DE102015209764A1 - Extrinsic calibration of an image capture device of a vehicle - Google Patents

Abstract

The extrinsic calibration of an image capture device in a stationary vehicle has not been sufficiently solved.
According to the invention, the problem is solved by a device for determining the extrinsic parameters of an image capture device (21), comprising:
A memory device (31);
- At least one image detection device (21) which is adapted to detect at least one environmental image of the vehicle (10);
- an identification device (32);
A comparison device (34); and
A calibration device (33),
in which
- at least one vehicle feature (16, 17) and at least one desired position of the at least one vehicle feature (16, 17) are stored in the memory device (31);
- The identification device (32) is adapted to identify an actual position of the at least one vehicle feature (16, 17) in the at least one environmental image of the vehicle (10);
- The comparison device (34) is adapted to compare the actual position with the desired position; and
- The calibration device (33) on the basis of a deviation of the actual position of the desired position determines the extrinsic parameters of the at least one image detection device (21).

Description

The invention relates to an apparatus and a method for extrinsic calibration of an image capture device of a vehicle, in particular of a passenger car, and in particular the determination of extrinsic camera parameters.

Modern cars are usually equipped with a variety of cameras. These cameras can be e.g. in the apron of the front bumper, in the side mirrors or in the tailgate. The cameras allow the driver to e.g. when reversing, an image display is displayed in the cockpit showing the environment behind the vehicle. It is also possible to combine the images of the individual cameras so as to provide an all-round view of the vehicle. Such all-round visibility systems can be used by the driver e.g. show a plan view of the vehicle, wherein the environment of the vehicle is displayed. This allows safe navigation of the vehicle, even in otherwise confusing situations.

For the composition of the individual images in a panoramic system, it is necessary that the cameras are calibrated correctly. Inadequate calibration may be i.a. lead to inaccuracies or double images at the transitions of the individual pictures.

The camera parameters to be determined are subdivided into extrinsic and intrinsic camera parameters. The intrinsic camera parameters include i.a. the focal length and the x and y values of the center of the image. These parameters can be calibrated at the factory during production.

The intrinsic parameters of the cameras are used, for example, to eliminate a distortion of the image occurring in fish-eye optics in order to be able to display a natural environment image to the driver.

The extrinsic camera parameters, on the other hand, indicate the position of the camera in the room. This position can be expressed by a "pose".

A pose is a combination of position and orientation in space. The orientation in space is defined by an angular offset of the coordinate axes of a coordinate system shifted by the translational parameters with respect to a base coordinate system, e.g. a vehicle coordinate system. A pose thus comprises a translational and a rotational component. A pose can be expressed in three-dimensional space as a 6D vector.

Like the intrinsic camera parameters, the extrinsic camera parameters can also be determined at the factory.

However, while intrinsic camera parameters generally do not change over the life cycle of a camera, extrinsic camera parameters are subject to many external influences and thus to fluctuations.

So can u.a. mechanical or thermal stresses cause the orientation and / or position of the cameras to change. Such a change in the pose of the cameras results in driver assistance systems, e.g. an all-round vision system, not working optimally.

To address this, it is common today for the extrinsic camera parameters to be redetermined while the vehicle is traveling.

In one common method, environmental characteristics are observed while driving the vehicle through a camera. Thus, e.g. a longer observation of the lanes of a roadway can be used to make a calibration.

With the aid of a vehicle model held in a vehicle computer, which has all relevant parameters, such as e.g. the current speed of the vehicle or the current steering angle, it can be predicted in which direction and at what distance the features will move on the ground. If there is a deviation of the prediction and a measurement, conclusions can be drawn on the current pose of the camera. Thus, the extrinsic parameters can be determined.

The method described is error-prone.

In addition, it is necessary for the cameras to observe the feature over an extended period of time so that the extrinsic camera parameters can converge toward certain values, thus minimizing the error. However, driver assist systems of a vehicle often need to operate optimally right at the start of the engine, e.g. to allow a safe and possibly completely automated parking.

Another disadvantage of the method described is that only the translational parameters of the cameras can be clearly determined, but not the rotational. This can lead to the already described inaccuracy of the driver assistance systems.

Based on this prior art, the object is to provide a device or a method which eliminates the aforementioned disadvantages. In particular, a device is to be created which determines (all) extrinsic parameters of a camera.

This object is achieved by a device according to claim 1 and a method according to claim 13.

• – eine Speichereinrichtung;
• – mindestens eine Bilderfassungseinrichtung, insbesondere mindestens eine Videokamera, die dazu ausgebildet ist, mindestens ein Umgebungsbild des Fahrzeugs zu erfassen;
• – eine Identifikationseinrichtung;
• – eine Vergleichseinrichtung; und
• – eine Kalibriereinrichtung.

In particular, the object is achieved by a device which comprises:

• A memory device;
• - At least one image capture device, in particular at least one video camera, which is adapted to detect at least one environmental image of the vehicle;
• An identification device;
• A comparison device; and
• - a calibration device.

In the memory device according to the invention at least one vehicle feature and a desired position of the at least one vehicle feature are stored. The identification device is designed to identify an actual position of the at least one vehicle feature in the at least one environmental image of the vehicle.

The comparison device is designed to compare the actual position with the desired position. Furthermore, the calibration device is designed to determine the extrinsic parameters of the at least one image capture device on the basis of a deviation of the actual position from the desired position.

The image capture devices of a vehicle are installed in many models so that they capture a part of the vehicle with. Thus, a built-in rearview mirror camera can detect the wheels, as well as other areas of the body. In a built-in front bumper camera parts of the front bumper are included in many cases. Rear view cameras and all-round vision cameras generally use fisheye optics, which always include the contour of their own vehicle.

The actual position of a stored in a memory device at least one vehicle feature can now be identified in the captured by the at least one image capture device image.

Such a vehicle feature may be, for example, a wheel, a part of the hubcap or a contour of the vehicle body. In the context of this application, it is assumed that a contour can represent a closed curve but also a non-closed curve. This curve can be curved or not curved. The contour may also contain areas that correspond to a kink, such as a corner.

If the actual position of the vehicle feature is identified, a desired position of the vehicle feature stored in the memory device can be compared with the measured actual position of the vehicle feature in the captured image of the image capture device. Based on the deviation of the positions of the features, it is possible to determine the extrinsic parameters of the at least one image capture device.

The extrinsic parameters of the at least one image capture device may be found, for example, by a perspective transformation that transforms the at least one vehicle feature so that the transformed at least one vehicle feature is in registration with the stored at least one vehicle feature.

The advantages of the present invention lie in a simple and rapid determination of the extrinsic camera parameters. Also, it is not necessary to attach additional markers or similar components to the vehicle to calibrate the image capture devices. In addition, the calibration can be performed while stationary, so that the image capture devices are calibrated before driving. Furthermore, all extrinsic parameters are determined, which ensures a high accuracy of the calibration.

In a further embodiment, the at least one vehicle feature may represent a geometric structure, in particular at least one contour, at least one edge or at least one corner of the vehicle or a combination thereof.

An advantage of choosing at least one vehicle feature as a geometric structure is that there are no other elements on the vehicle, such as a vehicle. Marker, must be attached. Furthermore, it is advantageous to choose a combination of different geometric vehicle features in order to achieve high accuracy in the determination of extrinsic parameters. Furthermore, it is advantageous to use a contour and at least one corner, as corners and contours can be found very quickly and efficiently in pictures. Furthermore, it is thus possible to clearly determine the orientation.

Corners are particularly well suited as features, since at corners a change of the gradient in all directions is very large. They thus indicate characteristic image areas that can easily be found in further images.

In one embodiment of the invention, the at least one vehicle feature may relate to a corner, wherein a position and an orientation of the corner are stored in the memory device.

Further advantages arise when as few vehicle features as possible are stored, but they already provide all the information necessary to determine the extrinsic parameters. So it is advantageous to save an alignment to a corner, because thus also a rotation can be determined.

In one embodiment, the at least one vehicle feature may be represented by a matrix describing a vertex, in particular, the matrix comprising a scaling invariant feature descriptor ("SIFT descriptor") or a multiscale oriented patches descriptor ("MOPS descriptor ").

Further advantages arise when a corner is represented as a matrix describing the corner. So it is possible to choose a representation that is geometrically invariant, for example against rotations and scaling. The probability of successfully identifying the corner in the captured at least one environment image is thus increased.

In one embodiment, the identification device may be configured to identify a profile of a contour of the vehicle in the at least one environmental image and to identify the actual position of a corner as the at least one vehicle characteristic based on the contour of the contour.

Contours of a vehicle can be identified by edge detection using known methods. It is advantageous to identify a corner from an identified trace of a contour. For example, it may be assumed that the corner to be identified is at one end of the course. Such a prior knowledge can be readily assumed, since the vehicle geometry is known. This speeds up the identification of the at least one vehicle feature and thus the extrinsic calibration of the at least one image capture device can be completed more quickly.

In one embodiment, the identification device may be designed to carry out the identification of the actual position of the at least one vehicle feature on the basis of a previously determined partial section of the environmental image.

It is advantageous to downsize the area used for identification to allow increased processing speed. For example, when incorporating image capture devices on a passenger car, it can be assumed that the errors of the extrinsic parameters are at certain intervals. Thus, it is possible to determine the image detail in which the at least one vehicle feature can be expected.

In one embodiment, the at least one image capture device may have a fish-eye optic which has a viewing angle of greater than 90 degrees, in particular greater than 170 degrees.

The use of fish-eye optics has the advantage that a larger part of the vehicle is detected. Thus, it is also possible to detect multiple vehicle features, resulting in increased accuracy of the extrinsic calibration.

In one embodiment, the memory device may store a first high priority vehicle feature and a second low priority vehicle feature. The identification device may be configured to determine whether the first vehicle feature lies in the at least one environmental image of the vehicle. In the case where the first vehicle feature is not located in the at least one environmental image of the vehicle, the identification device may be configured to determine whether the second vehicle feature is located in the at least one environmental image. In the case where the second vehicle feature lies in the at least one environmental image, the identification device may identify the actual position of the second vehicle feature in the at least one environmental image.

It is thus possible for the storage device to store two vehicle features, with one vehicle feature being assigned a high priority and a second vehicle feature having a low priority. The identification device can now first select the first vehicle feature due to the high priority. If the first vehicle feature can not be found in the at least one environment image, then the identification device can use the second vehicle feature. Additionally or alternatively, confidence levels may be stored for each or some features related to their identification.

This embodiment has the advantage that the method becomes more robust. For example, it is possible for a vehicle feature due to Pollution is not and / or not reliable can be found and thus a second vehicle feature must be used. The device is thus still operational even under difficult external conditions. In addition, it is possible to provide feedback to the driver that individual characteristics are not identifiable. This then has the opportunity to check his vehicle and remove any impurities.

In one embodiment, the calibrator may be configured to determine the extrinsic parameters using the least squares method. Advantages of using the least squares method are that very fast algorithms are available for them.

The stated object is also achieved by a vehicle which comprises a device for determining the extrinsic camera parameters of the at least one image capture device, as described above. In this case, the vehicle may comprise the at least one image capture device and the at least one image capture device may be oriented so that the at least one environmental image depicts the surroundings of the vehicle and a part of the vehicle.

The vehicle has similar or identical advantages as described in connection with the device.

In one embodiment, the vehicle may include an estimation device that estimates the distance between the actual position of the at least one vehicle feature and the highest point of a wheel rim of the vehicle based on the actual position of the at least one vehicle feature.

It is thus possible to individually perform an altitude estimation for each wheel of the vehicle. This can be used, among other things, to validate the measured height values of wheel mounted ride height sensors to detect faulty-working sensors.

• – Bremsen;
• – Lenkvibrationsmotor;
• – Signalleuchte; oder
• – Seitenspiegelstellmotor.

In one embodiment, the vehicle may include a driver assistance system that generates control signals for an actuator using the at least one image capture device and the extrinsic parameters, wherein the actuator preferably comprises:

• - brakes;
• - steering vibration motor;
• - Signal light; or
• - Side mirror actuator.

An image capture device that has been calibrated with a device described above may be used in a driver assistance system that sends control signals to actuators. One advantage is that the driver assistance system works with very accurately calibrated imaging systems and therefore has greater confidence in the data affecting their behavior. It is thus possible that the vehicle also safely realizes safety-critical functions via driver assistance systems, such as e.g. an automatic parking.

• – Erfassung mindestens eines Umgebungsbildes des Fahrzeugs durch mindestens eine Bilderfassungseinrichtung;
• – Identifizieren einer Ist-Position eines mindestens einen Fahrzeugmerkmals in dem mindestens einen Umgebungsbild;
• – Vergleichen der Ist-Position mit einer in einer Speichereinrichtung gespeicherten Soll-Position des mindestens einen Fahrzeugmerkmals;
• – extrinsisches Kalibrieren der mindestens einen Bilderfassungseinrichtung auf Basis der bestimmten Abweichung.

Furthermore, the object is achieved by a method for extrinsic calibration of an image capture device of a vehicle, comprising the following steps:

• - Detecting at least one environmental image of the vehicle by at least one image capture device;
• - identifying an actual position of at least one vehicle feature in the at least one environmental image;
• Comparing the actual position with a desired position of the at least one vehicle feature stored in a memory device;
• - Extrinsic calibration of the at least one image capture device based on the determined deviation.

In one embodiment, the identifying may include restricting to a predetermined portion of the environmental image, based on which the identification of the actual position of the at least one vehicle feature may be performed.

There are similar or identical advantages, as has already been described in connection with the device.

In an embodiment, the method may be performed while the vehicle is traveling.

The method described can be used during the life of a vehicle but also while driving. The calibration while driving has the advantage that a permanent check of the particular extrinsic parameters can be made. Thus, it can be guaranteed at any time that the at least one image capture device works optimally.

Furthermore, the method can be used in a vehicle all-round visibility system.

For an all-round visibility system, it is necessary to capture captured environmental images from different ones To combine image capture facilities into a single image. In the case of an insufficient extrinsic calibration of the image capture devices, overlaps and distorted representations occur in the edge regions of the individual images. This can cause the driver of a vehicle can not detect obstacles. An improved extrinsic calibration thus leads to increased safety of the vehicle and thus contributes to the protection of road users.

The object of the invention is further solved by a computer-readable storage medium having executable instructions that cause a computer to implement a method according to one of the methods described above when the instructions are executed.

The invention will be described by means of several embodiments, which are explained in more detail with reference to figures. Hereby show:

1 a plan view of a vehicle with multiple cameras;

2 the pose of a camera in a base coordinate system;

3 a front view of the vehicle with vehicle feature;

4 a side view of the vehicle with side camera and vehicle limit;

5 a perspective transformation of an object;

6 a schematic view of an on-board computer;

7 schematic representation of a use of the invention in an automatic parking system assistance system.

In the following description, the same reference numerals are used for identical and equivalent parts.

1 shows a schematic plan view of a vehicle 10 , which with four image capture devices 21 Is provided. The image capture devices 21 capture each part of the vehicle environment 15 and forward this to an on-board computer 30 further. In a first embodiment, all image capture devices are 21 as video cameras 21 formed with fisheye lenses, so that the angle of view is> 160 °. Thus, also parts of the vehicle 10 recorded with.

So capture, for example, the right side camera 21 and the left side camera 21 one front and one rear wheel each 12 , a part of the body and a part of each side door. The front camera 21 is oriented so that it covers part of the front apron.

In the vehicle 10 is a vehicle center 13 defining the origin of a base coordinate system X, Y, Z. With respect to this basic coordinate system X, Y, Z, the extrinsic calibration of the cameras 21 carried out.

An optimal installation of the cameras 21 is assumed to be known, for example from design drawings. Because the optimal installation of the cameras 21 are known are also optimal extrinsic parameters of the cameras 21 known. The pose of the optimal installation of a camera 21 is referred to below as the target pose.

• – die Position relativ zu dem Fahrzeugzentrum 13; und
• – die Orientierung des Koordinatensystems X‘, Y‘, Z‘ relativ zu dem Basiskoordinatensystem X, Y, Z.

2 is a representation of the base coordinate system X, Y, Z, which in the vehicle center 13 is spanned, as well as the camera coordinate system X ', Y', Z ', which in the right side camera 21 is spanned. The extrinsic parameters of the right side camera to be determined 21 include:

• The position relative to the vehicle center 13 ; and
• The orientation of the coordinate system X ', Y', Z 'relative to the base coordinate system X, Y, Z.

The extrinsic parameters therefore comprise a spatial coordinate (with three values), as well as three angles.

3 shows a front view of the vehicle 10 , In the right side mirror 11 is a camera 21 installed, which is an environmental image of the vehicle 10 detected. As indicated by the dashed line in 3 implied, is also the vehicle limit 16 and a bike 12 also detected.

4 shows this clearly. In the 4 is the right side camera 21 of the vehicle 10 shown. Dashed lines from the side camera 21 starting interpret the captured image detail. 4 also shows that the vehicle limit 16 between the wheels 12 essentially straight and on the wheels 12 a kink 17 power. This kink 17 may be in a captured image of the right side camera 21 as a corner 17 be interpreted.

Due to the known vehicle geometry and the known target pose of the right side camera 21 , a reference image of the image section can be generated before delivery of the vehicle. In one embodiment, it is possible for such a Image is rendered virtually in the construction. In the reference image can now be the target position of individual vehicle features 16 . 17 be determined.

vehicle features 16 . 17 are preferably characteristic features that are different from their environment. For example, the vehicle limit 16 in a gray image representation, a very characteristic gradient of pixel intensity. This means that the intensity values of the pixels of the image change over the vehicle boundary 16 change strongly. The vehicle limit 16 however, has such a characteristic gradient only in one direction. This direction is transverse to the course of the vehicle boundary 16 ,

corners 17 of a picture are also very characteristic, as well as with corners 17 a strong change in intensity is observed. At corners 17 However, this is to be observed in all directions.

The vehicle limit 16 can in a 2D image of the camera 21 as a contour 16 be interpreted. In the first embodiment, the contour runs 16 between the front and rear wheels 12 essentially straight.

The contour 16 may be stored as an ordered list of points, assuming that the first element of the list is the position of the contour 16 indicates. In other embodiments, another representation of a contour is 16 conceivable, such as by a start and an end point. This is particularly advantageous when the contour 16 straight.

A corner 17 may be stored in the first embodiment as a point with an orientation. In this case, the orientation may include two vectors, which are the corner 17 span. So it's possible the corner 17 in an array where the first two elements are the X and Y coordinates of the corner 17 in the picture and 4 other elements indicate the X and Y parts of two vectors.

A corner 17 In other embodiments, it may also be stored as a feature descriptor, such as a scaling invariant feature descriptor ("SIFT descriptor"). Here, a quadratic neighborhood of the vertex is stored, choosing a representation of the eigenvectors that is invariant to geometric transformations, such as rotations. This has the advantage that the feature can easily be found even in heavily twisted images.

The contour 16 is in the storage device 31 saved with a target position. Also, in the first embodiment, the kink becomes 17 the vehicle limit 16 on the front wheel 12 as a corner 16 with a desired position in a memory device 31 saved. The reference image itself is not necessary in the first embodiment.

When starting the vehicle 10 in the first embodiment, an extrinsic calibration of the cameras 21 of the vehicle 10 causes. To determine the extrinsic camera parameters of the side camera 21 , captures the side camera 21 an environmental image of the vehicle 10 , In the first embodiment, it is assumed that the deviation of the actual pose of the side camera 12 small from the target pose. It can therefore be assumed that the previously defined vehicle characteristics 16 . 17 with their actual positions in the storage device 31 are stored in the captured environmental image.

It can further be assumed that due to the small deviations of the actual pose from the target pose of the side camera 12 the approximate position of the vehicle features 16 . 17 is known in the captured environment image. Thus, in other embodiments, the search range may be for these vehicle features 16 . 17 be restricted.

An identification device 32 can the contour 16 for example, based on gradient-based methods, such as the Canny algorithm, in which captured environmental image is detected. The corner 17 For example, in one embodiment, it may be detected with the Harris operator in the environmental image.

Next represents the identification device 32 in the first embodiment, it is sure that the features found in the environment image 16 . 17 around the previously defined characteristics 16 . 17 is.

In one embodiment, the correspondence problem is solved. One way to resolve the correspondence in pairs, lies in it, first the correspondence of the corner 17 dissolve. In the case where the corner is stored as a SIFT descriptor, correspondence can be resolved with known SIFT based methods. Also, the spatial relation of the corner 17 to the contour 16 be exploited to the correspondence of the found contour 16 with the stored contour 16 sure.

Through a changed pose of the side camera 21 the perspective of the contour also changes 16 and the corner 17 , So it comes to one perspective distortion in the captured environmental image. 5 shows exemplarily how an object 9 is transformed exemplarily from an original space A into a target space B by means of a transformation T.

A comparison device 34 compares for this purpose the actual position of the vehicle features found 16 . 17 with the desired positions of the stored vehicle characteristics 16 . 17 ,

On the basis of the comparison device 34 determined deviation of the actual position of the target position, finds a calibration device 33 a transformation function T whose parameters are the parameters of a perspective transformation. With the aid of the method of least squares, the parameters can be determined such that the error, ie the deviation of the actual position from the nominal position of the vehicle characteristics 16 . 17 , is minimized.

In further embodiments, further methods such as "Random Sample Consensus (RANSAC)" are possible for this purpose. Other methods from the field of image registration are of course also possible in further embodiments.

On the basis of the known vehicle geometry, the known intrinsic camera parameters and the found transformation function T, the calibration device can now use the extrinsic parameters of the side camera 21 determine and thus the extrinsic calibration of the side camera 21 make.

6 shows the on-board computer 30 that the storage device 31 , the identification device 32 as well as the calibration device 33 having. The on-board computer 30 is in this embodiment in the vehicle 10 built-in, enabling extrinsic calibration of all cameras 21 in the state and while driving the vehicle 10 ,

In a further embodiment, a plurality of vehicle features may be used to determine the extrinsic parameters 16 . 17 be provided with a different priority. These can be in an ordered list in the storage device 31 held together with their target positions.

First, the vehicle feature 16 . 17 with the highest priority for identification in the captured environment image of the vehicle 10 used. Should the identification of the actual position of the vehicle feature 16 . 17 fail, eg because the corresponding part of the vehicle 10 is heavily polluted, so may be another attempt with a feature 16 . 17 be started with a lower priority.

It is thus also possible that vehicle features 16 . 17 be grouped by specifying their priority. Thus, an attempt is made to determine the actual positions of a group of vehicle features with the same priority.

7 shows another vehicle 10 , which is equipped with a driver assistance system for autonomous Ausparken. The parking assistant used in this embodiment uses, among other things, environmental images taken by a video camera 21 are included in the vehicle 10 are installed.

In the in 7 scenario presented is the vehicle 10 backwards follow the arrow movement. For this it is necessary to provide virtually a three-dimensional representation of the environment.

Since a Ausparkassistent should behave differently in different situations, it is also necessary to be able to recognize the objects in the environment. For example, a lower speed is necessary when close to a pedestrian.

To objects in the environment of the vehicle 10 to recognize, that is their type, such as the parked car 40 , it is necessary to project the acquired environmental images onto a three-dimensional representation of the environment.

In order to ensure the correct projection it is further necessary that the extrinsic parameters of the camera 21 are known exactly. Otherwise it can lead to error detection and the Ausparkassistent does not behave as it is necessary in the situation.

In one embodiment of the invention when starting the vehicle 10 the extrinsic parameters of all cameras 21 of the vehicle 10 certainly. This ensures that the data on which the parking assistant supports his calculations have the necessary accuracy.

LIST OF REFERENCE NUMBERS

9

object

10

vehicle

11

Mirrors

12

wheel

13

vehicle center

15

vehicle environment

16

Vehicle limit, contour

17

Corner, kink

21

Image capture device, side camera, camera

30

board computer

31

memory device

32

identification device

33

calibration

34

comparator

40

Parking car

X, Y, Z

Base coordinate system

X ', Y', Z '

Camera coordinate system

A

source space

B

Finish area

T

transformation

Claims (17)

Device for the extrinsic calibration of an image capture device ( 21 ) of a vehicle ( 10 ), in particular a passenger car, comprising: - a memory device ( 31 ); At least one image capture device ( 21 ), in particular at least one video camera, which is designed to display at least one environmental image of the vehicle ( 10 ) capture; An identification device ( 32 ); A comparison device ( 34 ); and - a calibration device ( 33 ); characterized in that - in the memory device ( 31 ) at least one vehicle feature ( 16 . 17 ) and at least one desired position of the at least one vehicle feature ( 16 . 17 ) are stored; - the identification device ( 32 ) is adapted to an actual position of the at least one vehicle feature ( 16 . 17 ) in the at least one environmental image of the vehicle ( 10 ) to identify; - the comparator ( 34 ) is adapted to compare the actual position with the desired position; and - the calibration device ( 33 ) based on a deviation of the actual position from the desired position, the extrinsic parameters of the at least one image acquisition device ( 21 ) certainly. Apparatus according to claim 1, characterized in that the at least one vehicle feature ( 16 . 17 ) a geometric structure, in particular at least one contour, at least one edge, at least one surface or at least one corner of the vehicle ( 10 ) or a combination of these. Device according to one of the preceding claims, characterized in that the at least one vehicle feature ( 16 . 17 ) relates to a corner, wherein a position and an orientation of the corner in the memory device ( 31 ) are stored. Apparatus according to claim 3, characterized in that the at least one vehicle feature ( 16 . 17 ) represented by a matrix having a corner ( 17 In particular, the matrix describes: - a scaling invariant feature descriptor; or - represents a multiscale oriented patches descriptor. Device according to one of the preceding claims, characterized in that the identification device ( 32 ) is designed to: - a course of a contour ( 16 ) of the vehicle ( 10 ) in the at least one environmental image; Based on the contour of the contour ( 16 ) the actual position of a corner ( 17 ) than the at least one vehicle characteristic ( 16 . 17 ) to identify. Device according to one of the preceding claims, characterized in that the identification device ( 32 ) is adapted to the identification of the actual position of the at least one vehicle feature ( 16 . 17 ) based on a previously determined partial section of the environmental image. Device according to one of the preceding claims, characterized in that the at least one image capture device ( 21 ) has a fisheye optics and this has a viewing angle greater than 90 degrees, in particular greater 170 degrees. Device according to one of the preceding claims, characterized in that - in the memory device ( 31 ) a first vehicle feature ( 16 . 17 ) is stored with a high priority; In the storage device ( 31 ) a second vehicle feature ( 16 . 17 ) is stored with a low priority; - the identification device ( 32 ) is designed to: - determine whether the first vehicle feature ( 16 . 17 ) in the at least one environmental image of the vehicle ( 10 ) lies; In the case where the first vehicle feature ( 16 . 17 ) not in the at least one environmental image of the vehicle ( 10 ), determine whether the second vehicle feature ( 16 . 17 ) is located in the at least one environmental image; and - in the case where the second vehicle feature ( 16 . 17 ) in the at least one environmental image of the vehicle ( 10 ), the actual position of the second vehicle feature ( 16 . 17 ) in the at least one environmental image. Device according to one of the preceding claims, characterized in that the calibration device ( 33 ) is adapted to determine the extrinsic parameters using the least squares method. Vehicle ( 10 ), characterized in that - the vehicle ( 10 ) an apparatus for extrinsic calibration of the at least one image capture device ( 21 ) of the vehicle ( 10 ) according to any one of the preceding claims; The vehicle is the at least one image capture device ( 21 ) and the at least one image capture device ( 21 ) is oriented such that the at least one surrounding image surrounds the surroundings of the vehicle ( 15 ) and part of the vehicle ( 10 ) maps; Vehicle ( 10 ) according to claim 10, characterized in that - the vehicle ( 10 ) comprises an estimation device which, on the basis of the actual position of the at least one vehicle feature ( 16 . 17 ) the distance between the actual position of the at least one vehicle feature ( 16 . 17 ) and the highest point of a wheel rim of the vehicle ( 10 ) appreciates. Vehicle ( 10 ) according to one of claims 10 or 11, characterized in that - a driver assistance system using the at least one image capture device ( 21 ) and the extrinsic parameters generate control signals for an actuator, the actuator preferably comprising: - brake; - steering wheel vibration motor; - Signal light; or - Side mirror actuator. Method for extrinsic calibration of an image capture device ( 21 ) of a vehicle ( 10 ), in particular a passenger car, in particular by means of a device according to one of claims 1-9, comprising the steps: - detecting at least one environmental image of the vehicle ( 10 ) by at least one image capture device ( 21 ); Identifying an actual position of at least one vehicle characteristic ( 16 . 17 ) in the at least one environmental image; Comparing the actual position with one in the memory device ( 31 ) stored desired position of the at least one vehicle feature ( 16 . 17 ); Extrinsic calibration of the at least one image capture device ( 21 ) based on the determined deviation. The method of claim 13, wherein the identifying comprises restricting to a predetermined portion of the environmental image based on which the vehicle feature is identified. 16 . 17 ) is carried out. Method according to one of claims 13 or 14, wherein the method during a journey of the vehicle ( 10 ) is carried out. Use of a method according to any one of claims 13-15 in an all-round visibility system of a vehicle ( 10 ).  A computer-readable storage medium having executable instructions that cause a computer to implement a method according to any one of claims 13-15 when the instructions are executed.

Images