KR102477480B1 - Apparatus for calibrating for around view camera and method thereof - Google Patents

Abstract

An apparatus and method for calibrating a camera of an around-view monitor system are provided. An apparatus for calibrating a camera of an around-view monitor system according to an embodiment of the present invention includes: a corner point detection unit receiving N×N checkerboard images from an around-view camera and detecting corner points on the checkerboard; a calibration unit for calibrating using a random number of corner points in the checker board; an error calculation unit that calculates a total reprojection error according to a calibration result; and comparing the currently computed total reprojection error with the previously computed total reprojection error, and if the currently computed total reprojection error is greater than or equal to the previously computed total reprojection error, the current number of corner points is optimally calculated. and a determination unit for determining a calibration parameter based on a number.

Description

Apparatus for calibrating for around view camera and method thereof}

The present invention relates to an around-view monitor system, and more particularly, to a calibration apparatus and method for an around-view camera that automatically detects corner points and determines the optimum number of corner points for calibration.

Recently, the use of a front camera, a rear camera, and a side view camera has been spread in vehicles. These cameras are often used to assist drivers and can improve vehicle stability.

Although these cameras are known to provide views that cannot be provided by typical rear and side view mirrors, they also have blind spots. In addition to blind spots, these cameras can be dangerous in that the view provided by these cameras often gives the driver that the space between the vehicle and other objects is larger than it actually is.

Accordingly, the around-view camera is calibrated to accurately provide a positional relationship between the vehicle and other objects.

However, the calibration of conventional around-view cameras requires manual marking of corner points in a checkerboard pattern in order to perform calibration of all cameras. Although this may have high accuracy, the error tendency due to user intervention is very high, and the number of inputs increases exponentially according to the size of the checkerboard pattern, so it takes a lot of time. Therefore, the entire calibration process is complicated and takes a lot of time.

KR 2016-0123411 A

In order to solve the problems of the prior art as described above, an embodiment of the present invention automatically detects a pattern for calibration and uses an optimized and simpler pattern to simplify the process and reduce the required time. It is intended to provide a camera calibration device and method.

According to one aspect of the present invention for solving the above problems, a corner point detection unit receiving an N×N checkerboard image from an around-view camera and detecting a corner point on the checkerboard; a calibration unit for calibrating using an arbitrary number of corner points in the checkboard; an error calculation unit that calculates a total reprojection error according to a result of the calibration; and comparing a currently computed total reprojection error with a previously computed total reprojection error, if the currently computed total reprojection error is greater than or equal to the previously computed total reprojection error, of the current corner point. An apparatus for calibrating an around-view camera including a determining unit determining a calibration parameter by using the number as an optimal number is provided.

In one embodiment, the determination unit controls the calibration unit and the operation unit to re-perform the calibration and the reprojection error calculation while increasing the number of corner points, and the currently calculated total reprojection error is the previously calculated If the current number of corner points is greater than or equal to the number of corner points detected in the checkerboard, the number of corner points detected in the checkerboard may be determined as the optimum number.

In one embodiment, the calibration unit may project the corner points of the checkerboard into a real coordinate system and then project them into an image coordinate system.

In one embodiment, the error calculation unit may calculate a difference between a position where the corner point of the checkerboard is projected in a real coordinate system and a position where it is projected again in an image coordinate system.

In an embodiment, the corner point detection unit may include an extraction unit extracting the corner point from the input image; and whether all corner points have been extracted from the input image is input through the input unit, and if all corner points are not extracted, corner points that have not been extracted on the input image are marked according to the input of the input unit. It may include a marking unit to.

In one embodiment, the corner point detection unit receives whether or not the direction of the corner point is inappropriate for all the corner points through the input unit, and if the direction of the corner point is inappropriate, the corner point is detected according to the input of the input unit. A correction unit for rotating the point may be further included.

According to another aspect of the present invention, detecting a corner point in the input N × N checkerboard image; Calibrating using an arbitrary number of corner points on the checkboard; calculating a total reprojection error according to a result of the calibration; comparing a currently computed total reprojection error with a previously computed total reprojection error; and if the currently calculated total reprojection error is greater than or equal to the previously calculated total reprojection error, determining a calibration parameter using the current number of corner points as an optimal number. is provided.

In one embodiment, the method for calibrating the around-view camera further includes re-performing the calibrating step to the comparing step while increasing the number of corner points, wherein the determining of the parameters is based on the current calculation If the total reprojection error calculated is less than the total reprojection error calculated previously, the number of corner points detected in the checkerboard is calculated if the current number of corner points is greater than or equal to the number of corner points detected in the checkerboard. It can be determined by the above optimal number.

In one embodiment, in the calibrating step, the corner points of the checkerboard may be projected to a real coordinate system and then projected to an image coordinate system.

In one embodiment, the calculating step may calculate a difference between a position where the corner point of the checkerboard is projected in a real coordinate system and a position where it is projected again in an image coordinate system.

In one embodiment, the detecting may include extracting the corner point from the input image; receiving an input through an input unit whether or not all corner points have been extracted from the input image; and if all of the corner points are not extracted, marking the corner points that are not extracted on the input image according to the input of the input unit.

In an embodiment, the detecting may include receiving an input through the input unit whether or not directions of corner points are inappropriate with respect to all the corner points; and rotating the corner point according to an input of the input unit when the direction of the corner point is inappropriate.

An apparatus and method for calibrating an around-view camera according to an embodiment of the present invention is optimized for calibration and uses a simpler pattern, thereby simplifying a calibration procedure and reducing required time.

In addition, since the present invention can minimize the user's intervention by automatically extracting corner points from the checkerboard image, the time required for marking the corner points can be shortened.

In addition, the present invention can accurately extract all corner points even when the image quality is poor by checking the corner point extraction result and manually marking the missing portion.

1 is a plan view of a vehicle equipped with an around-view monitoring system;
2 is a photograph showing an example of an image input to an around-view camera during calibration;
3 is a block diagram showing a calibration device for an around-view camera according to an embodiment of the present invention;
4 is a flowchart illustrating a calibration method of an around-view camera according to an embodiment of the present invention;
5 is a diagram showing corner points detected in an image input from a front camera;
6 is a diagram showing corner points detected in an image input from a right camera;
7 is a diagram showing corner points detected in an image input from a rear camera;
8 is a diagram showing corner points detected in an image input from a left camera;
9 is a diagram showing the results of reprojection errors according to the number of corner points, and
10 is a flowchart illustrating a process of detecting corner points in FIG. 4 .

Hereinafter, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily carry out the present invention. This invention may be embodied in many different forms and is not limited to the embodiments set forth herein. In order to clearly describe the present invention in the drawings, parts irrelevant to the description are omitted, and the same reference numerals are assigned to the same or similar components throughout the specification.

Hereinafter, an apparatus for calibrating an around-view camera according to an embodiment of the present invention will be described in more detail with reference to the drawings. 1 is a plan view of a vehicle equipped with an around-view monitoring system, and FIG. 2 is a photograph showing an example of an image input to an around-view camera during calibration.

Referring to FIG. 1 , a vehicle 10 equipped with an around-view monitoring system may include an around-view camera including a plurality of cameras 12a to 12d having fisheye lenses.

At this time, the front camera 12a is provided in the front of the vehicle 10, such as the front bumper, the right camera 12b is provided in the right side of the vehicle 10, such as the right side mirror, and the rear camera 12c is It may be provided at the rear of the vehicle 10, such as a rear bumper, and the left camera 12d may be provided at the left side of the vehicle 10, such as a left side mirror.

Here, the around view camera 12 has been illustrated and described as having one for each direction, but is not limited thereto and may include a plurality of cameras. In particular, since the field of view of the right camera 12b and the left camera 12c is limited depending on the installation location, a plurality of cameras installed in a plurality of locations may be included.

Also, the around-view monitoring system may include a brightness control logic to match a bird's eye view. Also, the around view monitoring system may include a parking guide that changes the view based on the gear position of the vehicle 10 .

In this case, the around-view monitoring system may support the driver to visually check the position of the vehicle 10 with respect to the line around the parking space and adjacent objects.

This can assist the driver to more easily park into a parking space by better understanding the surroundings of the vehicle 10 through a virtual bird's eye view from above the vehicle 10 .

The calibration device of the around-view monitoring system may be performed through calibration logic that may include software, hardware, firmware, or a combination thereof.

This calibration device performs calibration of the camera while positioning a set of markers in a plane at least partially surrounding the vehicle 10 .

Here, the plane may be a field of view of a camera of an around-view monitoring system. That is, as shown in FIG. 1, the plane surrounding the vehicle 10 has a front area F corresponding to the field of view of the front camera 12a and a right area R corresponding to the field of view of the right camera 12b. , a rear region B corresponding to the field of view of the rear camera 12c, and a left region L corresponding to the field of view of the left camera 12d.

At this time, as shown in FIG. 2 , checkboards may appear on both sides of input images acquired using the around-view cameras 12a to 12d such as fisheye cameras for calibration. That is, one image may include two checkboards arranged on both sides.

Meanwhile, camera calibration may include internal camera calibration and external camera calibration.

Here, the internal camera calibration is used to calculate the five internal parameters of the pinhole camera, and can obtain the center coordinates of the camera image, the focal length in the x and y directions, and the distortion parameter (the angle between the camera axes).

The extrinsic camera calibration is used for point conversion from a 3D-camera coordinate system to a 2D-pixel coordinate system, and an extrinsic calibration matrix including a rotation matrix and a translation vector can be obtained. By means of this extrinsic calibration matrix, an orientation between the camera and the object may be obtained, and an image may be obtained from the orientation.

At this time, since the internal camera calibration is determined in advance, the calibration apparatus of the present invention relates to external camera calibration.

The external camera calibration determines the optimal number of corner points in the pattern by considering:

1) The planar homography matrix H has 9 entities, but is defined to the degree of scale.

2) The number of degrees of freedom for 2D projective transformation is 8.

3) Each corresponding 2D point or line generates two constraints on the planar homography matrix H.

4) The correspondence of 4 points or 4 lines is sufficient to compute the planar homography matrix H.

5) For a planar affine transformation with 6 degrees of freedom, only 3 correspondence points or lines are required.

6) Set the limit of the minimum number of checkerboard corner points required to calculate the planar homography matrix H to 5.

However, since the number of corner points of the checkerboard is fixed, it may not be suitable depending on the around view camera 12 and the used checkerboard pattern.

Accordingly, the present invention can perform calibration with a simpler pattern by automatically determining the optimal number of corner points during calibration.

3 is a block diagram illustrating an apparatus for calibrating an around-view camera according to an embodiment of the present invention.

The around-view camera calibration apparatus 100 includes a corner point detection unit 110 and a parameter determination unit 120 .

Here, the around-view camera calibration apparatus 100 may determine a plurality of external parameters of the around-view camera 12 with respect to a set of markers such as a checkerboard. At this time, the around-view camera calibration apparatus 100 may repeat locating a set of markers and determining extrinsic parameters for all of the around-view cameras 12 .

The corner point detector 110 may receive an N×N checkerboard image from the around view camera 12 and detect corner points on the checkerboard. The corner point detection unit 110 may include an extraction unit 112 , a marking unit 114 , and a correction unit 116 .

The extraction unit 112 may automatically extract a corner point from an image input from the around view camera 12 . Here, a method of extracting a corner point from an image is not particularly limited, and since various image processing is possible, a detailed description thereof will be omitted.

At this time, an image obtained using the around-view camera 12 such as a fisheye or omnidirectional camera may have poor quality, blur, or severe distortion.

To compensate for this, the present invention can visually inspect the automatically detected corner points and manually mark the missing corner points.

More specifically, the marking unit 114 may receive input through an input unit (not shown) whether all corner points have been extracted from the input image. In this case, corner points that are omitted in the process of automatically extracting corner points by the extraction unit 112 may be visually confirmed by the user.

Also, the marking unit 114 may manually mark corner points that are not extracted on the input image according to the input of the input unit. That is, the marking unit 114 may manually mark corner points that are omitted in the process of automatically extracting corner points by a user through the input unit.

The correction unit 116 may receive input through the input unit whether or not the direction of the corner points is suitable for all the corner points extracted or marked by the extraction unit 112 and the marking unit 114 . At this time, the correction unit 116 may receive a result of determining whether the directions of all the corner points are suitable with the naked eye of the user through the input unit.

In addition, the correction unit 116 may rotate the corner point according to the input of the input unit when the direction of the corner point is not suitable. At this time, the correction unit 116 may receive rotation of the corner point from the user through the input unit.

This allows manually correcting the orientation of corner points to establish correspondence between cameras.

The parameter determiner 120 may determine the optimal number of corner points for calibration based on the corner points of the checkerboard detected by the corner point detector 110 . The parameter determination unit 120 may include a calibration unit 122 , an error operation unit 124 , and a determination unit 126 .

The calibration unit 122 may perform calibration using an arbitrary number of corner points on the checkerboard. Here, an arbitrary number M of corner points for calibration may be set in advance. Preferably, the arbitrary number (M) may be smaller than the total number (N) of corner points included in the checkerboard.

In this case, the calibration unit 122 may construct a bird's eye view from an image input from the around view camera 12 . For example, the calibration unit 122 may project M×M corner points into a real coordinate system and then project them into an image coordinate system using a homography matrix.

The error calculation unit 124 may calculate a total re-projection error (TRPE) according to a result of calibration by the calibration unit 122 . Here, the total reprojection error (TRPE) may be an absolute difference before and after calibration.

For example, the error calculation unit 124 may calculate a difference between a position where M×M corner points of the checkerboard are projected in real coordinates and a position projected back in image coordinates as a total reprojection error (TRPE). In this case, the error calculator 124 may calculate a total reprojection error (TRPE) for each camera of the around view camera 12 .

The determination unit 126 controls the calibration unit 122 and the error operation unit 124 to re-perform calibration and reprojection error calculations while increasing the number of corner points (M) in order to determine the optimal number of corner points during calibration. can do. That is, the determination unit 126 may repeatedly perform calibration until the total reprojection error (TRPE) no longer decreases or the number of corner points (M) no longer remains.

More specifically, the determiner 126 may compare a currently calculated total reprojection error (TRPE _M ) with a previously calculated total reprojection error (TRPE _{M - 1} ). Here, when the number M of corner points is initially set, this process may be omitted.

At this time, the determiner 126 determines the current number of corner points (M) as an optimum when the currently calculated total reprojection error (TRPE _M ) is greater than or equal to the previously calculated total reprojection error (TRPE _{M − 1} ). The number can be used to determine the calibration parameters.

Accordingly, the calibration procedure can be simplified with a simpler pattern by optimizing the number of corner points of the checkerboard.

In addition, the determination unit 126 determines that the current number of corner points (M) is determined from the checkerboard when the currently calculated total reprojection error (TRPE _M ) is smaller than the previously calculated total reprojection error (TRPE _{M - 1} ). If it is greater than or equal to the total number of detected corner points (N), the external camera calibration parameter may be determined by determining the total number of detected corner points (N) as the optimum number.

With this configuration, the calibration apparatus 100 for an around-view camera according to an embodiment of the present invention can shorten the time required by simplifying the calibration procedure, and can shorten the time required for corner point marking, All corner points can be accurately extracted even when the image quality is poor.

Hereinafter, a method for calibrating an around-view camera according to the present invention will be described with reference to FIGS. 4 to 10 . 4 is a flowchart illustrating a calibration method of an around-view camera according to an embodiment of the present invention.

The calibration method 200 of the around-view camera includes automatically detecting corner points (S201 and S202), calibrating using an arbitrary number of corner points (S203 and S204), and calculating total reprojection errors ( S205), re-performing calibration while increasing the number of corner points (S206 to S208), and determining parameters (S209).

More specifically, as shown in FIG. 4, first, the calibration apparatus 100 of the around-view camera automatically detects corner points in the N×N checkerboard image input from the around-view camera 12 (step S201 ).

At this time, corner points may be detected for each checker board included in images input from the front camera 12a, the right camera 12b, the rear camera 12c, and the left camera 12d. This will be described in more detail with reference to FIG. 10 .

5 is a diagram showing corner points detected in an image input from a front camera, FIG. 6 is a diagram showing corner points detected in an image input from a right camera, and FIG. 7 is a diagram showing corner points detected in an image input from a rear camera. It is a diagram showing corner points, and FIG. 8 is a diagram showing corner points detected in an image input from a left camera.

5 to 8, the checkerboard includes a total of 5×5 corner points. Here, the number for the corner point is displayed in the direction of the arrow indicated on one side of the checkerboard. The numbering order is from the reference corner point (1) in the direction of the arrow. Also, the numbering order is in the vertical direction of the arrows.

Next, the around-view camera calibration apparatus 100 stores the detected corner points in a storage unit (not shown) (step S202). Here, the detected corner point may be marked on the checkerboard image.

Next, the around-view camera calibration apparatus 100 sets an arbitrary number M of corner points (step S203). At this time, the arbitrary number (M) may be appropriately set according to the total number (N) of corner points included in the checkerboard included in the image input from the around view camera 12 . Here, the arbitrary number (M) may be the number of corner points in one direction on the checkerboard. For example, the checkerboard includes 5×5 corner points, and M can be set to 3.

Next, the around-view camera calibration apparatus 100 performs calibration using M×M corner points on the checkerboard (step S204). That is, in order to derive the optimal number of calibrations, calibration may be performed targeting an arbitrary number (M) of corner points smaller than the total number (N) of corner points included in the checkerboard.

In this case, the calibration may configure a bird's eye view from images input from each of the around view cameras 12 . For example, 3×3 corner points may be projected into a real coordinate system and then projected into an image coordinate system by a homography matrix.

Next, the calibration apparatus 100 of the around-view camera calculates a total reprojection error (TRPE) according to the calibration result (step S205). Here, the total reprojection error (TRPE) may be an absolute difference before and after calibration.

For example, the difference between the projected position of the 3x3 corner point of the checkerboard in the real coordinate system and the projected position in the image coordinate system can be calculated as the total reprojection error (TRPE).

As such, error factors affecting overall quality by calibration may include lens distortion, yaw error, sensor tilt, pitch error, baseline error, and focal length error.

At this time, a total reprojection error (TRPE) for each camera of the around-view camera 12 may be calculated.

Next, the apparatus 100 for calibrating the around-view camera compares the currently calculated total reprojection error (TRPE _M ) with the previously calculated total reprojection error (TRPE _{M - 1} ) (step S206 ). Here, in the case of the first comparison, that is, in the case where the previous total reprojection error (TRPE _{M - 1} ) has not been previously calculated, this step can be omitted.

As a result of comparison in step S206, if the currently calculated total reprojection error (TRPE _M ) is smaller than the previously calculated total reprojection error (TRPE _{M - 1} ), the calibration apparatus 100 of the around-view camera calculates the number of corner points (M) is increased (step S207). That is, M may be increased by 1 to repeat the calibration to determine the optimum number of corner points.

Next, the calibration device 100 of the around-view camera compares the current number of corner points (M) with the total number (N) of all corner points included in the checkerboard (step S208), and calculates the number of corner points (M). If is less than the total number (N) of all corner points included in the checkerboard, the process returns to step S204 and steps S204 to S207 are repeatedly performed.

That is, if the currently computed total reprojection error (TRPE _M ) is less than the previously computed total reprojection error (TRPE _M-1 ), and there are still more corner points remaining on the checkerboard, until no corner points remain. Alternatively, calibration and calculation of the total reprojection error TRPE may be repeatedly performed until the total reprojection error TRPE _M does not decrease any more.

As an example, when M = 3, the calculation of calibration and total reprojection error (TRPE) is first performed using 3 × 3 corner points, and then using 4 × 4 corner points, and All corner points, up to N×N corner points, can be performed.

As a result of comparison in step S208, if the current number of corner points (M) is greater than or equal to the total number of corner points (N) detected on the checkerboard, the calibration device 100 of the around-view camera calculates the number of corner points detected on the checkerboard. An external camera calibration parameter is determined by taking the total number N as the optimum number (step S209).

Again, as a result of the comparison in step S206, if the currently calculated total reprojection error (TRPE _M ) is greater than or equal to the previously calculated total reprojection error (TRPE _{M - 1} ), the calibration apparatus 100 of the around-view camera steps Returning to S209, external camera calibration parameters are determined by setting the current number of corner points (M) as the optimal number.

Accordingly, the calibration procedure can be simplified with a simpler pattern by optimizing the number of corner points of the checkerboard.

9 is a diagram showing results of reprojection errors according to the number of corner points.

9 shows the distribution of reprojection errors according to the number of corner points used for calibration. Here, "○" represents the position of the corner point detected from the image, and "*" represents the position of the corner point after re-projection.

In the case of using 3×3 corner points, it was found that the location of the detected corner point and the location of the corner point after re-projection had a significant error.

In the case of using the 4×4 corner point and the case of using the 5×5 corner point, there was almost no error between the location of the detected corner point and the location of the corner point after reprojection. At this time, it may be desirable to set the number of corner points to 4, which is the minimum number.

10 is a flowchart illustrating a process of detecting corner points in FIG. 4 .

The corner point detection method 300 includes automatically extracting corner points (steps S301 to S303), manually marking (steps S304 to S306), and correcting corner points (steps S307 to 309).

More specifically, as shown in FIG. 10 , first, the around-view camera calibration apparatus 100 receives the checkboard acquired by the around-view camera 12 (step S301). At this time, images input from each of the around view cameras 12 may be sequentially input.

Next, the calibration apparatus 100 of the around-view camera determines whether there is a mark for the corner point in the input image (step S302), and returns to step S301 to receive the next image if there is a mark (step S302). In this case, whether or not there is a mark may be determined whether detection or marking of a corner point in the image is completed.

As a result of the determination in step S302, when it is determined that there is no mark for the corner point in the input image, the calibration apparatus 100 of the around-view camera automatically extracts the corner point from the input image (step S303). A method of extracting a corner point from an image is not particularly limited, and since various image processing is possible, a detailed description thereof will be omitted.

Here, as the size of the checker board increases, the number of corner points to be detected or marked increases exponentially. Accordingly, the present invention can automatically detect corner points from the checkerboard image in order to exclude or minimize human intervention by manual work.

Next, the around-view camera calibration apparatus 100 receives an input through an input unit whether or not all corner points have been extracted from the input image (step S304). That is, the user may visually determine whether or not all corner points have been extracted, and input the determination result to the calibration apparatus 100 of the around-view camera.

Here, an image obtained using a fisheye or omnidirectional camera may often be of poor quality. For example, the image quality may be poor, a blur may exist in the image, or the image may have severe distortion.

To compensate for this, the present invention can visually inspect the automatically detected corner points and manually mark the missing corner points.

As a result of the determination in step S304, when it is determined that all corner points have not been extracted from the input image, the calibration apparatus 100 of the around-view camera marks the corner points that have not been extracted on the input image according to the input of the input unit ( Step S305). That is, the user can manually mark corner points that are omitted in the process of automatically extracting corner points through the input unit.

Next, it is determined whether all corner points are marked (step S306), and if all corner points are not marked in the input image, the process proceeds to step S305 to continuously perform manual marking. At this time, the calibration apparatus 100 of the around-view camera may receive manual marking from the user.

As a result of the determination in step S306, when it is determined that all corner points are marked in the input image, or as a result of determination in step S304, it is determined that all corner points are extracted from the input image, the calibration apparatus 100 of the around-view camera inputs Whether or not the direction of the corner points is suitable for all corner points of the image is received through the input unit (step S307). That is, the user may visually determine whether the directions of all the corner points are appropriate, and input the determination result to the calibration device 100 of the around-view camera.

As a result of the determination in step S307, when it is determined that the direction of the corner point is not suitable, the calibration apparatus 100 of the around-view camera rotates the detected corner point according to the input of the input unit (step S308). At this time, the calibration apparatus 100 of the around-view camera may receive rotation of the corner point from the user.

As such, the present invention can manually correct the orientation of corner points to establish correspondence between cameras.

As a result of the determination in step S307, when it is determined that the direction of the corner point is appropriate, the calibration apparatus 100 of the around-view camera determines whether processing has been completed for all input images (step S309), and processing of all images is performed. When finished, end the automatic detection of corner points.

As a result of the determination in step S309, when it is determined that processing of all input images has not been completed, the calibration apparatus 100 of the around-view camera returns to step S301 to receive input of the next image. In this case, the around-view camera calibration apparatus 100 may repeatedly perform steps S303 to S308 for all input images.

With this method, the present invention can simplify the calibration procedure, shorten the time required, shorten the time required for corner point marking, and accurately extract all corner points even when the image quality is poor. .

The methods described above may be implemented by the calibration apparatus 100 of an around-view camera as shown in FIG. 3, and in particular, may be implemented by a software program that performs these steps, in which case, these programs It may be stored in a readable recording medium or transmitted by a computer data signal combined with a carrier wave in a transmission medium or communication network.

At this time, the computer-readable recording medium includes all types of recording devices in which data readable by a computer system is stored, and includes, for example, ROM, RAM, CD-ROM, DVD-ROM, DVD-RAM, magnetic tape, It may be a floppy disk, a hard disk, an optical data storage device, and the like.

Although one embodiment of the present invention has been described above, the spirit of the present invention is not limited to the embodiments presented herein, and those skilled in the art who understand the spirit of the present invention may add elements within the scope of the same spirit. However, it will be possible to easily suggest other embodiments by means of changes, deletions, additions, etc., but this will also be said to fall within the scope of the present invention.

100: Calibration device for around view camera
110: corner point detection unit 112: extraction unit
114: marking unit 116: correction unit
120: parameter determination unit 122: calibration unit
124; Error calculation unit 126: decision unit

Claims (12)

a corner point detection unit receiving N×N checkerboard images from the around-view camera and detecting corner points on the checkboard;
a calibration unit for calibrating using an arbitrary number of corner points in the checkboard;
an error calculation unit that calculates a total reprojection error according to a result of the calibration; and
Comparing the currently computed total reprojection error with the previously computed total reprojection error, if the currently computed total reprojection error is greater than or equal to the previously computed total reprojection error, the number of the current corner points And a determination unit for determining a calibration parameter by using the optimal number of
The decision section
Controlling the calibration unit and the operation unit to re-perform the calibration and the reprojection error operation while increasing the number of corner points.
A calibration device for around-view cameras. According to claim 1,
The decision section
If the currently calculated total reprojection error is less than the previously calculated total reprojection error, if the current number of corner points is greater than or equal to the number of corner points detected in the checkerboard, the corner points detected in the checkerboard An apparatus for calibrating an around-view camera that determines the number of as the optimal number. According to claim 1,
The calibration unit of the around-view camera calibration device for projecting the corner points of the checkerboard into real coordinates and then projecting them back into image coordinates. According to claim 1,
The error calculation unit calculates a difference between a position where the corner point of the checkerboard is projected in a real coordinate system and a position where it is projected back in an image coordinate system. According to claim 1,
The corner point detection unit,
an extraction unit extracting the corner points from the input image; and
Receiving an input through an input unit whether all corner points have been extracted from the input image, and if all corner points are not extracted, marking corner points that are not extracted on the input image according to the input of the input unit A calibration device for an around-view camera including a marking unit. According to claim 5,
The corner point detecting unit receives an input as to whether the direction of the corner point is inappropriate for all the corner points through the input unit, and if the direction of the corner point is inappropriate, corrects rotation of the corner point according to the input of the input unit. An apparatus for calibrating an around-view camera further comprising a unit. Detecting corner points in the input N×N checkerboard image;
Calibrating using an arbitrary number of corner points on the checkboard;
calculating a total reprojection error according to a result of the calibration;
comparing a currently computed total reprojection error with a previously computed total reprojection error;
re-performing the calibrating step to the comparing step while increasing the number of corner points; and
determining a calibration parameter by taking the current number of corner points as an optimum number when the currently calculated total reprojection error is greater than or equal to the previously calculated total reprojection error;
Calibration method of an around-view camera comprising a. According to claim 7,
The step of determining the parameter may include: if the currently calculated total reprojection error is less than the previously calculated total reprojection error, and if the number of the current corner points is greater than or equal to the number of corner points detected in the checkerboard, A method for calibrating an around-view camera for determining the number of corner points detected in a checkerboard as the optimal number. According to claim 7,
The calibrating step is a method for calibrating an around view camera in which the corner points of the checkerboard are projected in real coordinates and then projected again in image coordinates. According to claim 7,
The calculating step calculates a difference between a projected position of the corner point of the checkerboard in real coordinates and a projected position in image coordinates again. According to claim 7,
The step of detecting
extracting the corner points from the input image;
receiving an input through an input unit whether or not all corner points have been extracted from the input image; and
and if all of the corner points are not extracted, marking the corner points that are not extracted on the input image according to the input of the input unit. According to claim 11,
In the detection step,
receiving, through the input unit, whether or not directions of corner points are inappropriate for all of the corner points; and
The method of calibrating an around-view camera, further comprising rotating the corner point according to an input of the input unit when the direction of the corner point is inappropriate.

Images