Automobile active bidirectional anti-collision side collision warning device and method

Abstract

The application discloses a car active bidirectional anti-collision side collision warning device and method, relates to the technical field of car active safety warning, and is used for solving the problem that when a rear target is transversely avoided, only according to the trend that the target center point is far away from, a dangerous side-avoiding behavior is easily mistaken for a safe side-avoiding behavior, leading to early degradation of collision warning and missing report of tail corner side collision risk; the method acquires relative state information of the car and the rear or side-rear target, judges the trend that the target center point is far away from the car, and judges the convergence trend of the side corner feature part of the target close to the car relative to the corresponding rear tail corner feature part of the car; when the far-away trend and the convergence trend are established at the same time, the rear target is identified as a side-avoiding target, and an anti-collision side collision warning is output; the application can identify the contradictory state that the center point is far away but the corner part converges, improve the accuracy of rear tail corner collision risk identification, and reduce the probability of collision warning misjudgment and side collision warning lag.

Description

Technical Field

[0001] This application relates to the field of vehicle safety warning technology, and in particular to an active two-way anti-collision and side-collision warning device and method for automobiles. Background Technology

[0002] With the development of automotive driver assistance technology, rear-end collision warning, side collision warning, and blind spot monitoring technologies have been gradually applied to vehicle safety systems. Existing rear-end collision warnings are usually based on information such as the distance, relative speed, and collision time between the rear target and the vehicle to determine the risk of a rear-end collision. Side collision warnings or blind spot monitoring are usually based on the position of the side and rear target, the lateral distance, whether the target has entered the blind spot, and the vehicle's steering state to determine the risk of a side collision. The above technologies can provide certain warning functions in scenarios such as rear vehicles approaching in a straight line, side vehicles driving sideways, or regular lane changes.

[0003] However, as a rear target shifts to the left or right rear to avoid the vehicle, the target's center point may gradually move away from the vehicle's longitudinal centerline or the collision detection area. Existing systems can easily assume that the rear target is leaving the vehicle's rear driving path, thereby reducing or canceling the collision warning. However, the actual collision may not occur at the target's center point. The front corner feature of the rear target, which is closer to the vehicle, may still continue to approach the corresponding rear corner feature of the vehicle, creating a risk of rear corner collision or side rear collision.

[0004] Therefore, existing technologies lack a mechanism to identify the contradictory state of "the target's center point moving away but the near front corner converging" when dealing with scenarios where rear targets are avoiding close proximity. This can easily lead to misjudging dangerous close proximity avoidance as safe avoidance, resulting in premature downgrading of rear-end collision warnings and delayed triggering of rear-end collision warnings, making it difficult to cover the risk of side-rear collisions during the rear target avoidance process in a timely manner. Therefore, it is necessary to propose a vehicle active two-way rear-end collision avoidance and side-end collision warning device and method that can comprehensively judge the trend of the rear target's center point moving away and the trend of the near front corner converging. Summary of the Invention

[0005] This application provides an active two-way rear-end collision avoidance and side-impact warning device and method for automobiles. When a rear target's center point moves further away from the vehicle, forming a potential safe avoidance candidate, the device further verifies the target based on the convergence relationship between the front corner feature of the rear target closest to the vehicle and the corresponding rear corner feature of the vehicle. If the verification is successful, the potential safe avoidance target is reclassified as a close-to-the-edge collision avoidance target, thereby reducing the likelihood of a dangerous close-to-the-edge collision avoidance being mistakenly identified as a safe avoidance. The technical solution is as follows: On one hand, a method for active two-way collision avoidance and side collision warning for automobiles is provided. The method includes: acquiring relative state information between the vehicle and a rear target located behind or to the side of the vehicle, wherein the relative state information is used to characterize the positional change of the target center point of the rear target relative to the vehicle, and the positional change of the front corner feature of the rear target near the vehicle relative to the corresponding rear corner feature of the vehicle; when the target center point shows a trend of moving away from the vehicle, identifying the rear target as a candidate target for safe avoidance; based on the relative state information, performing convergence verification on the relative positional change between the front corner feature of the candidate target and the corresponding rear corner feature of the vehicle; when the front corner feature shows a convergence trend relative to the corresponding rear corner feature, reclassifying the candidate target as a target for edge collision avoidance; and outputting a collision avoidance and side collision warning based on the target for edge collision avoidance.

[0006] In one possible implementation, the relative state information includes at least one of the following: target center point information of the rear target, target outline information, vehicle body pose information, vehicle body size information, vehicle rear tail angle information, target motion direction information, target heading information, and target speed information.

[0007] In one possible implementation, obtaining the relative state information between the vehicle and a rear target located behind or to the side of the vehicle includes: acquiring rear target detection data using at least one of the vehicle's rearward sensing unit, side-rearward sensing unit, or vehicle-to-vehicle communication unit; acquiring vehicle state data using the vehicle's body state acquisition unit; and performing time synchronization and coordinate transformation on the rear target detection data and the vehicle state data to obtain the relative state information under the same reference coordinate system.

[0008] In one possible implementation, the target center point shows a trend of moving away from the vehicle, including an increase in the lateral distance between the target center point and the longitudinal centerline of the vehicle, or the target center point gradually leaving the collision judgment area of ​​the vehicle, or the lateral distance sequence corresponding to the target center point satisfies a preset center moving away condition.

[0009] In one possible implementation, the front corner feature is determined based on the target contour information of the rear target; when the rear target is located on the left rear side of the vehicle, the front corner feature includes at least one of the right front corner point, right front corner region, or right front contour boundary of the rear target; when the rear target is located on the right rear side of the vehicle, the front corner feature includes at least one of the left front corner point, left front corner region, or left front contour boundary of the rear target.

[0010] In one possible implementation, the corresponding rear rear corner feature is determined based on the vehicle's body pose information and vehicle body size information; when the front corner feature on one side is located on the left rear side of the vehicle, at least one of the vehicle's left rear rear corner, left rear rear corner area, or left rear rear corner outward expansion area is determined as the corresponding rear rear corner feature; when the front corner feature on one side is located on the right rear side of the vehicle, at least one of the vehicle's right rear rear corner, right rear rear corner area, or right rear rear corner outward expansion area is determined as the corresponding rear rear corner feature.

[0011] In one possible implementation, the convergence verification includes: obtaining multiple relative distances between the front corner feature and the corresponding rear corner feature within a preset verification window; arranging the multiple relative distances in the order of sampling time to obtain a tail corner distance sequence; and determining that the front corner feature exhibits a convergence trend relative to the corresponding rear corner feature when the tail corner distance sequence satisfies a preset tail corner convergence condition.

[0012] In one possible implementation, when the front corner feature is a front corner point, the front corner point is used as a representative point for calculating the relative distance; when the front corner feature is a front corner region, the region center point, the nearest point of the region boundary, or the point within the region closest to the corresponding rear tail corner feature is used as a representative point; when the front corner feature is a contour boundary, the boundary point on the contour boundary that is closest to the corresponding rear tail corner feature is used as a representative point.

[0013] In one possible implementation, the step of outputting a rear-end collision avoidance warning based on the edge-to-edge collision avoidance target includes: generating an alarm control command corresponding to the edge-to-edge collision avoidance target; maintaining or increasing the current rear-end collision warning level, or outputting a rear corner side collision verification warning based on the alarm control command; wherein, the alarm control command is used to control the in-vehicle alarm unit to output an in-vehicle alarm to the driver of the vehicle, or to control the external alarm unit to output an external alarm to the outside of the vehicle.

[0014] On the other hand, an active two-way collision avoidance and side collision warning device for automobiles is provided. The device includes: an information acquisition module, a trend determination module, a target recognition module, and an alarm output module. The information acquisition module is used to acquire the relative state information between the vehicle and the target behind it. The trend determination module is used to identify the target behind it whose center point is moving away from the target as a candidate target for safe avoidance, and to perform convergence verification on the relative position change between the front corner feature of one side of the candidate target for safe avoidance and the corresponding rear corner feature of the vehicle. The target recognition module is used to change the candidate target for safe avoidance to a target for edge collision avoidance when the convergence verification is successful. The alarm output module is used to output a collision avoidance and side collision warning based on the target for edge collision avoidance.

[0015] The solution provided in this application does not simply cancel or reduce the alarm based on the distance of the rear target's center point from the vehicle. Instead, it further verifies the convergence relationship between the near front corner and the rear corner of the vehicle based on the safety avoidance candidate judgment formed by the distance of the center point. Thus, when the target appears to be far away but the corner is still close to the rear corner of the vehicle, the system can perform risk reassessment and output a rear-end collision and side-end collision warning, improving the accuracy of rear corner collision risk identification and reducing the probability of false alarm cancellation of rear-end collision warning and delayed side-end collision warning. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the implementation environment of an active two-way collision avoidance and side collision warning device for automobiles provided in this application embodiment; Figure 2 This is a flowchart of an active two-way collision avoidance and side collision warning method for automobiles provided in an embodiment of this application; Figure 3 This is a schematic diagram illustrating a state where the center point is far from and converges with a feature portion at one side of the front corner, as provided in an embodiment of this application. Figure 4 This is a schematic diagram of a tail corner collision area provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an active two-way collision avoidance and side collision warning device for automobiles provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0019] It is understood that the terms “first,” “second,” etc., used in this application may be used to describe various concepts herein, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another; for example, without departing from the scope of this application, a first distance may be referred to as a second distance, and similarly, a second distance may be referred to as a first distance.

[0020] Wherein, at least one refers to one or more, for example, at least one sensing unit can be one sensing unit, two sensing units or more sensing units; multiple refers to two or more, for example, multiple sampling times can be any integer number of sampling times greater than or equal to two, such as two sampling times, three sampling times or so; each refers to each of the multiple, for example, each sampling time refers to each of the multiple sampling times.

[0021] In this application, the target for edge collision avoidance refers to a target whose center point moves away from the vehicle, but whose front corner feature near the vehicle moves towards the corresponding rear corner feature of the vehicle, thus potentially posing a risk of collision or side-rear impact with the rear corner of the vehicle.

[0022] In this application, the front corner feature can be the front corner point, front corner region, or front contour boundary of a rear target; the rear rear corner feature can be the rear rear corner point, rear rear corner region, or rear rear corner outward expansion region of the vehicle. Unless otherwise specified, the rear rear corner, tail corner, and rear rear corner feature all refer to the reference object of the rear side corner of the vehicle used to judge the risk of a rear target grazing against the edge.

[0023] In this application, the trend of moving away refers to the lateral distance between the center point of the rear target and the longitudinal centerline of the vehicle, the collision judgment area, or the center area of ​​the rear of the vehicle increasing or moving away; the trend of convergence refers to the state in which the distance between the front corner feature of the rear target close to the vehicle and the corresponding rear corner feature of the vehicle decreases, enters the conflict area, or the probability of collision increases; the trend of moving away and the trend of convergence correspond to different judgment objects, the former corresponds to the center point of the target, and the latter corresponds to the risk relationship between the corner of the target and the rear corner of the vehicle.

[0024] Figure 1 This can be a schematic diagram of an implementation environment provided in the embodiments of this application; see also Figure 1 The implementation environment may include a rearward sensing unit, a side-rearward sensing unit, a vehicle status acquisition unit, an on-board alarm controller, an in-vehicle alarm unit, and an external alarm unit. The rearward sensing unit and the side-rearward sensing unit are used to acquire rear target detection data behind or to the side of the vehicle. The vehicle status acquisition unit is used to acquire vehicle status data. The on-board alarm controller is used to generate relative status information based on the rear target detection data and the vehicle status data, and to perform trend judgment, convergence verification, target reclassification, and alarm control based on the relative status information. The in-vehicle alarm unit and the external alarm unit are used to output alarms according to the alarm control commands output by the on-board alarm controller.

[0025] Optionally, the rearward sensing unit or the side-rearward sensing unit may include at least one of millimeter-wave radar, camera, ultrasonic radar, and lidar; the vehicle status acquisition unit may include at least one of vehicle speed sensor, steering wheel angle sensor, inertial measurement unit, and vehicle body controller; the in-vehicle alarm unit may include at least one of instrument display unit, buzzer, seat vibration unit, and steering wheel vibration unit; the external alarm unit may include at least one of rear warning light, side warning light, external audio-visual warning unit, or vehicle-to-vehicle communication unit; the specific type of each unit is not limited in the embodiments of this application.

[0026] Figure 2 This can be a flowchart of an active two-way collision avoidance and side collision warning method for automobiles provided in an embodiment of this application; see also Figure 2 This method can be executed by an onboard alarm controller, and includes the following steps: S201, The vehicle alarm controller acquires the relative status information between the vehicle and a target located behind or to the side of the vehicle; The relative state information can at least be used to determine the position change of the target center point relative to the vehicle, and the position change of the front corner feature of the rear target near the vehicle relative to the corresponding rear corner feature of the vehicle. The relative state information may include at least one of the following: target center point information of the rear target, target outline information, vehicle body pose information, vehicle body size information, vehicle rear corner information, target movement direction information, target heading information, and target speed information.

[0027] Optionally, the vehicle-mounted alarm controller acquires rear target detection data using at least one of the vehicle's rearward sensing unit, side-rearward sensing unit, or vehicle-to-vehicle communication unit, and acquires vehicle status data using the vehicle's body state acquisition unit. The vehicle-mounted alarm controller can perform time synchronization, coordinate transformation, target association, or trajectory smoothing based on the rear target detection data and vehicle status data to obtain relative status information. Among these, time synchronization is used to ensure that the rear target detection data and vehicle status data correspond to the same sampling time; coordinate transformation is used to transform data from different sources to the same reference coordinate system; target association is used to determine whether the detected targets in consecutive sampling times belong to the same rear target; and trajectory smoothing is used to reduce the impact of detection noise on trend judgment.

[0028] The target center point can be the center point of the visual detection box, the center point of the target's bounding rectangle, the millimeter-wave radar tracking point, the geometric center point of the fused tracking target, or the target location point carried in the vehicle-to-vehicle communication message; center points from different sources are unified as target center point information after time synchronization and coordinate transformation.

[0029] Target contour information can be obtained in at least one of the following ways: after a camera identifies a vehicle detection box, segments the contour or vehicle key points, the target contour information is determined based on the corner points of the detection box or the contour boundary; after a millimeter-wave radar or lidar outputs a target point cloud, the target contour information is estimated based on the point cloud bounding box, the target heading, and the vehicle size model; when the vehicle-to-vehicle communication message contains target position, heading, length, and width information, the target contour information is reconstructed based on the target geometric model.

[0030] S202. The vehicle alarm controller determines, based on relative status information, whether the center point of the target behind the vehicle is moving away from the vehicle. In one possible implementation, the vehicle alarm controller acquires the lateral position change of the target center point relative to the longitudinal centerline of the vehicle or the collision judgment area of ​​the vehicle; when the lateral distance of the target center point relative to the longitudinal centerline of the vehicle increases, or the target center point gradually moves away from the collision judgment area of ​​the vehicle, or the last lateral distance in the lateral distance sequence corresponding to the target center point is greater than the first lateral distance, the vehicle alarm controller can determine that the target center point is moving away from the vehicle.

[0031] Optionally, the vehicle alarm controller can acquire the target center point position at multiple sampling times within a preset time window, and calculate the lateral distance between the target center point and the longitudinal centerline of the vehicle to obtain a center lateral distance sequence; if the center lateral distance sequence increases as a whole, or the number of times the adjacent lateral distances in the center lateral distance sequence increase reaches a preset number of times they move away, it is determined that the target center point is moving away from the vehicle.

[0032] For example, the center lateral distance can be determined by the following formula: Dc(ti)=|xc(ti)-x0(ti)|; where Dc(ti) is the center lateral distance at the i-th sampling time, xc(ti) is the lateral coordinate of the target center point at the i-th sampling time, x0(ti) is the lateral coordinate of the longitudinal centerline of the vehicle at the i-th sampling time, and ti is the i-th sampling time; if Dc(tn)-Dc(t1)>Δc, then it can be determined that the target center point is moving away; where Dc(tn) is the center lateral distance at the last sampling time of the sampling window, Dc(t1) is the center lateral distance at the first sampling time of the sampling window, and Δc is the preset center moving away threshold.

[0033] S203. When the center point of the target is moving away from the vehicle, the vehicle alarm controller identifies the rear target as a candidate target for safe avoidance. In this embodiment, the fact that the center point of the target is moving away from the vehicle does not directly indicate that the target behind has been safely avoided, but rather that the target behind may be moving away from the collision judgment area of ​​the vehicle. In order to avoid canceling or reducing the alarm simply based on the trend of the target center point moving away, the vehicle alarm controller first determines the target behind as a candidate target for safe avoidance, and then performs a convergence verification between the near front corner and the rear rear corner of the vehicle on the candidate target for safe avoidance.

[0034] Optionally, the vehicle-mounted alarm controller can determine the front corner feature of one side of the rear target using different methods based on the data source of the rear target detection data. When the rear target detection data comes from a camera, the vehicle-mounted alarm controller can perform vehicle detection, vehicle key point recognition, or instance segmentation on the rear target to obtain the detection box, vehicle outline, or vehicle key points of the rear target. The vehicle-mounted alarm controller determines the left and right front corner points of the rear target based on the driving direction and vehicle outline of the rear target, and determines the front corner feature of the side closer to the vehicle from these two corner points based on the lateral position of the rear target relative to the vehicle.

[0035] Optionally, when the rear target detection data comes from millimeter-wave radar or lidar, the vehicle warning controller can use the point cloud data of the rear target, the target heading angle, the target length, and the target width to establish the target's circumscribed rectangle. The vehicle warning controller then determines the front corner of the rear target based on the four corner points of the target's circumscribed rectangle and the target's direction of movement, and selects the front corner point closest to the vehicle from the front corner points as a front corner feature. When the rear target detection data comes from vehicle-to-vehicle communication messages, the vehicle warning controller can reconstruct the vehicle body rectangle of the rear target based on the position, heading, length, width, and speed information sent by the rear target, and determine the front corner feature based on the vehicle body rectangle.

[0036] Optionally, if the vehicle warning controller cannot reliably acquire the precise front corner point of the rear target, the vehicle warning controller can use the front corner region or the front contour boundary as a front corner feature part on one side; if the front corner feature part is a front corner point, the vehicle warning controller directly uses the coordinates of the front corner point to calculate the distance between it and the corresponding rear corner feature part; if the front corner feature part is a front corner region, the vehicle warning controller can select the region point within the front corner region that is closest to the corresponding rear corner feature part, or select the center point of the front corner region, as the representative point for distance calculation; if the front corner feature part is a front contour boundary, the vehicle warning controller can use the boundary point on the front contour boundary that is closest to the corresponding rear corner feature part as the representative point for distance calculation.

[0037] Optionally, the vehicle alarm controller can set a contour confidence level for a feature area at one front corner. This contour confidence level can be determined based on at least one of target detection confidence level, target tracking stability, corner continuity, and sensor fusion consistency. If the contour confidence level is lower than a preset contour confidence level threshold, the vehicle alarm controller will not directly output a high-level rear corner side collision verification alarm, but will mark the rear target as a target to be confirmed and continue to acquire the contour information of the rear target in subsequent sampling windows.

[0038] S204. The vehicle-mounted alarm controller determines the front corner feature of the candidate target for safe avoidance that is close to the vehicle, and also determines the corresponding rear corner feature of the vehicle; In one possible implementation, the vehicle warning controller determines multiple target corner points, front corner regions, or front contour boundaries of the rear target based on the target contour information of the rear target, and determines whether the rear target is located on the left or right rear side of the vehicle based on the lateral position of the rear target relative to the vehicle; if the rear target is located on the left rear side of the vehicle, the vehicle warning controller determines at least one of the right front corner point, right front corner region, or right front contour boundary of the rear target as a front corner feature near the vehicle; if the rear target is located on the right rear side of the vehicle, the vehicle warning controller determines at least one of the left front corner point, left front corner region, or left front contour boundary of the rear target as a front corner feature near the vehicle.

[0039] In one possible implementation, the vehicle alarm controller determines the corresponding rear rear corner feature based on the vehicle's body pose information and vehicle body size information; when the rear target is located on the left rear side of the vehicle, and its front corner feature on the side closest to the vehicle is the right front corner feature, the corresponding rear rear corner feature is the left rear rear corner feature of the vehicle; when the rear target is located on the right rear side of the vehicle, and its front corner feature on the side closest to the vehicle is the left front corner feature, the corresponding rear rear corner feature is the right rear rear corner feature of the vehicle.

[0040] Figure 4 This could be a schematic diagram of a tail-angle conflict scenario provided in an embodiment of this application, where the center point is far away but the leading angle converges; see also Figure 4 When a target is located to the left rear of the vehicle, the target's center point can move towards the left lane, thus appearing to move away from the vehicle's longitudinal centerline. However, a rear target that is close to the front corner of the vehicle, such as the right front corner, may still approach the left rear corner and enter or approach the rear corner collision zone corresponding to the left rear corner. In this case, judging that the rear target has been safely avoided simply based on the trend of the target's center point moving away could easily cause the rear collision warning to be reduced or canceled prematurely.

[0041] Optionally, the vehicle alarm controller can determine the trend of the target center point moving away and the convergence trend of the front corner feature part on one side within a preset sampling window; the preset sampling window can be set to 0.3s to 2s, such as 1s; this preset sampling window can include 3 to 10 sampling times; the length of the preset sampling window and the number of sampling times can be calibrated according to the vehicle speed, the relative speed of the target behind, the sensor sampling period and the processing period of the vehicle alarm controller.

[0042] Optionally, the vehicle alarm controller can calculate the center lateral distance of the target center point in the following manner: Dc(ti)=|xc(ti)-x0(ti)|; where Dc(ti) represents the center lateral distance at the i-th sampling time; xc(ti) represents the lateral coordinate of the target center point of the target behind at the i-th sampling time; x0(ti) represents the lateral coordinate of the longitudinal centerline of the vehicle at the i-th sampling time; ti represents the i-th sampling time; if Dc(tn)-Dc(t1)>Δc, or the number of times the adjacent center lateral distances in the center lateral distance sequence increase reaches the preset number of times they move away, then the vehicle alarm controller determines that the target center point is moving away from the vehicle; where tn represents the last sampling time in the sampling window, t1 represents the first sampling time in the sampling window, and Δc represents the preset center moving away threshold.

[0043] Optionally, the vehicle alarm controller can calculate the tail angle distance as follows: Dh(ti)=sqrt((xf(ti)-xr(ti))^2+(yf(ti)-yr(ti))^2); where Dh(ti) represents the tail angle distance between the front corner feature and the corresponding rear tail angle feature at the i-th sampling time; xf(ti) and yf(ti) represent the coordinates of the representative point of the front corner feature at the i-th sampling time; xr(ti) and yr(ti) represent the coordinates of the representative point of the corresponding rear tail angle feature at the i-th sampling time; if Dh(tn)-Dh(t1)<-Δh, or the number of consecutive decreases in the tail angle distance sequence of adjacent tail angles reaches the preset convergence number, then the vehicle alarm controller determines that the front corner feature shows a convergence trend relative to the corresponding rear tail angle feature; where Δh represents the preset tail angle convergence threshold.

[0044] Optionally, the preset center distance threshold Δc and the preset tail angle convergence threshold Δh can be calibrated based on the vehicle width, the width of the target behind, the sensor ranging error, the sampling period, and the vehicle speed; for example, Δc can be 0.1m to 0.8m, and Δh can also be 0.1m to 0.8m; the preset distance number and the preset convergence number can be 2 to 5 times; the above values ​​are only examples and are not intended to limit the scope of protection of this application.

[0045] Optionally, the vehicle warning controller can use the rear corner feature as a reference to establish a rear corner conflict zone. The rear corner conflict zone can be formed by extending longitudinally behind the vehicle and laterally outward. The shape of the extension can be rectangular, trapezoidal, fan-shaped, or polygonal. The longitudinal extension distance of the rear corner conflict zone can be determined based on the vehicle speed, the relative speed of the rear target, or the longitudinal collision time. The lateral extension distance of the rear corner conflict zone can be determined based on the vehicle width, lane width, rear target width, or safety clearance. If one front corner feature enters the rear corner conflict zone, the vehicle warning controller can determine that the one front corner feature meets the convergence trend judgment condition relative to the corresponding rear corner feature.

[0046] Optionally, the vehicle warning controller may perform a misjudgment protection judgment before reclassifying a candidate target for safe avoidance as a target for edge collision avoidance. If the center point of the target is moving away from the vehicle and the front corner feature on one side is also moving away from the corresponding rear corner feature, then the vehicle warning controller will maintain the safe avoidance judgment and will not reclassify the target behind as a target for edge collision avoidance.

[0047] Optionally, if the tail angle distance sequence decreases only at a single sampling moment, but does not meet the continuous decrease condition within the preset sampling window, or if the last distance relative to the first distance does not reach the preset tail angle convergence threshold, then the vehicle alarm controller will not change the rear target to a target that is close to the edge and needs to be avoided. In this way, false alarms caused by single-frame detection noise, target contour jitter, or sensor instantaneous error can be reduced.

[0048] Optionally, if the target identifier of the rear target changes within the preset sampling window, or if the rear target is lost and then reappears within the preset sampling window, the vehicle alarm controller will not directly trigger the rear corner collision verification alarm based on the corner position of a single reappearance. Instead, it will re-establish the tracking sequence of the rear target and re-determine the trend of the target center point moving away and the convergence trend of the front corner feature on one side within a new sampling window.

[0049] Optionally, if the longitudinal collision time of the rear target relative to the vehicle is greater than a preset safe time threshold, and the front corner feature of one side does not enter the rear corner collision area, then the vehicle warning controller may not trigger a high-level rear corner side collision verification alarm; if the corresponding rear corner feature of the vehicle and the front corner feature of the rear target do not overlap longitudinally or overlap in the predicted time range, then the vehicle warning controller may maintain the normal risk judgment and not change the rear target to a target for edge collision avoidance.

[0050] S205. The vehicle-mounted alarm controller performs convergence verification on the relative positional change between the front corner feature of a candidate target for safe avoidance and the corresponding rear corner feature of the vehicle. In one possible implementation, the vehicle alarm controller acquires multiple relative distances between a front corner feature and a corresponding rear corner feature within a preset verification window; arranges these relative distances according to the sampling time sequence to obtain a rear corner distance sequence; and determines that the front corner feature shows a convergence trend relative to the corresponding rear corner feature if the rear corner distance sequence meets a preset rear corner convergence condition. For example, the tail angle distance can be determined by the following formula: Dh(ti)=sqrt((xf(ti)-xr(ti))^2+(yf(ti)-yr(ti))^2); where Dh(ti) is the distance between the front corner feature and the corresponding rear tail angle feature at the i-th sampling time, (xf(ti),yf(ti)) are the coordinates of the representative point of the front corner feature at the i-th sampling time, and (xr(ti),yr(ti)) are the coordinates of the representative point of the corresponding rear tail angle feature at the i-th sampling time; if Dh(tn)-Dh(t1)<-Δh, then it can be determined that the front corner feature shows a convergence trend relative to the corresponding rear tail angle feature; where Dh(tn) is the tail angle distance at the last sampling time of the sampling window, Dh(t1) is the tail angle distance at the first sampling time of the sampling window, and Δh is the preset tail angle convergence threshold.

[0051] When the front corner feature is a front corner point, the vehicle alarm controller can use the coordinates of that front corner point as the representative point coordinates; when the front corner feature is a front corner region, the vehicle alarm controller can use the center point of the front corner region, the nearest point on the region boundary, or the point within the region closest to the corresponding rear corner feature as the representative point coordinates; when the front corner feature is a contour boundary, the vehicle alarm controller can use the boundary point on that contour boundary that is closest to the corresponding rear corner feature as the representative point coordinates.

[0052] The rear corner collision zone can be formed by extending it outwards along the longitudinal rear of the vehicle and laterally outwards, based on the rear corner point. The outward shape can be rectangular, trapezoidal, fan-shaped, or polygonal. The longitudinal outward extension distance can be determined based on the vehicle speed, the relative speed of the target behind, or the longitudinal collision time. The lateral outward extension distance can be determined based on the vehicle width, lane width, target width, or the calibrated safety clearance. For example, the rear corner collision zone can be formed by extending it outwards by 0.5m to 3m along the rear of the vehicle and 0.2m to 1.5m along the lateral outwards of the vehicle, based on the rear corner point.

[0053] The preset calibration window can be 0.3s to 2s, and the number of sampling points can be 3 to 10. The preset center distance threshold Δc and the preset tail angle convergence threshold Δh can be calibrated according to vehicle speed, vehicle width, target width, sensor error and sampling period. The preset distance number and preset convergence number can be 2 or 3 consecutive times, or can be calibrated according to vehicle speed and sensor refresh rate.

[0054] Figure 5 This can be a schematic diagram illustrating the trend determination of the center lateral distance sequence and the tail angle distance sequence provided in the embodiments of this application; see also Figure 5 The lateral distance sequence of the center shows an increasing trend over time, indicating that the target center point is moving further away from the vehicle. The tail angle distance sequence shows a decreasing trend over time, indicating that the front corner feature on one side is converging relative to the corresponding rear tail corner feature of the vehicle. When both trends are true, the target behind should not be judged as a safe avoidance simply because the target's center point is far away, but should enter the revised judgment process of avoiding the target by brushing against the edge.

[0055] S206. When the front corner feature on one side shows a convergence trend relative to the corresponding rear corner feature, the vehicle warning controller changes the safety avoidance candidate target to a side-impact collision avoidance target; In one possible implementation, if the target's center point moves away from the vehicle and one of its front corner features also moves away from the corresponding rear corner feature, the vehicle warning controller can maintain a safe avoidance judgment. If the target's center point moves away from the vehicle but one of its front corner features converges towards the corresponding rear corner feature, the vehicle warning controller will change the safe avoidance candidate target to a close-to-the-edge collision avoidance target. This change is used to prevent the rear target from being judged as a safe avoidance target solely based on the target's center point moving away, or to prevent the rear collision warning level from being deactivated or reduced solely based on the target's center point moving away.

[0056] In one possible implementation, the vehicle alarm controller can set false alarm protection rules; if the confidence level of the target outline is lower than the preset confidence level threshold, the vehicle alarm controller will not directly trigger a high-level alarm, but will instead set the rear target to a pending confirmation state; if the target identifier changes or the target is lost within the continuous sampling window, the vehicle alarm controller will not perform a re-judgment based on the corner position of a single frame; if the rear corner distance decreases briefly but the longitudinal collision time of the rear target relative to the vehicle is greater than the safety threshold, or if there is no predicted overlap between the corresponding rear corner of the vehicle and the near front corner of the target, the vehicle alarm controller may not trigger a rear corner side collision verification alarm.

[0057] Optionally, after identifying the target for edge collision avoidance, the vehicle warning controller can determine the warning direction based on the corresponding rear corner feature. If the corresponding rear corner feature is the left rear corner feature of the vehicle, the vehicle warning controller generates a left rear corner risk direction; if the corresponding rear corner feature is the right rear corner feature of the vehicle, the vehicle warning controller generates a right rear corner risk direction. The vehicle warning controller can write the warning direction into the warning control command so that the in-vehicle warning unit or the external warning unit can output a warning prompt corresponding to the risk direction.

[0058] Optionally, the vehicle alarm controller can determine the alarm level based on the rear corner distance, the rate of change of the rear corner distance, the longitudinal collision time, and whether the front corner feature of one side enters the rear corner conflict zone. If the rear corner distance shows a convergence trend but has not yet fallen below the preset rear corner distance threshold, the vehicle alarm controller outputs a low-level rear corner side collision verification alarm. If the rear corner distance is lower than the preset rear corner distance threshold, or the longitudinal collision time is less than the preset alarm time threshold, the vehicle alarm controller maintains or increases the current rear-end collision alarm level and outputs a medium-level or high-level rear corner side collision verification alarm. If the front corner feature of one side enters the rear corner conflict zone and the rear corner distance continues to decrease, the vehicle alarm controller outputs a high-level rear-end collision avoidance side collision alarm.

[0059] Optionally, the vehicle alarm controller can determine the alarm target based on the vehicle's avoidance status and the respondability status of the rear target. If the vehicle still has avoidance space, the vehicle alarm controller prioritizes controlling the in-vehicle alarm unit to output an alarm to the driver. If the avoidance space in front of or to the side of the vehicle is insufficient, and the rear target still has room to decelerate or avoid, the vehicle alarm controller can control the external alarm unit to output an alarm to the outside of the vehicle. In-vehicle alarms can include at least one of instrument alarms, audible alarms, seat vibration alarms, or steering wheel vibration alarms. External alarms can include at least one of rear warning light alarms, side warning light alarms, external audible alarms, external light signal alarms, or vehicle-to-vehicle communication message alarms.

[0060] Optionally, the vehicle alarm controller can send alarm control commands to the in-vehicle alarm unit or the external alarm unit; the alarm control commands can include at least one of the following: alarm direction, alarm level, alarm object, alarm duration, and alarm output method; after receiving the alarm control commands, the in-vehicle alarm unit and the external alarm unit will execute the corresponding alarm output according to the alarm control commands.

[0061] S207, the vehicle-mounted alarm controller outputs a rear-end collision and side-impact warning based on the target of edge-to-edge collision avoidance; In one possible implementation, the on-board warning controller generates a warning control command corresponding to the edge collision avoidance target; the warning control command may include at least one of a warning direction, a warning level, and a warning object; the warning direction is used to indicate that there is a risk of rear corner collision at the left or right rear corner of the vehicle; the warning level is used to maintain or increase the current rear-end collision warning level, or to output a rear corner collision verification warning; the warning object is used to determine whether to output a warning to the driver of the vehicle, or to output a warning to the outside of the vehicle.

[0062] Optionally, the vehicle alarm controller can determine the alarm object and alarm method based on the alarm control command, and control the corresponding alarm unit to perform alarm output according to the determination result; when the alarm object is the driver of the vehicle, the vehicle alarm controller controls the in-vehicle alarm unit to output an in-vehicle alarm to the driver of the vehicle; wherein, the in-vehicle alarm may include at least one of the following: instrument alarm, audible alarm, seat vibration alarm, or steering wheel vibration alarm; when the alarm object is a road user outside the vehicle, the vehicle alarm controller controls the external alarm unit to output an external alarm to the outside of the vehicle; wherein, the external alarm may include at least one of the following: rear warning light alarm, side warning light alarm, external audible alarm, external light signal alarm, or vehicle-to-vehicle communication message alarm.

[0063] Figure 3 This can be an internal functional structure diagram of an active two-way collision avoidance and side collision warning device for automobiles provided in the embodiments of this application; see also Figure 3 The device may include an information acquisition module, a trend determination module, a target recognition module, and an alarm output module; The information acquisition module acquires information from millimeter-wave radar, cameras, ultrasonic radar, vehicle status information, and optional V2X information, and generates relative status information between the vehicle and the target behind it. The trend determination module determines whether the target center point of the target behind it is moving away from the vehicle based on the relative status information, and verifies whether the front corner feature of one side of the candidate target for safe avoidance is converging relative to the corresponding rear corner feature of the vehicle. The target recognition module changes the candidate target for safe avoidance to a target for edge collision avoidance when the convergence verification is successful. The alarm output module generates alarm control commands based on the target for edge collision avoidance and outputs in-vehicle alarms or external alarms.

[0064] The method provided in this application does not directly take the trend of the target's center point moving away from the vehicle as a safe avoidance conclusion when avoiding a target behind it. Instead, it identifies the target as a candidate target for safe avoidance and continues to determine whether the front corner feature of the target on the side closest to the vehicle and the corresponding rear corner feature of the vehicle show a convergence trend. If the target's center point shows a moving away trend while the front corner feature shows a convergence trend, then the vehicle warning controller identifies the target behind it as a target for edge collision avoidance and outputs a side collision warning. Therefore, this application embodiment can identify the contradictory state of "center point moving away but corner convergence", reducing the situation where dangerous edge avoidance is misjudged as safe avoidance.

[0065] In this embodiment, the side collision judgment is not based solely on the vehicle's geometric contours. Instead, when the target's center point is far away, causing the rear target to be judged as a candidate for safe avoidance, the convergence trend between the front corner feature and the corresponding rear corner feature is used to verify the misjudgment. Based on the verification result, the target is re-judged and the warning is mapped. This can reduce the probability of false cancellation of rear-end collision warnings, missed detection of rear corner side collision risks, and misjudgment of dangerous edge avoidance as safe avoidance.

Claims

1. A method for active two-way rear-end collision avoidance and side-impact warning for automobiles, characterized in that, The method includes: Acquire relative state information between the vehicle and a target located behind or to the side of the vehicle. The relative state information is used to characterize the position change of the target center point of the target relative to the vehicle, and the position change of the front corner feature of the target on the side closer to the vehicle relative to the corresponding rear corner feature of the vehicle. When the target center point is moving away from the vehicle, the rear target is identified as a candidate target for safe avoidance; based on the relative state information, the relative position change between the front corner feature of one side of the candidate target for safe avoidance and the corresponding rear corner feature of the vehicle is converged and verified. If the front corner feature on one side shows a convergence trend relative to the corresponding rear corner feature, the candidate target for safe avoidance is reclassified as a target for edge-to-edge collision avoidance; based on the target for edge-to-edge collision avoidance, a side-impact warning is output to prevent rear-end collisions.

2. The method according to claim 1, characterized in that, The relative state information includes at least one of the following: target center point information of the rear target, target outline information, vehicle body posture information, vehicle body size information, vehicle rear corner information, target movement direction information, target heading information, and target speed information.

3. The method according to claim 1, characterized in that, The acquisition of relative state information between the vehicle and a target located behind or to the side of the vehicle includes: Rear target detection data is acquired through at least one of the vehicle's rearward sensing unit, side-rearward sensing unit, or vehicle-to-vehicle communication unit; vehicle status data is acquired through the vehicle's body status acquisition unit. The rear target detection data and the vehicle status data are synchronized in time and transformed in coordinate to obtain the relative status information under the same reference coordinate system.

4. The method according to claim 1, characterized in that, The target center point moving away from the vehicle includes: the lateral distance between the target center point and the longitudinal centerline of the vehicle increases; or the target center point gradually moves away from the collision judgment area of ​​the vehicle; or the lateral distance sequence corresponding to the target center point satisfies the preset center moving away condition. The preset center distance condition includes the last horizontal distance of the center horizontal distance sequence being greater than the first horizontal distance, or the number of times the adjacent horizontal distances in the center horizontal distance sequence increase reaches a preset distance number.

5. The method according to claim 1, characterized in that, The front corner feature on one side is determined based on the target contour information of the target behind it; When the rear target is located on the left rear side of the vehicle, the front corner feature includes at least one of the right front corner point, right front corner region or right front contour boundary of the rear target; When the rear target is located on the right rear side of the vehicle, the front corner feature of one side includes at least one of the left front corner point, left front corner region or left front contour boundary of the rear target.

6. The method according to claim 1, characterized in that, The corresponding rear tail corner feature is determined based on the vehicle's body pose information and vehicle body size information; When the front corner feature is located on the left rear side of the vehicle, at least one of the left rear corner, the left rear corner area, or the extended area of ​​the left rear corner is identified as the corresponding rear corner feature. When the front corner feature is located on the right rear side of the vehicle, at least one of the right rear corner, the right rear corner area, or the right rear corner outward expansion area is identified as the corresponding rear corner feature.

7. The method according to claim 1, characterized in that, The convergence verification of the relative positional change between the front corner feature of one side of the candidate target for safe avoidance and the corresponding rear corner feature of the vehicle includes: The relative distances between the front corner feature and the corresponding rear corner feature within a preset verification window are obtained; the relative distances are arranged in the order of sampling time to obtain a tail corner distance sequence; if the tail corner distance sequence satisfies a preset tail corner convergence condition, it is determined that the front corner feature is converging with respect to the corresponding rear corner feature. The preset tail angle convergence condition includes at least one of the following: the last distance of the tail angle distance sequence is less than the first distance; the number of consecutive decreases in the adjacent distances in the tail angle distance sequence reaches a preset convergence number; the slope of the distance change in the tail angle distance sequence is less than a preset slope threshold; or the front corner feature on one side enters the tail angle conflict region established with the rear tail angle as the reference.

8. The method according to claim 7, characterized in that, When the front corner feature is a front corner point, the front corner point is used as a representative point for calculating the relative distance; when the front corner feature is a front corner region, the region center point, the nearest point of the region boundary, or the point within the region closest to the corresponding rear tail corner feature is used as a representative point for calculating the relative distance; when the front corner feature is a contour boundary, the boundary point on the contour boundary closest to the corresponding rear tail corner feature is used as a representative point for calculating the relative distance.

9. The method according to any one of claims 1 to 8, characterized in that, The step of outputting a collision avoidance and side collision warning based on the edge-grabbing collision avoidance target includes: generating an alarm control command corresponding to the edge-grabbing collision avoidance target; Based on the alarm control command, maintain or increase the current rear-end collision alarm level, or output a rear corner side collision verification alarm; wherein, the alarm control command is used to control the in-vehicle alarm unit to output an in-vehicle alarm to the driver of the vehicle, or to control the external alarm unit to output an external alarm to the outside of the vehicle.

10. A vehicle active two-way collision avoidance and side collision warning device, characterized in that, The device includes: The information acquisition module is used to acquire relative state information between the vehicle and a target located behind or to the side of the vehicle. The relative state information is used to characterize the position change of the target center point of the target relative to the vehicle, and the position change of the front corner feature of the target on the side closer to the vehicle relative to the corresponding rear corner feature of the vehicle. The trend determination module is used to determine the rear target as a candidate target for safe avoidance when the target center point is moving away from the vehicle, and to perform convergence verification on the relative position change between the front corner feature of one side of the candidate target for safe avoidance and the corresponding rear corner feature of the vehicle. The target recognition module is used to change the candidate target for safe avoidance to a target that is close to the edge and can be avoided when the front corner feature on one side shows a convergence trend relative to the corresponding rear corner feature; the alarm output module is used to output a side collision avoidance alarm based on the target that is close to the edge and can be avoided.

Images