CN107633536A - A kind of camera calibration method and system based on two-dimensional planar template - Google Patents

Abstract

The invention discloses a camera calibration method and system based on a two-dimensional plane template, wherein the method includes: obtaining a checkerboard calibration image, the calibration image contains a plurality of corner points; Among the corner points, the first error value of the corner point is greater than the first preset value, and the target corner point is obtained; according to the coordinates of the target corner point and the corresponding world coordinates, a homography matrix is obtained; a random sampling consensus algorithm is used Eliminate the homography matrix whose second error value is greater than the second preset value to obtain the target homography matrix; obtain the target conic curve model according to the target homography matrix, and obtain the target conic curve model according to the target conic curve model The initial parameters of the camera; the initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, and the camera is calibrated by using the target parameters. The invention solves the technical problem that the accuracy of camera calibration is not high in Zhang Zhengyou's calibration method.

Description

A camera calibration method and system based on a two-dimensional plane template

technical field

The invention relates to the technical field of visual measurement, in particular to a camera calibration method and system based on a two-dimensional plane template.

Background technique

In visual measurement, the calibration of camera parameters is a very critical link. The calibration accuracy and the stability of the calibration algorithm directly affect the accuracy of the results produced by the camera. The purpose of camera calibration is to estimate the parameters of the camera lens and image sensor, including internal parameters, external parameters, and lens distortion coefficients. These parameters are widely used in the field of computer vision, such as correcting lens distortion, measuring the actual size of objects, and determining the camera in the scene. location etc. Camera calibration is also widely used in robotics, navigation systems, and 3D reconstruction.

At present, camera calibration methods are roughly divided into two categories: traditional calibration methods and camera self-calibration methods. The traditional calibration method is to use a calibration block of known size to establish the corresponding relationship between object coordinates and image coordinates, and then obtain the internal and external parameters of the camera through the corresponding algorithm. The calibration accuracy is high, but the processing and maintenance of the three-dimensional calibration block is difficult and the cost is high. ; Camera self-calibration mainly uses camera motion constraints or scene constraints to estimate parameters, which is more flexible, but the self-calibration method is too strong for camera motion and scene constraints, and is less robust and less accurate in actual use. These factors also limit the scope of use of self-calibration.

In order to solve the above technical problems, Zhang Zhengyou proposed a flexible planar calibration method, which is between the traditional calibration and the camera self-calibration method. The calibration method requires high equipment requirements, and the self-calibration method has low precision. However, Zhang Zhengyou’s calibration method uses multiple checkerboard images, and uses the corresponding relationship of corner points to calculate the corresponding homography matrix of each image, and then uses the closed solution to find the internal and external parameters of the camera, and finally considers the lens distortion. The obtained closed solution is used as the initial value, and a nonlinear search algorithm is used to estimate all parameters, including internal parameters, external parameters and distortion coefficients. However, the nonlinear search algorithm has strict requirements on the accuracy of the given initial value. Since the camera parameters are coupled, if the accuracy of the initial value provided is not high, the performance of nonlinear optimization will be poor, and it will fall into a local optimal solution. As a result, the accuracy and robustness of solving camera parameters and distortion coefficients are not enough.

It can be seen that the existing Zhang Zhengyou calibration method has the technical problem of low camera calibration accuracy.

Contents of the invention

Embodiments of the present invention provide a camera calibration method and system based on a two-dimensional plane template to solve the technical problem of low camera calibration accuracy in the existing Zhang Zhengyou calibration method.

The invention discloses a camera calibration method based on a two-dimensional plane template. The method includes:

Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points;

Using a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point;

Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates;

Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix;

Obtain a target conic model according to the target homography matrix, and obtain initial parameters of the camera according to the target conic model, where the initial parameters include a first internal parameter, a first external parameter, and a first distortion coefficient;

The initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated using the target parameters.

In the method provided by the present invention, said adopting the random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points to obtain the target corner point includes:

Get the reprojected coordinates of the corner points;

Obtaining a first distance between the reprojected coordinates of the corner point and the initial coordinates of the corner point;

A random sampling consensus algorithm is used to eliminate the corner points corresponding to the first distance greater than the first preset value, so as to obtain the target corner points.

In the method provided by the present invention, according to the coordinates of the target corners and the corresponding world coordinates, obtaining the homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system includes:

Randomly select four corner points from the checkerboard calibration image, and obtain the coordinates of the four corner points;

According to the coordinates of the four corner points and the world coordinates corresponding to the four corner points, a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained.

In the method provided by the present invention, the random sampling consensus algorithm is used to eliminate the homography matrix whose second error value is greater than the second preset value, and obtain the target homography matrix, including:

Obtain a preset quadratic curve model according to the homography matrix;

Obtaining a second distance between the homography matrix and the conic model, using the second distance as a difference;

If the difference is greater than the second preset value, it is an unqualified homography matrix;

The unqualified homography matrix is eliminated from the original homography matrix to obtain a target homography matrix.

In the method provided by the present invention, the checkerboard calibration images include multiple pieces, and different calibration images have different angles from calibration cameras.

Based on the same inventive concept, the second aspect of the present invention provides a camera calibration system based on a two-dimensional planar template, the system includes: an acquisition module, used to acquire a checkerboard calibration image, the calibration image contains a plurality of corner points ;

The first obtaining module is configured to use a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, and obtain the target corner point;

The second obtaining module is used to obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates;

The third obtaining module is used to eliminate the homography matrix whose second error value is greater than the second preset value by adopting the random sampling consensus algorithm to obtain the target homography matrix;

The fourth obtaining module is used to obtain the target conic curve model according to the target homography matrix, and obtain the initial parameters of the camera according to the target conic curve model, and the initial parameters include the first internal parameter, the first external parameter Parameter and the first distortion coefficient;

A calibration module, configured to estimate the initial parameters using a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter, and a second distortion coefficient, and use the target parameters to estimate the camera Calibrate.

In the system provided by the present invention, the first obtaining module is also used for:

Get the reprojected coordinates of the corner points;

Obtaining a first distance between the reprojected coordinates of the corner point and the initial coordinates of the corner point;

A random sampling consensus algorithm is used to eliminate the corner points corresponding to the first distance greater than the first preset value, so as to obtain the target corner points.

In the system provided by the present invention, the second obtaining module is also used for:

Randomly select four corner points from the checkerboard calibration image, and obtain the coordinates of the four corner points;

According to the coordinates of the four corner points and the world coordinates corresponding to the four corner points, a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained.

In the system provided by the present invention, the third obtaining module is also used for:

Obtain a preset quadratic curve model according to the homography matrix;

Obtaining a second distance between the homography matrix and the conic model, using the second distance as a difference;

If the difference is greater than the second preset value, it is an unqualified homography matrix;

Eliminate the unqualified homography matrix from the original homography matrix to obtain the target homography matrix

In the system provided by the present invention, the checkerboard calibration images include multiple pieces, and different calibration images have different angles from calibration cameras.

Based on the same inventive concept, the third aspect of the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the following steps are implemented:

Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points;

Using a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point;

Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates;

Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix;

Obtain a target conic model according to the target homography matrix, and obtain initial parameters of the camera according to the target conic model, where the initial parameters include a first internal parameter, a first external parameter, and a first distortion coefficient;

The initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated using the target parameters.

Based on the same inventive concept, the fourth aspect of the present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and operable on the processor, and the processor implements the program when executing the program. The following steps:

Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points;

Using a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point;

Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates;

Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix;

Obtain a target conic model according to the target homography matrix, and obtain initial parameters of the camera according to the target conic model, where the initial parameters include a first internal parameter, a first external parameter, and a first distortion coefficient;

The initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated using the target parameters.

One or more technical solutions provided in the embodiments of the present invention have at least the following technical effects or advantages:

An embodiment of the present application provides a camera calibration method based on a two-dimensional plane template, the method comprising: acquiring a checkerboard calibration image, the calibration image contains multiple corner points; Among the corner points, the first error value of the corner point is greater than the first preset value, and the target corner point is obtained; according to the coordinates of the target corner point and the corresponding world coordinates, the world coordinate system where the checkerboard calibration image is located is obtained. To the homography matrix of the pixel coordinate system; using the random sampling consensus algorithm to eliminate the homography matrix whose second error value is greater than the second preset value, and obtain the target homography matrix; according to the target homography matrix, obtain The target quadratic curve model, and obtain the initial parameters of the camera according to the target quadratic curve model, the initial parameters include the first internal parameter, the first external parameter and the first distortion coefficient; the method of maximum likelihood estimation is used for the The initial parameters are estimated to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated by using the target parameters. In the above method provided by the present invention, on the one hand, when obtaining the initial value, the random sampling consensus algorithm is used to eliminate the corner points whose corner point error value is greater than the first preset value among the multiple corner points, thereby obtaining the target corner point, Then calculate the homography matrix according to the target corner points. Since the corner points with low precision are eliminated, the accuracy of the homography matrix calculation can be improved, thereby improving the calibration accuracy. On the other hand, the random sampling consistency algorithm is used to eliminate The homography matrix whose error value is greater than the second preset value is obtained to obtain the target homography matrix, and then the calibration parameters are solved according to the target homography matrix. Since the low-quality calibration image is eliminated, it can be The accuracy of the calibration is further improved, and the technical problem of low camera calibration accuracy and robustness in the existing Zhang Zhengyou calibration method is solved.

The above description is only an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable , the specific embodiments of the present invention are enumerated below.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is a flow chart of a camera calibration method based on a two-dimensional plane template in an embodiment of the present invention;

2 is a structural diagram of a camera calibration system based on a two-dimensional plane template in an embodiment of the present invention;

FIG. 3 is a structural diagram of a computer-readable medium in an embodiment of the present invention;

Fig. 4 is a structural diagram of a computer device in an embodiment of the present invention.

detailed description

Embodiments of the present invention provide a camera calibration method and system based on a two-dimensional planar template to solve the technical problem of low camera calibration accuracy in the existing Zhang Zhengyou calibration method.

The general idea of the technical solution in the embodiment of the application is as follows:

A camera calibration method based on a two-dimensional plane template, the method comprising: obtaining a checkerboard calibration image, the calibration image including a plurality of corner points; using a random sampling consensus algorithm to eliminate the first corner point among the plurality of corner points A corner point whose error value is greater than the first preset value is obtained to obtain the target corner point; according to the coordinates of the target corner point and the corresponding world coordinates, the unitary distance from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained a corresponding matrix; a random sampling consistency algorithm is used to eliminate a homography matrix whose second error value is greater than a second preset value to obtain a target homography matrix; according to the target homography matrix, a target quadratic curve model is obtained, And obtain the initial parameters of the camera according to the target quadratic curve model, the initial parameters include the first internal parameter, the first external parameter and the first distortion coefficient; adopt the method of maximum likelihood estimation to estimate the initial parameter, obtain Target parameters, where the target parameters include a second internal parameter, a second external parameter, and a second distortion coefficient, and the camera is calibrated by using the target parameters.

In the above method, on the one hand, when obtaining the initial value, the random sampling consensus algorithm is used to eliminate the corner points whose corner point error value is greater than the first preset value among the multiple corner points, so as to obtain the target corner point, and then according to The homography matrix is calculated by the target corner point. Since the corner points with low precision are eliminated, the accuracy of the homography matrix calculation can be improved, thereby improving the calibration accuracy. On the other hand, the random sampling consistency algorithm is used to eliminate the error value A homography matrix greater than the second preset value, so as to obtain the target homography matrix, and then solve the calibration parameters according to the target homography matrix, because the low-quality calibration images are eliminated, which can further improve the calibration It solves the technical problem that the accuracy and robustness of camera calibration in the existing Zhang Zhengyou calibration method are not high.

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Embodiment one

This embodiment provides a camera calibration method based on a two-dimensional plane template, please refer to FIG. 1, the method includes:

Step S101: Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points;

Step S102: Use the random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point.

Step S103: Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates.

Step S104: Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix.

Step S105: According to the target homography matrix, obtain the target conic curve model, and obtain the initial parameters of the camera according to the target conic curve model, the initial parameters include the first internal parameter, the first external parameter and the first Distortion coefficient;

Step S106: Estimate the initial parameters using the method of maximum likelihood estimation to obtain target parameters, the target parameters include the second internal parameters, the second external parameters and the second distortion coefficient, and use the target parameters to calibrate the camera .

It should be noted that the applicant of the present invention has found through long-term practice that in Zhang Zhengyou's method based on plane calibration, the accuracy of solving the initial value is mainly related to the following two aspects, first, the extraction accuracy of the checkerboard corner points, and second , the quality of the collected calibration image. However, in practical applications, the camera calibration process will be affected by noise, image blur, illumination changes, and checkerboard viewing angles, etc., resulting in low extraction accuracy of checkerboard corners. In addition, only high-quality calibration images that meet the specifications can accurately calibrate the camera parameters. For example, if the angle between the checkerboard plane and the camera image plane is greater than 45 degrees, the calibration accuracy will be reduced, the ratio of the checkerboard in the image and the distance between the checkerboard and the camera The distance of the optical center will affect the accuracy of solving the initial value. Based on the above knowledge, the method of this application proposes a camera calibration method based on a two-dimensional plane template. The present invention uses a double-random sampling consensus algorithm (D-RANSAC) to start with two aspects that affect the accuracy of the initial value solution , by eliminating low-precision corner points and substandard low-quality images, thereby improving the solution accuracy of the initial value. Specifically: use the reprojection error size of the corner points of each checkerboard, and use the random sampling consistency algorithm to eliminate the corner points with large errors, and then recalculate the homography matrix of each image. And based on the theory that the preset quadratic curve model is invariant in a single camera, the random sampling consistency algorithm is used again for the homography matrix corresponding to each image to eliminate low-quality calibration images that do not meet the specifications. Finally, the closed solution is used to obtain a relatively accurate initial value, that is, the initial internal and external parameters of the camera (namely the first internal reference and the second internal reference), and combined with the lens distortion factor, the nonlinear search algorithm is used to obtain the precise second value of the camera. The internal reference, the second external reference and the second distortion coefficient are used to calibrate the camera, thereby improving the accuracy and precision of the calibration.

Below, a camera calibration method based on a two-dimensional plane template provided by the present application is introduced in detail in conjunction with FIG. 1:

First perform step S101: acquire a checkerboard calibration image, the calibration image includes a plurality of corner points;

In a specific implementation process, multiple checkerboard calibration images can be taken by the camera, where each checkerboard calibration image contains multiple corner points. Preferably, different calibration images have different angles from the calibration camera. For example, the calibration images can be taken from multiple angles to ensure that the calibration images have sufficient angle changes relative to the camera, thereby ensuring the calibration accuracy. The number of calibration images can be taken according to actual conditions. For example, it can be 20, 25, 30 and so on.

Then step S102 is executed: using the random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point.

In the specific implementation process, since the accuracy of the corner points of the checkerboard affects the final calibration accuracy, it is necessary to remove the corner points with low precision. The corner points whose error value is larger than the first preset value are obtained to obtain the target corner points.

The method of removing low-precision corner points can be achieved through the following steps:

Get the reprojected coordinates of the corner points;

Obtaining a first distance between the reprojected coordinates of the corner point and the initial coordinates of the corner point;

A random sampling consensus algorithm is used to eliminate the corner points corresponding to the first distance greater than the first preset value, so as to obtain the target corner points.

In the specific implementation process, the camera model is first introduced. The camera model is divided into linear model and nonlinear model. The linear model based on the two-dimensional plane template is as follows:

in

In the above formula, s represents the scale factor, (u, v) is the pixel coordinate of the camera imaging point corresponding to the corner point of the checkerboard image (that is, the corner point coordinate), (X, Y) is the world coordinate of the corner point of the checkerboard image, K is the internal parameters of the camera, r ₁ and r ₂ are the column vectors of the camera external parameter rotation matrix R _3×3 , t represents the camera external parameter translation matrix, R _3×3 and t represent the world coordinate system where the checkerboard calibration board is located Rotation and translation matrices to the camera coordinate system.

Next, the homography matrix is introduced. For example, if the four pairs of mapping points (u, v) and (X, Y) in each calibration image and camera imaging are known, the position of the checkerboard calibration board can be obtained by formula (2). The homography matrix H from world coordinates to pixel coordinates, a homography matrix can be obtained for each calibration image, as follows:

in

Among them, the homography matrix H has 8 degrees of freedom.

For the acquired checkerboard image, its corner coordinates (u, v) can be extracted, and then its corresponding world coordinates (X, Y) and formula (2) can be used to calculate the homography matrix H of the checkerboard image. Due to noise, image Effects such as blurring and lighting changes will cause the extracted corner coordinates to be inaccurate, so the calculated matrix obtained above is not accurate, so the coordinates of the corner reprojection can be obtained by formula (2) And compared with the initial coordinates (u, v) of the corner point, the Euclidean distance between the two is the reprojection error The smaller the value of d ₁ is, the higher the extraction accuracy of the corner point is, and if the error value is greater than the first preset value, it indicates that the corner point is not high in accuracy, and thus the corner point is eliminated. For example, set the first preset value to δ ₁ , if d ₁ <δ ₁ , then the corner meets the requirements of the model, called an interior corner, and record the number t ₁ of corners that meet the calibration requirements; otherwise, the corner Points that do not meet the model requirements are called exterior corner points. In order to ensure the accuracy, the above process is repeated many times, and the number of repetitions is determined by the following formula:

N=log(1-p)/log(1-w ^s )(4)

The above formula means that in N iterations, the minimum number of samples is randomly selected each time as s, then the probability that there is no outlier at least once is p, and the value of p is 0.99, and w represents the probability that any selected sample is an inlier. Then retain all the corner points corresponding to the model when t ₁ is the largest, and then use these corner points to re-estimate the homography matrix by the least square method to get the optimal model.

Next, step S103 is executed: according to the coordinates of the target corners and the corresponding world coordinates, a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained.

In a specific implementation process, the homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system can be obtained by formula (2).

Specifically, according to the coordinates of the target corner points and the corresponding world coordinates, obtaining the homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system includes:

Randomly select four corner points from the checkerboard calibration image, and obtain the coordinates of the four corner points;

According to the coordinates of the four corner points and the world coordinates corresponding to the four corner points, a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained.

In the specific implementation process, four corner coordinates (u, v) are randomly selected in each calibration image, and the formula (2) and the corresponding world coordinates (X, Y) are used to obtain the calibration image The corresponding homography matrix

Then step S104 is executed: using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value, to obtain a target homography matrix.

Specifically, the random sampling consensus algorithm is used to eliminate the homography matrix whose second error value is greater than the second preset value, and obtain the target homography matrix, including:

Obtain a preset quadratic curve model according to the homography matrix;

Obtaining a second distance between the homography matrix and the conic model, using the second distance as a difference;

If the difference is greater than the second preset value, it is an unqualified homography matrix;

The unqualified homography matrix is eliminated from the original homography matrix to obtain a target homography matrix.

In the specific implementation process, the internal parameter matrix K and the preset conic curve model B are firstly introduced. In the Zhang Zhengyou calibration method, using the known condition that r ₁ and r ₂ are unit orthogonal, the following two constraints can be obtained:

h ₁ ^T K ^-T K ^-1 h ₂ =0

h ₁ ^T K ^-T K ^-1 h ₁ ＝h ₂ ^T K ^-T K ^-1 h ₂ (5)

Among them, h ₁ , h ₂ are the column vectors of the homography matrix H, K is the internal reference matrix of the camera, B=K ^-T K ^-1 is the preset conic curve model (abbreviated as ICA), from the expression of B, we can know , B is only related to the internal parameter matrix K of the camera, but has nothing to do with the direction and position of the camera.

The method of obtaining the target homography matrix can be achieved through the following steps. First, two homography matrices are randomly selected , calculate ICA by formula (5), that is, B=K ^-T K ^-1 , and then calculate each homography matrix separately Distance d ₂ from ICA, where d ₂ =(h ₁ ^T K ^-T K ^-1 h ₂ ) ² +(h ₁ ^T K ^-T K ^-1 h ₁ -1) ² +(h ₂ ^T K ^-T K ^- 1h ₂ -1) ² (6); the second preset value is δ ₂ , if d ₂ <δ ₂ , the homography matrix meets the requirements of the model, it is called the inner homography matrix, and it is recorded that The number t ₂ of the homography matrix of the model; otherwise, the homography matrix does not conform to the model, which is called the outer homography matrix. In order to ensure accuracy, repeat the above process N ₂ times (N ₂ is determined by formula 4), and retain all homography matrices corresponding to the model when t ₂ is the largest Let these homography matrices be the target homography matrices.

Next, step S105 is performed: according to the target homography matrix, the target conic curve model is obtained, and the initial parameters of the camera are obtained according to the target conic curve model, and the initial parameters include the first internal parameter and the first external parameter .

In the specific implementation process, the target homography matrix obtained in the previous steps is used, and the least square method is used to re-estimate the model to obtain an accurate ICA model, and then the initial parameters of the camera are obtained by linear calculation.

Finally, step S106 is executed: using the method of maximum likelihood estimation to estimate the initial parameters to obtain the target parameters, the target parameters include the second internal parameter, the second external parameter and the second distortion coefficient, and use the target parameter to The camera is calibrated.

In the specific implementation process, the initial parameters obtained in step S105 are used as closed solutions, and these closed solutions are used as initial values, and then all initial parameters are estimated by the method of maximum likelihood estimation, including the first internal parameter, the first external parameter and the first distortion coefficient. And the initial values of the first distortion coefficients k ₁ and k ₂ are set to zero. Use formula (7) to optimize and obtain accurate internal reference A, external reference k ₁ , k ₂ , second distortion coefficient R _i , t _i

Among them, n represents the number of images finally used for calibration after removing low-quality calibration images, m _i represents the number of corner points used to calculate the homography matrix after eliminating low-precision corner points in the i-th calibration image, m _ij represents the number of corner points in the i-th calibration image The coordinates of the jth corner point in the image, M _j represents the known world coordinates corresponding to m _ij , denote the reprojected coordinates of _Mj . Using the closed solution A, R _i and k ₁ =0, k ₂ =0 obtained in the previous steps as initial values, the formula (7) is solved by the Levenberg-Marquarat iterative optimization algorithm, so as to obtain accurate target parameters, and the target The parameters include A, R _i , t _i , k ₁ , k ₂ , namely the second internal parameter, the second external parameter and the second distortion coefficient of the camera, and then the camera is calibrated through the above target parameters to complete the high-precision calibration of the camera.

Based on the same inventive concept, an embodiment of the present invention also provides a system corresponding to a camera calibration method based on a two-dimensional plane template, see Embodiment 2 for details.

Embodiment two

Embodiment 2 of the present invention provides a camera calibration system based on a two-dimensional plane template, please refer to FIG. 2 , the system includes:

An acquisition module 201, configured to acquire a checkerboard calibration image, the calibration image including a plurality of corner points;

The first obtaining module 202 is configured to use a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, and obtain the target corner point;

The second obtaining module 203 is configured to obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner point and the corresponding world coordinates;

The third obtaining module 204 is configured to use a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value, so as to obtain a target homography matrix;

The fourth obtaining module 205 is used to obtain the target conic curve model according to the target homography matrix, and obtain the initial parameters of the camera according to the target conic curve model, and the initial parameters include the first internal parameter, the first External parameters and the first distortion coefficient;

The calibration module 206 is configured to estimate the initial parameters by means of maximum likelihood estimation to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and use the target parameters to The camera is calibrated.

In the system provided in this embodiment, the first obtaining module 202 is also used for:

Get the reprojected coordinates of the corner points;

Obtaining a first distance between the reprojected coordinates of the corner point and the initial coordinates of the corner point;

A random sampling consensus algorithm is used to eliminate the corner points corresponding to the first distance greater than the first preset value, so as to obtain the target corner points.

In the system provided in this embodiment, the second obtaining module 203 is further configured to:

Randomly select four corner points from the checkerboard calibration image, and obtain the coordinates of the four corner points;

According to the coordinates of the four corner points and the world coordinates corresponding to the four corner points, a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained.

In the system provided in this embodiment, the third obtaining module 204 is also used for:

Obtain a preset quadratic curve model according to the homography matrix;

Obtaining a second distance between the homography matrix and the conic model, using the second distance as a difference;

If the difference is greater than the second preset value, it is an unqualified homography matrix;

The unqualified homography matrix is eliminated from the original homography matrix to obtain a target homography matrix.

In the system provided in this embodiment, the checkerboard calibration image includes multiple pieces, and different calibration images have different angles from the calibration camera.

The various variations and specific examples based on the first embodiment are also applicable to the system of this embodiment. Those skilled in the art can clearly understand the above-mentioned detailed descriptions in this embodiment. Therefore, for the sake of brevity in the description, It will not be described in detail here.

Based on the same inventive concept, an embodiment of the present invention also provides a computer-readable medium corresponding to a camera calibration method based on a two-dimensional planar template, see Embodiment 3 for details.

Embodiment three

Embodiment 3 of the present invention provides a computer-readable storage medium 300, referring to FIG. 3 , on which a computer program 301 is stored. When the program is executed by a processor, the following steps are implemented:

Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points;

Using a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point;

Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates;

Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix;

Obtain a target conic model according to the target homography matrix, and obtain initial parameters of the camera according to the target conic model, where the initial parameters include a first internal parameter, a first external parameter, and a first distortion coefficient;

The initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated using the target parameters.

Various variations and specific examples of the calibration method in Embodiment 1 are also applicable to the computer-readable medium of this embodiment. Through the foregoing detailed description of the camera calibration method, those skilled in the art can clearly know the computer-readable medium, so for the sake of brevity of the description, it will not be described in detail here.

Based on the same inventive concept, an embodiment of the present invention also provides a computer device corresponding to a camera calibration method based on a two-dimensional plane template, see Embodiment 4 for details.

Embodiment four

Embodiment 3 of the present invention provides a computer device, please refer to FIG. 4 , which includes a memory 401, a processor, and a computer program stored in the memory and operable on the processor. When the processor executes the program, it implements the following step:

Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points;

Using a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point;

Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates;

Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix;

Obtain a target conic model according to the target homography matrix, and obtain initial parameters of the camera according to the target conic model, where the initial parameters include a first internal parameter, a first external parameter, and a first distortion coefficient;

The initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated using the target parameters.

For ease of description, FIG. 4 only shows the parts related to the embodiment of the present invention. For specific technical details not disclosed, please refer to the method part of the embodiment of the present invention. Wherein, the memory 401 can be used to store software programs and modules, and the processor 402 executes various functional applications and data processing of the mobile terminal by running and executing the software programs and modules stored in the memory 401 .

The memory 401 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function, etc.; the data storage area may store data created according to the use of the computer device, etc. The processor 402 is the control center of the mobile communication terminal, which uses various interfaces and lines to connect various parts of the entire mobile communication terminal, and runs or executes software programs and/or modules stored in the memory 401, and calls stored in the memory 401. Data, execute various functions of the mobile terminal and process data, so as to monitor the mobile terminal as a whole. Optionally, the processor 402 may include one or more processing units.

Various variations and specific examples of the calibration method in Embodiment 1 are also applicable to the computer device in this embodiment. Through the foregoing detailed description of the camera calibration method, those skilled in the art can clearly know that the computer device in this embodiment , so for the sake of brevity of the description, it will not be described in detail here.

One or more technical solutions provided in the embodiments of the present invention have at least the following technical effects or advantages:

An embodiment of the present application provides a camera calibration method based on a two-dimensional plane template, the method comprising: acquiring a checkerboard calibration image, the calibration image contains multiple corner points; Among the corner points, the first error value of the corner point is greater than the first preset value, and the target corner point is obtained; according to the coordinates of the target corner point and the corresponding world coordinates, the world coordinate system where the checkerboard calibration image is located is obtained. To the homography matrix of the pixel coordinate system; using the random sampling consensus algorithm to eliminate the homography matrix whose second error value is greater than the second preset value, and obtain the target homography matrix; according to the target homography matrix, obtain A target quadratic curve model, and obtain initial parameters of the camera according to the target quadratic curve model, the initial parameters include a first internal parameter and a first external parameter; the initial parameters are estimated by a method of maximum likelihood estimation, A target parameter is obtained, the target parameter includes a second internal parameter, a second external parameter and a distortion coefficient, and the camera is calibrated by using the target parameter. In the above method provided by the present invention, on the one hand, when obtaining the initial value, the random sampling consensus algorithm is used to eliminate the corner points whose corner point error value is greater than the first preset value among the multiple corner points, thereby obtaining the target corner point, Then calculate the homography matrix according to the target corner points. Since the corner points with low precision are eliminated, the accuracy of the homography matrix calculation can be improved, thereby improving the calibration accuracy. On the other hand, the random sampling consistency algorithm is used to eliminate The homography matrix whose error value is greater than the second preset value is obtained to obtain the target homography matrix, and then the calibration parameters are solved according to the target homography matrix. Since the low-quality calibration image is eliminated, it can be The accuracy of the calibration is further improved, and the technical problem of low camera calibration accuracy and robustness in the existing Zhang Zhengyou calibration method is solved.

While preferred embodiments of the invention have been described, additional changes and modifications to these embodiments can be made by those skilled in the art once the basic inventive concept is appreciated. Therefore, it is intended that the appended claims be construed to cover the preferred embodiment as well as all changes and modifications which fall within the scope of the invention.

Apparently, those skilled in the art can make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. In this way, if the modifications and variations of the embodiments of the present invention fall within the scope of the claims of the present invention and equivalent technologies, the present invention also intends to include these modifications and variations.

Claims (10)

1. A camera calibration method based on a two-dimensional plane template, characterized in that the method comprises: Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points; Using a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point; Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates; Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix; Obtain a target conic model according to the target homography matrix, and obtain initial parameters of the camera according to the target conic model, where the initial parameters include a first internal parameter, a first external parameter, and a first distortion coefficient; The initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated using the target parameters. 2. The method according to claim 1, wherein the random sampling consensus algorithm is used to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points to obtain the target angle points, including: Obtain the reprojected coordinates of the corner points; Obtaining a first distance between the reprojected coordinates of the corner point and the initial coordinates of the corner point; A random sampling consensus algorithm is used to eliminate the corner points corresponding to the first distance greater than the first preset value, so as to obtain the target corner points. 3. The method according to claim 1, wherein the homography from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained according to the coordinates of the target corner point and the corresponding world coordinates sex matrix, including: Randomly select four corner points from the checkerboard calibration image, and obtain the coordinates of the four corner points; According to the coordinates of the four corner points and the world coordinates corresponding to the four corner points, a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained. 4. The method according to claim 1, wherein said adopting a random sampling consensus algorithm to eliminate a homography matrix whose second error value is greater than a second preset value to obtain a target homography matrix comprises: Obtain a preset quadratic curve model according to the homography matrix; Obtaining a second distance between the homography matrix and the conic model, using the second distance as a difference; If the difference is greater than the second preset value, it is an unqualified homography matrix; The unqualified homography matrix is eliminated from the original homography matrix to obtain a target homography matrix. 5. The method according to any one of claims 1-4, wherein the checkerboard calibration image comprises multiple pieces, and different calibration images have different angles from the calibration camera. 6. A camera calibration system based on a two-dimensional plane template, characterized in that the system includes: An acquisition module, configured to acquire a checkerboard calibration image, where the calibration image includes a plurality of corner points; The first obtaining module is configured to use a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, and obtain the target corner point; The second obtaining module is used to obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates; The third obtaining module is used to eliminate the homography matrix whose second error value is greater than the second preset value by adopting the random sampling consensus algorithm to obtain the target homography matrix; The fourth obtaining module is used to obtain the target conic curve model according to the target homography matrix, and obtain the initial parameters of the camera according to the target conic curve model, and the initial parameters include the first internal parameter, the first external parameter Parameter and the first distortion coefficient; A calibration module, configured to estimate the initial parameters using a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter, and a second distortion coefficient, and use the target parameters to estimate the camera Calibrate. 7. The system according to claim 6, wherein the first obtaining module is further configured to: Obtain the reprojected coordinates of the corner points; Obtaining a first distance between the reprojected coordinates of the corner point and the initial coordinates of the corner point; A random sampling consensus algorithm is used to eliminate the corner points corresponding to the first distance greater than the first preset value, so as to obtain the target corner points. 8. The method according to claim 6, wherein the second obtaining module is also used for: Randomly select four corner points from the checkerboard calibration image, and obtain the coordinates of the four corner points; According to the coordinates of the four corner points and the world coordinates corresponding to the four corner points, a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system is obtained. 9. A computer-readable storage medium, on which a computer program is stored, characterized in that, when the program is executed by a processor, the following steps are implemented: Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points; Using a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point; Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates; Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix; Obtain a target conic model according to the target homography matrix, and obtain initial parameters of the camera according to the target conic model, where the initial parameters include a first internal parameter, a first external parameter, and a first distortion coefficient; The initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated using the target parameters. 10. A computer device comprising a memory, a processor and a computer program stored on the memory and operable on the processor, wherein the processor implements the following steps when executing the program: Obtain a checkerboard calibration image, the calibration image includes a plurality of corner points; Using a random sampling consensus algorithm to eliminate the corner points whose first error value of the corner point is greater than the first preset value among the plurality of corner points, to obtain the target corner point; Obtain a homography matrix from the world coordinate system where the checkerboard calibration image is located to the pixel coordinate system according to the coordinates of the target corner points and the corresponding world coordinates; Using a random sampling consensus algorithm to eliminate homography matrices whose second error value is greater than a second preset value to obtain a target homography matrix; Obtain a target conic model according to the target homography matrix, and obtain initial parameters of the camera according to the target conic model, where the initial parameters include a first internal parameter, a first external parameter, and a first distortion coefficient; The initial parameters are estimated by a maximum likelihood estimation method to obtain target parameters, the target parameters include a second internal parameter, a second external parameter and a second distortion coefficient, and the camera is calibrated using the target parameters.