KR101697229B1 - Automatic calibration apparatus based on lane information for the vehicle image registration and the method thereof - Google Patents

Abstract

The present invention relates to an automatic calibration device based on lane information for vehicle image registration for multi-channel synthesis around a vehicle. The automatic calibration device of the present invention includes: a photographing unit which receives and outputs photographed images of a vehicle surrounding area including lanes or guard rails on the road from a plurality of cameras installed on the vehicle; and an image processing unit which registers the images by extracting lane information from the photographed images, correcting image distortion, and estimating installation location and angles of the cameras using the lane information and then outputs the registered image.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic correction apparatus based on lane information for vehicle image registration,

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic correction device based on lane information for vehicle image matching and, more particularly, to a method and apparatus for correcting a mounting error of a camera using an image of a lane on a road or a guard rail, The present invention relates to a lane-information-based automatic correction device for vehicle image matching capable of matching a multi-channel image of a vehicle, and a method thereof.

Recently, various kinds of cameras such as a rear camera for parking support have been used in vehicles. Among them, a top view system has been recently developed which can completely remove the front, rear, left, and right rectangles of a vehicle. This system has been developed using a wide angle camera mounted at front, rear, left, And then converts the normal view into a bird view. This top-view system allows the driver to see exactly the distance and situation between the vehicle and surrounding objects, as if the birds were looking down from the sky.

However, in order to implement such a top view system, there is a problem that image parallax caused by an input sequence in an image acquired from a camera at each position, image distortion caused by a lens used in a wide angle camera, and perspective projection distortion must be corrected.

In particular, when synthesizing images acquired from a plurality of cameras mounted at various positions, each camera has a mounting error, and since each image is obtained in a different coordinate system, each camera in the final synthesized image The images of the images are not coincident with each other.

Therefore, it is necessary to correct the mounting error of the camera due to the prerequisite for the image registration. For this, information about the installation height and the installation angle for each installed camera is required.

A common method for obtaining such information is to use a pattern image captured after a predetermined reference pattern such as a cherkerboard is installed on the ground. This method has an advantage that precise results can be obtained because the relative positions of specific markers on the pattern can be precisely known in advance. However, in order to install such a pattern around the vehicle, it is necessary to have a specific space having a predetermined width or more. Due to the difficulty in installing the pattern due to the area and weight of the pattern, it is troublesome .

A related prior art is Korean Patent Laid-Open Publication No. 2015-0028532 (published on May 3, 2015).

In order to visualize the surroundings of a vehicle, a plurality of cameras are mounted around the vehicle, and a mounting error of a camera mounted on the vehicle is corrected through an image acquired from the camera, and a plurality of camera images are registered And an automatic correction device based on lane information for vehicle image registration for vehicle image registration and a method therefor.

In addition, when correcting an image, it is possible to perform mounting error correction and image correction of a camera mounted on a vehicle using information on lanes or guard rails on an actual road without a specific reference pattern or structure such as a checkerboard A lane-based automatic correction device for image matching, and a method therefor.

The technical objects to be achieved by the present invention are not limited to the above-mentioned technical problems.

According to an aspect of the present invention, there is provided an apparatus for automatically correcting a lane information based on a lane information of a vehicle, comprising: a plurality of cameras mounted on a vehicle, An image processing unit for extracting the line information from the photographed image to output and correcting the image distortion, estimating the installation position and the installation angle of the camera using the line information, and outputting the matched image by matching the images .

Specifically, the image processor may use at least one of a source image, a perspective image, a topview image, and a cylinder view image in extracting line information from the photographed image . ≪ / RTI >

Specifically, the image processing unit performs image distortion correction of the photographed image using a known camera internal parameter, wherein the internal parameter includes at least one of an optical center, an aspect ratio, a projection type, and a focal length can do.

Specifically, the image processing unit rotates the image so that the lane is parallel to the driving direction using the lane information, moves the image so that the width of the lane in the image coincides with the image, And the image is moved so as to match the position of the camera and the installation angle of the camera.

Specifically, the image processing unit uses an optimization algorithm for rotating and moving the image, and the optimization algorithm is one of PSO (Particle Swarm Optimization), GD (Gradient Descent), and LMSE (Least Mean Square Error) . ≪ / RTI >

Specifically, the image processing unit may further include a storage unit for storing results after moving and rotating the image.

Specifically, the image processor may perform image matching using the movement and rotation information of the camera stored in the storage unit when the image matching is continuously performed in a subsequent process.

More specifically, a lane-based information-based automatic correction method for vehicle image matching in a multi-channel image synthesis around a vehicle includes steps of: acquiring a surrounding image while the vehicle is traveling in parallel with a lane; A step of correcting the image distortion of the peripheral image of the vehicle, a step of calculating and storing the moving and rotating information of the camera, a step of matching and storing the moving image of the camera, And performing the steps of:

Specifically, the step of acquiring the surrounding image while the vehicle travels in parallel with the lane includes the step of extracting the line information by including an image such as a lane or a guardrail on the road in the acquired peripheral image .

Specifically, the step of extracting line information such as a lane or a guardrail from the vehicle surroundings image may include extracting a line image such as a source image, a perspective image, a topview image, a cylinder view image, Screen mode may be used.

Specifically, the step of correcting the distortion of the peripheral image of the vehicle performs image distortion correction of the photographed image using a known camera internal parameter, wherein the internal parameter includes at least one of an optical center, an aspect ratio, a projection type, And at least one of them may be included.

Specifically, the moving and rotating information calculation, storage, and image matching steps of the camera may include rotating the image to align the lane with the driving direction, moving the image to match the width of the lane, A step of moving an image, and a step of storing movement and rotation information of a camera image and outputting a matching image.

Specifically, when the image is moved and rotated, the image of the camera is mapped to a virtual three-dimensional sphere, and the angle and distance are changed until the shape of the desired lane is repeated, .

 Specifically, an optimization algorithm is used to move and rotate the image, wherein the optimization algorithm is any one of PSO (Particle Swarm Optimization), GD (Gradient Descent), and LMSE (Least Mean Square Error) have.

More specifically, the step of performing image matching using the stored movement and rotation information of the camera may include using a change parameter corresponding to the stored movement and rotation information of the camera, So that the image matching is performed immediately.

As described above, according to the present invention, it is possible to perform a mounting error correction and an image correction operation of a camera mounted on a vehicle using information about a lane or a guardrail on an actual road without using a specific reference pattern or structure such as a checkerboard Therefore, it is possible to easily and easily obtain the camera external parameters without using a specific pattern tool such as a heavy checker board, and to install the pattern in a wide area, thereby facilitating the matching of images obtained from a plurality of cameras There is an effect that can be made.

1 is a block diagram illustrating an automatic lane information-based automatic correction apparatus for vehicle image matching according to an embodiment of the present invention.
2 is a view illustrating an image input through a plurality of cameras shown in FIG.
3 to 6 are diagrams illustrating a process of extracting line information from the photographed image.
FIG. 7 is a photograph showing various screen modes applied for extracting line information from a photographed image.
8 is a flowchart illustrating an automatic correction method based on lane information for vehicle image matching according to an exemplary embodiment of the present invention.
FIG. 9 is a diagram illustrating the image matching process shown in FIG.
FIG. 10 shows the camera movement and rotation information calculation, storage, and image registration steps shown in FIG. 8 in more detail.
11 is a photograph showing before and after automatic correction for vehicle image matching according to the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same components are denoted by the same reference symbols whenever possible. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail.

FIG. 1 is a block diagram illustrating an automatic lane information-based automatic correction apparatus for vehicle image matching according to an embodiment of the present invention. The lane information-based automatic correction apparatus 100 for vehicle image matching includes a photographing unit 110, And an image processing unit 120.

First, the photographed image should include an image such as a lane or guardrail on an actual road so that the line information can be extracted from the image. For this purpose, the vehicle obtains the image while traveling in parallel with the lane or the guard rail. Preferably, the vehicle speed is 30 km / h or less, and at the same time, the height of the camera and the change To be minimized.

The photographing unit 110 may include at least one camera for outputting a photographed image of a surrounding area of the vehicle. Preferably, the camera may include first to fourth cameras 111, 113, 115, and 117 for visualizing the surroundings of the vehicle mounted on the front, rear, left and right sides, You may.

The camera used in the photographing unit 110 may be a wide-angle lens having a larger viewing angle than that of a normal lens, or a fisheye lens having an extra-wide angle of 180 degrees or more.

Referring to FIG. 2, the image of the surroundings of the vehicle taken by the photographing unit 110 is shown in one image. The images of the front and rear portions and the images of the left and right side portions are synthesized, It can be seen that the joints of the images are shifted.

The image processing unit 120 has a function of extracting line information such as a lane or a guardrail appearing in a photographed image, a function of correcting distortion of a camera image using a camera internal parameter such as a light center, And performs a function of matching a multi-channel image by performing a function of estimating an installation position and an installation angle of the camera, matching and outputting images, and storing information on the images.

The image processing unit 120 uses various lane detection algorithms to extract line information such as a lane or a guard rail displayed on the photographed image.

3 to 6, FIG. 3 illustrates a B-spline matching based high-speed lane detection method, FIG. 4 illustrates a curve template matching based lane and curvature detection algorithm using a top view image FIG. 5 shows a low-level image processing method for lane detection and tracking, and FIG. 6 shows a real-time lane mark detection method in a city center. The image processing section 120 uses a lane- One or more of the following lane detection algorithms may be used.

Table 1 below shows an example of an algorithm for detecting patterns such as the above-described lane or guard rail.

// DETECTION (PATTERN EXTRACTION)
// 1. CAPTURE IMAGES
1. camera installation
2. for i ← 1 to n do // i: camera index, n: the number of cameras
3. C _i ← CaptureCamera (i) // C _i : captured image of i-th camera
4. end for
// 2. LANE DETECTION
5. for i ← 1 to n do
6. L _i ← DetectLane (C _i ) // detecting lane marking or guardrail information
7. if L _i is invalid do
8. go to 2
9. end if
10. <c _i ^p , c _i ^q > ← DetectCorrespondPoint (C _i ) // <c _i ^p , c _i ^q >
11. end for

7, there are shown various screen modes used when the image processing unit 120 detects line information in an image including a lane or a guardrail. In this case, the original image, Perspective image, Topview image, and Cylindrical image.

Specifically, the original image is not suitable for detecting a lane because the shape of the radial distortion remains intact and the image distributed in the area has a large deviation according to the area, . In addition, the depth information of the image is strongly affected by the lens distortion, and it is very difficult to interpret the actual perspective information. That is, when the original image is used in lane detection, the actual lane detection becomes difficult.

Next, the perspective image is an image that cancels the radial distortion of the fisheye lens, which is advantageous in that the shape of the lane in the image is linearly preserved. However, the distortion of the depth information of the image is significant, It is not effective for lane detection. Further, there arises a problem that the edge portion is elongated in a state where the perspective distortion remains.

Nevertheless, the perspective image is an image generally used when detecting a lane, and the range of expression of the lane is limited by the calibration information, which is the most ideal form for lane detection.

Next, the topview image is a homography-converted image from the perspective image. Since the lecture is expressed in the Euclidean space, the image can be intuitively recognized and the search area can be reduced. However, There is a possibility that the detection of the lane may become somewhat difficult due to the step artifact in the conversion.

That is, the top view image is a realistic screen mode most likely to be implemented in lane detection, and the lane display range is largely limited, so that it is easy to detect lanes but the lane is corrected.

Cylindrical image is an image that aligns the eye height of the image with the horizon. It has an advantage that there is no additional problem such that the edge part of the image in the image does not increase. In addition, However, since the detected lane includes the perspective distortion, it is necessary to switch to the top view image and re-interpret it.

Therefore, using only the cylinder image is more efficient for vehicle detection than lane detection.

An evaluation table is prepared for various screen modes used in detecting the line information in the above-described image, and is shown in Table 2 below.

Plane Image distortion Lane deviation Distortion Distortion Pixel Artifact difficulty Synthesis Original High (1) High (1) High (1) Low (3) High (1) 7 Perspective Middle (2) High (1) Highest (0) Middle (2) Middle (2) 7 Cylindrical Middle (2) Middle (2) Middle (2) Low (3) Middle (2) 11 Topview Low (3) Low (3) Low (3) High (1) Low (3) 13

As a result, the most suitable screen mode (Plane) for detecting the line information in the image including the lane or the guard rail becomes the top view mode. This is because the top view image has the lowest evaluation value for image distortion, lane deviation, and distance distortion, and the difficulty in actual implementation is also the most satisfactory. However, the pixel artifact is so severe that it is necessary to devise a filter for stable line detection.

Top View Mode The next best image is the cylinder mode, which is a typical screen mode used in the LDW algorithm. If the top view mode filter is not stable, it can be considered as a next best solution Do.

Next, the image processing unit 120 performs image distortion correction using a known camera internal parameter in an image photographed by a camera of the photographing unit 110, wherein the internal parameters include an optical center, an aspect ratio, a projection type, Or more.

Next, the image processing unit 120 estimates the installation position and the installation angle of the camera using the line information extracted from the image taken by the camera, registers the images, outputs the matched images, and stores information about the matched images .

That is, as described above, when the line information such as the lane of the road or the guard rail displayed on the photographed image photographed by the photographing unit 110 is extracted, the image processing unit 120 calculates the movement and rotation information of the camera, And the images are registered.

Specifically, first, the image is rotated (Orientation Estimation) so that the lane is parallel to the driving direction by using the direction information of the lane of the image, and then the image is rotated so that the width of the lane defined in the pre- (Alignment Estimation). Then, the image is moved so that the images are aligned using the region common to the adjacent cameras (Correspondence Estimation).

In this way, an optimization algorithm for estimating change parameters when moving and rotating the camera image can be used. Any of optimization algorithms such as PSO (Particle Swarm Optimization), GD (Gradient Descent), and LMSE It can be one. An example of such an algorithm is shown in Table 3 below.

// e_i: correspondence error
21. end for
// 3. SELECT OPTIMAL PARAMETERS
22.

← merge error calculation
23. if e > T do // T: threshold for correspondence error
24. check camera location and re-install
25. go to 2
26. end if
27. return (r_i ^x,r_i ^y,r_i ^z), (t_i ^x,t_i ^y,t_i ^z), i={1,…,n} // return and update rotation and translation parameters for each camera // ESTIMATION (OPTIMIZE GEOMETRY TRANSFORM)
// 1. ESTIMATE ORIENTATION AND ALIGNMENT
12. for i ← 1 to n do
13. (r _i ^x, _i ^y r, ^z r _i), ^(x _i t, _i t ^y, t ^z _i) ← 0 // initializing rotation parameters (r _i ^x, _i ^y r, ^z r _i) and translation parameters (t _i ^x , t _i ^y , t _i ^z ) of i-th camera
14. (r _i ^x, _i ^y r, ^z r _i) ← EstimateOrientation (C _i, L _i, (r _i ^x, _i ^y r, ^z r _{i), (i} t ^x, t ^y _i, t _i ^z )) // Optimization for Orientation Estimation to get rotation parameters which fit to lane L _i orientation
15. _(i t ^x, t ^y _i, t _i ^z) EstimateAlignment ← (C _i, L _i, (r _i ^x, _i ^y r, ^z r _{i), (i} t ^x, t ^y _i, t _i ^z )) // Optimization for Alignment Estimation to get translation parameters which fit to lane L _i width
16. end for
// 2. ESTIMATE CORRESPONDENCE
17. for i ← 1 to n do
18. _(i t ^x, t ^y _i, t _i ^z) ← EstimateCorrespondence (C _i, L _i, (r _i ^x, _i ^y r, ^z r _{i), (i} t ^x, t ^y _i, t _i ^{z _{), <c i p, c}} (i + 1)% n q>) // Optimization for correspondence estimation to get translation parameters which fit to correspond point pairs <c i p, c (i + 1)% n q> of i-th camera and (i + 1)% n-th adjacent (clock-wise) camera
19. c _i ^p ← UpdateGeometry ((r _i ^x , r _i ^y , r _i ^z ), (t _i ^x , t _i ^y , t _i ^z ), c _i ^p ) // Transform correspondence point c _i ^p to world coordinate
20.

// e _i : correspondence error
21. end for
// 3. SELECT OPTIMAL PARAMETERS
22.

← merge error calculation
23. if e> T do // T: threshold for correspondence error
24. check camera location and re-install
25. go to 2
26. end if
27. return (r _i ^x , r _i ^y , r _i ^z ), (t _i ^x , t _i ^y , t _i ^z ), i = {1, ... , n} // return and update rotation parameters for each camera

When the movement and rotation information of the camera is calculated through the above process, the image processing unit 120 outputs the matched images and can match the plurality of images. Then, the image processing unit 120 stores information about the movement and rotation of the camera in the storage unit.

When the image processing unit 120 subsequently performs the image matching, the image matching can be performed easily and quickly using the movement and rotation information of the camera stored in the storage unit.

The image processing unit 120 may include a microcontroller unit (MCU) or an image processing processor that executes a corresponding function together with a memory in which software for performing the above-described algorithm is stored. These devices may be implemented as a chip-on- Or may be mounted on an electronic control unit (ECU) of a vehicle or a separate vehicle control unit.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a method for automatically correcting lane information for vehicle image matching according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 8 is a flowchart illustrating an automatic correction method based on lane information for vehicle image matching according to an embodiment of the present invention. FIG. 9 is a drawing showing such a process as an illustration. The automatic correction method based on lane information for vehicle image matching A step S210 of obtaining a surrounding image while the vehicle is traveling in parallel with the lane, a step S220 of extracting line information such as a lane or a guardrail from the vehicle periphery image, a step of correcting distortion of the vehicle periphery image (S230), calculating and storing the camera movement and rotation information (S240), and performing the image matching using the movement and rotation information of the stored camera (S250).

First, you need to know the parameters inside the camera, and you should also know the initial camera parameters in advance. It should be considered that the camera mounting information may change due to circumstances such as the height of the road when the vehicle travels on the road.

The step S210 of acquiring the surrounding image while the vehicle travels in parallel with the lane includes acquiring the lane information on the road or the image such as the guard rail so that the vehicle can be extracted from the lane or the guard rail And acquiring an image while traveling in parallel with the vehicle.

At this time, preferably, the speed of the vehicle is set to 30 km / h or less, and the change of the height and the rotation information of the camera mounted on the vehicle by the road condition is also considered. In addition, the image around the vehicle allows visualization of all the images of the front, rear, left, and right sides during traveling (see Fig. 2).

A step S220 of extracting line information such as a lane or a guard rail from the vehicle surroundings image is a step of extracting line information such as a lane or a guard rail from the image around the vehicle acquired in step S210, Or the guard rail, and extracts the line information from the detected lane or guardrail image (see the third column of FIG. 9).

As described above, various lane detection algorithms can be used in extracting line information from the surroundings of the vehicle, and at the same time, such a series of processes can be performed in various screen modes. The screen mode used herein may be at least one of an original image, a perspective image, a topview image, and a cylindrical view image.

The step of correcting the image distortion of the peripheral image of the vehicle (S230) is a step of correcting the image distortion such as the lens distortion of the camera using the known camera internal parameters as described above, wherein the internal parameters include the optical center, A projection type, and a focal length.

The movement and rotation information calculation, storage, and image matching step (S240) of the camera are steps of matching the images after estimating the installation position and the installation angle of the camera using the images performed up to step S230 and the extracted line information, And includes the following steps.

Referring to FIG. 10, in step S310, the moving and rotating information calculation, storage, and image matching operations of the camera S240 may include rotating an image in order to align the lane with the driving direction S310, moving the image to match the width of the lane A step S330 of moving an image for image matching, and a step S340 of storing movement and rotation information of a camera image and outputting a matching image.

The step of adjusting the lane to the driving direction (S310) is a step of rotating the direction of the lane of the image according to the running direction of the vehicle using the line information extracted in the step S220. As shown in the third column of Fig. 9 It can be seen that the lane shown irregularly is shown in parallel with the running direction of the vehicle in the fourth column of Fig.

The step of adjusting the width of the lane (S320) is a step of moving the camera image so that the lane in the image is expressed in accordance with the width of the lane. As shown in the fourth column of Fig. 9, Thereby matching the widths of the lanes in the image as shown in the fifth column of FIG.

In the sixth column of FIG. 9, the image is displayed on one screen.

The step of moving the image for image matching (S330) is a step of moving the camera image so that the shifted parts can be matched by using the area in the image common to the adjacent camera in the camera image, Shows a picture arranged in one screen without deviating the image of the lane after this process is finished.

Particularly, in steps S310 to S330, a method of mapping a camera image to a virtual three-dimensional sphere in order to move and rotate the camera image, and then arranging the camera image in the three-dimensional virtual space by changing the angle and distance until the desired lane shape is obtained Can be used.

In addition, we can use an optimization algorithm for estimating the change parameters when moving and rotating the camera image. The optimization algorithm uses the optimization algorithm for movement and rotation, such as PSO (Particle Swarm Optimization), GD Descent), Least Mean Square Error (LMSE), and the like.

The step S340 of storing the movement and rotation information of the camera image and outputting the matching image may include moving and rotating the changed camera after the optimization of the movement and rotation of the image is completed in the previous step, And matching the camera image by the movement and rotation information of the camera and outputting the matching image.

In the upper part of the seventh column of FIG. 9, the movement and rotation information of the camera is outputted as a change parameter and the matching image is outputted at the lower end.

In operation S250, image matching using the movement and rotation information of the stored camera is performed using the change parameters corresponding to the camera movement and rotation information stored in the storage unit in step S240, And matching is performed immediately.

FIG. 11 shows the results before and after image matching with reference to horizontal and vertical seam-lines. That is, the photograph on the left side shows the top view mode image based on the front and rear sides of the vehicle, and shows the photographs before and after the image registration, and the photograph on the right side shows the top view mode image based on the left and right sides of the vehicle. And a photograph before and after the execution. It can be seen that the lanes in the image where the images are misaligned and displaced appear as a single straight line after the image matching.

The automatic correction based on the lane information for the vehicle image matching and the method thereof are not limited to the configuration and the operation manner of the embodiments described above. The embodiments may be configured so that all or some of the embodiments may be selectively combined so that various modifications may be made.

100: Automatic correction device based on lane information for vehicle image registration
110: photographing units 111, 113, 115, and 117: first to fourth cameras
120:

Claims (15)

An automatic correction device based on lane information for vehicle image matching in multi-channel image synthesis around a vehicle,
A photographing unit for receiving and outputting a photographed image including a lane or a guardrail on a road from a plurality of cameras mounted on the vehicle and photographing the area around the vehicle; And
And an image processor for extracting line information from the photographed image, correcting image distortion, estimating an installation position and an installation angle of the camera using the line information, and matching the images and outputting the matched image,
The image processing unit may rotate the image so that the lane is parallel to the driving direction using the line information, move the image so that the width of the lane in the image coincides with each other, And estimating an installation position and an installation angle of the camera based on the lane information.
The method according to claim 1,
The image processing unit may use at least one of an original image, a perspective image, a topview image, and a cylindrical image when extracting line information from the captured image. Lane information based automatic correction device for vehicle image matching.
The method according to claim 1,
Wherein the image processing unit performs image distortion correction of the photographed image using a known camera internal parameter including at least one of a light center, an aspect ratio, a projection type, and a focal length, Based automatic compensation device.
delete The method according to claim 1,
Wherein the image processing unit uses an algorithm for rotating and moving the image, wherein the algorithm is one of PSO (Particle Swarm Optimization), GD (Gradient Descent) and LMSE (Least Mean Square Error) Lane - based Automatic Compensation System for Image Matching.
The method according to claim 1,
Wherein the image processing unit further comprises a storage unit for storing a result after moving and rotating the image.
The method according to claim 1,
Wherein the image processing unit is capable of performing image matching using movement and rotation information of a camera stored in a storage unit.
An automatic correction method based on lane information for vehicle image matching in a multi-channel image synthesis around a vehicle,
Acquiring a surrounding image while the vehicle travels in parallel with the lane;
Extracting line information of a lane or a guard rail from a vehicle surroundings image;
Correcting an image distortion of a peripheral image of the vehicle;
Calculating movement and rotation information of the camera, and storing and matching the images;
And immediately performing image matching of a camera image to be obtained subsequently using movement and rotation information of the stored camera,
Wherein the step of calculating and storing movement and rotation information of the camera and matching the images comprises:
Using the line information, the image is rotated so that the lane is parallel to the running direction, the image is moved so that the width of the lane in the image is coincident, the image is moved so that the images are aligned using the area common to the adjacent cameras, And the image is matched after estimating the installation position and the installation angle of the lane information.
The method of claim 8,
Wherein the step of acquiring a peripheral image while the vehicle travels in parallel with the lane,
And the line information is extracted from the image on the road in the acquired peripheral image.
The method of claim 8,
The step of extracting line information of a lane or a guard rail from the vehicle periphery image includes:
Wherein at least one of a screen mode such as an original image, a perspective image, a topview image, and a cylinder view image is used. Automatic calibration method.
The method of claim 8,
Wherein the step of correcting the image distortion of the vehicle peripheral image comprises:
An automatic correction method based on a lane information for a vehicle image registration, characterized in that image distortion correction of a photographed image is performed using a known camera internal parameter including at least one of an optical center, an aspect ratio, a projection type, and a focal length .
delete The method of claim 8,
Wherein when the image is moved and rotated, the image of the camera is mapped to a virtual three-dimensional sphere, and the angle and the distance are changed, and then the image is arranged in a three-dimensional virtual space. .
The method of claim 8,
Wherein an algorithm is used to move and rotate the image, wherein the algorithm is one of PSO (Particle Swarm Optimization), GD (Gradient Descent), and LMSE (Least Mean Square Error) Information based automatic correction method.
The method of claim 8,
The step of directly performing image matching of a camera image to be obtained subsequently using the stored movement and rotation information of the camera,
And the image matching is performed using the change parameters corresponding to the stored moving and rotating information of the camera.

Images