Camera calibration method and system based on solid circle feature points

Abstract

The invention relates to a camera calibration method and system based on solid circle feature points, and belongs to the field of machine vision. The method comprises the following steps: firstly, acquiring a plurality of images of different poses of a calibration plate, then sorting a plurality of large solid circles on the calibration plate from large to small according to the circle center spacing between the large solid circles, and then calculating a first perspective projection transformation matrix according to the circle center coordinates of the sorted large solid circles and the center coordinates of an ellipse; and obtaining internal and external parameters of the to-be-calibrated camera according to the first projection transformation matrix so as to realize calibration of the to-be-calibrated camera. The feature points are obtained by using the sorted large solid circles, so that the point location quality and the matching accuracy are improved, and the precision of calibration parameters is remarkably improved.

Description

Technical Field

[0001] The present invention relates to a camera calibration method and system based on solid circle feature points, belonging to the field of machine vision. Background Art

[0002] With the rapid development of sensing, optical, computer, computational geometry, and image processing technologies, camera calibration, a key technology in computer vision, has been widely applied in numerous fields. Camera calibration corrects lens-induced image distortion, enabling the camera to capture images as realistically as possible. Furthermore, as a preprocessing step, camera calibration provides support for subsequent 2D image processing and 3D vision tasks, making it highly sought after in industrial and commercial production applications.

[0003] A classic camera calibration system consists of a camera, a computer, a target, a workbench with a fixture, and acquisition software. The camera's lens is a convex lens. Light from an object passes through the lens and is captured by a photosensitive device. The lens, aperture, and shutter control the amount of light entering. Through photoelectric conversion, the optical signal is converted into an electrical signal, which is then transformed into an image.

[0004] However, during camera calibration, the accuracy of feature points affects the accuracy of the camera's intrinsic and extrinsic parameters, thereby affecting the quality of the calibration. Furthermore, camera calibration is subject to nonlinear parameters caused by lens nonlinear distortion, which limits the calibration speed. Traditional camera calibration methods primarily rely on optimization algorithms to solve for calibration parameters, which are highly complex and slow. Summary of the Invention

[0005] The purpose of the present invention is to provide a camera calibration method and system based on solid circle feature points, so as to solve the problem of insufficient accuracy of calibration parameters determined when calibrating a camera.

[0006] To achieve the above object, the solution of the present invention includes:

[0007] A camera calibration method based on solid circle feature points of the present invention includes the following steps: 1) collecting multiple images of different postures of a calibration plate taken by a camera to be calibrated, the calibration plate including solid circles, the solid circles including multiple large solid circles with the same radius and multiple small solid circles with the same radius, and the distances between the centers of adjacent solid circles in the horizontal and vertical directions are equal; 2) collecting the center coordinates of the solid circles in a world coordinate system, sorting the multiple large solid circles, the first rank of the multiple large solid circles after sorting is any one of the multiple large solid circles, and the large solid circles corresponding to other ranks are the large solid circles with the largest center distances with the large solid circles corresponding to the previous rank; 3) collecting the center coordinates of ellipses corresponding to the multiple large solid circles in the image in an image coordinate system; calculating a first perspective projection transformation matrix based on the center coordinates of the sorted multiple large solid circles and the center coordinates of the ellipses corresponding to the sorted multiple large solid circles; 4) obtaining calibration parameters including intrinsic and extrinsic parameters of the camera to be calibrated based on the first perspective projection transformation matrix to achieve calibration of the camera to be calibrated.

[0008] Furthermore, in step 4), the center coordinates of multiple small solid circles in the world coordinate system are mapped to the image according to the first perspective projection transformation matrix, thereby obtaining the center coordinates of the ellipses corresponding to the multiple small solid circles in the image coordinate system; the second perspective projection transformation matrix is ​​calculated according to the center coordinates of the solid circles in the world coordinate system and the center coordinates of the ellipses corresponding to the solid circles in the image coordinate system; according to the second perspective projection transformation matrix, two vanishing points corresponding to each ellipse in two different directions are obtained; for each vanishing point of each ellipse, two tangent points are obtained according to the two tangent lines passing through the vanishing point and located on both sides of the corresponding ellipse and the ellipse equation corresponding to the ellipse, and a line equation passing through the two tangent points is established to obtain two line equations passing through the ellipse, and the two line equations are combined to obtain the intersection coordinates, and the intersection coordinates are used as the feature point coordinates; the intrinsic and extrinsic parameters of the camera to be calibrated are calculated according to the coordinates of all feature points to realize the calibration of the camera to be calibrated.

[0009] Furthermore, in step 3), a greedy algorithm is used to sort the multiple large solid circles.

[0010] Furthermore, in step 3), ellipses are screened out from the image based on the circularity of the graphics in the image. The screening condition is whether the circularity corresponding to the graphics is greater than the set circularity threshold. If it is greater than the circularity threshold, the graphics is determined to be an ellipse.

[0011] Furthermore, in step 3), the collected center coordinates of the ellipse are subjected to sub-pixel processing, and then a first perspective projection transformation matrix is ​​calculated based on the center coordinates after the sub-pixel processing.

[0012] Furthermore, after the intrinsic and extrinsic parameters of the camera to be calibrated are obtained, the elements in the intrinsic and extrinsic parameters that have not yet been obtained are assigned corresponding set coefficients to achieve calibration of the camera to be calibrated.

[0013] Furthermore, the coefficient setting process includes: setting the corresponding step size, obtaining the variation range of the element through the step size, determining the optimal value within the variation range, updating the calibration parameters according to the optimal value, calculating the predicted center coordinates of the solid circle in the image coordinate system according to the updated calibration parameters, and subtracting the predicted center coordinates from the true center coordinates of the corresponding solid circle in the image coordinate system calculated according to the calibration parameters before the update. When the difference is less than the set threshold, the coefficient is obtained according to the optimal value and the element.

[0014] Furthermore, the calibration plate includes 5 large solid circles and 94 small solid circles, and is arranged in 9 rows and 11 columns. The diameter of the large solid circle is smaller than the center distance between adjacent solid circles. There are no more than 2 large solid circles in each row or column. The color of the solid circles is black, and the surface color of the calibration plate is white.

[0015] A camera calibration system based on solid circle feature points includes a processor that executes a computer program to implement the steps of the above method.

[0016] The beneficial effects of the present invention are as follows: the present invention is a pioneering invention. The present invention first collects images of a plurality of different postures of the calibration plate, and collects the center coordinates of the ellipse corresponding to the large solid circle in the image, and then sorts the plurality of large solid circles on the calibration plate from large to small according to the size of the center distance between the large solid circles, and then calculates the first perspective projection transformation matrix based on the center coordinates of the sorted large solid circles and the center coordinates of the ellipse corresponding to the large solid circle, and then obtains the internal and external parameters of the calibration camera based on the first projection transformation matrix to realize the calibration of the camera to be calibrated. Since the large solid circles are sorted according to the center distance, it is ensured that the distance between the center coordinates of the ellipse selected in sequence after imaging and the center coordinates of other ellipses is the local maximum, thereby improving the point quality and matching accuracy, and thus improving the accuracy of the calibration parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the camera calibration system architecture of the present invention;

[0018] Figure 2 It is a schematic diagram of a calibration plate of the present invention;

[0019] Figure 3 It is a schematic diagram of a camera calibration process of the present invention;

[0020] Figure 4 It is a schematic diagram of a calibration plate with markings according to the present invention;

[0021] Figure 5 This is a schematic diagram of imaging of a large solid circle before parameter calibration of the present invention;

[0022] Figure 6 This is a schematic diagram of imaging after solid circle parameter calibration of the present invention;

[0023] Figure 7 It is a perspective projection diagram of a calibration plate of the present invention;

[0024] Figure 8 This is a schematic diagram of a flow chart for solving calibration parameters of the present invention. DETAILED DESCRIPTION

[0025] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and implementation methods.

[0026] The idea of ​​the present invention is that, in camera calibration, the selected feature points may be unevenly distributed or have weak responses, thereby affecting the matching stability. That is, in the process of calculating the intrinsic and extrinsic parameters of the camera based on the feature points selected in sequence, the spacing between the two adjacent feature points selected will affect the accuracy of the intrinsic and extrinsic parameters. Therefore, before determining the feature points, the present invention sorts the center spacing between the solid circles selected on the calibration plate, extracts feature points with strong responses and uniform distribution, so each selected solid circle is sorted according to the current local maximum, thereby ensuring that the spacing between the center coordinates of the ellipse selected in sequence after imaging and the center coordinates of other ellipses is the local maximum, which can improve the point quality and matching accuracy, thereby improving the accuracy of the feature points finally determined, and further improving the accuracy of the calibration parameters.

[0027] Method implementation method:

[0028] This embodiment provides a camera calibration method based on solid circle feature points, such as Figure 1 As shown in FIG, it includes a calibration plate, a camera to be calibrated, and a computer running a camera acquisition software system. Among them, the black background circular target as the calibration plate is as follows Figure 2 As shown in the figure, the target includes 5 reference solid circles (large solid circles) with the same radius and 94 standard solid circles (small solid circles) with the same radius. These 99 solid circles are arranged in 9 rows and 11 columns. The center distance between adjacent solid circles is d. The radius of the standard solid circle is smaller than the radius of the reference solid circle. The diameter of the reference solid circle is smaller than d. There are no more than two reference solid circles in each row or column, and the solid circles are black, while the target surface is white.

[0029] like Figure 3As shown, the camera is fixed to the workbench with the lens facing downward, and the target is placed horizontally on the workbench. The target position is adjusted so that the solid circle in the effective area on the target is within the camera's visual range. The camera then captures and saves the corresponding image. The target position is changed multiple times to obtain multiple target images. The visible range is the area enclosed by rows 1 to 9 and columns 1 to 11 of the target. This area is defined as the visible boundary of the target. The effective area of ​​the target is the distance from the visible boundary to the center distance of the circle.

[0030] The preprocessing module performs image denoising, filtering, image segmentation and other processing on the images captured by the camera, extracts edge contours, and selects valid contours. Since the solid circle on the target degenerates into an ellipse after being photographed by the camera to be calibrated, the effective edge contour is fitted with an ellipse to obtain the center coordinates of the ellipse. The process of selecting valid contours includes: on the target plane, the shape and size of the solid circle are fixed, and the target is photographed within a fixed distance range. The range of the contour circumference is P∈(P min ,P max ), P min is the minimum perimeter of the contour (the perimeter of the small solid circle), P max is the maximum perimeter of the contour (the perimeter of the large solid circle); the range of the contour area is A∈(A min ,A max ), where A min is the minimum area of ​​the contour (the area of ​​the small solid circle), A max is the maximum area of ​​the contour (the area of ​​the large solid circle). Then, define The circularity C is used to describe the degree of circular annular degradation after solid circle imaging. When C = 1, the contour is a perfect circle, while when C = 0, it degenerates into an irregular shape. Therefore, when the circularity of the contour is greater than a set circularity threshold (e.g., 0), it is determined to be an ellipse, and the center coordinates of the ellipse corresponding to the contour are collected.

[0031] The feature extraction module also performs sub-pixel processing on the center coordinates of the ellipse to obtain more accurate center coordinates of the ellipse. The sub-pixel processing process includes:

[0032] The difference vector between the center coordinate of the ellipse and the center coordinate of the sub-pixelized ellipse is calculated, and then the sub-pixelization of the center coordinate of the ellipse is achieved by using the orthogonal relationship between the difference vector and the gradient vector within a 3*3 range of the center coordinate of the ellipse.

[0033] Assume that q0 is the center coordinate of an ellipse. In order to obtain the corresponding sub-pixel coordinate q, select the integer coordinate closest to q0, select a 3*3 window with it as the center, and calculate the integer coordinate p in the window.i With the gradient G in this area i , if p i Located on the edge of the black circular target, G i Orthogonal to p i -q; if p i Not on the edge, then p i The pixel values ​​around the coordinates are the same and the gradient value is 0. For the selected single pixel point p i , subject to:

[0034] G i ·(p i -q)=0 (1)

[0035] Then, the pseudo-inverse is calculated to obtain the new sub-pixel coordinate q1, which is expressed as:

[0036]

[0037] Then, q1 is used to iterate continuously to obtain a more accurate sub-pixel coordinate q. In order to avoid introducing more errors from other pixels, a threshold ε is specified as the iteration termination condition, that is:

[0038] ||q t -q t-1 ||2≤ε (3)

[0039] When the difference between two adjacent sub-pixel coordinates is less than a specified threshold, the iteration is terminated and the final sub-pixel center coordinate is obtained.

[0040] Mark and sort the five solid reference circles on the target, and further improve the accuracy of the feature points by numbering them. This process includes:

[0041] A world coordinate system is established for the target, with the X axis along the horizontal direction of the target, the Y axis verifying the vertical direction of the target, and the Z axis perpendicular to the XOY plane. Figure 4 As shown, the Z axis is taken as 0, the center coordinates of the reference solid circle A are (2d, 4d, 0), the center coordinates of the reference solid circle B are (8d, 4d, 0), the center coordinates of the reference solid circle V are (4d, 6d, 0), the center coordinates of the reference solid circle D are (5d, 6d, 0), and the center coordinates of the reference solid circle E are (5d, 2d, 0). Then, the five reference solid circles are sorted using the set sorting rule. After the sorting is completed using the sorting rule, the first rank of the multiple reference solid circles is any one of the multiple reference solid circles, and the reference solid circles corresponding to the other ranks are the large solid circles with the largest center distance from the reference solid circle corresponding to the previous rank. As a specific embodiment of the present invention, the sorting rule is a greedy algorithm. The sorting process of the five reference solid circles using the greedy algorithm is explained with a specific example:

[0042] Choose one of the five reference solid circles (for example, reference solid circle A) as the starting visit point and mark it as I. Calculate the distance from the center of this starting visit point to the centers of the other four reference solid circles (which have not been visited). Since the distance from the center of reference solid circle A to the center of reference solid circle B is the largest, mark reference solid circle B as II. Then, using reference solid circle B as the starting visit point, calculate the distance from the center of reference solid circle B to the centers of the other three circles (reference solid circles C, D, and E). Since the distance from the center of reference solid circle B to the center of reference solid circle C is the largest, mark reference solid circle C as III. Similarly, mark reference solid circle D as IV and reference solid circle E as V. Finally, the sorted order is {I, II, III, IV, V}.

[0043] Then use the center coordinates of the five sorted concentric circles and the center coordinates of the five ellipses corresponding to the sorted concentric circles, such as Figure 7 As shown, a perspective projection transformation relationship is established to calculate the perspective projection transformation matrix H. The center coordinates of the ellipse are located in the image coordinate system. A specific example will illustrate this process: from the center coordinates of the five sorted reference concentric circles, the center coordinates of four consecutive reference concentric circles and the center coordinates of the four ellipses corresponding to these four reference concentric circles are selected. At this point, selecting the center coordinates of the four consecutive reference concentric circles will result in two sets of values: I, II, III, IV and II, III, IV, V. These two sets of values ​​also correspond to two sets of center coordinates of four ellipses, resulting in two pairs of perspective projection change relationships. Then, based on these two pairs of perspective projection change relationships, two perspective projection transformation matrices are calculated, and one is selected as the first perspective projection transformation matrix. The above process is also applicable to selecting three consecutive and five consecutive center coordinates. The principles are the same, so it will not be described in detail here.

[0044] Since the first perspective projection transformation matrix contains the intrinsic and extrinsic parameter matrices, the intrinsic and extrinsic parameters of the camera to be calibrated can be obtained through the first perspective projection transformation matrix, thereby realizing the calibration of the camera.

[0045] Furthermore, in order to improve the accuracy of the internal and external parameters of the camera to be calibrated. The center coordinates of the other 94 small solid circles are also mapped to the image according to the first perspective projection transformation matrix, so as to obtain the center coordinates of the ellipse corresponding to the 94 small solid circles in the image coordinate system. At this time, the center coordinates of 99 solid circles in the world coordinate system and the center coordinates of the ellipse corresponding to these 99 solid circles in the image coordinate system can be obtained. Then, 5 center coordinates are randomly selected from the center coordinates of the 99 solid circles each time, and the center coordinates corresponding to these 5 are selected from the center coordinates of the 99 ellipses, and then the perspective projection transformation matrix at this time is calculated. Since there are multiple groups of choices for randomly selecting 5 center coordinates from the center coordinates of the 99 solid circles, there are also multiple perspective projection transformation matrices calculated, and one is selected from these multiple perspective projection transformation matrices as the second perspective projection transformation matrix. Because the second perspective projection transformation matrix is ​​selected from multiple projection transformation matrices, it can more accurately reflect the mapping relationship between the target plane and the image plane compared to the first perspective projection transformation matrix. The above process is also applicable to the selection of three consecutive and four consecutive circle center coordinates. The principle is the same, so it will not be repeated here.

[0046] According to the second perspective projection transformation matrix H', calculate the coordinates of the center of the ellipse. Let the general equation of the ellipse be: Ax 2 +Bxy+Cy 2 +Dx+Ey+F=0, assuming that there is a vanishing point V1 (h x ,h y ), the straight line along the vanishing point is tangent to both sides of the ellipse, thus obtaining two tangent points (x, y). Assume that the slopes of the lines between the vanishing point and the two tangent points are k1 and k2 respectively (if the vanishing point is far enough, then k1 = k2). Substituting the tangent point into the general equation of the ellipse, the simultaneous solution of the straight line equations forms a quadratic system of two variables:

[0047]

[0048] After estimation, the coordinates of the tangent points A1′(x1,y1) and A′2(x2,y2) can be obtained.

[0049] Similarly, a set of tangent point coordinates in another direction (for example, vertical direction) of the current ellipse can be solved, and then two sets of tangent point coordinates B1′(x3,y3) and B2′(x4,y4) on the ellipse can be obtained. The process of solving the tangent point coordinates of the remaining ellipses is the same as described above.

[0050] Assume that A′1(x1,y1), A′2(x2,y2), B′1(x3,y3), B′2(x4,y4) are the coordinates of the tangent points of each ellipse contour, establish the general equations of A′1A′2, B′1B′2, and solve the coordinates of the central image point of the ellipse (x c ,y c ):

[0051]

[0052] The coordinates of the central image point are then used as a feature point coordinate. The coordinates of the central image point of all ellipses can be solved according to the same process as above, thereby obtaining the coordinates of the feature points corresponding to all ellipses. The camera perspective projection module uses the pinhole imaging model and the perspective transformation relationship to establish the projection equation for all extracted feature points. The solid circle target is projected from the world coordinate system to the image coordinate system, which is affected by radial distortion and tangential distortion. The projection equation is:

[0053]

[0054] Among them, δ r is the radial distortion parameter, δ d is the tangential distortion parameter.

[0055]

[0056] in,

[0057] Let R and T be the identity matrix and further simplify the projection equation:

[0058]

[0059] Among them, f u , f v are the scale factors in the u-axis and v-axis directions respectively; du, dv are the physical sizes of the pixels in the u-axis and v-axis directions respectively; u0 and v0 are the coordinates of the imaging center respectively; K(f u ,f v ,u0,v0,k1,k2,p1,p2) is the camera's intrinsic parameter matrix; R is the camera's rotation matrix; t is the camera's position translation matrix; [R|t] is the camera's extrinsic parameter matrix; and f is the focal length.

[0060] Calibration modules, such as Figure 8 As shown in the figure, the projection equation is solved according to all the feature points obtained, and then the intrinsic parameters, extrinsic parameters and distortion coefficients of the camera are determined, and finally the camera is quickly calibrated.

[0061] First, suppose that in n-dimensional space, vector X t =(x1,x2,x3,…,xt ). During the calibration process, some interference may occur (for example, the workbench is not level), which may lead to inaccurate calculated camera parameters. Therefore, the present invention artificially adds a disturbance of μ i , add disturbance to the calibration parameters as follows:

[0062] d=(μ1f x ,μ2f y ,μ3μ0,μ4v0,μ5k1,μ6k2,μ7p1,μ8p2,μ9Rt) (11)

[0063] Among them, f x and f y f u and f v The coordinate values ​​converted from the pixel coordinate system to the image coordinate system, u0 and v0 are the principal point coordinates, k1 and k2 are the radial distortion parameters, and p1 and p2 are the tangential distortion parameters.

[0064] To solve the calibration parameters after adding perturbations, the perturbation search algorithm features low computational complexity, high speed, and high accuracy, particularly effective for optimizing high-dimensional vectors. Perturbation search works by making small perturbations to the intrinsic and extrinsic parameters, as well as the distortion coefficients, based on the calibration parameters of the camera being measured. It then gradually evaluates the changes in the reprojection error after the perturbations and retains the parameter updates that minimize the error, thereby achieving local optimization of the obtained calibration parameters.

[0065] In order to improve the calibration accuracy, the calibration parameters are the 9 parameters in formula (11), μ i =Uniform(-μ,μ),1≤i≤9, the optimization process settings include the following steps:

[0066] Step S1: Set the disturbance parameters and define the disturbance step length δ. The range of the disturbance step length is: δ = 10 -6 ~10 -3 .

[0067] Step S2: Perturbation search, with internal parameter f x , f y , u0, v0 as an example, assuming that each element of the internal parameter is recorded as K ij , then, the perturbation search process is as follows:

[0068]

[0069] By for each element K ij After adding or subtracting the perturbation step, the corresponding perturbation range (variation range) is obtained, and an optimal value is found within the range. Then, the optimal value (for example, f xThe corresponding optimal value f x ′) Calculate the reprojection error after perturbation:

[0070]

[0071] in, are the predicted center coordinates of the ellipse.

[0072] Use formula (13) to recalculate the center coordinates (predicted center coordinates) of the ellipse corresponding to a solid circle in the image coordinate system, and then compare the recalculated predicted center coordinates of the ellipse with the original true center coordinates of the ellipse:

[0073]

[0074] in, The original true center coordinates of the ellipse are the center coordinates (true center coordinates) of the ellipse corresponding to the solid circle in the image coordinate system, obtained using the calibration parameters before the perturbation is added. When E is less than a set threshold (for example, 0.005), the perturbation update is accepted, and the obtained optimal value is assigned to the corresponding calibration parameter. Otherwise, the original calibration parameter remains unchanged.

[0075] Step S3: Repeat the perturbation process for the distortion parameters and extrinsic parameters. The distortion parameters and extrinsic parameters are denoted as D = (k1, k2, p1, p2, Rt). As in step 2, the perturbation search is performed element by element.

[0076] Step S4: Iteration and convergence. Repeat steps 2 and 3 until the total error decreases by less than the set total threshold (for example, 0.008), or the number of iterations reaches 50. The search process is terminated, and the coefficient corresponding to each element is obtained based on the obtained optimal value and the original value of the element, and the optimized coefficients μ1 to μ9 are output. The calculation formula for the total error is:

[0077]

[0078] Where N is the total number of elements.

[0079] The image of the large solid circle captured by the camera before calibration is as follows: Figure 5 As shown, after the camera is calibrated using the present invention, the large solid circle is photographed and the resulting image is as follows: Figure 6 shown.

[0080] System implementation method:

[0081] This embodiment provides a camera calibration system based on solid circle feature points. The computer program executed by the processor in the system is designed using the camera calibration method based on solid circle feature points as described in the method implementation. Since the introduction of this method is clear enough, it will not be repeated here.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific embodiments of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A camera calibration method based on solid circle feature points, characterized in that: The method comprises the following steps: 1) acquiring images of a calibration plate in different postures taken by a camera to be calibrated, wherein the calibration plate comprises solid circles, and the solid circles comprise multiple large solid circles with the same radius and multiple small solid circles with the same radius, and the distances between the centers of adjacent solid circles in the horizontal and vertical directions are equal; 2) acquiring the center coordinates of the solid circles in a world coordinate system, and sorting the multiple large solid circles, wherein the first order of the multiple large solid circles after sorting is any one of the multiple large solid circles, and the large solid circles corresponding to the other orders are the large solid circles with the largest center distances with the large solid circles corresponding to the previous order; 3) acquiring the center coordinates of the ellipses corresponding to the multiple large solid circles in the image in an image coordinate system; and calculating a first perspective projection transformation matrix according to the center coordinates of the multiple large solid circles after sorting and the center coordinates of the ellipses corresponding to the multiple large solid circles after sorting; and 4) obtaining calibration parameters including the intrinsic and extrinsic parameters of the camera to be calibrated according to the first perspective projection transformation matrix, so as to realize calibration of the camera to be calibrated.

2. The camera calibration method based on solid circle feature points according to claim 1, characterized in that: In the step 4), the center coordinates of multiple small solid circles in the world coordinate system are mapped to the image according to the first perspective projection transformation matrix, thereby obtaining the center coordinates of the ellipses corresponding to the multiple small solid circles in the image coordinate system; the second perspective projection transformation matrix is ​​calculated according to the center coordinates of the solid circles in the world coordinate system and the center coordinates of the ellipses corresponding to the solid circles in the image coordinate system; according to the second perspective projection transformation matrix, two vanishing points corresponding to each ellipse in two different directions are obtained; for each vanishing point of each ellipse, two tangent points are obtained according to two tangent lines passing through the vanishing point and located on both sides of the corresponding ellipse and the ellipse equation corresponding to the ellipse, and a line equation passing through the two tangent points is established to obtain two line equations passing through the ellipse, and the two line equations are combined to obtain the intersection coordinates, and the intersection coordinates are used as the feature point coordinates; the intrinsic and extrinsic parameters of the camera to be calibrated are calculated according to the coordinates of all feature points to realize the calibration of the camera to be calibrated.

3. The camera calibration method based on solid circle feature points according to claim 1, characterized in that: In step 3), a greedy algorithm is used to sort the multiple large solid circles.

4. The camera calibration method based on solid circle feature points according to claim 1, characterized in that: In the step 3), ellipses are also screened out from the image based on the circularity of the graphics in the image. The screening condition is whether the circularity corresponding to the graphics is greater than the set circularity threshold. If it is greater than the circularity threshold, the graphics is determined to be an ellipse.

5. The camera calibration method based on solid circle feature points according to claim 1, characterized in that: In the step 3), the collected center coordinates of the ellipse are further subjected to sub-pixel processing, and then a first perspective projection transformation matrix is ​​calculated based on the center coordinates after the sub-pixel processing.

6. The camera calibration method based on solid circle feature points according to claim 1 or 2, characterized in that: After obtaining the intrinsic and extrinsic parameters of the camera to be calibrated, corresponding set coefficients are allocated to the elements in the obtained intrinsic and extrinsic parameters to achieve calibration of the camera to be calibrated.

7. The camera calibration method based on solid circle feature points according to claim 6, characterized in that: The coefficient setting process includes: setting a corresponding step size, obtaining a variation range of the element through the step size, determining an optimal value within the variation range, updating a calibration parameter according to the optimal value, calculating a predicted center coordinate of a solid circle in an image coordinate system according to the updated calibration parameter, subtracting the predicted center coordinate from a true center coordinate of a corresponding solid circle in an image coordinate system calculated according to the calibration parameter before the update, and when the difference is less than a set threshold, obtaining a coefficient according to the optimal value and the element.

8. The camera calibration method based on solid circle feature points according to claim 1, characterized in that: The calibration plate includes 5 large solid circles and 94 small solid circles, and is arranged in 9 rows and 11 columns. The diameter of the large solid circle is smaller than the center distance between adjacent solid circles. There are no more than 2 large solid circles in each row or column. The color of the solid circles is black, and the surface color of the calibration plate is white.

9. A camera calibration system based on solid circle feature points, comprising a processor, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 8.

Images