical dangerous scene automatic identification device for autonomous driving test, and a readable storage medium, which have the following beneficial effects:

[0159] (1) It can serve the testing and evaluation of autonomous driving functions in different application scenarios, connect scene recognition methods with testing requirements, and provide technical support for autonomous driving testing;

[0160] (2) An automatic identification and classification method for typical dangerous scenarios is proposed. It can automatically identify and select typical scenarios that meet the requirements from the scenario library according to the test requirements, reduce the degree of manual participation, and improve the efficiency of scenario testing.

[0161] (3) The Frenet coordinate transformation is used to decouple the relationship between the longitudinal and lateral behaviors of the vehicle in the scene and the road curvature, and then the driving behavior characteristics of the vehicle are accurately characterized from the longitudinal and lateral driving behavior parameters of the vehicle, thereby improving the accuracy of the analysis of the vehicle driving behavior in the scene.

[0162] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0163] The various illustrative logic modules and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0164] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information from / to the storage medium. In an alternative, a storage medium may be integrated into a processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components.

[0165] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented as a computer program product in software, each function may be stored on or transmitted by a computer-readable medium as one or more instructions or codes. Computer-readable media include both computer storage media and communication media, including any medium that facilitates the transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a computer. As an example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer. Any connection is also properly referred to as a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwaves, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwaves are included in the definition of the medium. Disk and disc as used herein include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, wherein disk often reproduces data magnetically, while disc reproduces data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0166] It will be apparent to those skilled in the art that various modifications and variations may be made to the above exemplary embodiments of the present invention without departing from the spirit and scope of the present invention. Therefore, it is intended that the present invention covers modifications and variations of the present invention that fall within the scope of the appended claims and their equivalent technical solutions.

Claims

1. A typical dangerous scene autonomous identification method for autonomous driving testing, comprising the steps of: S1, obtain the original accident scene dataset; S2, converting the vehicle coordinate information in the accident scene data set from a Cartesian coordinate system to a Frenet coordinate system through coordinate transformation; S3, determining the accident type of each piece of accident scene data in the accident scene data set based on the accident analysis tool, and marking the accident scene data set; S4, calculating longitudinal behavior characteristic parameters of the vehicle in the accident scene based on the longitudinal driving behavior characteristics of the accident scene data; S5, calculating a lateral behavior characteristic parameter of the vehicle in the corresponding accident scene based on the lateral driving behavior characteristics of the accident scene data, and combining the longitudinal behavior characteristic parameter and the corresponding lateral behavior characteristic parameter into a scene key characteristic parameter; S6, constructing a labeled data set based on the key feature parameters of the scene and the labeling result of step S3; S7, based on the labeled data set, constructing a typical dangerous scene autonomous recognition model based on a support vector machine; S8, automatically identifying typical dangerous scenes through the typical dangerous scene autonomous identification model.

2. The typical dangerous scene autonomous identification method according to claim 1, characterized in that: In step S1, the acquired accident scene data set is preprocessed, and the preprocessing includes removing redundant data and missing data.

3. The typical dangerous scene autonomous identification method according to claim 1, characterized in that: In step S2, the vehicle coordinate information in the accident scene is transformed from the Cartesian coordinate system (x t ,y t ,v x ,a x ,θ t ,k t ) is converted to the Frenet coordinate system The following steps are involved: S21, using the center line of the road as the reference line of the Frenet coordinate system, based on the normal vector of the reference line and the tangent vector Establish a coordinate system, and record (l, s) as the coordinate position in the Frenet coordinate system; S22, calculating the relative difference between the coordinates of the vehicle's driving trajectory line and the coordinates of the reference line in the Cartesian coordinate system: d x =x t -x r ; d y =y t -y r ; d θ =θ t -θ r ; Among them, (x t ,y t ) is the vehicle’s driving trajectory point in the Cartesian coordinate system, x t is the horizontal coordinate, y t is the vertical coordinate; θ t is the heading angle of the vehicle in the Cartesian coordinate system; (x r ,y r ) is the trajectory point (x t ,y t ) is the projection point on the reference line, θ r is the tangent vector The angle with the horizontal axis; S23, calculate the lateral coordinate l of the vehicle in the Frenet coordinate system: Among them, I(·) is the indicator function, which means: ||(d x ,d y )||2 is represented as a vector (d x ,d y )’s second norm; S24, calculate the derivative l of the horizontal coordinate l with respect to the vertical coordinate in the Frenet coordinate system ′ , the calculation formula is: Among them, k r is the reference line curvature; S25, calculate l in Frenet coordinate system ′ Derivative with respect to the ordinate: Among them, k t is the curvature of the vehicle trajectory; S26, calculate the longitudinal coordinate s of the vehicle in the Frenet coordinate system: s=s r ; Among them, s r From the starting point of the reference line to the projection point (x r ,y r ) between the reference lines; S27, calculate the rate of change of the longitudinal coordinates of the vehicle in the Frenet coordinate system The calculation formula is: Among them, v x is the vehicle's speed in the Cartesian coordinate system; S28, calculate the second-order derivative of the vehicle's ordinate with respect to time in the Frenet coordinate system The calculation formula is: Among them, a x is the vehicle's acceleration in the Cartesian coordinate system.

4. The typical dangerous scene autonomous identification method according to claim 1, characterized in that: In step S3, based on the PC-crash traffic accident analysis tool, the accident scenes corresponding to the accident scene data are replayed, the accident type is determined, and the accident scene data set is labeled.

5. The typical dangerous scene autonomous identification method according to claim 3, characterized in that: In step S4, the longitudinal behavior characteristic parameters of the vehicle in the accident scene are calculated based on the longitudinal driving behavior characteristics, and the longitudinal behavior characteristic parameters include the collision time TTC, and the calculation thereof includes the following steps: S41, calculating the headway distance between the front and rear vehicles, the headway distance H is calculated as follows: H = s1-s2; Wherein, s1 is the longitudinal coordinate position of the center of mass of the front vehicle in the Frenet coordinate system, and s2 is the longitudinal coordinate position of the center of mass of the rear vehicle in the Frenet coordinate system; S42, calculate the speed difference between the front and rear vehicles, the speed difference The calculation formula is as follows: in, is the speed of the front vehicle in the Frenet coordinate system, is the speed of the following vehicle in the Frenet coordinate system; S43, calculating the collision time TTC of each frame of trajectory data in each group of dangerous accident scenarios, and the calculation formula is as follows: S44, set TTC threshold TTC m , set TTC≤TTC in the scene m The fragments are extracted as the longitudinal behavior feature parameters of the scene data.

6. The typical dangerous scene autonomous identification method according to claim 5, characterized in that: In step S4, based on the lateral driving behavior characteristics, the lateral behavior characteristic parameters of the vehicle in the accident scene are calculated, wherein the lateral behavior characteristic parameters include lateral spacing and trajectory slope parameters, including the following steps: S41', calculating the lateral distance l between vehicles in the scene, the calculation formula of the lateral distance l is as follows: L＝l1-l2； l1 and l2 represent the lateral coordinate positions of the centroids of two adjacent vehicles in the Frenet coordinate system; S42', calculate the heading angle of each vehicle in the scene in the Frenet coordinate system; S43', calculate the lateral spacing L and the trajectory slope parameter l of each frame of trajectory data in each group of dangerous accident scenarios ′ As a parameter of horizontal behavior characteristics; S44', based on the extracted longitudinal behavior feature parameters, extract the transverse behavior feature parameters corresponding to the segment, and combine the longitudinal behavior feature parameters and the transverse behavior feature parameters of the segment to form a complete scene key feature parameter.

7. The typical dangerous scene autonomous identification method according to claim 6, characterized in that: Step S7 specifically includes the following steps: S71, dividing the labeled data set into a model training set and a model test set; S72, gridding the hyperparameters in the support vector machine to obtain multiple groups of parameters, training and testing the model with each group of parameters, and outputting parameters corresponding to the best test accuracy as the selected parameters of the typical dangerous scene autonomous recognition model based on the support vector machine; S73, inputting a model training set to train the typical danger scene autonomous recognition model parameters of the support vector machine; S74, input the test set data to verify the accuracy of the typical dangerous scene autonomous recognition model of the support vector machine. The verification indicators include precision rate P, recall rate R, accuracy rate A, and F1 score. The corresponding indicator calculation formula is as follows: Among them, TP is the number of positive classes predicted as positive classes; TN is the number of negative classes predicted as negative classes; FP is the number of negative classes predicted as positive classes; FN is the number of positive classes predicted as negative classes.

8. The typical dangerous scene autonomous identification method according to claim 1, characterized in that: Step S8 specifically includes the following steps: S81, extracting longitudinal key feature parameters and transverse key feature parameters of an accident scene data set of a typical dangerous scene to form scene key feature parameters; S82, placing the scene key feature parameters into the typical dangerous scene autonomous identification model for classification calculation; S83, determine the accident types of typical dangerous scenarios.

9. A typical dangerous scene automatic identification device for autonomous driving testing, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for autonomously identifying typical dangerous scenarios as described in any one of claims 1-8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for autonomously identifying typical dangerous scenarios as described in any one of claims 1 to 8 are implemented.

Images