Calibration method for camera parameter calibration

Abstract

The present invention relates to the technical field of camera parameter calibration, and particularly relates to a calibration board and a calibration method for camera parameter calibration, including a substrate, on the surface of which black and white squares are arranged in an array; first identification codes are arranged in the white squares of the odd rows on the substrate, and second identification codes are arranged in the white squares of the even rows on the substrate; the number of the first identification codes in the white squares of the odd rows on the substrate is exactly the same along the row direction, and the number of the second identification codes in the white squares of the even rows on the substrate increases sequentially along the row direction; the number of the first identification codes in the white squares of the odd columns on the substrate increases sequentially along the column direction, and the number of the second identification codes in the white squares of the even columns on the substrate is exactly the same along the column direction; the present invention provides a calibration board and a calibration method for camera parameter calibration with high accuracy of camera parameter calibration, simple coding and decoding principles, and allowing partial occlusion of the calibration board.

Description

Technical Field

[0001] The present invention relates to the technical field of camera parameter calibration, and particularly to a calibration method for camera parameter calibration. Background Art

[0002] Machine vision has wide applications in fields such as measurement and detection. As the main image acquisition execution component, the calibration of the internal and external parameters of a camera has always been the most important basic work in the field of machine vision applications. The accuracy of camera parameter calibration will directly affect the accuracy of subsequent measurement, detection, etc. Therefore, how to improve the calibration accuracy of the internal and external parameters of a camera has always been the focus of research in the industry.

[0003] Generally, different camera calibration methods need to rely on different types of calibration plates. Currently, the commonly used planar calibration plates for camera calibration mainly include checkerboard calibration plates (as shown in Figure 1 ), ChArUco calibration plates (as shown in Figure 2 ), and other various calibration plates. For the commonly used checkerboard calibration plate, it cannot provide a unique number for each detected corner point used for camera calibration, which easily leads to the problem of incorrect matching between the corner points extracted from the pictures taken by the camera and their physical size coordinates in the world coordinate system, resulting in relatively low accuracy of the parameters calibrated by the camera. In addition, the checkerboard calibration plate does not allow any occlusion. If there is partial occlusion, the pixel coordinates of the corner points cannot be accurately obtained, so that the corner points cannot be accurately matched with their corresponding physical sizes, which brings inconvenience to the calibration of the camera. For the ChArUco calibration plate, it overcomes the deficiencies of the checkerboard calibration plate to a certain extent. However, when using the ChArUco calibration plate for camera calibration, it is necessary to accurately identify the ArUco codes in each square of the calibration plate. When the similarity between two ArUco codes is relatively high or the resolution of the captured ArUco codes is low due to the excessive rotation angle of the ChArUco calibration plate, it is easy to cause decoding errors of the ArUco codes, and then lead to the problem of incorrect matching between the corner points extracted from the pictures taken by the camera and their physical size coordinates in the world coordinate system, resulting in relatively low accuracy of the parameters calibrated by the camera. In addition, the encoding and decoding processes of the ChArUco calibration plate are complex. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: to overcome the deficiencies in the prior art and provide a calibration method for camera parameter calibration with high accuracy, simple encoding and decoding principles, and allowing partial occlusion of the calibration plate.

[0005] The technical solution adopted by the present invention to solve its technical problems is: a calibration board for camera parameter calibration, including a substrate, on the surface of which black and white squares are arranged in an array; first identification codes are arranged in the white squares of the odd rows on the substrate, and second identification codes are arranged in the white squares of the even rows on the substrate; the number of the first identification codes in the white squares of the odd rows on the substrate is exactly the same along the row direction, and the number of the second identification codes in the white squares of the even rows on the substrate increases sequentially along the row direction; the number of the first identification codes in the white squares of the odd columns on the substrate increases sequentially along the column direction, and the number of the second identification codes in the white squares of the even columns on the substrate is exactly the same along the column direction.

[0006] Further, the area of the black region of the first identification code is equal to times the area of the black region of the second identification code, where , is an integer.

[0007] Further, the first identification code is circular, and the second identification code is annular.

[0008] Further, the outer diameter of the first identification code is equal to the outer diameter of the second identification code .

[0009] Further, the sizes of the white squares and the black squares can be freely set.

[0010] Further, the number of the second identification codes in the white squares of the even rows on the substrate increases sequentially by 1 along the row direction, and the number of the first identification codes in the white squares of the odd columns on the substrate increases sequentially by 1 along the column direction.

[0011] A calibration method for camera parameter calibration, using the above calibration board for camera parameter calibration, includes the following steps:

[0012] S1. Use a camera to take a picture of the calibration board, and then use the Shi-Tomasi algorithm to identify all the corner points in the picture;

[0013] S2. Taking each corner point as the center, extract the two white squares connected to it as the decoding area of the current corner point number, and obtain the number uniquely corresponding to the current corner point by decoding the identification code in the decoding area;

[0014] S3. According to the number corresponding to each corner point, accurately assign the physical size coordinates corresponding to it in the world coordinate system to each corner point, and then use the Zhang-Zhengyou calibration method to complete the camera parameter calibration.

[0015] Further, the steps of decoding the identification code in the decoding area in step S2 are as follows:

[0016] S21. Identification and quantity calculation of identification codes: Extract one identification code from each of the two white squares in the decoding area, calculate the pixel area of the black areas of the two identification codes, identify the first identification code and the second identification code according to the size of the pixel area of the black areas, then count the number of all enclosed black areas within the white square where the first identification code is located, and denote it as , count the number of all enclosed black areas within the white square where the second identification code is located, and denote it as ;

[0017] S22. Decoding of corner point numbers: Extract the pixel coordinates of the center points of any one first identification code and any one second identification code in the decoding area in the picture, and implement the decoding of the current corner point number through the decoding rules. Among them, the corner point number includes the corner point row number and the corner point column number .

[0018] Furthermore, the decoding rules in step S22 are as follows:

[0019] The decoding rule of the corner point row number is as follows:

[0020] When , then the row number of the current corner point number ;

[0021] When , then the row number of the current corner point number ;

[0022] The decoding rule of the corner point column number is as follows:

[0023] When and , then the column number of the current corner point number ;

[0024] When and , then the column number of the current corner point number ;

[0025] When and , then the column number of the current corner point number ;

[0026] When and , then the column number of the current corner point number ;

[0027] Among them, represents the pixel coordinates of the center point of any one first identification code; Represents the pixel coordinates of the center point of any second identification code.

[0028] Further, the specific process of step S1 is as follows:

[0029] Set a window and a preset threshold ;

[0030] Assume that the pixel coordinates of a certain point in the window are , the displacement is , and the grayscale is , then the grayscale change function is:

[0031]

[0032] Among them, the window function ;

[0033] Expand the grayscale change function by Taylor series and omit the high-order terms to obtain:

[0034]

[0035] Among them, , , ;

[0036] In the formula, , respectively represent the gradient values of the image grayscale in the , directions;

[0037] Define the corner response function as:

[0038]

[0039] In the formula, and are the two eigenvalues of the matrix ;

[0040] When the window traverses all pixel points in the picture, if the value of a certain pixel point is greater than the threshold and is a local maximum in the neighborhood of this pixel point, then this pixel point is a Shi-Tomasi feature corner point, and output the coordinates of this pixel point.

[0041] The beneficial effects of the present invention are:

[0042] (1) The present invention extracts the decoding area of the corner point numbers, then decodes the identification codes in the decoding area to obtain the numbers uniquely corresponding to the corner points. Furthermore, through the numbers, the physical size coordinates corresponding to the corner points in the world coordinate system can always be accurately matched, thus completely avoiding the problem of mis-matching, improving the accuracy and stability of camera parameter calibration. At the same time, it also allows the calibration board to be partially occluded, thereby improving the convenience in the process of parameter calibration.

[0043] (2) On the substrate of the present invention, the first identification codes are arranged in the white squares of the odd rows. The number of the first identification codes in the white squares of the odd rows is exactly the same along the row direction, and the number of the first identification codes in the white squares of the odd columns increases sequentially along the column direction. At the same time, the second identification codes are arranged in the white squares of the even rows. The number of the second identification codes in the white squares of the even rows increases sequentially along the row direction, and the number of the second identification codes in the white squares of the even columns is exactly the same along the column direction. With such a design, the layout of the entire calibration board is beautiful, the utilization rate can be improved, and at the same time, the distinguishability of the identification codes for corner point numbers is large and easy to identify. Moreover, the identification codes for corner point numbers have a simple structure, are easy to encode and decode, and reduce the usage threshold.

[0044] (3) The first identification code of the present invention is circular, and the second identification code is annular. The distinguishability between the two is further enlarged, making it easier to identify, reducing the resolution requirement for the captured calibration board images, and ensuring that the decoding is always correct. In addition, the setting of the circular identification code and the annular identification code makes the identification code itself have no angular features, avoiding the possibility of introducing additional corner points. Furthermore, it ensures the accuracy rate of the matching between the detected corner points and their corresponding physical size coordinates in the world coordinate system, thus greatly improving the accuracy of camera parameter calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The present invention will be further described below in conjunction with the drawings and embodiments.

[0046] Figure 1 is a schematic diagram of a checkerboard calibration board;

[0047] Figure 2 is a schematic diagram of a ChArUco calibration board;

[0048] Figure 3 is a schematic diagram of the calibration board in the present invention;

[0049] Figure 4 is a schematic diagram of the identification code in the present invention;

[0050] Figure 5 is a flowchart of the calibration method in the present invention;

[0051] Figure 6 is a schematic diagram of the corner point in the present invention;

[0052] Figure 7 It is a schematic diagram of the current corner point decoding area in the present invention. Detailed implementation manners

[0053] Now, the present invention will be further described in conjunction with the accompanying drawings and preferred embodiments. These drawings are all simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic way, so they only show the components related to the present invention. Embodiment

[0054] As Figure 3 and Figure 4 shown, a calibration board for camera parameter calibration includes a substrate, and black and white square grids are arrayed on the surface of the substrate; first identification codes are arranged in the white square grids of the odd rows on the substrate, and second identification codes are arranged in the white square grids of the even rows on the substrate; the number of the first identification codes in the white square grids of the odd rows on the substrate is exactly the same along the row direction, and the number of the second identification codes in the white square grids of the even rows on the substrate increases sequentially along the row direction; the number of the first identification codes in the white square grids of the odd columns on the substrate increases sequentially along the column direction, and the number of the second identification codes in the white square grids of the even columns on the substrate is exactly the same along the column direction.

[0055] Specifically, the sizes of the white square grids and the black square grids can be freely set. Usually, the sizes of the white square grids and the black square grids are exactly the same, but this is not limited thereto, and they can also be completely different; the first identification code is circular, and the second identification code is annular, further expanding the distinction between the two, making it easier to identify, reducing the resolution requirement for the captured calibration board picture, ensuring that the subsequent decoding is always correct. At the same time, the setting of the circular identification code and the annular identification code makes the identification code itself have no corner features, avoiding the possibility of introducing additional corner points, and then ensuring the accuracy rate of the matching between the detected corner points and their corresponding physical size coordinates in the world coordinate system, thereby greatly improving the accuracy of camera parameter calibration; the outer diameter of the first identification code is equal to the outer diameter of the second identification code , and the outer diameter of the first identification code, the outer diameter of the second identification code and the inner diameter can all be freely defined according to the side length of the white square grid; the number of the second identification codes in the white square grids of the even rows on the substrate increases by 1 sequentially along the row direction, and the number of the first identification codes in the white square grids of the odd columns on the substrate increases by 1 sequentially along the column direction.

[0056] The first identification code is set in the white squares of the odd rows on the substrate. The number of the first identification codes in the white squares of the odd rows is exactly the same along the row direction, and the number of the first identification codes in the white squares of the odd columns increases sequentially along the column direction. At the same time, the second identification code is set in the white squares of the even rows on the substrate. The number of the second identification codes in the white squares of the even rows increases sequentially along the row direction, and the number of the second identification codes in the white squares of the even columns is exactly the same along the column direction. With such a design, the layout of the entire calibration plate is beautiful, the utilization rate can be improved, and at the same time, the discrimination degree of the identification codes for corner numbering is large and easy to identify. Moreover, the identification codes for corner numbering have a simple structure, are easy to encode and decode, and reduce the usage threshold.

[0057] The area of the black region of the first identification code is equal to times the area of the black region of the second identification code, where , is an integer. With such a design, it is convenient to identify the first identification code and the second identification code. Embodiment

[0058] As Figure 5 shown, a calibration method for camera parameters uses the calibration plate for camera parameter calibration in Embodiment 1, and includes the following steps:

[0059] S1. Use a camera to take a picture of the calibration plate, and then use the Shi-Tomasi algorithm (corner detection algorithm) to identify all the corners in the picture. The specific process is as follows:

[0060] Set a window and preset a threshold ;

[0061] Assume that the pixel coordinates of a point in the window are , the movement amount is , and the gray level is , then the gray level change function is:

[0062]

[0063] Among them, the window function ;

[0064] Expand the gray level change function by Taylor series and omit the high-order terms to obtain:

[0065]

[0066] Among them, , , ;

[0067] In the formula, , respectively represent the gradient values of the image gray level in the , directions;

[0068] Define the corner response function as:

[0069]

[0070] In the formula, and are the two eigenvalues of the matrix ;

[0071] The window traverses all pixel points in the picture. When the value of a certain pixel point is greater than the threshold and it is a local maximum in the neighborhood of this point, then this pixel point is a Shi-Tomasi feature corner point, and the coordinates of this pixel point are output.

[0072] Using the Shi-Tomasi algorithm to identify the corner points in the picture can quickly identify all the corner points in the picture and improve the accuracy of corner point identification, thus ensuring that the identified corner points are the best.

[0073] As Figure 6 shown, the " " in the figure are all the detected corner points, is the pixel coordinate system of the camera shooting the calibration board picture.

[0074] S2. Taking each corner point as the center, extract the two white squares connected to it as the decoding area of the current corner point number. By decoding the identification code in the decoding area, obtain the number uniquely corresponding to the current corner point.

[0075] S3. According to the number corresponding to each corner point, accurately assign the physical size coordinates corresponding to it in the world coordinate system to each corner point, and then use the Zhang Zhengyou calibration method to complete the camera parameter calibration. Specifically, using the Zhang Zhengyou calibration method to calibrate the camera parameters belongs to the prior art and will not be described in detail here.

[0076] By extracting the decoding area of the corner point number, then decoding the identification code in the decoding area, obtaining the number uniquely corresponding to this corner point, and further being able to always accurately match the physical size coordinates corresponding to this corner point in the world coordinate system through the number, thus completely avoiding the problem of mis-matching, improving the accuracy and stability of camera parameter calibration; at the same time, it also enables the calibration board to be partially blocked, thereby improving the convenience in the process of parameter calibration.

[0077] The steps of decoding the identification code in the decoding area in step S2 are as follows:

[0078] S21. Identification and Quantity Calculation of Identification Codes: Extract one identification code from each of the two white squares in the decoding area, calculate the pixel area of the black areas of the two identification codes, identify the first identification code and the second identification code according to the size of the pixel area of the black areas, then count the number of all enclosed black areas within the white square where the first identification code is located, and denote it as , count the number of all enclosed black areas within the white square where the second identification code is located, and denote it as . Among them, is the quantity of the first identification code; is the quantity of the second identification code.

[0079] It should be noted that the pixel area of the black area of the first identification code must be much larger than that of the black area of the second identification code. Therefore, the one with a larger pixel area of the black area is the first identification code, and the one with a smaller pixel area of the black area is the second identification code.

[0080] S22. Decoding of Corner Point Numbers: Extract the pixel coordinates in the picture of the center points of any one first identification code and any one second identification code in the decoding area, and implement the decoding of the current corner point number through the decoding rules. Among them, the corner point number includes the corner point row number and the corner point column number .

[0081] The decoding rules in step S22 are as follows:

[0082] The decoding rule of the corner point row number is as follows:

[0083] When , then the row number of the current corner point number;

[0084] When , then the row number of the current corner point number;

[0085] The decoding rule of the corner point column number is as follows:

[0086] When and , then the column number of the current corner point number;

[0087] When and , then the column number of the current corner point number;

[0088] When and , then the column number of the current corner point number;

[0089] When and at this time, the column number of the current corner point number ;

[0090] Among them, represents the pixel coordinates of the center point of any first identification code; represents the pixel coordinates of the center point of any second identification code.

[0091] Through the decoding rule, it is always ensured that the corner has a unique corresponding number, and even when the calibration plate is partially blocked, as long as the decoding area corresponding to the corner remains intact, the corner can still be decoded, so as to ensure the precise matching between the corner and its corresponding physical size coordinates.

[0092] The following takes decoding Figure 6 the number of the corner point inside the circle in

[0093] to illustrate the decoding rule in detail: Figure 7 Extract the two white squares connected to the current corner point as the decoding area, as shown in Figure 6 and Figure 7 It can be seen that Figure 7 the pixel coordinates of the center point of any circular identification code in compared with the pixel coordinates and then , , that is, the current corner point is located in the 4th column of the 6th row.

[0094] The above embodiments are only for explaining the technical concept and characteristics of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it, and cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit of the present invention should be covered within the protection scope of the present invention.

Claims

1. A calibration method for camera parameter calibration, using a calibration board, characterized in that The calibration board includes a substrate, and black and white checkerboards are arrayed on the surface of the substrate; first identification codes are arranged in the white checkerboards of odd rows on the substrate, and second identification codes are arranged in the white checkerboards of even rows on the substrate; the number of the first identification codes in the white checkerboards of odd rows on the substrate is exactly the same along the row direction, and the number of the second identification codes in the white checkerboards of even rows on the substrate increases sequentially along the row direction; the number of the first identification codes in the white checkerboards of odd columns on the substrate increases sequentially along the column direction, and the number of the second identification codes in the white checkerboards of even columns on the substrate is exactly the same along the column direction; The specific steps are as follows: S1. Use a camera to take a picture of the calibration board, and then use the Shi-Tomasi algorithm to identify all the corner points in the picture; S2. With each corner point as the center, extract the two connected white checkerboards as the decoding area of the current corner point number, and obtain the number uniquely corresponding to the current corner point by decoding the identification code in the decoding area; S3. According to the number corresponding to each corner point, accurately assign the physical size coordinates corresponding to it in the world coordinate system to each corner point, and then use the Zhang Zhengyou calibration method to complete the camera parameter calibration; Among them, the steps of decoding the identification code in the decoding area in step S2 are as follows: S21. Identification and quantity calculation of identification codes: Extract one identification code from each of the two white squares in the decoding area, calculate the pixel areas of the black areas of the two identification codes, identify the first identification code and the second identification code according to the sizes of the pixel areas of the black areas, then count the number of all enclosed black areas within the white square where the first identification code is located, and record it as , count the number of all enclosed black areas within the white square where the second identification code is located, and record it as ; S22. Decoding of corner point numbers: Extract the pixel coordinates of the centers of any one first identification code and any one second identification code in the decoding area in the picture respectively, and implement the decoding of the current corner point number through the decoding rule, where the corner point number includes the corner point row number and the corner point column number .

2. The calibration method for camera parameter calibration according to claim 1, wherein, The area of the black region of the first identification code is equal to times the area of the black region of the second identification code, where , rounded to an integer.

3. The calibration method for camera parameter calibration according to claim 1 or 2, characterized in that, The first identification code is circular, and the second identification code is annular.

4. The calibration method for camera parameter calibration according to claim 3, wherein The outer diameter of the first identification code is equal to the outer diameter of the second identification code .

5. The calibration method for camera parameter calibration according to claim 1, characterized in that, The sizes of the white checkerboards and the black checkerboards can be freely set.

6. The calibration method for camera parameter calibration according to claim 1, wherein The number of the second identification codes in the white checkerboards of even rows on the substrate increases sequentially by 1 along the row direction, and the number of the first identification codes in the white checkerboards of odd columns on the substrate increases sequentially by 1 along the column direction.

7. The calibration method for camera parameter calibration according to claim 1, characterized in that, The decoding rules in step S22 are as follows: Corner point line number The decoding rules are as follows: When the line number of the current corner point number ; When the line number of the current corner point number ; Corner point serial number The decoding rules are as follows: When and then the column number of the current corner point number ; When and then the column number of the current corner point number ; When and then the column number of the current corner point number ; When and then the column number of the current corner point number ; Among them, represents the pixel coordinates of the center point of any first identification code; represents the pixel coordinates of the center point of any second identification code.

8. The calibration method for camera parameter calibration according to claim 1, characterized in that, The specific process of step S1 is as follows: Set window, preset threshold ; Assume that the pixel coordinates of a certain point in the window are , the movement amount is , and the gray level is . Then the gray level change function is: ； Among them, the window function ; The gray-scale change function is Taylor-expanded and high-order terms are omitted, resulting in: ； Among them, , , ; In the formula, , respectively represent the gradient values of the image gray level in the , directions; Define the corner response function as follows: ； Wherein, and are two eigenvalues of the matrix ; The window traverses all the pixel points in the image. When the value of a certain pixel point is greater than the threshold and it is a local maximum within the neighborhood of this pixel point, then this pixel point is a Shi-Tomasi feature corner point, and the coordinates of this pixel point are output .

Images