Recognize a vehicle profile based on measurement data from a surrounding sensor

Polygon chain envelopes are generated through surrounding environmental sensors and classified edges. Combined with Ge Liheng scanning algorithm and corner point screening, the problem of inaccurate recognition of rectangular borders is solved, and more accurate vehicle profile recognition and accident prediction are achieved.

CN112041886BActive Publication Date: 2024-11-29CONTINENTAL AUTOMOTIVE TECHNOLOGIES GMBH
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN201980029192.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-04-16
Publication Date
2024-11-29
Estimated Expiration
2039-04-16

AI Technical Summary

Technical Problem

The prior art uses rectangular border methods to identify vehicle profiles, which may lead to unnecessary or incorrect triggering of passive safety devices, affecting the accuracy of accident prediction.

Method used

Through the measurement data of the surrounding environment sensor, a polygon chain envelope is generated, irrelevant space points are removed, the edges are classified as front and side of the vehicle, and the complete shape is supplemented through the central axis estimation. The Ge Liheng scanning algorithm and corner point screening method are used to improve the accuracy of contour recognition .

Benefits of technology

It realizes more accurate vehicle profile recognition, reduces unnecessary or wrong safety device triggers, and improves the reliability of accident prediction and control accuracy of passive safety devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112041886B_ABST
    Figure CN112041886B_ABST
Patent Text Reader

Abstract

A method for identifying a vehicle (12) contour based on measurement data of an environment sensor is disclosed, wherein the measurement data has a certain number of spatial points of the target detected by the environment sensor. The method comprises the following contents: generating an envelope around the target in the form of a polygon chain according to the number of spatial points; reducing the number of spatial points by deleting all spatial points within the envelope; removing spatial points irrelevant to the identification of a vehicle contour from the polygon chain and determining spatial points of the polygon chain relevant to the identification of a vehicle contour; and classifying one or more edges as belonging to the front and side of a vehicle according to the reduced number of spatial points. Here, removing spatial points irrelevant to the identification of a vehicle contour from the polygon chain comprises the following contents: deleting spatial points (26) on the envelope whose connecting straight lines (28) and adjacent spatial points (26) of the polygon chain (24) form an angle less than or equal to a predetermined angle, in particular 2 degrees, and connecting (30) spatial points (26) connected to the deleted spatial points in the polygon chain (24').
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present patent application publication relates to the recognition of a vehicle profile based on a surrounding sensor measurement data. Background Art

[0002] In modern vehicles that are at least partially designed to drive autonomously, sensors and sensor systems are increasingly used to detect other traffic participants. In particular, cameras, radar sensors and sometimes lidar sensors are used as sensors. A lidar sensor scans a specific spatial detection range with the aid of laser radiation and can determine spatial points of objects based on the laser radiation reflected by the objects in the detection range. Since the spatial points have spatial coordinates, they are also referred to as 3D (three-dimensional) measurement points below.

[0003] The data generated by such sensors are processed by algorithms in the driver assistance system, in particular for identifying objects in a vehicle's driving path and avoiding accidents that may occur with the identified objects. The algorithms usually used in the driver assistance system are configured to mark the identified objects with a so-called "rectangular frame", i.e., a box-shaped envelope in space or a rectangular envelope in a plane, for subsequent algorithm processing.

[0004] For example, DE 10 2013 008 828 A1 discloses a method for identifying a vehicle contour based on measurement data of an ambient sensor, wherein the measurement data has a plurality of contour points of the detected vehicle. A rectangle or polygon surrounding a vehicle is formed based on the contour points, wherein contour points relevant to the identification of the vehicle contour are determined for further processing. To this end, the number of contour points located in the row or column is determined for each row and each column of a formed grid diagram. For example, the column located in the left half of the grid diagram and having the most contour points constitutes the left edge of the rectangle describing the vehicle contour, and the row located in the lower half of the grid diagram and having the most contour points constitutes the lower edge of the rectangle describing the vehicle contour. Summary of the invention

[0005] In the following, a method according to the invention for recognizing a vehicle contour based on measurement data of a surroundings sensor and a safety system according to the invention for a vehicle are described.

[0006] According to a first aspect, a method for identifying a vehicle contour based on measurement data of an environment sensor is disclosed, wherein the measurement data has a plurality of spatial points of an object detected by the environment sensor. The disclosed method has the following contents: generating an envelope around the object in the form of a polygon chain based on the number of spatial points; reducing the number of spatial points by deleting all spatial points within the envelope; removing spatial points irrelevant to the identification of a vehicle contour from the polygon chain and determining spatial points of the polygon chain relevant to the identification of the vehicle contour; classifying one or more edges as belonging to the front and side of a vehicle based on the reduced number of spatial points. The above processing can be performed in particular as an algorithm step in a given sequence. The envelope can be, for example, a convex hull or formed by a cubic Hermite matrix spline. A vehicle contour identified by the method is different from a relatively inaccurate rectangular frame, which is particularly useful, for example, in the case of predicting an accident as reliably as possible and controlling passive safety devices. For example, the probability of a collision between the vehicle and another vehicle can be estimated more accurately based on the identified other vehicle contour, and the control of the passive safety device can be improved by, in particular, avoiding unnecessary or incorrect triggering of the passive safety device or more accurately adjusting the setting of the so-called pre-collision triggering.

[0007] In addition, the method may also include the following: the edges obtained according to the classification are supplemented with an estimate of a vehicle center axis to form a complete shape of the vehicle. By obtaining a more accurate vehicle contour, the recognition of the contour can be further improved.

[0008] In particular, the generation of an envelope around the object in the form of a polygonal chain based on the number of spatial points allows a Graham scanning algorithm to be performed on the number of spatial points. The Graham scanning algorithm can effectively calculate the envelope around the object even in the case of relatively large amounts of data, such as can be generated by a LiDAR-based surroundings sensor.

[0009] Removing spatial points irrelevant to the identification of a vehicle contour from the polygon chain includes: deleting spatial points on the envelope whose connecting straight lines and adjacent spatial points of the polygon chain mutually form an angle less than or equal to a predetermined angle, in particular 2 degrees, and connecting spatial points connected to the deleted spatial points in the polygon chain. Thus, spatial points that do not represent corner points of the vehicle contour and are therefore insignificant in principle for accident prediction can be removed from the polygon chain.

[0010] Determining a spatial point of a polygon chain related to a vehicle contour recognition has at least one of the following contents: if the spatial point connecting line and the adjacent spatial point of the polygon chain are greater than a predetermined angle, especially 20 degrees, then a spatial point is determined to be related to the vehicle contour recognition; if the sum of the angles between the connecting line and the adjacent spatial point of the polygon chain is greater than a predetermined angle between the spatial point and multiple adjacent spatial points in the polygon chain, especially 17.5 degrees, then a spatial point is determined to be related to the vehicle contour recognition; if the angle between the connecting line and the adjacent spatial point of the polygon chain multiplied by the length of at least one connecting line exceeds a predetermined value, especially exceeds 8°m, then a spatial point is determined to be related to the vehicle contour recognition. Each of these three determination methods can effectively determine the spatial point related to contour recognition.

[0011] Classifying one or more edges as belonging to the front and side of a vehicle according to the reduced number of spatial points may include the following steps: sorting the connecting straight lines of the polygon chain whose number of spatial points is reduced according to their lengths, and connecting the two longest connecting straight lines with a front parameter l f and a side parameter l s A comparison is made and a determination is made as to whether the front and / or sides of a vehicle are visible.

[0012] In particular, determining whether the front and / or side of a vehicle is visible includes at least one of the following: if the longest connected straight line is longer than the front parameter l f , and the second longest connecting line is in front of the parameter l f and side parameters l s If the longest connecting line is shorter than the side parameter l s , but longer than the previous parameter l f , then only the front edge is visible; if the longest connecting line is longer than the previous parameter l f , and the second longest connecting line is shorter than the side parameter l s , then it is determined that only the side faces are visible.

[0013] In addition, the following steps can be taken to supplement the complete shape of a vehicle by estimating a vehicle center axis based on the classified edges: a center axis is generated by setting a straight line passing through the center of a classified front face of the vehicle, whereby the straight line is roughly parallel to the classified sides of the vehicle, the classified front and sides of the vehicle are mirrored on the center axis, and the spatial points that have been mirrored onto the center axis plane are covered by the mirrored spatial points.

[0014] The method can be executed in a loop, and a trajectory of the vehicle can be determined by evaluating the supplementation of the complete vehicle shape. For example, a time change of a plurality of supplemented complete vehicle shapes can be determined, in particular a change of a vehicle longitudinal axis angle of the supplemented complete vehicle shape can be determined, and then a vehicle trajectory can be derived based on the change, for example, whether the vehicle is traveling on the trajectory of the vehicle passing by itself. Then, an accident prediction can be evaluated by a safety assistance system or a driver assistance system.

[0015] According to a second aspect, a safety system for a vehicle is disclosed, wherein the safety system is configured to execute the method disclosed herein and control at least one of the following components based on an identified vehicle contour: at least one passive safety component, in particular an airbag, a seat belt tensioner; at least one active safety component, in particular a brake assist system, an active steering device, a vehicle anti-lock braking system, and a driving dynamics control system.

[0016] In particular, the safety system can be designed to estimate the probability of an accident with the relevant vehicle based on the detected vehicle contour and to control the activation of a minimum number of components based on this estimate.

[0017] Further features emerge from the following description in conjunction with the embodiments shown in the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The following figures show

[0019] Figure 1 A vehicle having a surrounding environment sensor for object detection and an object vehicle approaching from the vehicle are shown;

[0020] Figure 2 Displays a vehicle and a target vehicle coming in the opposite direction of the vehicle, as well as the trajectory of the vehicle when performing an evasive maneuver to avoid an accident;

[0021] Figures 3 to 5 is an example of a polygon chain with corner points; and

[0022] Figure 6 An example of a vehicle contour identified using the method is shown. DETAILED DESCRIPTION

[0023] In the following description, identical, functionally identical and functionally related elements are provided with the same reference numerals. The absolute values ​​below are given only as examples and should not be understood as limiting.

[0024] The embodiments explained below are basically classified into the field of vehicle safety for all steps from automated driving to fully autonomous driving, wherein in this field, the focus is especially on early triggering of passive safety devices such as airbags and seat belt tensioners by means of sensors for detecting the surrounding environment and the collision partner. However, in general, the methods and systems described here are suitable for all types of driver assistance systems that use object recognition for vehicle safety, for example, they are also suitable for active safety systems such as automatic steering intervention, brake intervention, and driving dynamics control intervention that can affect vehicle control and avoid accidents.

[0025] The driver assistance systems currently used in vehicles mainly rely on radar and camera sensors for detecting the surroundings. This sensor information can be used, for example, to reduce the ignition threshold of the actuators of passive safety devices. However, the activation of the actuators has always been at the time t of the first contact between the two parties involved in the accident. 0 Then start.

[0026] By triggering protective devices such as seat belt tensioners or airbags and seat belt retractors, the 0 In order to minimize the impact of the accident on the occupants, it is necessary to have as much information as possible about the other party to the collision. This includes the speed and position that can be determined by the sensor. Other relevant parameters can only be determined within a very limited range or cannot be determined at all. Therefore, a lidar sensor (or laser scanner) will be used in the vehicle as the third surrounding environment detection sensor in the future. Since the sensor data of a lidar sensor can provide a large number of measurement points or 3D (three-dimensional) measurement points in space, the vehicle contour of the other party to the accident can be calculated with the help of the sensor. As a result, the impact point position of the other party to the accident on the vehicle and the impact range of the other party to the accident can be estimated with higher accuracy. The following describes a method with the help of which the estimated value of a simple rectangular frame (an envelope in the form of a cuboid in space or a rectangular form in a plane) can be raised to a detailed application for identifying the contour of an oncoming vehicle in a pre-collision situation.

[0027] If objects in the surroundings of the vehicle are detected today by a vehicle surroundings sensor, the contours and exact dimensions of these objects cannot be determined. A rectangular frame (in the smallest possible rectangular plane parallel to the axis, enclosing a predetermined number of objects or measuring points) is simply placed around each individual object without taking into account the exact object shape. In this way, valuable information about a detected vehicle object is lost, which information could be relevant for triggering the ignition of passive safety devices such as seat belt tensioners.

[0028] Figure 1The problem is explained based on a simplified illustration of a recognized vehicle. The target vehicle 12 is in the field of view (circular arc segment surrounded by a dotted line) or sensor detection area 14 of the host vehicle 10. The sensor detection area 14 corresponds to the detection area of ​​the surrounding environment sensor in the host vehicle 10 for detecting objects in front of the vehicle, that is, detecting objects along the usual driving direction. As sensors, the surrounding environment sensor has, for example, a radar sensor, a lidar sensor and / or a camera sensor. The target vehicle 12 is detected by the surrounding environment sensor as an oncoming vehicle and is integrated into a surrounding environment model of a driver assistance system of the host vehicle 10. However, this integration is not carried out in the form of an actual target vehicle, but in the form of a rectangular frame 16 (in Figure 1 The target vehicle 12 is represented by a rectangular border 16 around the target vehicle 12 in the system. Accordingly, the target vehicle 12 is stored as a rectangular border 16 that is larger and less accurate than the actual one in the system. This error is displayed as a shadow area 18 around the target vehicle 12.

[0029] If a passive safety device is to be triggered, the assumption that the target vehicle 12 is too large may lead to an erroneous triggering. Due to the overlap of the outlines between the rectangular frame 16 and the host vehicle 10, an impending collision is predicted, including the collision situation and severity, so the passive safety devices, especially the seat belt tensioner, are activated. However, in reality, completely different collision situations may occur, resulting in completely different collision severities, until the accident is completely avoided when passing. Therefore, although the host vehicle 10 and the target vehicle 12 both meet in the prediction (the rectangular frame 16 of the target vehicle 12 overlaps with the host vehicle 10, but the real vehicle outlines of the host vehicle 10 and the target vehicle 12 do not overlap), it is also possible that no collision occurs. Figure 2 This is illustrated by way of example in FIG: the trajectory 20 of the host vehicle 10 passes through the trajectory 22 of the target vehicle 12, so that no collision actually occurs between the host vehicle 10 and the target vehicle 12. However, the driver assistance system predicts a collision because it only takes into account the rectangular frame 16 of the target vehicle 12 and assumes that the front of the target vehicle 12 collides with the left fender of the host vehicle 10 in the prediction.

[0030] Using this simplified "rectangular frame" method without precise contour determination can lead to false triggering, which makes it impossible to safely trigger the actuators of the passive safety devices before a collision. Therefore, it is necessary to ensure that the precise contour of the target vehicle 12 is known. The following describes how to solve this technical problem in a pre-collision situation by the contour recognition method of the oncoming vehicle introduced in this article.

[0031] As mentioned above, the goal is to estimate the contour of an oncoming vehicle as accurately as possible during the pre-collision phase. The algorithm described below performs this accurate calculation in particular through four consecutive steps, namely "generating a convex hull around the object", "determining key corner points", "classifying the front and side of the vehicle", and finally "completing the complete vehicle shape by estimating a central axis". The last supplementary step is optional and is mainly used to improve the accuracy of contour recognition.

[0032] Generate a convex hull around the target

[0033] As the basis for the convex hull, the well-known Graham scanning algorithm is used, which is characterized by its short running time and high efficiency. The algorithm was proposed by mathematician Ronald Lewis Graham in 1972. Here, a starting point P must be defined first. 0 , where the point with the smallest ordinate and abscissa is always used, since in the method described this point must represent a corner point. Now, all other points will be calculated according to P 0 The angles between the convex hull and the x-axis are sorted counterclockwise in ascending order. It is now tested whether the three points form a positive triangle and this is checked by means of a row matrix. If a positive triangle is formed, the outer line of the triangle is defined as the outer envelope. If a triangle is negative, the calculation is restarted, with the intermediate points being skipped. This process is repeated until the starting point is reached again, thereby fully completing the convex hull. Finally, the connection between the point with the lowest x-coordinate and the highest x-coordinate is deleted in order to obtain only shapes that can also be seen by the lidar. The convex hull obtained in this way corresponds to a polygonal chain in which pairs of spatial points that define the envelope are connected to each other by connecting paths or connecting straight lines.

[0034] Determine relevant corners or key corners

[0035] In the following steps, corner points are formed from the polygonal chain, from which an accurate estimate of the vehicle profile is established and the longitudinal axis of the vehicle (the relevant corner points or key corner points or spatial points) is derived. Since the LiDAR point cloud represents a large amount of data, most of which is not used, the measurement data is reduced here. On the one hand, all points inside the convex hull are deleted, and on the other hand, all points where a path or a straight line connecting the spatial points of a polygonal chain has only a negligibly small angle are deleted, i.e. all points with a negligible angle to each other are deleted.

[0036] This is Figure 3 This is illustrated by the example of a short polygon chain before 24 and after 24' in which the relevant corner points or key corner points are determined: Figure 3A polygon chain 24 with three corner points 26 is shown on the left. The connecting straight lines 28 of the corner points are at an angle of 20 degrees to each other at the left corner point, at an angle of 2 degrees to each other at the middle corner point, and at an angle of 17 degrees to each other at the right corner point. In this example, an angle limit of 2 degrees is specified. This means that the algorithm only classifies a corner point as a critical corner point or a relevant corner point if the angle between the straight lines connecting a corner point is greater than 2 degrees to each other. All corner points whose connecting straight lines are at an angle of less than or equal to 2 degrees to each other are thus deleted. Therefore, Figure 3 In the example shown, the middle corner point is deleted by the algorithm, and the left corner point and the right corner point are connected to each other again by a new straight line 30, thereby forming Figure 3 The transformed polygon chain 24' is shown on the right side of the figure.

[0037] Next, the relevant corner point is defined from the other measured points. For this purpose, it is checked whether the measured points meet one of the following conditions, in particular in descending order from a) to c). If a point has already been defined as a corner point, it can be predefined that it does not have to meet the other conditions. It should also be noted that the two points with the smallest x-coordinate and the largest x-coordinate automatically represent corner points.

[0038] a) Regarding large angle changes

[0039] Large angle changes in vehicle structure are closely related to significant changes in radius, such as the transition from lights to fenders. Therefore, a corner point is generated at a location with an angle transition of, for example, greater than 20 degrees.

[0040] b) Addition of small angles

[0041] The focus is primarily on the transition between the vehicle radiator grille and the lights. To this end, the angular variation of all points within a defined length (e.g. 20 cm) is checked and the individual values ​​are added up. If, for example, the sum exceeds a limit value of 17.5 degrees, a corner point is defined on this section. The coordinates of this point are calculated from the arithmetic mean of all the points involved. Figure 4 An example is shown: In the polygon chain 24'' shown, the middle of the three corner points 32, 32' and 32' has a specified length. The connecting straight lines connecting these corner points make angles of 7 degrees, 5 degrees and 6 degrees to each other. The sum of the angles is 18 degrees, which is higher than the limit value of 17.5 degrees given as an example above, so that a corner point can be defined on the portion of the polygon chain 24'' that is to surround the three corner points 32, 32', 32'.

[0042] c) Weighting with respect to line length

[0043] For the side structure of a vehicle, it is usually not sufficient to define a corner point based solely on the angle between previously filtered straight lines. Therefore, it may be necessary to also take the length of the adjacent straight lines into account in the determination. To this end, the angle is multiplied by the straight line and then weighted. If a limit value (e.g. 8°m) is exceeded again here, the point is determined to be a corner point. Figure 5 An example is shown: the length of the straight line 36 between two points 34 and 34' is multiplied by the angle 2.5 degrees enclosed by the straight line connecting the corner point 34. The value thus obtained is compared with a limit value; if the limit value is exceeded, the point 34 is classified as a corner point, otherwise it is deleted.

[0044] If the target vehicle is absolutely head-on towards the host vehicle, the front structure of the vehicle can be classified. If the target vehicle is approaching the host vehicle at an angle, the side of the vehicle can also be classified. This is done while taking into account the movement history of the tracked vehicle.

[0045] Classification of the front and side of the vehicle

[0046] After the corner points are determined, the contour is simplified based on, for example, six to eight points. From this, the front, side and headlights of the vehicle are derived. The background is that it is desired to determine the orientation angle of the target vehicle relative to the host vehicle.

[0047] To do this, first sort all lines in ascending order according to their length. Then, compare the two longest lines with the two parameters, the positive l f (e.g. = 1.60 m) and lateral l s (e.g. = 0.55 m) for comparison. Here, four situations can be distinguished that clearly determine the visible part of the vehicle. The specific situations are shown in the following table.

[0048] Case significance illustrate 1 Visible front and side <![CDATA[The longest line is longer than l f and the second longest line is between l f and l s > 2 Only the front is visible <![CDATA[The longest line is shorter than l s , but longer than l f . If this line meets the two longest lines, select the line with a larger angle relative to this vehicle.]]> 3 Only the sides are visible <![CDATA[The longest line is longer than l f , and the second longest line is shorter than l s .]]> 4 No side can be recognized / interfered by LiDAR Does not meet the conditions of cases 1, 2, and 3

[0049] With the aid of the steps described above, a vehicle contour, in particular a vehicle front contour, can be identified with relatively high accuracy and, if necessary, a vehicle side contour can be identified as a digital representation of a vehicle with relatively high accuracy in an improved display method compared to a simple rectangular frame or a cuboid, because the contour identified in the method corresponds more accurately to the actual vehicle contour.

[0050] Completing the vehicle shape by estimating the mid-axis

[0051] This last step further improves the recognition of the contour and allows an almost complete vehicle shape or vehicle contour to be determined, even if the vehicle is partially obscured or cannot be recognized by the LiDAR sensor. To this end, the longitudinal axis of the vehicle is first determined. If part of the target vehicle is obscured or part of the vehicle is not visible to the surroundings sensor, the longitudinal axis of the vehicle must be determined in order to obtain the entire vehicle contour. For this purpose, a straight line is set through the center of the detected front structure, parallel to the visible vehicle side. Subsequently, the vehicle contour facing the host vehicle is mirrored onto the center axis. The points on the other side of the center axis are covered by the new, mirrored corner points.

[0052] In the case of a target vehicle approaching the host vehicle head-on, the longitudinal axis of the vehicle can also be determined to improve the accuracy of the coverage estimate. In this case, an orthogonal line defines the longitudinal axis by identifying the center of the front of the vehicle. In addition, the accuracy of the center axis position can be gradually improved over time by continuously observing the target or by determining the width and deviation angle of the target parameters.

[0053] The back side of the target vehicle cannot be specifically determined in this way, but can only be established by connecting the corner points at the back. This is due to the invisibility of the laser radar. However, this does not pose a problem here, because the exact shape of the back side of the vehicle is not important for the occurrence of an accident.

[0054] Thus, a complete outline estimation of the target vehicle is finally formed, such as Figure 6 The example is shown by an outer contour line 38 surrounding the target vehicle 12 .

[0055] at the same time, Figure 6 A significant advantage over current systems is also demonstrated. With the new method, the target vehicle 12 is identified more accurately than with previous methods, thereby omitting the overhanging shadow surface area 18 estimated by a rectangular frame 16. The uncertainty caused by the shadow surface area 18, which could lead to false triggering, is completely eliminated, so that the reliability of the accident prediction can be significantly improved. Thus, this algorithm for improving the recognition of the contour of the other party in the accident in a pre-crash phase can solve the main problems when activating passive safety actuators (unnecessary triggering, false triggering, etc.).

Claims

1. A method for identifying a vehicle (12) contour based on measurement data from a surroundings sensor, wherein: The measurement data has a set of spatial points of a target detected by a surrounding environment sensor, wherein the method has the following contents: Generate an envelope around the target in the form of a polygon chain according to the spatial point set; Simplifying the spatial point set by: deleting all spatial points within the envelope; deleting spatial points irrelevant to a vehicle contour recognition from the polygon chain; and determining spatial points of the polygon chain relevant to a vehicle contour recognition; and According to the simplified spatial point set, one or more edges are classified as belonging to the front of a vehicle and the side of a vehicle, wherein, Deleting a spatial point irrelevant to a vehicle contour recognition from a polygon chain comprises the following steps: deleting a spatial point (26) on an envelope, wherein the angle between the spatial point (26) and a connecting straight line (28) of an adjacent spatial point (26) in the polygon chain (24) is less than or equal to a first predetermined angle, and connecting (30) the spatial point (26) in the polygon chain (24) connected to the deleted spatial point.

2. The method according to claim 1, characterized in that The first predetermined angle is 2 degrees.

3. The method according to claim 1 or 2, It is characterized in that It has the following content: A vehicle center axis is estimated based on the classified edges, thereby completing a complete shape (38) of the vehicle.

4. The method according to claim 1 or 2, It is characterized in that An envelope around the object is generated in the form of a polygon chain according to the spatial point set, which includes executing a Graham scanning algorithm on the spatial point set.

5. The method according to claim 1 or 2, It is characterized in that Determining a polygonal chain of spatial points associated with a vehicle contour recognition has at least one of the following contents: If the angle between a spatial point and a connecting straight line of adjacent spatial points of the polygon chain is greater than a second predetermined angle, the spatial point is determined to be relevant to vehicle contour recognition; For a spatial point and a plurality of spatial points adjacent to the spatial point in the polygon chain, if the sum of angles between connecting straight lines connecting the spatial point and the plurality of adjacent spatial points is greater than a third predetermined angle, the spatial point is determined to be relevant to vehicle contour recognition; A spatial point is determined to be relevant to vehicle contour recognition if the angle between a spatial point and a connecting line of adjacent spatial points in the polygon chain multiplied by the length of at least one connecting line exceeds a predetermined value.

6. The method according to claim 5, characterized in that The second predetermined angle is 20 degrees.

7. The method according to claim 5, characterized in that The third predetermined angle is 17.5 degrees.

8. The method according to claim 5, characterized in that The predetermined value is 8°m.

9. The method according to claim 1 or 2, It is characterized in that According to the simplified spatial point set, one or more edges are classified as belonging to the front of a vehicle and the side of a vehicle, which has the following steps: Sort the spatial point connecting straight lines after the polygon chain simplification according to their length, Will have a preceding parameter l f and a side parameter l s Compare the two longest connecting lines of Determine if the front and / or sides of the vehicle are visible.

10. The method according to claim 9, It is characterized in that Determines whether the front and / or sides of a vehicle are visible, with one of the following: If the longest connection line is longer than the previous parameter l f , and the second longest connecting line is located in the front parameter l f and side parameters l s If the distance between the front and the side of the vehicle is smaller than 100 km, only the front and the side of the vehicle are visible; If the longest connecting line is shorter than the side parameter l s , but longer than the previous parameter l f , it is determined that only the front of the vehicle is visible; If the longest connection line is longer than the previous parameter l f , the second longest connecting line is shorter than the side parameter l s , it is determined that only the side of the vehicle is visible.

11. The method according to claim 9, It is characterized in that The method estimates a vehicle center axis based on the classified edges, thereby supplementing a complete shape of a vehicle, which includes the following steps: A median axis is created by arranging a line through the center of a classified front face of a vehicle so that the line is substantially parallel to the classified sides of the vehicle, mirroring the classified front and side images of the vehicle onto the central axis, and The spatial point reflected by the mirror covers the spatial point that has been reflected by the mirror on the medial axis plane.

12. The method according to claim 10, It is characterized in that The method estimates a vehicle center axis based on the classified edges, thereby supplementing a complete shape of a vehicle, which includes the following steps: A median axis is created by arranging a line through the center of a classified front face of a vehicle so that the line is substantially parallel to the classified sides of the vehicle, mirroring the classified front and side images of the vehicle onto the central axis, and The spatial point reflected by the mirror covers the spatial point that has been reflected by the mirror on the medial axis plane.

13. The method according to claim 1 or 2, It is characterized in that The method is executed cyclically and a trajectory of the vehicle can be determined by analyzing and evaluating the complete shape supplement of the vehicle.

14. A safety system for a vehicle, wherein: The safety system for implementing the method described in any one of the above claims and controlling at least one of the following devices according to the identified vehicle contour is designed as: at least one passive safety device; at least one active safety device.

15. The security system according to claim 14, It is characterized in that The passive safety device is an airbag and a seat belt tensioner.

16. The security system according to claim 14, It is characterized in that The active safety components are a brake assist system, an active steering device, a vehicle anti-lock braking system, and a driving dynamics control system.

17. The security system according to claim 14, It is characterized in that It is designed to estimate the possibility of an accident with the relevant vehicle based on the recognized vehicle profile and to control the triggering of at least one device based on the estimation.

Citation Information

Patent Citations

  • Method for creating a model of a vehicle's environment and a correspondingly designed driver assistance system

    DE102013008828A1

  • Method for displaying the surroundings of a vehicle

    CN102576075A

  • Measurement of a dimension on a surface

    CN107003409A