CN100470590C - Camera calibration method and calibration apparatus thereof - Google Patents

Abstract

The invention involves a calibration device for camera calibration, which consists of two vertical grid boards that have distribution grid and the number of signs marked with a unique code point. The invention also provides a method of the camera calibration, including the following steps: (a) Acquisition of images of the calibration device; (2) corresponding coordinates of the signs point to the coordinates of object point; (3) projection the plane grid to get the initial point coordinates of grid point; (4) the use of line extraction intersection to get the precise coordinates of points; (5) The use of 3-D direct linear transformation for the initial value of calibration parameters; (6) iterative solution with beam adjustment to get the camera's internal and external orientation elements. After shooting a picture of calibration device in the invention, the follow-up calibration process can be done automatically by computer. The advantage of the invention is automatic and quickness.

Description

Camera calibration method and used caliberating device

Technical field

The invention belongs to the camera calibration technical field, particularly relate to camera calibration method and used caliberating device.

Background technology

Ask the process of camera inner parameter to be called demarcation.Utilize the image of central projection to recover 3D shape, generally carrying out camera calibration is a necessary step.The method of camera calibration has multiple.Method (the Z.Zhang:Flexible Camera Calibration by Viewing a Plane from Unknown Orientations.Proc of 7 that demarcates based on the plane reference field is arranged ^ThInt.Conference on Computer Vision, Kerkyra, Greece.pp.666-673, Sept.1999.), this method need be taken the image of at least two different angles, therefore needs plane of motion to demarcate field or camera, is difficult to realize robotization; Demarcation based on three-dimensional controlling filed is arranged, and three-dimensional controlling filed is difficult to safeguard, and carries inconvenience; Also have without any need for the self-calibrating method of demarcating the field, self-calibrating method calculates unstable, and that adopts in actual applications is few.In addition, said method generally all needs manually to specify the correspondence of several points in order to realize the corresponding of picpointed coordinate and object space point coordinate.

Summary of the invention

The present invention is directed to the problems referred to above, proposed a kind of easy camera calibration method and used caliberating device, this method is easy and can realize the automatic corresponding of picpointed coordinate and object space point coordinate, finally realizes the automatic demarcation of camera.

Technical scheme provided by the invention is: a kind of caliberating device that is used for camera calibration, form by two mutually perpendicular planar grid plates, planar grid and equally distributed six or six above monumented points are arranged on the planar grid plate, and each monumented point has unique encoding.

Above-mentioned monumented point is made up of center circle and annulus, and annulus is made up of some five equilibrium segmental arcs, and each segmental arc identifies with black or white, and constitutes the unique encoding of each monumented point by the arrangement of black and white.

The present invention also provides the method for utilizing above-mentioned caliberating device to demarcate camera, may further comprise the steps:

(1) image of the above-mentioned caliberating device of collection

With the image of camera shooting caliberating device to be calibrated, by computing machine image is lined by line scan then, obtain the picpointed coordinate of six or six above monumented points on the image;

(2) picpointed coordinate of the monumented point that step (1) is obtained carries out corresponding one by one with object coordinates

The coding of the monumented point that obtains according to step (1) and the corresponding relation of predefined coding and object coordinates are determined the object coordinates of each monumented point correspondence;

(3) the projection plane grid points obtains the initial value of the picpointed coordinate of planar grid point

The picpointed coordinate of the monumented point that obtains according to step (2) and the one-to-one relationship of object coordinates are set up the plane perspective conversion, calculate the initial value of the picpointed coordinate of other grid points on the image, and are projected on the image;

(4) utilize straight line to extract and ask friendship, obtain the exact value of the picpointed coordinate of grid points;

(5) utilize three-dimensional direct linear transformation to find the solution the initial value of calibrating parameters;

(6) carry out iterative with bundle adjustment, obtain the s internal and external orientation of camera.

According to the present invention, in above-mentioned steps (1), take an image of demarcating with camera to be calibrated after, by computing machine image is lined by line scan, find out black segmental arc on the image to determine the monumented point under these black segmental arcs; Counterclockwise or clockwise direction is carried out the scannings of 360 degree around the center of monumented point to the annulus of monumented point, runs into the black segmental arc and just the coding of this segmental arc is made as 1, otherwise be made as 0; Behind the end of scan, optional one 0 and 1 separation begin order to be listed, and obtains binary code; By transferring the decimal system to, get wherein maximal value as the coding of this monumented point to the binary code cyclic ordering and with these binary codes; Carry out the operation of step (2) again.

The inventive method only need be taken an image to caliberating device, and all the other processes (extraction of monumented point and identification, parameter calculation etc.) are all finished automatically by computing machine.Caliberating device of the present invention is made up of mutually perpendicular two planes, and is cheap, and structure has easily overcome the shortcoming that surface plate does not have depth information, is easy to safeguard than three-dimensional controlling filed simultaneously.Arranged the monumented point with unique coding of some on the caliberating device, can make things convenient for, automatically identification fast, enough accuracy of identification are provided, make the precision at picture point center reach sub-pixel, can avoid the interference of most of image noise, can realize the automatic corresponding of picpointed coordinate and object space point coordinate, reduce manual intervention.

Description of drawings

Fig. 1 is the inventive method FB(flow block);

Fig. 2 is a caliberating device synoptic diagram of the present invention;

Fig. 3 has binary-coded monumented point synoptic diagram for the mark among Fig. 2.

Embodiment

One, caliberating device

Referring to Fig. 2, the present invention relates to the caliberating device that two mutually perpendicular planes are formed, and on two planes, stamp uniform graticule mesh (graticule mesh is made up of orthogonal two groups of parallel lines, and the intersection point of straight line is a grid points).In order to make picpointed coordinate corresponding one by one with the object space point coordinate, each grid points is numbered: with the O point be the center from the bottom up, from the inside to surface with a position and ten bit representations row number, represent row number with kilobit for hundred, plane, left side myriabit is 1, and plane, the right myriabit is 0.For example, the left plane of 11211 expressions the 12nd row 11 row, the right plane of 706 expressions the 7th row the 6th row.Then on planar grid, put the monumented point that has coding (below abbreviation monumented point) of some, purpose be to carry out monumented point automatic identification and and the automatic correspondence of object space point coordinate, the shape of monumented point and distributing as shown in Figure 2.

Two, monumented point

1. specific design.The design proposal that realizes monumented point is: the center of monumented point is a circle, around round dot, there is a circle to have certain width, and by the annulus of n equal portions (n=4～12 are got in suggestion), each segmental arc is with deceiving or representing in vain, in order to be numbered to annulus, the black segmental arc represents that with 1 white segmental arc is represented with 0.Arranged the monumented point of some (more than or equal to 6) on two planes, each monumented point all has a well-determined coding, the regular following introduction of coding.Fig. 3 has shown the structure of a monumented point, is a circle at the center of monumented point, and we can ask for the centre coordinate of round center as this monumented point very easily.One circle is arranged by the annulus of 8 equal portions around circle, each segmental arc is with to deceive or to represent that in vain in order being numbered annulus, the black segmental arc represents that with 1 white segmental arc is with representing with 0.Read these codings then counterclockwise,, 8 kinds of different possibilities are arranged: 10101011,01010111,10101110,01011101,10111010,01110101,11101010,11010101 obviously according to the difference of starting position.In these 8 binary numbers, 11101010 (11101010 ₂=234 ₁₀) be maximal value, we are just with 234 unique number as this monumented point.It should be noted that not all positive integer all corresponding a such coding collar, we just with in all codings maximum one identify this monumented point.Another one it should be noted that, here the coding of the monumented point of Que Dinging not is the numbering of foregoing grid points, need pre-setly for the numbering of the grid points of each monumented point correspondence, be numbered 11202 (referring to Fig. 2) such as the grid points that is encoded to 234 monumented point correspondence.

2. identification.According to the characteristics of image, utilize the variation of gradient that image is lined by line scan, can obtain several candidate's black regions easily and fast about monumented point.Because monumented point all can be scanned out with graticule mesh,, get rid of the interference of graticule mesh to mark point recognition so before landmark identification, preferably this zone is judged according to following order:

(1) number of pixel:, get rid of this zone if the number of pixel very little.

(2) width:, get rid of zone less than this width to minimum threshold of width setup of the center of circle and annulus.

(3) distance:, get rid of zone greater than this distance to the center of circle max-thresholds of distance setting to annulus.

Under the situation in remaining monumented point zone only, to several zones of monumented point according to determining the central area of this monumented point to the shortest principle of other region distances, and the center that to get this regional center of gravity be monumented point.Then, one week of center is divided into 360 scales, and draws a ray, begin to carry out around the center counterclockwise the scanning of 360 degree, run into mark region and just this scale is made as 1, otherwise be made as 0 from certain direction from the center.Behind the end of scan, 360 scales being unified into an annulus and beginning to add up from one 0,1 separation, is a unit-sized with 1/8 circle, if in a unit length, scale is 1, thinks that then this segmental arc is 1, otherwise just is 0.Like this, just whole annulus is converted to one 8 binary code.By transferring the decimal system to as this coding corresponding codes value to the binary code cyclic ordering and with these binary codings, according to the reference position difference that reads, different encoded radios is arranged, get the coding of the maximal value of all encoded radios as this monumented point, obviously, each monumented point corresponding codes value is unique.According to the corresponding relation of this monumented point coding and predefined coding dative number node, can determine the numbering of the object space point of this monumented point correspondence.Like this, just finished the corresponding one by one of picpointed coordinate and object space point coordinate.

3. obtain coordinate.According to the picpointed coordinate and the corresponding object coordinates value of these monumented points that obtain, set up the plane perspective conversion, calculate the initial value of the picpointed coordinate of other grid points on the image, and be projected on the image.The picpointed coordinate that comes out through the plane perspective transformation calculations not very accurately, in order to obtain the accurate coordinates of grid points, adopt straight line to extract the method (as the straight line template matching method) of asking friendship, obtain near the exact position of 2 sections straight lines that intersect the graticule mesh initial value point, and with their intersection point as the accurate coordinates of grid points.

Three scaling methods:

1 theoretical foundation

Direct linear transformation (directlineartransformation DLT) is an algorithm of setting up picpointed coordinate and the direct linear relationship of object point coordinate, does not need the initial value of video camera s internal and external orientation during processing.Therefore utilize three-dimensional direct linear transformation to ask for the initial value of video camera, 3D DLT model can be expressed as:

$x=\frac{L_{1}X+L_{2}Y+L_{3}Z+L_4}{L_{9}X+L_{10}Y+L_{11}Z+1}$

$y=\frac{L_{5}X+L_{6}Y+L_{7}Z+L_8}{L_{9}X+L_{10}Y+L_{11}Z+1}$ (1)

In the formula (1), L ₁, L ₂, L ₃, L ₄, L ₅, L ₆, L ₇, L ₈, L ₉, L ₁₀, L ₁₁11 transformation parameters for three-dimensional direct linear transformation; X, Y, Z are the volume coordinates of grid points in the three-dimensional controlling filed; X, y look like coordinate accordingly.In order to find the solution this 11 parameters, need the volume coordinate of the object space point of known 6 picture points and correspondence thereof at least.When number of dots during greater than 6, available least square is found the solution this overdetermined equation group.

The elements of interior orientation x of camera during photography ₀.y ₀.f and elements of exterior orientation

ω, κ, X _s, Y _s, Z _sAll be included in L ₁To L ₁₁Coefficient in, can draw through inverse:

Elements of interior orientation:

x ₀＝(L ₁L ₉+L ₂L ₁₀+L ₃L ₁₁)L′

y ₀＝(L ₅L ₉+L ₆L ₁₀+L ₇L ₁₁)L′

$f_x=-x_0^2+({\mathrm{<figure-callout\; id="1"\; label="L"\; filenames="C200710051485E00082.png,C200710051485E00083.png"\; state="\{\{state\}\}">L</figure-callout>}}_1^2+{\mathrm{<figure-callout\; id="2"\; label="L"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">L</figure-callout>}}_2^2+L_3^2)L\&prime;---\left(2\right)$

$f_y=-y_0^2+(L_5^2+L_6^2+{\mathrm{<figure-callout\; id="7"\; label="L"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">L</figure-callout>}}_7^2)L\&prime;$

f＝(f _x+f _y)/2

Wherein ${L\&prime;}^2=1/(L_9^2+L_{10}^2+{\mathrm{<figure-callout\; id="11"\; label="L"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">L</figure-callout>}}_{11}^2);$

Elements of exterior orientation;

$\left.\begin{array}{c}{L}_1{X}_s+L_2{Y}_s+L_3{Z}_s=-L_4\\ L_5{X}_s+L_6{Y}_s+L_7{Z}_s=-L_8\\ L_9{X}_s+L_{10}{Y}_s+L_{11}{Z}_s=-1\end{array}\right\}---\left(3\right)$

a ₃＝L ₉L′

b ₃＝L ₁₀L′

(4)

c ₃＝L ₁₁L′

a ₂＝(L ₅+L ₉y ₀)L′/f _x

After solving each L coefficient of photo, can separate 9 independent parameter asking corresponding photo according to above-mentioned each relational expression.And its initial value as bundle adjustment.

After calculating the s internal and external orientation initial value of camera, follow available photogrammetric in bundle adjustment commonly used carry out the demarcation of camera.To amateur camera, the distortion of camera lens is general bigger, considers the influence of radial distortion and decentering distortion, adopts following correction model:

Δx＝(x-x ₀)(k ₁r ²+k ₂r ⁴)+p ₁(r ²+2(x-x ₀) ²)+2p ₂(x-x ₀)(y-y ₀)

(6)

Δy＝(y-y ₀)(k ₁r ²+k ₂r ⁴)+p ₁(r ²+2(y-y ₀) ²)+2p ₂(x-x ₀)(y-y ₀)

Wherein, r ²=(x-x ₀) ²+ (y-y ₀) ²k ₁, k ₂It is poor to be called radial distortion, p ₁, p ₂Be that the decentering distortion difference is brought this distortion into collinearity condition equation [7,8]:

$x-x_0-\mathrm{\&Delta;x}=-f_x\frac{a_1(X-X_s)+b_1(Y-Y_s)+c_1(Z-Z_s)}{a_3(X-X_s)+b_3(Y-Y_s)+c_3(Z-Z_s)}$

$y-y_0-\mathrm{\&Delta;y}=-f_x\frac{a_2(X-X_s)+b_2(Y-Y_s)+c_2(Z-Z_s)}{a_3(X-X_s)+b_3(Y-Y_s)+c_3(Z-Z_s)}$ (7)

Following formula is obtained error equation with the Taylor's formula linearization:

$+\frac{\mathrm{\&delta;x}}{\mathrm{\&delta;}{Z}_s}\mathrm{\&Delta;}{Z}_s+\frac{\mathrm{\&delta;x}}{\mathrm{\&delta;}{k}_1}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="1"\; label="k"\; filenames="C200710051485E00082.png,C200710051485E00083.png"\; state="\{\{state\}\}">k</figure-callout>}}_1+\frac{\mathrm{\&delta;x}}{\mathrm{\&delta;}{k}_2}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="2"\; label="k"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">k</figure-callout>}}_2+\frac{\mathrm{\&delta;x}}{\mathrm{\&delta;}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C200710051485E00082.png,C200710051485E00083.png"\; state="\{\{state\}\}">p</figure-callout>}}_1}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C200710051485E00082.png,C200710051485E00083.png"\; state="\{\{state\}\}">p</figure-callout>}}_1+\frac{\mathrm{\&delta;x}}{\mathrm{\&delta;}{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">p</figure-callout>}}_2}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">p</figure-callout>}}_2+\frac{\mathrm{\&delta;x}}{\mathrm{\&delta;X}}\mathrm{\&Delta;X}+\frac{\mathrm{\&delta;x}}{\mathrm{\&delta;Y}}\mathrm{\&Delta;Y}+\frac{\mathrm{\&delta;x}}{\mathrm{\&delta;Z}}\mathrm{\&Delta;Z}-l_x$

Each coefficient

$+\frac{\mathrm{\&delta;y}}{\mathrm{\&delta;}{Z}_s}\mathrm{\&Delta;}{Z}_s+\frac{\mathrm{\&delta;y}}{\mathrm{\&delta;}{k}_1}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="1"\; label="k"\; filenames="C200710051485E00082.png,C200710051485E00083.png"\; state="\{\{state\}\}">k</figure-callout>}}_1+\frac{\mathrm{\&delta;y}}{\mathrm{\&delta;}{k}_2}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="2"\; label="k"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">k</figure-callout>}}_2+\frac{\mathrm{\&delta;y}}{\mathrm{\&delta;}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C200710051485E00082.png,C200710051485E00083.png"\; state="\{\{state\}\}">p</figure-callout>}}_1}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="C200710051485E00082.png,C200710051485E00083.png"\; state="\{\{state\}\}">p</figure-callout>}}_1+\frac{\mathrm{\&delta;y}}{\mathrm{\&delta;}{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">p</figure-callout>}}_2}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C200710051485E00082.png"\; state="\{\{state\}\}">p</figure-callout>}}_2+\frac{\mathrm{\&delta;y}}{\mathrm{\&delta;X}}\mathrm{\&Delta;X}+\frac{\mathrm{\&delta;y}}{\mathrm{\&delta;Y}}\mathrm{\&Delta;Y}+\frac{\mathrm{\&delta;y}}{\mathrm{\&delta;Z}}\mathrm{\&Delta;Z}-l_y$

Value and constant value are by asking local derviation to obtain to collinearity equation.

2 implementation procedures (referring to Fig. 1)

(1) caliberating device of the present invention is 2 mutually perpendicular dull and stereotyped compositions, has arranged planar grid on each plane, in addition, also has some equally distributed monumented points.

(2) gather image before, adjust the distance of caliberating device and camera, make imaging clear, and the film size of camera can comprise all monumented points, gather an image and import computing machine.

(3) by Computer Automatic Extraction monumented point and decoding.

(4) projection grid's point, and carry out grid points and accurately aim at (referring to embodiment two, 3).

(5) the demarcation initial value is asked in three-dimensional DLT conversion, carries out iterative with bundle adjustment then, obtains the s internal and external orientation of camera, has promptly realized the demarcation of camera.

Obviously, those skilled in the art can carry out suitable adjustment and distortion to the present invention and not break away from the spirit and scope of the present invention.Like this, if these adjustment and distortion belong to the category of requirement of the present invention and equivalent technologies thereof, then the present invention also is intended to comprise these adjustment and is out of shape interior.

Claims (4)

1. caliberating device that is used for camera calibration, it is characterized in that: be made up of two mutually perpendicular planar grid plates, planar grid and equally distributed six or six above monumented points are arranged on the planar grid plate, each monumented point has unique encoding.

2. caliberating device according to claim 1, it is characterized in that: monumented point is made up of center circle and annulus, annulus is made up of some five equilibrium segmental arcs, and each segmental arc identifies with black or white, and constitutes the unique encoding of each monumented point by the arrangement of black and white.

3. the method for utilizing the described caliberating device of claim 1 to demarcate camera is characterized in that may further comprise the steps:

(1) image of the described caliberating device of collection claim 1

With the image of camera shooting caliberating device to be calibrated, by computing machine image is lined by line scan then, obtain the picpointed coordinate of six or six above monumented points on the image;

(2) picpointed coordinate of the monumented point that step (1) is obtained carries out corresponding one by one with object coordinates

The coding of the monumented point that obtains according to step (1) and the corresponding relation of predefined coding and object coordinates are determined the object coordinates of each monumented point correspondence;

(3) the projection plane grid points obtains the initial value of the picpointed coordinate of planar grid point

The picpointed coordinate of the monumented point that obtains according to step (2) and the one-to-one relationship of object coordinates are set up the plane perspective conversion, calculate the initial value of the picpointed coordinate of other grid points on the image, and are projected on the image;

(4) utilize straight line to extract and ask friendship, obtain the exact value of the picpointed coordinate of grid points;

(5) utilize three-dimensional direct linear transformation to find the solution the initial value of calibrating parameters;

(6) carry out iterative with bundle adjustment, obtain the s internal and external orientation of camera.

4. the method for utilizing the described caliberating device of claim 2 to demarcate camera is characterized in that may further comprise the steps:

(1) image of the described caliberating device of collection claim 2

Take the image of caliberating device with camera to be calibrated, by computing machine image is lined by line scan then, find out black segmental arc on the image to determine the monumented point under these black segmental arcs; Counterclockwise or clockwise direction is carried out the scannings of 360 degree around the center of monumented point to the annulus of monumented point, runs into the black segmental arc and just the coding of this segmental arc is made as 1, otherwise be made as 0; Behind the end of scan, optional one 0 and 1 separation begin order to be listed, and obtains binary code; By transferring the decimal system to, get wherein maximal value as the coding of this monumented point to the binary code cyclic ordering and with these binary codes;

(2) picpointed coordinate of the monumented point that step (1) is obtained carries out corresponding one by one with the object space point coordinate

The coding of the monumented point that obtains according to step (1) and the corresponding relation of predefined coding and object space point coordinate are determined the object space point coordinate of each monumented point correspondence;

(3) the projection plane grid points obtains the initial value of the picpointed coordinate of planar grid point

The picpointed coordinate of the monumented point that obtains according to step (2) and the one-to-one relationship of object space point coordinate are set up the plane perspective conversion, calculate the initial value of the picpointed coordinate of other grid points on the image, and are projected on the image;

(4) utilize straight line to extract and ask friendship, obtain the exact value of the picpointed coordinate of grid points;

(5) utilize three-dimensional direct linear transformation to find the solution the initial value of calibrating parameters;

(6) carry out iterative with bundle adjustment, obtain the s internal and external orientation of camera.

Images