Methods, devices, electronic equipment, and storage media for predicting vehicle side collision time.
By acquiring vehicle images through the vehicle's own camera and fusing the information, combined with geometric analysis, the problem of high cost and low accuracy of existing vehicle side collision warning systems has been solved, achieving low-cost and accurate collision time prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GAC AION NEW ENERGY AUTOMOBILE CO LTD
- Filing Date
- 2024-02-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing vehicle side collision warning methods are costly and have inaccurate prediction results. They are also difficult to accurately identify the geometric features and driving trajectory of the target vehicle, resulting in insufficient accuracy in collision time prediction.
By capturing vehicle images using the vehicle's own camera, information fusion methods are used to obtain the pose information and driving data of surrounding vehicles. Combined with geometric analysis, the vehicle is simplified into a direction vector, and collision time is predicted, avoiding the use of expensive LiDAR and large model training.
It achieves low-cost, accurate prediction of vehicle side collision times, reducing reliance on expensive sensors and complex calculations, and improving prediction accuracy.
Smart Images
Figure CN117901858B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of vehicle assisted driving technology, and more specifically, to a method, device, electronic device, and storage medium for predicting vehicle side collision time. Background Technology
[0002] Vehicle side collision warning is an important technology related to driving safety and driver assistance. It requires predicting the driving trajectories of both the vehicle and the target vehicle based on the acquired driving data, calculating the time or distance at which a collision may occur, and making a warning judgment based on the corresponding safety threshold.
[0003] The methods for acquiring target vehicle driving data include vehicle radar (LiDAR, ultrasonic radar, etc.), visual perception, and vehicle-to-everything (V2X) communication. Among these, collision algorithms can estimate the potential collision time between the vehicle and a target vehicle on the predicted path. For example, using LiDAR to identify the target vehicle is costly; millimeter-wave radar has limited accuracy and cannot accurately identify the target vehicle's geometric features; ultrasonic radar has a small ranging range and provides little help for trajectory planning. Data-driven solutions require extensive model training under standard operating conditions, which is costly. Driving state models built based on vehicle dynamics typically simplify the vehicle as a point mass for collision detection, neglecting to consider the vehicle's geometric features, resulting in insufficient accuracy in obtaining collision times. Bounding box collision algorithms generally include AABB and OBB. AABB has low detection accuracy, while OBB involves extensive trigonometric function calculations, resulting in high computational costs and requiring precise positional information of the four vertices of the vehicle's bounding box, making perception difficult. Summary of the Invention
[0004] The purpose of this application is to provide a method, device, electronic device, and storage medium for predicting vehicle side collision time. It can obtain the pose information and driving data of surrounding vehicles through vehicle images captured by the vehicle's camera, which is low-cost and accurately predicts the collision time through geometric analysis, thus solving the problems of high cost and inaccurate prediction results of existing methods.
[0005] This application provides a method for predicting vehicle side collision time, the method comprising:
[0006] Periodically acquire continuous multi-frame images of the vehicle's surrounding environment;
[0007] The side collision risk is determined based on the vehicle image. If there is a collision risk, the corresponding target vehicle is obtained.
[0008] Obtain the initial driving status information of the vehicle and the target vehicle;
[0009] The driving trajectories of the vehicle and the target vehicle are predicted based on the initial driving state information.
[0010] Collision time is predicted based on the predicted driving trajectory.
[0011] In the above implementation process, the vehicle's image captured by the vehicle's camera can be used to obtain the pose information and driving data of surrounding vehicles through information fusion. This method is low-cost, avoids the use of expensive LiDAR, and does not require large model training or vehicle-to-everything (V2X) connectivity. It can accurately predict collision time through geometric analysis, thus solving the problems of high cost and inaccurate prediction results of existing methods.
[0012] Furthermore, the step of determining the lateral collision risk based on the vehicle image, and obtaining the corresponding target vehicle if there is a collision risk, includes:
[0013] The vehicle in the vehicle image is identified using a pre-trained dynamic object detection model;
[0014] Assign an ID to the vehicle and obtain the heading angle of the vehicle relative to the self-vehicle;
[0015] If the heading angle of a vehicle with the same ID is greater than a set threshold in multiple consecutive vehicle images, there is a risk of collision, and the vehicle is marked as the target vehicle.
[0016] In the above implementation process, the risk of collision is determined based on the heading angle relative to the vehicle itself, thereby identifying the target vehicle.
[0017] Furthermore, obtaining the initial driving state information of the vehicle and the target vehicle includes:
[0018] Within the initialization time T, the vehicle speed message information sent by the vehicle bus is obtained, and the total number of messages is K1 frames. At the same time, K2 frames of vehicle images are obtained, and the timestamp of each frame of vehicle image is recorded.
[0019] Calculate the vehicle's average longitudinal acceleration based on the vehicle speed message information. ;
[0020] The pre-trained dynamic target detection model is used to obtain the position and heading angle information of the target vehicle relative to the vehicle in each frame of the vehicle image;
[0021] The vehicle speed information at the same time in each frame of the vehicle image is obtained by interpolation from the vehicle speed message information.
[0022] Based on the relative positional changes between the target vehicle and the driver vehicle, the lateral and longitudinal velocities of the target vehicle at each frame of the vehicle image, as well as the average lateral acceleration of the target vehicle within the initialization time T, are calculated. Mean longitudinal acceleration and the average heading angle relative to the vehicle ;
[0023] The vehicle's driving state information at the moment of the last vehicle image is used as the vehicle's initial driving state information: longitudinal speed. Mean longitudinal acceleration ;
[0024] The target vehicle's driving state information at the moment of the last vehicle image is used as the target vehicle's initial driving state information: lateral velocity. Longitudinal velocity Mean lateral acceleration Mean longitudinal acceleration The average heading angle of the target vehicle relative to its own vehicle and the target vehicle's position relative to its own vehicle. .
[0025] In the above implementation process, since the message period and the vehicle image period are different, the initial driving information of the vehicle and the target vehicle can be obtained by aligning the time axis.
[0026] Furthermore, the prediction of the driving trajectories of the self-vehicle and the target vehicle based on the initial driving state information includes:
[0027] Using the first M Predicting the driving trajectory based on the position information of the vehicle and the target vehicle during the prediction period:
[0028] Longitudinal position of the vehicle:
[0029] ;
[0030] Longitudinal position of the target vehicle:
[0031] ;
[0032] Lateral position of the target vehicle:
[0033] ;
[0034] in, This indicates the set prediction time step.
[0035] In the above implementation process, the driving trajectory and position of the vehicle and the target vehicle over a future period of time can be predicted based on the initial driving information.
[0036] Furthermore, the collision time prediction based on the predicted driving trajectory includes:
[0037] Obtain the predicted position of the vehicle at the current moment. Predicted location of the target vehicle ;
[0038] The length of the target vehicle is obtained based on a pre-trained dynamic target detection model. And the length of the vehicle has always been known to be ;
[0039] Obtain the first direction vector corresponding to the vehicle. ,in, Indicates the vehicle's heading angle;
[0040] Obtain the second direction vector corresponding to the target vehicle. ;
[0041] Any point on the first direction vector and any point on the second direction vector are represented as follows:
[0042] ;
[0043] ;
[0044] in, and Indicates the proportionality coefficient;
[0045] Obtain the direction vector from any point on the first direction vector to any point on the second direction vector. :
[0046] ;
[0047] Obtain the projections of the direction vectors onto the first and second direction vectors respectively:
[0048] ;
[0049] ;
[0050] Calculate the shortest distance between the first direction vector and the second direction vector;
[0051] If the set safe distance is greater than the nearest distance, there is a risk of collision, based on the vehicle's predicted position at the current moment. Predicted collision time .
[0052] In the above implementation process, the closest distance between the self-vehicle and the current vehicle is obtained, and the collision time can be predicted based on the closest distance.
[0053] Further, calculating the closest distance between the first direction vector and the second direction vector includes:
[0054] Obtain the second direction vector Distance from the first direction vector The nearest point P0 on the line:
[0055] First direction vector Second direction vector The lines intersect at point O, satisfying... According to the projection formula, the scaling factor corresponding to the intersection point O can be obtained. :
[0056] like Then the nearest point P0 is the second direction vector. The finish line, making We can obtain P0;
[0057] like If the nearest point P0 is the intersection point O;
[0058] like Then the nearest point P0 is the second direction vector. The starting point, We can obtain P0;
[0059] Based on P0, obtain the first direction vector. Distance from the second direction vector The nearest point P1:
[0060] From P0 to the first direction vector Draw a perpendicular line from the line containing the line, with the foot of the perpendicular at V1. The proportionality coefficient corresponding to V1 is calculated. :
[0061] like Then the nearest point P1 is the first direction vector. The finish line, making P1 can be obtained;
[0062] like Then the nearest point P1 is the foot of the perpendicular V1, according to P1 can be obtained;
[0063] like Then the nearest point P1 is the first direction vector. The starting point, We can obtain P1.
[0064] Based on P1, obtain the second direction vector. Distance from the first direction vector The nearest point P2:
[0065] From P1 to the second direction vector Draw a perpendicular line from the line containing the line, with the foot of the perpendicular at V2. The proportionality coefficient corresponding to V2 can be calculated. :
[0066] like Then the nearest point P2 is the second direction vector. The finish line, making P2 can be obtained;
[0067] like Then the nearest point P2 is the foot of the perpendicular V2, according to P2 can be obtained;
[0068] like Then the nearest point P2 is the second direction vector. The starting point, P2 can be obtained;
[0069] The distance between P1 and P2 is the shortest distance between the first direction vector and the second direction vector.
[0070] In the above implementation process, the vehicle and the target vehicle are simplified to direction vectors, which can accurately calculate the shortest distance between them.
[0071] This application embodiment also provides a vehicle side collision time prediction device, the device comprising:
[0072] The vehicle image acquisition module is used to periodically acquire multiple consecutive frames of vehicle images of the surrounding environment;
[0073] The side collision risk assessment module is used to assess the side collision risk based on the vehicle image. If there is a collision risk, the corresponding target vehicle is obtained.
[0074] The trajectory prediction initialization module is used to obtain the initial driving state information of the vehicle and the target vehicle;
[0075] The trajectory prediction module is used to predict the driving trajectories of the vehicle and the target vehicle based on the initial driving state information.
[0076] The side collision time prediction module is used to predict the collision time based on the predicted driving trajectory.
[0077] In the above implementation process, the vehicle's image captured by the vehicle's camera can be used to obtain the pose information and driving data of surrounding vehicles through information fusion. This method is low-cost, avoids the use of expensive LiDAR, and does not require large model training or vehicle-to-everything (V2X) connectivity. It can accurately predict collision time through geometric analysis, thus solving the problems of high cost and inaccurate prediction results of existing methods.
[0078] Furthermore, the side collision risk assessment module includes:
[0079] A vehicle recognition module is used to identify vehicles in the vehicle image using a pre-trained dynamic target detection model;
[0080] The heading angle calculation module is used to assign an ID to the vehicle and obtain the heading angle of the vehicle relative to itself.
[0081] The risk assessment module is used to identify vehicles with the same ID as the target vehicle if their heading angles are greater than a set threshold in multiple consecutive frames of vehicle images.
[0082] In the above implementation process, the risk of collision is determined based on the heading angle relative to the vehicle itself, thereby identifying the target vehicle.
[0083] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor runs the computer program to enable the electronic device to perform the vehicle side collision time prediction method described above.
[0084] This application also provides a readable storage medium storing computer program instructions, which, when read and executed by a processor, perform the vehicle side collision time prediction method described above. Attached Figure Description
[0085] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0086] Figure 1 A flowchart of a vehicle side collision time prediction method provided in this application embodiment;
[0087] Figure 2 A flowchart for determining the side collision risk provided in this application embodiment;
[0088] Figure 3 A flowchart for obtaining initial driving status information provided in this application embodiment;
[0089] Figure 4 A flowchart for collision time prediction provided in an embodiment of this application;
[0090] Figure 5 A flowchart of near-range calculation provided for embodiments of this application;
[0091] Figure 6The second direction vector is provided for the embodiments of this application. Distance from the first direction vector A schematic diagram of the nearest point P0 on the line;
[0092] Figure 7 The first direction vector is provided for the embodiments of this application. Distance from the second direction vector A schematic diagram of the nearest point P1;
[0093] Figure 8 The second direction vector is provided for the embodiments of this application. Distance from the first direction vector A schematic diagram of the nearest point P2;
[0094] Figure 9 A structural block diagram of a vehicle side collision time prediction device provided in this application embodiment;
[0095] Figure 10 This application provides a timing diagram for predicting vehicle side collision time in an embodiment.
[0096] Figure 11 A structural block diagram of another vehicle side collision time prediction device provided in an embodiment of this application.
[0097] icon:
[0098] 100 - Vehicle image acquisition module; 200 - Side collision risk assessment module; 201 - Vehicle recognition module; 202 - Heading angle calculation module; 203 - Risk assessment module; 300 - Trajectory prediction initialization module; 400 - Trajectory prediction module; 500 - Side collision time prediction module. Detailed Implementation
[0099] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.
[0100] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0101] Example 1
[0102] Please refer to Figure 1 , Figure 1This is a flowchart illustrating a method for predicting vehicle side collision time according to an embodiment of this application. This application obtains driving data of the vehicle and the target vehicle through information fusion and uses a novel lightweight collision detection algorithm to accurately predict the collision time, while minimizing reliance on perception information and trigonometric function calculations.
[0103] The method specifically includes the following steps:
[0104] Step S100: Periodically acquire continuous multi-frame vehicle images of the surrounding environment;
[0105] The vehicle's driving status and the surrounding road environment are constantly changing, so it is necessary to conduct a side collision risk assessment every certain period of time.
[0106] In this case, for multi-frame vehicle images, such as continuously acquiring N frames of images taken by the vehicle camera, N needs to be greater than a certain value to avoid misjudgment.
[0107] Step S200: Determine the side collision risk based on the vehicle image. If there is a collision risk, obtain the corresponding target vehicle.
[0108] Step S300: Obtain the initial driving status information of the vehicle and the target vehicle;
[0109] Step S400: Predict the driving trajectories of the vehicle and the target vehicle based on the initial driving state information;
[0110] Step S500: Predict the collision time based on the predicted driving trajectory.
[0111] like Figure 2 The diagram shown is a flowchart for determining the side collision risk. Step S200 specifically includes the following steps:
[0112] Step S201: Identify vehicles in the vehicle image using a pre-trained dynamic target detection model;
[0113] Step S202: Assign an ID to the vehicle and obtain the heading angle of the vehicle relative to the self-vehicle;
[0114] Step S203: If the heading angle of a vehicle with the same ID is greater than a set threshold in multiple consecutive vehicle images, there is a risk of collision, and the vehicle is marked as the target vehicle.
[0115] If the heading angle of the same vehicle ID is greater than a certain threshold in N frames of vehicle images, it indicates that the vehicle's trajectory will intersect with the vehicle's trajectory in the future, meaning there is a possibility of collision between the two vehicles. In this case, the vehicle is set as the target vehicle and marked, and the side collision prediction function is triggered. Otherwise, it indicates that there is no risk of collision for the vehicle, and the next timing cycle is directly entered.
[0116] In step S300, the position of the vehicle is represented by the position of the rear axle center point. A reference coordinate system is established with the current rear axle center point as the origin and the vehicle's heading as the X-axis. The initialization time is T. Since the initialization time is short, the headings of the vehicle and the target vehicle are set to not change significantly within the initialization time T. That is, it is simplified to the vehicle moving in a straight line along the X-axis and the target vehicle moving in a straight line with a trend of gradually approaching the X-axis.
[0117] like Figure 3 The diagram shows the flowchart for obtaining initial driving status information. Step S300 specifically includes the following steps:
[0118] Step S301: Within the initialization time T, acquire the vehicle speed message information sent by the vehicle bus and the total number of messages is K1 frames. At the same time, acquire K2 frames of vehicle images and record the timestamp of each frame of vehicle image.
[0119] Step S302: Calculate the average longitudinal acceleration of the vehicle based on the vehicle speed message information. ;
[0120] Step S303: Use a pre-trained dynamic target detection model to obtain the position information and heading angle information of the target vehicle relative to the vehicle in each frame of the vehicle image;
[0121] Step S304: Interpolate the vehicle speed information at the same time in each frame of the vehicle image to obtain the vehicle speed information at the same time from the vehicle speed message information;
[0122] Since the image processing cycle is longer than the message sending cycle, the number of K2 will be less than that of K1. Based on the acquired vehicle speed message information of K1 frame, the vehicle speed information at the same time as each frame image is interpolated to align the vehicle data with the target vehicle data in time.
[0123] Step S305: Based on the relative positional changes between the target vehicle and the self-vehicle, calculate the lateral velocity and longitudinal velocity of the target vehicle at each frame of the vehicle image, as well as the average lateral acceleration of the target vehicle within the initialization time T. Mean longitudinal acceleration and the average heading angle relative to the vehicle ;
[0124] Step S306: Use the vehicle's driving state information at the moment of the last vehicle image frame as the vehicle's initial driving state information: longitudinal speed Mean longitudinal acceleration ;
[0125] Step S307: Use the target vehicle's driving state information at the moment of the last vehicle image frame as the target vehicle's initial driving state information: lateral velocity. Longitudinal velocity Mean lateral acceleration Mean longitudinal acceleration The average heading angle of the target vehicle relative to its own vehicle and the target vehicle's position relative to its own vehicle. .
[0126] In this application, the initial driving state information of the vehicle and the target vehicle is set to remain unchanged during trajectory prediction, which is simplified to the vehicle moving in a straight line along the X-axis and the target vehicle moving in a straight line with a trend of gradually approaching the X-axis.
[0127] In step S400, the driving trajectories of the vehicle and the target vehicle are predicted, specifically:
[0128] No. M The position information of the vehicle and the target vehicle during the prediction period is represented as follows:
[0129] Longitudinal position of the vehicle:
[0130] ;
[0131] Longitudinal position of the target vehicle:
[0132] ;
[0133] Lateral position of the target vehicle:
[0134] ;
[0135] in, This indicates the set prediction time step.
[0136] In step S500, collision detection is performed based on the predicted trajectories of the self-vehicle and the target vehicle and the vehicle's geometric features. If a collision is confirmed, the collision time is obtained based on the predicted location of the collision.
[0137] like Figure 4 The diagram shown is a flowchart for collision time prediction. Step S500 specifically includes the following steps:
[0138] Step S510: Obtain the predicted position of the vehicle at the current moment. Predicted location of the target vehicle ;
[0139] Step S520: Obtain the length of the target vehicle based on the pre-trained dynamic target detection model. And the length of the vehicle has always been known to be ;
[0140] Step S530: Obtain the first direction vector corresponding to the vehicle. ,in, Indicates the vehicle's heading angle;
[0141] Since the reference coordinate system is the vehicle itself, then It is 0.
[0142] Step S540: Obtain the second direction vector corresponding to the target vehicle. ;
[0143] Any point on the first direction vector and any point on the second direction vector are represented as follows:
[0144] ;
[0145] ;
[0146] in, and Indicates the proportionality coefficient;
[0147] Obtain the direction vector from any point on the first direction vector to any point on the second direction vector. :
[0148] ;
[0149] Obtain the projections of the direction vectors onto the first and second direction vectors respectively:
[0150] ;
[0151] ;
[0152] Step S550: Calculate the closest distance between the first direction vector and the second direction vector;
[0153] Step S560: If the set safe distance is greater than the nearest distance, there is a risk of collision. Based on the predicted position of the vehicle at the current moment... Predicted collision time .
[0154] Specifically, based on the different lengths of target vehicles, the safe distance between the vehicle and the target vehicle varies. A one-dimensional table is set up to map different safe distances according to the length of the target vehicle. If the safe distance is greater than L, it indicates that there is a risk of collision.
[0155] Among them, such as Figure 5The diagram shows the flowchart for calculating the closest distance. Specifically, the calculation of the closest distance between the first direction vector and the second direction vector is as follows:
[0156] Step S551: Obtain the second direction vector Distance from the first direction vector The nearest point P0 on the line:
[0157] First direction vector Second direction vector The lines intersect at point O, satisfying... According to the projection formula, the scaling factor corresponding to the intersection point O can be obtained. The solution for P0 can be divided into three scenarios:
[0158] like This indicates that the intersection point O is in the second direction vector. If the endpoint is ahead, then the nearest point P0 is the second direction vector. The finish line, making P0 can be obtained, such as Figure 6 As shown, this is to find the second direction vector. Distance from the first direction vector A schematic diagram of the nearest point P0 on the line;
[0159] like This indicates that the intersection point O is in the second direction vector. If the points are above, then the nearest point P0 is the intersection point O;
[0160] like This indicates that the intersection point O is in the second direction vector. If the point is behind the starting point, then the nearest point P0 is the second direction vector. The starting point, We can obtain P0;
[0161] Step S552: Obtain the first direction vector based on P0. Distance from the second direction vector The nearest point P1:
[0162] From P0 to the first direction vector Draw a perpendicular line from the line containing the line, with the foot of the perpendicular at V1. The proportionality coefficient corresponding to V1 is calculated. The solution for P1 can be divided into three scenarios:
[0163] like This indicates that the foot of the perpendicular V1 is in the first direction vector. If the endpoint is ahead, then the nearest point P1 is the first direction vector. The finish line, making P1 can be obtained, such as Figure 7As shown, this is to find the first direction vector. Distance from the second direction vector A schematic diagram of the nearest point P1;
[0164] like This indicates that the foot of the perpendicular V1 is in the first direction vector. If the point is above, then the nearest point P1 is the foot of the perpendicular V1, according to... P1 can be obtained;
[0165] like This indicates that the foot of the perpendicular V1 is in the first direction vector. If the point is behind the starting point, then the nearest point P1 is the first direction vector. The starting point, We can obtain P1.
[0166] Step S553: Obtain the second direction vector based on P1. Distance from the first direction vector The nearest point P2:
[0167] From P1 to the second direction vector Draw a perpendicular line from the line containing the line, with the foot of the perpendicular at V2. The proportionality coefficient corresponding to V2 can be calculated. P2 can be solved in three scenarios:
[0168] like This indicates that the foot of the perpendicular V2 is in the second direction vector. If the endpoint is ahead, then the nearest point P2 is the second direction vector. The finish line, making P2 can be obtained;
[0169] like This indicates that the foot of the perpendicular V2 is in the second direction vector. If the point is above, then the nearest point P2 is the foot of the perpendicular V2, according to... P2 can be obtained, such as Figure 8 As shown, this is to find the second direction vector. Distance from the first direction vector A schematic diagram of the nearest point P2;
[0170] like This indicates that the foot of the perpendicular V2 is in the second direction vector. If the point is behind the starting point, then the nearest point P2 is the second direction vector. The starting point, P2 can be obtained;
[0171] Step S554: Calculate the distance between P1 and P2, which is the shortest distance between the first direction vector and the second direction vector.
[0172] This application uses an information fusion method to obtain driving information of the self-vehicle and the target vehicle. By using continuous frame images captured by a visual camera and based on a trained model, the pose information of surrounding vehicles can be obtained. Combined with the vehicle's speed information, and by aligning the time axis, the driving information of the self-vehicle and the target vehicle can be obtained. This solution is low-cost, avoids the use of expensive LiDAR, does not require large model training, and does not require vehicle-to-everything (V2X) connectivity.
[0173] This application does not use traditional collision algorithms. Instead, it simplifies the vehicle into a direction vector. It only needs to obtain the vehicle's length and heading angle information through visual perception. Then, through driving data, it can obtain the shortest distance between the vehicle and the target vehicle through geometric analysis. Combined with trajectory prediction, it can estimate the lateral collision time relatively accurately.
[0174] The collision detection method of this application can also be applied to other scenarios involving collision detection, such as autonomous driving and driver assistance, and no limitation is made on the scope of application here.
[0175] Example 2
[0176] This application provides a vehicle side collision time prediction device, such as... Figure 9 The diagram shown is a structural block diagram of a vehicle side collision time prediction device, which includes, but is not limited to:
[0177] The vehicle image acquisition module 100 is used to periodically acquire multiple consecutive frames of vehicle images of the surrounding environment;
[0178] The side collision risk assessment module 200 is used to assess the side collision risk based on the vehicle image. If there is a collision risk, the corresponding target vehicle is obtained.
[0179] The trajectory prediction initialization module 300 is used to obtain the initial driving state information of the vehicle and the target vehicle;
[0180] The trajectory prediction module 400 is used to predict the driving trajectories of the vehicle and the target vehicle based on the initial driving state information.
[0181] The lateral collision time prediction module 500 is used to predict the collision time based on the predicted driving trajectory.
[0182] like Figure 10 As shown, this is a time series diagram for predicting the time of a vehicle side collision, where, as Figure 11 The diagram shown is a structural block diagram of another vehicle side collision time prediction device. The side collision risk determination module 200 includes:
[0183] The vehicle recognition module 201 is used to identify vehicles in the vehicle image using a pre-trained dynamic target detection model;
[0184] The heading angle calculation module 202 is used to assign an ID to the vehicle and obtain the heading angle of the vehicle relative to itself.
[0185] The risk determination module 203 is used to identify a collision risk if the heading angle of a vehicle with the same ID is greater than a set threshold in multiple consecutive frames of vehicle images, and to mark the vehicle as a target vehicle.
[0186] The specific implementation process of the trajectory prediction initialization module 300, the trajectory prediction module 400, and the side collision time prediction module 500 has been described in detail in Example 1, and will not be repeated here.
[0187] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor runs the computer program to enable the electronic device to perform the vehicle side collision time prediction method described in Embodiment 1.
[0188] This application also provides a readable storage medium storing computer program instructions. When the computer program instructions are read and executed by a processor, the vehicle side collision time prediction method described in Embodiment 1 is performed.
[0189] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0190] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0191] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0192] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0193] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0194] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
Claims
1. A method for predicting vehicle side collision time, characterized in that, The method includes: Periodically acquire continuous multi-frame images of the vehicle's surrounding environment; The side collision risk is determined based on the vehicle image. If there is a collision risk, the corresponding target vehicle is obtained. Obtain the initial driving status information of the vehicle and the target vehicle; The driving trajectories of the self-vehicle and the target vehicle are predicted based on their initial driving state information, respectively. The collision time is predicted based on the predicted driving trajectory, specifically including: obtaining the predicted position of the vehicle at the current moment. Predicted location of the target vehicle The length of the target vehicle is obtained based on a pre-trained dynamic target detection model. And the length of the vehicle is known to be Obtain the first direction vector corresponding to the vehicle. ,in, Represents the vehicle's heading angle; obtains the second direction vector corresponding to the target vehicle. Any point on the first direction vector and any point on the second direction vector are represented as follows: ; ;in, and Represents the scaling factor; obtains the direction vector from any point on the first direction vector to any point on the second direction vector. : ; Obtain the projections of the direction vectors onto the first and second direction vectors respectively: ; Calculate the closest distance between the first direction vector and the second direction vector; if the set safety distance is greater than the closest distance, there is a risk of collision, based on the predicted position of the vehicle at the current moment. Predicted collision time .
2. The vehicle side collision time prediction method according to claim 1, characterized in that, The step of determining the lateral collision risk based on the vehicle image, and obtaining the corresponding target vehicle if there is a collision risk, includes: The vehicle in the vehicle image is identified using a pre-trained dynamic object detection model; Assign an ID to the vehicle and obtain the heading angle of the vehicle relative to the self-vehicle; If the heading angle of a vehicle with the same ID is greater than a set threshold in multiple consecutive vehicle images, there is a risk of collision, and the vehicle is marked as the target vehicle.
3. The vehicle side collision time prediction method according to claim 1, characterized in that, The acquisition of the initial driving status information of the vehicle and the target vehicle includes: Within the initialization time T, the vehicle speed message information sent by the vehicle bus is obtained, and the total number of messages is K1 frames. At the same time, K2 frames of vehicle images are obtained, and the timestamp of each frame of vehicle image is recorded. Calculate the vehicle's average longitudinal acceleration based on the vehicle speed message information. ; The pre-trained dynamic target detection model is used to obtain the position and heading angle information of the target vehicle relative to the vehicle in each frame of the vehicle image; The vehicle speed information at the same time in each frame of the vehicle image is obtained by interpolation from the vehicle speed message information. Based on the relative positional changes between the target vehicle and the driver vehicle, the lateral velocity and longitudinal velocity of the target vehicle at each frame of the vehicle image, as well as the average lateral acceleration of the target vehicle within the initialization time T, are calculated. Mean longitudinal acceleration and the average heading angle relative to the vehicle ; The vehicle's driving state information at the moment of the last vehicle image is used as the vehicle's initial driving state information: longitudinal speed. Mean longitudinal acceleration ; The target vehicle's driving state information at the moment of the last vehicle image is used as the target vehicle's initial driving state information: lateral velocity. Longitudinal velocity Mean lateral acceleration Mean longitudinal acceleration The average heading angle of the target vehicle relative to its own vehicle and the target vehicle's position relative to its own vehicle. .
4. The vehicle side collision time prediction method according to claim 3, characterized in that, The prediction of the driving trajectories of the self-vehicle and the target vehicle based on the initial driving state information includes: Using the first M Predicting the driving trajectory based on the position information of the vehicle and the target vehicle during the prediction period: Longitudinal position of the vehicle: ； Longitudinal position of the target vehicle: ； Lateral position of the target vehicle: ； in, This indicates the set prediction time step.
5. The vehicle side collision time prediction method according to claim 1, characterized in that, The calculation of the closest distance between the first direction vector and the second direction vector includes: Obtain the second direction vector Distance from the first direction vector The nearest point P0 on the line: First direction vector Second direction vector The lines intersect at point O, satisfying... According to the projection formula, the scaling factor corresponding to the intersection point O can be obtained. : like Then the nearest point P0 is the second direction vector. The finish line, making We can obtain P0; like If the nearest point P0 is the intersection point O; like Then the nearest point P0 is the second direction vector. The starting point, making We can obtain P0; Based on P0, obtain the first direction vector. Distance from the second direction vector The nearest point P1: From P0 to the first direction vector Draw a perpendicular line from the line containing the line, with the foot of the perpendicular at V1. The proportionality coefficient corresponding to V1 is calculated. : like Then the nearest point P1 is the first direction vector. The finish line, making P1 can be obtained; like Then the nearest point P1 is the foot of the perpendicular V1, according to P1 can be obtained; like Then the nearest point P1 is the first direction vector. The starting point, making P1 can be obtained; Based on P1, obtain the second direction vector. Distance from the first direction vector The nearest point P2: From P1 to the second direction vector Draw a perpendicular line from the line containing the line, with the foot of the perpendicular at V2. The proportionality coefficient corresponding to V2 can be calculated. : like Then the nearest point P2 is the second direction vector. The finish line, making P2 can be obtained; like Then the nearest point P2 is the foot of the perpendicular V2, according to P2 can be obtained; like Then the nearest point P2 is the second direction vector. The starting point, making P2 can be obtained; The distance between P1 and P2 is the shortest distance between the first direction vector and the second direction vector.
6. A vehicle side collision time prediction device, characterized in that, The device includes: The vehicle image acquisition module is used to periodically acquire multiple consecutive frames of vehicle images of the surrounding environment; The side collision risk assessment module is used to assess the side collision risk based on the vehicle image. If there is a collision risk, the corresponding target vehicle is obtained. The trajectory prediction initialization module is used to obtain the initial driving state information of the vehicle and the target vehicle; The trajectory prediction module is used to predict the driving trajectories of the vehicle and the target vehicle based on the initial driving state information. The side collision time prediction module is used to predict the collision time based on the predicted driving trajectory. Specifically, it includes: obtaining the predicted position of the vehicle at the current moment. Predicted location of the target vehicle The length of the target vehicle is obtained based on a pre-trained dynamic target detection model. And the length of the vehicle is known to be Obtain the first direction vector corresponding to the vehicle. ,in, Represents the vehicle's heading angle; obtains the second direction vector corresponding to the target vehicle. Any point on the first direction vector and any point on the second direction vector are represented as follows: ; ;in, and Represents the scaling factor; obtains the direction vector from any point on the first direction vector to any point on the second direction vector. : ; Obtain the projections of the direction vectors onto the first and second direction vectors respectively: ; Calculate the closest distance between the first direction vector and the second direction vector; if the set safety distance is greater than the closest distance, there is a risk of collision, based on the predicted position of the vehicle at the current moment. Predicted collision time .
7. The vehicle side collision time prediction device according to claim 6, characterized in that, The lateral collision risk assessment module includes: A vehicle recognition module is used to identify vehicles in the vehicle image using a pre-trained dynamic target detection model; The heading angle calculation module is used to assign an ID to the vehicle and obtain the heading angle of the vehicle relative to itself. The risk assessment module is used to identify vehicles with the same ID as the target vehicle if their heading angles are greater than a set threshold in multiple consecutive frames of vehicle images.
8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program and the processor running the computer program to cause the electronic device to perform the vehicle side collision time prediction method according to any one of claims 1 to 5.
9. A readable storage medium, characterized in that, The readable storage medium stores computer program instructions, which, when read and executed by a processor, perform the vehicle side collision time prediction method according to any one of claims 1 to 5.