Vehicle collision warning methods, equipment and storage media
By using onboard stereo vision sensors in autonomous vehicles to obtain the coordinates of the rear corner points of target objects and performing nonlinear dynamic trajectory modeling, the problem of inaccurate perception in existing technologies is solved, enabling accurate collision warnings and decision support, and improving the safety of autonomous driving.
Patent Information
- Application Number
- CN202510595733.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2026-11-10
AI Technical Summary
Current environmental perception technologies for autonomous vehicles mainly rely on radar scanning, which cannot accurately predict the movement trends of target objects. This makes it difficult to establish forward-looking collision risk prediction models and cannot provide a complete basis for decision-making for autonomous driving systems.
Based on a coordinate system with the vehicle as the origin, the coordinates of the rear corner of the target object are obtained through an onboard stereo vision sensor, time series data is generated, nonlinear dynamic trajectory modeling is performed, the motion trajectory of the target object is predicted, and the trajectory intersection is analyzed to trigger a collision warning.
It improves the accuracy and real-time performance of collision warnings, provides comprehensive decision-making support, and enhances the safety of autonomous driving systems.
Smart Images

Figure CN123072227A_ABST
Abstract
Description
Technical Field
[0001] This specification relates to the field of autonomous driving technology, and in particular to vehicle collision warning methods, devices and storage media. Background Technology
[0002] During the operation of an autonomous driving system, continuous and accurate environmental perception is a core function to ensure driving safety. When there are pedestrians or other moving vehicles (such as other vehicles) in the road scene and there is relative displacement with the autonomous vehicle, the autonomous vehicle needs to monitor its relative position with these dynamic targets in real time to establish an effective obstacle avoidance strategy, thereby avoiding the risk of collision.
[0003] Currently, most autonomous vehicles rely on lidar or millimeter-wave radar for their autonomous driving environment perception solutions. By actively emitting detection signals and receiving reflected echoes, they can achieve spatial positioning and identification of target objects within a safe range, thereby sensing whether there are target objects within the safe range that may affect the vehicle's driving.
[0004] However, this radar-scanning-based perception technology has significant limitations, as its perception logic primarily focuses on instantaneous static data acquisition. This approach, which relies solely on perception data at the current moment, may not only misjudge the movement trend of target objects due to inaccurate perception, but also makes it difficult to establish forward-looking collision risk prediction models, thus failing to provide a complete decision-making basis for autonomous driving systems. Summary of the Invention
[0005] To overcome the problems existing in related technologies, this specification provides a vehicle collision warning method, device and storage medium.
[0006] According to a first aspect of the embodiments of this specification, a method is provided, the method comprising:
[0007] Based on a coordinate system with the vehicle as the origin, the coordinates of the tail corner of the target object at each moment in a continuous time period are obtained to generate time series data.
[0008] Based on the time series data, nonlinear dynamic trajectory modeling is performed to predict the motion trajectory of the corner points of the target object;
[0009] The motion trajectory is analyzed, and an early warning is triggered when the corner of the target object collides with the path of the vehicle in the current dynamic coordinate system.
[0010] According to the vehicle collision warning method provided in this specification, the nonlinear dynamic trajectory modeling employs polynomial fitting.
[0011] The step of performing nonlinear dynamic trajectory modeling based on the time series data to predict the motion trajectory of the corner points of the target object includes:
[0012] The time series data is fitted with an Nth-order polynomial to generate a trajectory equation representing the motion trajectory of the corner point of the target object, wherein the order N is positively correlated with the motion complexity of the target object.
[0013] According to a vehicle collision warning method provided in this specification, the step of parsing the motion trajectory, triggering a warning when a corner point of the target object collides with the path of the vehicle in the current dynamic coordinate system, includes:
[0014] Analyze the motion trajectory to determine the intersection point between the motion trajectory and the path of the vehicle in the current dynamic coordinate system;
[0015] Determine whether the intersection point is located within the lateral width range of the vehicle;
[0016] When the intersection point is within the lateral width range of the vehicle, it is determined that there is a collision point between the corner point of the target object and the path of the vehicle in the current dynamic coordinate system, and an early warning is triggered.
[0017] According to a vehicle collision warning method provided in this specification, the step of parsing the motion trajectory and determining the intersection point of the motion trajectory and the path of the vehicle in the current dynamic coordinate system includes:
[0018] The trajectory equation corresponding to the motion trajectory is solved, and the position corresponding to the real solution obtained when the trajectory equation is zero is determined as the intersection point. The trajectory equation is obtained by fitting the time series data with an Nth-order polynomial.
[0019] According to the vehicle collision warning method provided in this specification, the method further includes:
[0020] The intersection points located within the lateral width of the vehicle are determined as the collision points where there is a risk of collision between the target object and the vehicle;
[0021] Based on the location of the collision point, the collision warning time when the vehicle collides with the target object is predicted.
[0022] According to a vehicle collision warning method provided in this specification, the step of predicting the collision warning time when the vehicle collides with the target object based on the location of the collision point includes:
[0023] Determine the path length from the current position of the corner point of the target object to the collision point along the motion trajectory;
[0024] The collision warning time is obtained based on the path length and the instantaneous relative speed between the vehicle and the target object.
[0025] According to the vehicle collision warning method provided in this specification, the vehicle includes an onboard stereo vision sensor, the onboard stereo vision sensor being aligned with the origin of the coordinate system of the vehicle.
[0026] The method involves obtaining the coordinates of the tail corner of the target object at each moment in a continuous time period based on a coordinate system with the vehicle as the origin, and generating time series data, including:
[0027] For each moment in a continuous time period, perform the following steps:
[0028] The vehicle-mounted stereo vision sensor acquires a dense point cloud image of the rear of the target object, which is located in front of the vehicle in the direction of travel.
[0029] The point cloud of the dense point cloud image is vertically projected onto the ground to determine the projected coordinates of the tail corner of the target object;
[0030] Time series data is generated based on the projected coordinates of the tail corner points at multiple times.
[0031] According to the vehicle collision warning method provided in this specification, the vehicle-mounted stereo vision sensor includes a binocular camera.
[0032] According to the vehicle collision warning method provided in this specification, the corner points include the left and right rear corners of the target object's rear end.
[0033] The step of vertically projecting the point cloud of the dense point cloud image onto the ground to determine the projected coordinates of the tail corner point of the target object includes:
[0034] The dense point cloud image is vertically projected onto the ground, and the projected points are fitted with straight lines to obtain the slope of the line connecting the left rear corner and the right rear corner after projection.
[0035] Determine the projected coordinates of the center point of the tail of the target object;
[0036] Based on the tail width, slope, and projected coordinates of the center point of the target object, the projected coordinates of the left rear corner and the right rear corner of the target object are determined.
[0037] According to a vehicle collision warning method provided in this specification, the rear width of the target object is obtained in the following way:
[0038] The vehicle-mounted stereo vision sensor acquires an RGB image of the rear of the target object.
[0039] When the target object is a vehicle, the RGB image is identified to determine the vehicle information of the target object;
[0040] Based on the vehicle information, the size information of the target object is found, and the tail width of the target object is obtained;
[0041] And / or,
[0042] Determine the smallest bounding rectangle covering the projection point along the direction of the slope;
[0043] The horizontal length of the minimum bounding rectangle is determined to be the tail width of the target object.
[0044] According to the vehicle collision warning method provided in this specification, the continuous time period includes n sampling times, and the value of n is negatively correlated with the speed of the target object relative to the vehicle.
[0045] Specifically, when the relative speed decreases, the value of n increases accordingly; when the relative speed increases, the value of n decreases accordingly.
[0046] According to a second aspect of the embodiments of this specification, an apparatus is provided, comprising:
[0047] The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements a vehicle collision warning method as described above.
[0048] According to a third aspect of the embodiments of this specification, a computer-readable storage medium is provided, comprising:
[0049] When the computer program is executed by the processor, it implements a vehicle collision warning method as described above.
[0050] The technical solutions provided in the embodiments of this specification may include the following beneficial effects:
[0051] In this embodiment, based on a coordinate system with the vehicle as the origin, the coordinates of the tail corner of the target object at each moment in a continuous time period are obtained to generate time series data. This time series data fully considers the dynamic interaction characteristics between the target object and the autonomous vehicle, accurately perceiving the relative positional changes of the vehicle and other vehicles. When performing nonlinear dynamic trajectory modeling based on the time series data, the motion trajectory of the corner of the target object in complex motion scenarios can be predicted, improving the accuracy of trajectory prediction and the reliability of collision warning. Subsequently, the motion trajectory is analyzed, and when a collision point exists between the corner of the target object and the path of the vehicle in the current dynamic coordinate system, a warning is quickly triggered, improving the real-time performance of the warning.
[0052] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description
[0053] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this specification and, together with the description, serve to explain the principles of this specification.
[0054] Figure 1 This is a flowchart illustrating a method according to an exemplary embodiment of this specification.
[0055] Figure 2 This is a schematic diagram of a collision warning system illustrated in this specification according to an exemplary embodiment.
[0056] Figure 3 This is a block diagram illustrating an apparatus according to an exemplary embodiment of this specification.
[0057] Figure 4 This is a schematic diagram of a device illustrated in this specification according to an exemplary embodiment. Detailed Implementation
[0058] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this specification as detailed in the appended claims.
[0059] The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of this specification. The singular forms “a,” “the,” and “the” as used in this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.
[0060] It should be understood that although the terms first, second, third, etc., may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this specification, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."
[0061] This specification provides a vehicle collision warning method, device, and computer-readable storage medium. The embodiments of this specification are described in detail below with reference to the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0062] Currently, most autonomous vehicles use LiDAR or millimeter-wave radar for environmental perception, which is used to detect the presence of objects affecting the vehicle's movement within a safe distance. However, this approach only considers the current state and ignores the dynamic changes of the target objects and the autonomous vehicle. Therefore, the perception is inaccurate, and it cannot predict in advance whether a collision will occur between the autonomous vehicle and an obstacle, failing to provide a complete basis for the autonomous driving system's decision-making.
[0063] To address the aforementioned technical problems, this specification provides a method for vehicle collision warning.
[0064] Aimed at fully considering the dynamic interaction characteristics between autonomous vehicles and target objects, by fitting a coordinate system updated in real time and data sequences at multiple time points, it can predict the trajectory of autonomous vehicles, accurately perceive the relative position changes between autonomous vehicles and target objects, realize collision warning, and improve the accuracy and real-time performance of collision warning.
[0065] By using such technologies to establish forward-looking collision risk prediction models, we can not only improve the accuracy of perception, the accuracy and real-time performance of collision warnings, but also provide a complete decision-making basis for autonomous driving systems and improve the safety of autonomous driving.
[0066] It should be noted that the vehicle collision warning solution in this manual can be applied to any device or system in terminals, vehicles, independent servers, server clusters, cloud computing, vehicle network platforms, IoT platforms, etc.
[0067] This specification provides an embodiment of a vehicle collision warning method.
[0068] like Figure 1 As shown, Figure 1 This is a flowchart illustrating a method according to an exemplary embodiment, comprising the following steps:
[0069] In step 102, based on the coordinate system with the vehicle as the origin, the coordinates of the tail corner of the target object at each moment in the continuous time period are obtained to generate time series data.
[0070] The "autonomous vehicle" mentioned in this manual refers to an autonomous driving vehicle, which has the conventional structure and functions of vehicles on the market. For details, please refer to the descriptions in the automotive field, which will not be repeated here.
[0071] Autonomous vehicles can use various sensors installed on the vehicle to perceive their surroundings in real time during operation, acquiring motion information of target objects in the environment. This motion information is then used to predict collision risks. The motion information includes the coordinates of the corner points of the target object's rear end. The target object refers to pedestrians or other moving vehicles (such as other vehicles) that may affect the vehicle's movement. The following explanation uses a vehicle as the target object (hereinafter referred to as "other vehicle") as an example.
[0072] When perceiving the surrounding environment, the first step is to create the vehicle's coordinate system. The specific process is as follows:
[0073] A two-dimensional coordinate system is established by selecting the center point of the vehicle as the origin, with the positive x-axis pointing horizontally to the right and the positive y-axis pointing horizontally forward. Since the vehicle is in a dynamic state during the perception process, this two-dimensional coordinate system is updated in real time as the vehicle moves, meaning it is a dynamic coordinate system. This can be understood as follows: when acquiring the coordinates of the rear corner points of a target object at different times to generate time-series data, the origin of the coordinate system is different at each moment, and the obtained corner point coordinates are determined relative to the instantaneous position of the origin. Therefore, when the vehicle reaches a new origin, it is necessary to unify the historical data through coordinate transformation, converting the historical coordinates to the latest coordinate system to reflect the true trajectory of the preceding vehicle. Furthermore, subsequent accumulation of multi-time-moment data and trajectory fitting are also performed under a unified coordinate system to ensure the accuracy of trajectory fitting.
[0074] By using the dynamic coordinate system of the vehicle, the dynamic interaction characteristics between the vehicle and other vehicles can be captured, and the future unknowns of other vehicles on the path of the vehicle can be accurately predicted, thereby achieving accurate prediction of dynamic collision risks.
[0075] As an example, a two-dimensional coordinate system is established with the center point of the part of the vehicle that may collide as the origin, based on the relative position of the target vehicle and the vehicle itself. Collision risk warning is achieved by predicting the movement trajectory of the corner points of the corresponding parts of the target vehicle.
[0076] For example, refer to Figure 2 If another vehicle is located in front of the vehicle in its direction of travel, a collision may occur between the front and rear of the other vehicle. Therefore, a two-dimensional coordinate system is created with the center point of the front of the vehicle as the origin, and the trajectory of the corner points of the other vehicle's rear is predicted to provide a collision risk warning. Similarly, if another vehicle is located behind the vehicle in its direction of travel, a collision may occur between the rear and front of the other vehicle. Therefore, a two-dimensional coordinate system is created with the center point of the rear of the vehicle as the origin, and the trajectory of the corner points of the other vehicle's front is predicted to provide a collision risk warning. Other examples of relative positional relationships between the vehicle and other vehicles are essentially the same and will not be described further.
[0077] After creating the dynamic coordinate system of the vehicle, the vehicle-mounted stereo vision sensor is installed at the position corresponding to the origin of the vehicle's dynamic coordinate system. To avoid secondary coordinate conversion during the prediction process, the origin of the coordinate system of the vehicle-mounted stereo vision sensor is aligned with the origin of the vehicle's dynamic coordinate system during installation, simplifying the operation process.
[0078] It should be noted that the vehicle-mounted stereo vision sensor described in this specification has stereo vision and depth perception attributes. It can acquire RGB images of other vehicles and perceive the distance between other vehicles and its own vehicle, providing a data basis for subsequent collision risk prediction.
[0079] Vehicle-mounted stereo vision sensors include, but are not limited to, devices with depth measurement capabilities such as binocular cameras, combinations of monocular cameras and depth sensors, and RGB-D cameras (such as Kinect). In practical applications, a suitable vehicle-mounted stereo vision sensor can be selected based on factors such as data acquisition accuracy, real-time performance, and convenience. Preferably, a binocular camera is used to acquire motion information of other vehicles because it can simultaneously output RGB images and point cloud data in a single acquisition, eliminating the need for additional sensors and meeting real-time requirements.
[0080] As an example, continue to refer to Figure 2 The origin is selected as the center point of the foremost point of the vehicle in the direction of travel. A binocular camera with depth measurement function is installed at this origin, with the camera pointing horizontally forward. Subsequently, over a continuous time period, the coordinates of the rear corner points of other vehicles at multiple times are acquired using the binocular camera to generate time-series data.
[0081] In some embodiments, step 104, which involves obtaining the coordinates of the tail corner of the target object at each moment in a continuous time period based on a coordinate system with the vehicle as the origin, and generating time series data, includes steps a11 to a13:
[0082] Step a11: Based on each moment in the continuous time period, perform the following steps:
[0083] The vehicle-mounted stereo vision sensor acquires a dense point cloud image of the rear of the target object, which is located in front of the vehicle in the direction of travel.
[0084] Step a12: Vertically project the point cloud of the dense point cloud image onto the ground to determine the projection coordinates of the tail corner point of the target object;
[0085] Step a13: Generate time series data based on the projected coordinates of the tail corner points at multiple times.
[0086] The corner point of the target object's rear end, as described in this specification, refers to the vehicle boundary point on the other vehicle that first contacts the vehicle when the collision occurs. When the other vehicle is traveling in front of the vehicle, this vehicle boundary point is the left and right rear corners of the rear of the other vehicle.
[0087] At time T0, a dense point cloud image of the rear of the vehicle in front is acquired using a binocular camera, and the dense point cloud of the rear vertical surface is projected vertically downwards onto the ground (xOy plane) to obtain the projected coordinates P of the left rear corner C of the vehicle's rear. C0 The projected coordinates P of the right rear corner D D0 It should be noted that the point cloud data for the rear corner is generated from the left and right RGB images of the binocular camera using a stereo matching algorithm. Specifically, the left and right cameras of the binocular camera simultaneously capture RGB images of the rear of other vehicles in front, and use stereo matching algorithms (such as SGBM or deep learning) to calculate the disparity map of the left and right images to determine the depth information. This depth information is then converted into a 3D point cloud, forming dense 3D point cloud data of the vehicle's rear.
[0088] The projected coordinates of the two corner points, left rear corner C and right rear corner D, at time T0 are determined through the above process.
[0089] Continue to refer to Figure 2 Repeat the above process to obtain the projected coordinates of the left rear corner C corresponding to the n nearest sampling times acquired in chronological order within a continuous time period, and generate the time series data {P} corresponding to the left rear corner C. Cn}, and the projected coordinates of the right rear corner D at n sampling times, to generate the time series data {P} corresponding to the right rear corner D. Dn}. Among them, {P Cn}={P C0 P C1 P C2 ...P Cn-1}、{P Dn}={P D0 P D1 P D2 ...P Dn-1 The value of n is negatively correlated with the target object's speed relative to the vehicle's speed. It can be understood that when the relative speed between the target and vehicle decreases, the value of n increases accordingly; conversely, when the relative speed increases, the value of n decreases accordingly. For example, n can range from 8 to 20. It should be noted that continuous, equally spaced sampling is preferred within a continuous time period, with the sampling interval determined by the sensor frequency; non-uniform interval sampling is also allowed, but temporal continuity must be ensured to accurately model the dynamic trajectory.
[0090] In other embodiments, the actual point cloud may have irregular projection point distribution due to occlusion, noise, or non-rectangular tail shapes (such as trapezoids or arcs). In these cases, the outermost point is directly selected as P. C0 P D0 This may be inaccurate (for example, the outermost point may belong to a headlight rather than a true corner point), therefore, it is necessary to modify the P determined by the aforementioned discrete points. C0 P D0 Make corrections to ensure the physical accuracy of the corner coordinates.
[0091] As an example, step a12, which involves vertically projecting the point cloud of the dense point cloud image onto the ground to determine the projected coordinates of the tail corner point of the target object, includes the following steps a121-a123:
[0092] In step a121, the point cloud of the dense point cloud image is vertically projected onto the ground, and the projection points are fitted with straight lines to obtain the slope of the line connecting the left rear corner and the right rear corner after projection.
[0093] In step a122, determine the projected coordinates of the center point of the tail of the target object;
[0094] In step a123, the projection coordinates of the left rear corner and the right rear corner of the target object are determined based on the tail width, slope and projection coordinates of the center point of the target object.
[0095] The dense point cloud image of the rear facade is projected vertically downwards onto the ground (xOy plane). Straight line fitting is performed on the projected points to obtain the straight line P where the left rear corner C and the right rear corner D are projected. C0 P D0 The slope k is calculated, and the projected coordinates of the tail center point P0 are determined.
[0096] Specifically, by performing least-squares line fitting on the ground projection points, the equation of the line y = kx + b is obtained.
[0097] The center point P0 is the midpoint of the fitted line, obtained using the formula shown in the example below:
[0098]
[0099] Where j represents the number of discrete points in the dense point cloud image projected vertically downwards onto the ground.
[0100] Taking into account the rear width w and the straight slope k of the other vehicle, the width is allocated along the straight direction starting from P0, and the precise P is recalculated using the formula shown in the example below. C0 P D0 .
[0101]
[0102] in, It refers to the actual width component along the straight line (taking into account the slope).
[0103] In the above embodiments, the tail width w is obtained in the following two ways.
[0104] As an example, the vehicle-mounted stereo vision sensor acquires an RGB image of the rear of the target object; when the target object is a vehicle, the RGB image is identified to determine the vehicle information of the target object; based on the vehicle information, the size information of the target object is found to obtain the width of the rear of the target object.
[0105] The system uses a binocular camera to capture RGB images of the rear of another vehicle in front, and then uses an image recognition algorithm to determine the vehicle's information, including its brand and model. In other words, the system identifies the brand and model of the vehicle through image recognition.
[0106] Then, by searching the vehicle size information in the database, the precise rear width *w* of the vehicle in front can be obtained. Since the rear width of each model is precise data released by the automaker, it can be used directly.
[0107] As another example, if the vehicle brand and model cannot be identified from the RGB image, the projection points are analyzed to calculate the rear width w of the vehicle.
[0108] For example, along the straight line P where the left rear corner C and the right rear corner D are projected. C0 P D0 The direction of the slope is used to determine the minimum bounding rectangle covering the projection point; the horizontal length of the minimum bounding rectangle is determined to be the tail width of the target object.
[0109] Through the above embodiments, the error of discrete point projection after dense point cloud projection is corrected according to the actual rear width of the other vehicle, ensuring that the corner point coordinates P C0 and P D0 This improves the accuracy of fitting the motion trajectory of other vehicles in subsequent tests.
[0110] In step 104, nonlinear dynamic trajectory modeling is performed based on the time series data to predict the motion trajectory of the corner points of the target object.
[0111] After obtaining the projection coordinate sequence of the left and right rear corners of the tail at n different times in step 102, dynamic trajectory modeling is performed on the motion of the other vehicle to generate the motion trajectory of the other vehicle relative to the own vehicle in the dynamic relative motion scenario.
[0112] As an example, nonlinear dynamic trajectory modeling refers to the process of performing nonlinear regression analysis on time series data. This nonlinear regression analysis includes, but is not limited to, polynomial fitting, spline interpolation, and machine learning models (such as LSTM). Preferably, polynomial fitting is used to describe nonlinear motion trajectories, such as cornering, acceleration, and deceleration, which can handle complex dynamic scenarios.
[0113] For example, step 104, which involves performing nonlinear dynamic trajectory modeling based on the time series data to predict the motion trajectory of the corner points of the target object, includes:
[0114] In step b11, the time series data is fitted with an Nth-order polynomial to generate a trajectory equation representing the motion trajectory of the corner point of the target object, wherein the order N is positively correlated with the motion complexity of the target object.
[0115] By fitting the coordinates of n elements in the time series data {PCn} of the left rear corner using an Nth-order polynomial, the trajectory equation y = f of the left rear corner is obtained. C (x).
[0116] By fitting the coordinates of the n elements in {PDn} of the time series data in the right rear corner using an Nth-order polynomial, we obtain y = f D (x).
[0117] The order N is positively correlated with the motion complexity of the target vehicle. By quantifying the motion complexity, the value of N is adaptively adjusted to improve the trajectory fitting accuracy.
[0118] In some embodiments, the motion complexity is quantified by the rate of change of trajectory curvature or the standard deviation of acceleration. When the rate of change of curvature or the standard deviation of acceleration exceeds a preset threshold, the value of N is increased to improve the trajectory fitting accuracy. For example, the order N ranges from 3 to 5. For instance, when another vehicle is moving at a constant speed in a straight line, N = 3; when another vehicle is accelerating uniformly or turning gently, N = 4; when another vehicle is frequently changing lanes or accelerating / decelerating rapidly, N = 5.
[0119] In other examples, N reverts to its default value when the complexity drops below a threshold.
[0120] Through the above embodiments, by using high-order polynomial fitting, the trajectory of a vehicle in complex motion can be predicted more accurately, meeting the requirements for predicting the dynamic trajectory of a vehicle in complex environments and improving the reliability of collision warning.
[0121] In step 106, the motion trajectory is analyzed. When the corner of the target object collides with the path of the vehicle in the current dynamic coordinate system, an early warning is triggered.
[0122] The motion trajectory of other vehicles is analyzed to determine whether the corner points of other vehicles will collide with the vehicle itself during driving. When a collision is predicted, an early warning is triggered.
[0123] In some embodiments, the step 106 of parsing the motion trajectory and triggering an early warning when a corner of the target object collides with the path of the vehicle in the current dynamic coordinate system includes the following steps c11 to c13:
[0124] In step c11, the motion trajectory is analyzed to determine the intersection point of the motion trajectory and the path of the vehicle in the current dynamic coordinate system.
[0125] The origin of the vehicle's dynamic coordinate system is updated in real time as the vehicle moves, and the trajectory equation of the target object is y = f. C (x) and y = f D (x) is the coordinate system established based on the current position of the vehicle. In the dynamic coordinate system, the vehicle's path is simplified to travel along the positive y-axis (i.e., the origin of the coordinate system moves with the vehicle, and the y-axis is the direction of the vehicle's movement). Therefore, the analysis of the other vehicle's trajectory (f) C (x) = 0 and f D (x) = 0 represents the intersection point of its lateral position (x-coordinate) in the vehicle's coordinate system and the vehicle's current position (y = 0). Subsequently, based on the relationship between the intersection point coordinates and the lateral width range of the vehicle, it is determined whether there is a collision risk.
[0126] As an example, the step of parsing the motion trajectory and determining the intersection point of the motion trajectory and the path of the vehicle in the current dynamic coordinate system includes:
[0127] The trajectory equation corresponding to the motion trajectory is solved, and the position corresponding to the real solution obtained when the trajectory equation is zero is determined as the intersection point. The trajectory equation is obtained by fitting the time series data with an Nth-order polynomial.
[0128] Specifically, solve equation f C Given (x) = 0, solve for the value of x. C We obtain the intersection point of the left rear corner of his car and the path of our own car. Similarly, we solve the equation f. D Given (f) = 0, the value of x is x. D We obtain the intersection point between the right rear corner of his car and the path of our own car.
[0129] In step c12, determine whether the intersection point is located within the lateral width range of the vehicle;
[0130] The lateral width range of the vehicle described in this specification represents the left and right boundaries of the vehicle body in the direction of travel. When the solution x of the vehicle's trajectory equation... C or xD Falling into this area is considered a collision risk.
[0131] As an example, the lateral width of the vehicle ranges from [-0.5b, 0.5b], where b is the width of the vehicle body.
[0132] In step c13, when the intersection point is within the lateral width range of the vehicle, it is determined that the corner point of the target object and the path of the vehicle in the current dynamic coordinate system have a collision point, and an early warning is triggered.
[0133] Solving the aforementioned polynomial fitting equation, the following situations may occur:
[0134] No solution: The other vehicle's trajectory does not intersect with the vehicle's path, so there will be no collision.
[0135] A single valid solution: The other car will collide with the vehicle at one of its side corners (left rear corner or right rear corner).
[0136] Multiple solutions: For example, if a vehicle sways while cornering, there is a risk of multiple collisions, which may trigger multiple warnings. When calculating subsequent collision warnings, the solution with the most recent time should be selected.
[0137] As an example, if -0.5b <x C <0.5b means that the rear left corner of his car will collide with your car. If -0.5b <x D <0.5b indicates that a collision is imminent at the right rear corner of the other vehicle, triggering a warning. Where x C x D The value is the expected collision point.
[0138] In other embodiments, if calculations determine that a collision will occur at the left or right rear corner of the vehicle ahead, the time of the collision is further calculated to warn the user and improve vehicle safety. As an example, the intersection points within the lateral width of the vehicle are identified as the collision points where a collision risk exists between the target object and the vehicle; based on the location of the collision point, the collision warning time when the vehicle collides with the target object is predicted.
[0139] The step of predicting the collision warning time when the vehicle collides with the target object based on the location of the collision point includes the following steps d11 to d12:
[0140] In step d11, determine the path length from the current position of the corner point of the target object to the collision point along the motion trajectory.
[0141] The path length or arc length of the left rear corner moving from its current position to the collision point along the trajectory can be obtained using the following exemplary formula:
[0142]
[0143] Among them, P C0x P represents C0 The x-coordinate of the point.
[0144] The path length or arc length of the right rear corner moving from its current position to the collision point along the trajectory can be obtained using the following exemplary formula:
[0145]
[0146] Among them, P D0x P represents D0 The x-coordinate of the point.
[0147] In step d12, the collision warning time is obtained based on the path length and the instantaneous relative speed between the vehicle and the target object.
[0148] The calculation of collision warning time requires path integration (arc length) of the target vehicle's trajectory equation, combined with the relative speed v.
[0149] As an example, the collision warning time t for the rear left corner C. C The following exemplary formula is used to obtain:
[0150]
[0151] Collision warning time t for the rear right corner D D The following exemplary formula is used to obtain:
[0152]
[0153] Where v is the instantaneous relative velocity between the autonomous vehicle and other vehicles in front, which can be calculated from the position changes of points C at the left rear corner and D at the right rear corner.
[0154] In this embodiment, the collision time is calculated based on the integral displacement of the fitted trajectory, which is more suitable for dynamic relative motion scenarios than the traditional static distance-velocity ratio.
[0155] In other examples, tiered warning signals are triggered based on the location of the collision, the timing of the collision warning, and the relative motion of the other vehicle to the vehicle. The greater the risk, the more obvious the feedback of the warning signal.
[0156] Through the above embodiments, accurate prediction of dynamic collision risks can be achieved by using dynamic trajectory prediction and integral calculation.
[0157] This specification provides a vehicle collision warning method, device, and computer-readable storage medium. Based on a coordinate system with the vehicle as the origin, it acquires the coordinates of the tail corner of a target object at each moment in a continuous time period, generating time-series data. This time-series data fully considers the dynamic interaction characteristics between the target object and the autonomous vehicle, accurately perceiving the relative positional changes of the vehicle and other vehicles. When performing nonlinear dynamic trajectory modeling based on the time-series data, it can predict the motion trajectory of the corner of the target object in complex motion scenarios, improving the accuracy of trajectory prediction and the reliability of collision warning. Subsequently, the motion trajectory is analyzed, and when a collision point exists between the corner of the target object and the path of the vehicle in the current dynamic coordinate system, a warning is quickly triggered, improving the real-time performance of the warning.
[0158] like Figure 3 As shown, Figure 3 This is a block diagram illustrating an apparatus according to an exemplary embodiment, the apparatus comprising:
[0159] The data acquisition module is used to acquire the coordinates of the tail corner of the target object at each moment in a continuous time period based on a coordinate system with the vehicle as the origin, and generate time series data.
[0160] The trajectory prediction module is used to perform nonlinear dynamic trajectory modeling based on the time series data and predict the motion trajectory of the corner points of the target object.
[0161] The collision warning module is used to analyze the motion trajectory and trigger a warning when the corner of the target object collides with the path of the vehicle in the current dynamic coordinate system.
[0162] The specific implementation process of the functions and roles of each module / submodule / unit in the above device can be found in the implementation process of the corresponding steps in the above method, which can achieve the same technical effect, and will not be repeated here.
[0163] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the solution in this specification according to actual needs. Those skilled in the art can understand and implement this without creative effort.
[0164] Figure 4 An example is a schematic diagram of the physical structure of a vehicle collision warning device, such as... Figure 4As shown, the vehicle collision warning device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840. The processor 810, communication interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions stored in the memory 830 to execute the vehicle collision warning method.
[0165] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, 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.
[0166] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the vehicle collision warning method provided by the above methods.
[0167] In another aspect, this application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the vehicle collision warning methods provided by the above methods.
[0168] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0169] Other embodiments of this specification will readily occur to those skilled in the art upon consideration of the specification and practice of the invention claimed herein. This specification is intended to cover any variations, uses, or adaptations that follow the general principles of this specification and include common knowledge or customary techniques in the art not claimed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this specification are indicated by the following claims.
[0170] It should be understood that this specification is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this specification is limited only by the appended claims.
[0171] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.
Claims
1. A vehicle collision warning method, characterized in that, The method includes: Based on a coordinate system with the vehicle as the origin, the coordinates of the tail corner of the target object at each moment in a continuous time period are obtained to generate time series data. Based on the time series data, nonlinear dynamic trajectory modeling is performed to predict the motion trajectory of the corner points of the target object; The motion trajectory is analyzed, and an early warning is triggered when the corner of the target object collides with the path of the vehicle in the current dynamic coordinate system.
2. The vehicle collision warning method as described in claim 1, characterized in that, The nonlinear dynamic trajectory modeling employs polynomial fitting. The step of performing nonlinear dynamic trajectory modeling based on the time series data to predict the motion trajectory of the corner points of the target object includes: The time series data is fitted with an Nth-order polynomial to generate a trajectory equation representing the motion trajectory of the corner point of the target object, wherein the order N is positively correlated with the motion complexity of the target object.
3. The vehicle collision warning method as described in claim 1, characterized in that, The process of analyzing the motion trajectory involves triggering a warning when a corner of the target object collides with the path of the vehicle in the current dynamic coordinate system. This warning includes: Analyze the motion trajectory to determine the intersection point between the motion trajectory and the path of the vehicle in the current dynamic coordinate system; Determine whether the intersection point is located within the lateral width range of the vehicle; When the intersection point is within the lateral width range of the vehicle, it is determined that there is a collision point between the corner point of the target object and the path of the vehicle in the current dynamic coordinate system, and an early warning is triggered.
4. The vehicle collision warning method as described in claim 3, characterized in that, The step of analyzing the motion trajectory and determining the intersection point of the motion trajectory and the path of the vehicle in the current dynamic coordinate system includes: The trajectory equation corresponding to the motion trajectory is solved, and the position corresponding to the real solution obtained when the trajectory equation is zero is determined as the intersection point. The trajectory equation is obtained by fitting the time series data with an Nth-order polynomial.
5. The vehicle collision warning method as described in claim 3, characterized in that, The method further includes: The intersection points located within the lateral width of the vehicle are determined as the collision points where there is a risk of collision between the target object and the vehicle; Based on the location of the collision point, the collision warning time when the vehicle collides with the target object is predicted.
6. The vehicle collision warning method as described in claim 5, characterized in that, The method of predicting the collision warning time when the vehicle collides with the target object based on the location of the collision point includes: Determine the path length from the current position of the corner point of the target object to the collision point along the motion trajectory; The collision warning time is obtained based on the path length and the instantaneous relative speed between the vehicle and the target object.
7. The vehicle collision warning method as described in claim 1, characterized in that, The vehicle includes an onboard stereo vision sensor, which coincides with the origin of the vehicle's coordinate system. The method involves obtaining the coordinates of the tail corner of the target object at each moment in a continuous time period based on a coordinate system with the vehicle as the origin, and generating time series data, including: For each moment in a continuous time period, perform the following steps: The vehicle-mounted stereo vision sensor acquires a dense point cloud image of the rear of the target object, which is located in front of the vehicle in the direction of travel. The point cloud of the dense point cloud image is vertically projected onto the ground to determine the projected coordinates of the tail corner of the target object; Time series data is generated based on the projected coordinates of the tail corner points at multiple times.
8. The vehicle collision warning method as described in claim 7, characterized in that, The vehicle-mounted stereo vision sensor includes a binocular camera.
9. The vehicle collision warning method as described in claim 7, characterized in that, The corner points include the left and right rear corners of the target object's tail. The step of vertically projecting the point cloud of the dense point cloud image onto the ground to determine the projected coordinates of the tail corner point of the target object includes: The dense point cloud image is vertically projected onto the ground, and the projected points are fitted with straight lines to obtain the slope of the line connecting the left rear corner and the right rear corner after projection. Determine the projected coordinates of the center point of the tail of the target object; Based on the tail width, slope, and projected coordinates of the center point of the target object, the projected coordinates of the left rear corner and the right rear corner of the target object are determined.
10. The vehicle collision warning method as described in claim 9, characterized in that, The tail width of the target object is obtained in the following way: The vehicle-mounted stereo vision sensor acquires an RGB image of the rear of the target object. When the target object is a vehicle, the RGB image is identified to determine the vehicle information of the target object; Based on the vehicle information, the size information of the target object is found, and the tail width of the target object is obtained; And / or, Determine the smallest bounding rectangle covering the projection point along the direction of the slope; The horizontal length of the minimum bounding rectangle is determined to be the tail width of the target object.
11. The vehicle collision warning method as described in claim 7, characterized in that, The continuous time period includes n sampling moments, and the value of n is negatively correlated with the speed of the target object relative to the vehicle. Specifically, when the relative speed decreases, the value of n increases accordingly; when the relative speed increases, the value of n decreases accordingly.
12. A computer device, characterized in that, The system includes a memory, a processor, and a vehicle collision warning program stored in the memory and executable on the processor, wherein the processor, when executing the vehicle collision warning program, implements the steps of the vehicle collision warning method as described in any one of claims 1-11.
13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a vehicle collision warning program, which, when executed, implements the steps of the vehicle collision warning method as described in any one of claims 1-11.
