KR100831452B1 - Image Distortion Correction Using DLT algorithm - Google Patents

Abstract

The present invention relates to a method for correcting the distortion of the image imaged through the camera, the method for correcting the distortion of the screen image imaged through the front or rear camera of the vehicle, the front of the vehicle with the camera as the origin Is a Xa axis and the full width direction is the Ya axis. step; A second step of acquiring measured data which is a grid intersection value of the X axis Ya axis by acquiring an image from the generated grid and regenerating a distance reference line; A third step of acquiring X and Yd coordinates from the measured data to calculate a function of fixing the Ya axis and fitting in the Xa axis direction; A fourth step of generating a pixel Xp function by fitting a pixel Xp to the Xa axis using a sigmoid function, calculating a coefficient value of a Sigmoid function, generating a coefficient function, and generating a total function of the pixel Xp accordingly; And a fifth step of generating a coefficient function by calculating a polynomial coefficient value by fitting the pixel Yp to the Xa axis, and generating a total function of the pixel Yp accordingly. do.

Tsai's Equation Algorithm, Driect Linear Transform Algorithm, Camera Calibration, Pin-hole Camera, Boltzman-Sigmoid Function, Polynomail Function

Description

Image distortion correction method using DLT algorithm {CORRECTING METHOD OF IMAGE DISTORTION USING THE DIRECT LINEAR TRANSFORM ALGORITHM}

1 is a view showing the distortion phenomenon of the camera according to the prior art.

2 and 3 is a view showing the distortion of the trajectory line shown in the real image and the screen image provided through the camera according to the same technology.

4 is a diagram illustrating Xa-axis and Ya-axis setting for mapping of the world image and the screen image according to an embodiment of the present invention.

5 is a flowchart illustrating a process for mapping a world image to a screen image according to an embodiment of the present invention.

6 is a view showing a grid and a distance reference line generated according to an embodiment of the present invention.

7 illustrates a pixel Xp function according to an embodiment of the present invention.

8 and 9 are diagrams illustrating a coefficient function of X according to an embodiment of the present invention.

10 is a diagram showing coefficients of a coefficient function according to an embodiment of the present invention.

11 and 12 show pixel Yp functions according to Xw values.

13 and 14 show various coefficients of the pixel Yp function depending on the Xw value.

15 and 16 are diagrams showing the coefficient of the coefficient function according to the value of Xw.

Fig. 17 is a view showing the distortion correction result by DLT.

18 illustrates a lookup table generated based on a generated pixel Xp function and a pixel Yp function.

19 illustrates a screen image to which the DLT algorithm is applied according to an embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

50: screen trajectory 100: grid horizontal axis

200: vertical axis of the grid 300: distance reference line

The present invention relates to a method for correcting distortion of an image reflected through a front or rear camera of an automobile, and in particular, an image for minimizing an error by maximally applying measured data measured through a camera using a direct linear transform algorithm. It relates to a distortion correction method.

The development of image reproducing technology and applied technology using cameras is applied to the car, and the front or rear camera is attached to the vehicle to detect the distance between the front and rear spaces when parking and provide it to the driver. The accident can be prevented.

Camera calibration is the position of the camera and the internal parameters of the camera to detect whether the points of the actual world coordinates (X, Y, Z) are located at the image coordinates (X, Y) of the image. Focal Length, View Angle, Image Center, etc.) and camera lens distortion are detected to convert the world coordinates to image coordinates. However, in general camera modeling, ideal road modeling is applied symmetrically to the focal length as in the case of a pin-hole camera, as shown in FIG. Distortion occurs. In other words, the pinhole camera is a camera in which the world plane shape of the perspective is projected proportionally without distortion. However, because the angle of view of the pinhole camera is too narrow to be applied to a car, a lens for wide angle is mounted. In the case of wide-lens cameras for vehicles, the shape of the world plane is distorted according to the lens manufacturing method. The larger the angle of view, the smaller the lens aperture, the higher the image distortion. Tsai's Equation algorithm is a method of correcting the camera distortion. This method produces a lot of distortion in the camera using a lens as it moves away from the view axis, and a representative equation of camera distortion correction that solves the amount of distance from the view center in a sequential order. to be. As such, camera calibration is performed using the existing Tsai's Equation algorithm.

However, the conventional method has a disadvantage in that too many errors occur in the case of a camera having a small lens aperture and a small focal length, such as a camera used for a vehicle rear image. Therefore, when the existing method is used in a system requiring precise position correction values such as parking guidance and automatic parking assist system by steering linkage, as shown in FIGS. 2 and 3, the trajectory 50 displayed on the screen and the actual trajectory are displayed. There is a lot of error.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above problems, and provides a method for correcting image distortion using a DLT algorithm for correcting image distortion by applying measurement data measured by a camera to the maximum by using a direct linear transform algorithm. There is.

In order to achieve the above object, the image distortion correction method using the DLT algorithm according to an embodiment of the present invention, a method for correcting the distortion of the screen image is reflected through the front or rear camera of the vehicle,

In order to map Xw, Yw, Zw, the coordinate values of X, Y, Z of the real image to the Xd coordinate Yd coordinate of the screen image, with the camera as the origin and the front direction of the vehicle as the Xa axis and the full width direction as the Ya axis Displaying a grid on a real image; A second step of acquiring measured data which is a grid intersection value of the X axis Ya axis by acquiring an image from the generated grid and regenerating a distance reference line; A third step of acquiring X and Yd coordinates from the measured data to calculate a function of fixing the Ya axis and fitting in the Xa axis direction; A fourth step of generating a pixel Xp function by fitting a pixel Xp to the Xa axis using a sigmoid function, calculating a coefficient value of a Sigmoid function, generating a coefficient function, and generating a total function of the pixel Xp accordingly; And a fifth step of generating a coefficient function by calculating a polynomial coefficient value by fitting the pixel Yp to the Xa axis, and generating a total function of the pixel Yp accordingly. do.

The fourth step of generating the pixel Xp function may include a first substep of displaying the sampled pixel Xp on the graph while changing the Xd coordinate with respect to the fixed Yd coordinate; Generating a sigmoid function by connecting and fitting the displayed pixels Xp; A third substep of repeating the first substep and the second substep while changing the Yd coordinate to obtain all coefficients according to the Yd coordinate change; A fourth substep of displaying a value of each coefficient obtained in the third substep in a graph; A fifth substep of generating a Polynomial function by fitting and fitting the displayed coefficients; A sixth substep of selecting the Polynomial function and obtaining coefficients of the function; And a seventh substep of obtaining the overall function of the pixel Xp by obtaining the coefficients of the polynomial function.

The fifth step of generating the pixel Yp function may include a first substep of displaying the sampled pixel Yp on a graph while changing the Xd coordinate with respect to the fixed Yd coordinate; A second substep of obtaining a fitting function for connecting the displayed pixels Yp and a coefficient function of the fitting function; A third substep of obtaining the coefficients according to the change of the Yd coordinate by changing the Yd coordinate in the first substep and the second substep; A fourth substep of displaying a coefficient value according to the change of the Y value on a graph; A fifth substep of generating a polynomial fitting function as a fitting function connecting the displayed coefficients and obtaining coefficients of the polynomial function; And a sixth substep of obtaining the overall function of the pixel Xp by obtaining the coefficient of the coefficient function of X.

In addition, the image distortion correction method using the DLT algorithm according to an embodiment of the present invention, while generating a look-up table using the pixel Xp function or the pixel Yp function, the actual distance from the reference point to any point and corresponds to the arbitrary point And a sixth step of verifying the distance of the pixels by comparing the distance values of the lookup table.

The above and other objects and features and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. Hereinafter, with reference to the accompanying drawings will be described an embodiment of the present invention; First, in adding reference numerals to the components of each drawing, the same components will be given the same reference numerals as much as possible, even if displayed on different drawings. In addition, in describing the present invention, when it is determined that the detailed description of the related well-known configuration or function may obscure the gist of the present invention, the detailed description thereof will be omitted.

In the image distortion correction method using the DLT algorithm according to an embodiment of the present invention, as shown in FIG. 4, the full length direction of the vehicle is Xa-axis and full-width direction in order to make a world coordinate for measuring the actual world image. Set to Ya axis and derive equation for mapping image distance to display coordinate for display of screen image. In other words, the screen image is based on the camera of the car, so the horizontal direction reflected on the camera is Xa axis and the vertical direction is Ya axis. Looking at the distance mapping calculation step for applying the above-described Direct Linear Transform (hereinafter referred to as DLT) algorithm as follows.

The process for mapping the world image to the screen image,

First create a grid. The grid is displayed at a distance of 40 cm at a short distance of 20 cm within the visible area of the rear camera of the vehicle. The narrower the interval, the smaller the error range of the image distortion.

  In this way, the rear image is acquired, and then, after checking all the intersection pixels of the grid, a fitting is prepared. In other words, the pixels sampled from the grid (pixel Xp, pixel Yp) are marked on the graph and fit. Then, the pixel Xp and pixel Yp functions are calculated by fixing the Ya axis of the world coordinate, changing the Xa axis, and fitting the function. As a result, the pixel Xp function is composed of Sigmoid function and coefficient function, and the pixel Yp function is composed of cubic function and its coefficient function. Based on the generated function, a lookup table is created. The method of generating the pixel functions Xp and Yp will be described later in detail with reference to FIGS. 7 to 18.

5 is a flowchart illustrating a process for mapping a world image to a screen image according to an embodiment of the present invention.

Referring to FIG. 5, in the method for mapping a world image to a screen image, a grid is first generated (S101), and a distance reference line is regenerated based on this (S102), and a grid is obtained from the grid coordinates (S103). A generation step.

Referring to the grid generation step in more detail, in order to make the best use of the measured data, the grid is generated in the horizontal (100) or vertical (200) as shown in FIG. At this time, the grid spacing is irrelevant, and the narrower the spacing, the better. Then, in order to acquire an image from the generated grid to obtain accurate sampling data, as shown in FIG. 6, the distance reference line 300 is regenerated to accurately acquire the measured data value. From the regenerated distance reference line 300, the pixel Xp and the pixel Yp which are actual measurement data about the grid intersection of the Xa-axis Ya axis | shaft are acquired using the vehicle front and rear bumpers with a camera as the origin.

Subsequently, referring to FIG. 5, the Sigmoid function is used to fit the pixel Xp to the Xa axis (S104), and the coefficient value of Sigmoid is calculated by fitting to the Xa axis while changing the Yd coordinate value (S105). After that, the coefficient function is calculated using the Polynomial Fit method (S106), and the pixel Xp function is generated by calculating the overall function of the pixel Xp (S107).

The generation step of the pixel Xp function will be described in more detail. The Ya-axis is fixed to a specific Yd coordinate value, and the sampled pixel Xp value is displayed on the graph while changing the Xd coordinate value with respect to the fixed Yd coordinate value. As shown in FIG. 7, the Ya axis is a pixel Xp value in the image, and the Xa axis is a real coordinate Xw (mm).

Fitting each of the displayed pixels can be represented by a Sigmoid function as shown in FIG. 7. Therefore, the sigmoid function as in Equation 1 is obtained and the coefficients (A, B, C, D) of the function are obtained.

Thereafter, the above-described process is performed by changing each Yd coordinate value, and all coefficients according to the change of the Yd coordinate value are obtained. The coefficient function is calculated as shown in FIGS. 8 and 9 by fitting the obtained coefficients. Here, the calculation of the coefficient function displays the coefficient (A, B, C, D) value according to the change of the Yd coordinate value on the graph. In addition, when the fitting coefficients are connected together, they are expressed as a polynomial function. Therefore, we select the polynomial fitting function and find the coefficients of that function. Also, the overall function of pixel Xp can be obtained by obtaining the coefficients of the coefficient function (A, B, C, D) of X. The coefficients of the coefficient function are summarized as in Equation 2 below, and the coefficient values are as shown in FIG.

Subsequently, referring to FIG. 5, next, the pixel Yp is fitted to the X axis (S108), and the pixel Yp is fitted to the Xa axis (S109). Then, the coefficient value of the polynomial function is calculated (S110) and the coefficient function is calculated according to the criteria of the X value (S111, S112). Finally, the overall function of pixel Yp is calculated (S113) to generate the pixel Yp function.

The generation of the pixel Yp function will be described in more detail. The sampled pixel Yp values are displayed on the graph by fitting the respective values based on Xw = 0 while changing the X values with respect to the fixed Y values. Then, as a fitting function for connecting the displayed pixels, a third-order polynomial such as Equation 3 is selected and the coefficient functions E-H and K-N of the function are obtained.

As shown in FIGS. 11 and 12, the above process is performed while varying the Yd coordinate values to obtain all the coefficients according to the Y value change.

Thereafter, as shown in FIGS. 13 and 14, a coefficient function is calculated. The coefficients E-H and K-N according to the change of the Yd coordinate value are displayed on a graph, and the mathematical function is a fitting function that connects the displayed coefficients. Select the Polynomial fit function as shown in equation 4 and find the coefficients of that function.

Thus, the overall function of the pixel Xp is obtained by obtaining the coefficients of the coefficient functions (E ~ H, K ~ N) of X. The coefficients of the coefficient function of Equation 4 are summarized as shown in FIGS. 15 and 16.

5, a lookup table is generated as shown in FIG. 18 based on a function generated after the grid generation step, the pixel Xp function generation step, and the pixel Yp function generation step as described above (S114). . Then, the real coordinate value and the pixel value are verified using the lookup table, so that the world image is mapped to the screen image. That is, the pixel value by the DLT method is verified by substituting the real coordinates into the sampling pixel and the fitting function and comparing the pixel values. As a result, as shown in FIG. 17, the error is reduced to ± 4 pixels in the ROI (parking area) than the result of the distance mapping by the conventional method. Also, to verify the actual coordinate values, the actual distance from the reference point to an arbitrary point and the pixel corresponding to the arbitrary point are compared with the distance value of the lookup table.

FIG. 19 shows that the distortion image is corrected according to the application of the DLT algorithm of the present invention to provide the driver with a more accurate parking guideline than the existing image.

As described above, in the image distortion correction method using the DLT algorithm according to the embodiment of the present invention, the distortion of the screen image provided through the camera is adjusted by maximally applying the measured data measured by the camera by using the direct linear transform algorithm. You can correct it.

In addition, the image distortion correction method using the DLT algorithm according to an embodiment of the present invention, by applying to the existing parking support system can set the precise parking area, by applying to the system to represent the parking trajectory in conjunction with the steering screen of the existing screen The distortion of the image can be improved.

In addition, preferred embodiments of the present invention are disclosed for the purpose of illustration, those skilled in the art will be able to make various modifications, changes, additions, etc. within the spirit and scope of the present invention, such modifications and modifications belong to the scope of the claims You will have to look.

Claims (4)

As a method for correcting distortion of a screen image reflected by a front or rear camera of a vehicle, With the camera as the origin, the front direction of the vehicle is the Xa axis and the full width direction is the Ya axis. A first step of displaying a grid on a screen displaying an image; A second step of acquiring actual data of pixels Xp and Yp which are grid intersection pixel values of the Xa and Ya axes by regenerating a distance reference line by acquiring an image from the generated grid; A third step of acquiring Xd and Yd coordinates from the measured data and calculating a function of fixing the Ya axis and fitting in the Xa axis direction; Fitting a pixel Xp corresponding to the Xd coordinate to the Xa axis by using a Sigmoid function, calculating a coefficient value of a Sigmoid function, creating a coefficient function, and generating a total function of the pixel Xp for the coefficient function A fourth step of generating a pixel Xp function; And A pixel Yp function is generated by fitting a pixel Yp corresponding to the Yd coordinate with the Xa axis, calculating a coefficient value of a polynomial function, generating a coefficient function, and generating an overall function of the pixel Yp for the coefficient function. Image distortion correction method using a DLT algorithm, characterized in that it comprises a fifth step. The method of claim 1, wherein the fourth step of generating the pixel Xp function A first substep of displaying the sampled pixel Xp on the graph while changing the Xd coordinate with respect to the fixed Yd coordinate; Generating a sigmoid function by connecting and fitting the displayed pixels Xp; A third substep of repeating the first substep and the second substep while changing the Yd coordinate to obtain all coefficients according to the Yd coordinate change; A fourth substep of displaying a value of each coefficient obtained in the third substep in a graph; A fifth substep of generating a Polynomial function by fitting and fitting the displayed coefficients; A sixth substep of selecting the Polynomial function and obtaining coefficients of the function; And a seventh substep of obtaining the entire function of the pixel Xp by obtaining the coefficients of the polynomial function. The method of claim 1, wherein the fifth step of generating the pixel Yp function A first substep of displaying, on a graph, the pixel Yp sampled while changing an Xd coordinate with respect to the fixed Yd coordinate; A second substep of obtaining a fitting function for connecting the displayed pixels Yp and a coefficient function of the fitting function; A third substep of determining the coefficients according to the change of the Yd coordinate by changing the Yd coordinate in the first substep and the second substep; A fourth substep of displaying a coefficient value according to the change of the Yd coordinate on a graph; A fifth substep of generating a polynomial fitting function as a fitting function connecting the displayed coefficients and obtaining coefficients of the polynomial function; And a sixth substep of obtaining an overall function of the pixel Xp by obtaining a coefficient of the coefficient function with respect to the pixel Xp. The method of claim 1, A sixth step of generating a lookup table using the pixel Xp function or the pixel Yp function, and verifying a real distance from a reference point to an arbitrary point and a distance of a pixel corresponding to the arbitrary point by comparing the distance value of the lookup table Image distortion correction method using a DLT algorithm, characterized in that it further comprises.

Images