CN101425185A - Method for demarcating small-scale vision measuring video camera based on composite planar target drone - Google Patents

Abstract

The invention belongs to the field of measurement technology, which relates to a video camera calibration method of small size visual measurement based on a composite plane target. The composite plane target designed by the invention comprises a black square block and a white circular speckle, the center of the white circular speckle is coincident with the center of the black square block, and the semidiameter of the white circular speckle is less than that of the black square block; the composite plane target generates no less than 12 high precision gauge points having projective invariant by utilizing the characteristic that a point and a tangent point of a circle have the projective invariant; and a nonlinear optimization method is adopted to gauge the video camera by utilizing generated world coordinate and image coordinate of the 12 gauge points. The target has simple structure and low cost, and can be suitable for gauge to the video camera of various size visual measurements.

Description

A kind of camera marking method of the small-scale vision measuring based on compound plane target drone

Technical field

The invention belongs to technical field of visual measurement, relate to a kind of camera marking method of the small-scale vision measuring based on compound plane target drone.

Background technology

Camera calibration is meant according to camera model, utilize the extraction image coordinate and the known world coordinate solving model parameter of calibration point, thereby set up the mapping relations between image pixel positions and the space three-dimensional point, being key and the prerequisite in the vision measurement, is an important step setting up vision measurement system.

The camera calibration of vision measurement needs to provide the object of reference of standard calibration point, is called target.Calibration point on the target requires world coordinates known, and the image coordinate of calibration point is extracted easily.The form of target mainly contains two kinds of three-dimensional and plane formulas.Stereo target can provide the non-coplane space three-dimensional point coordinate of some, is made of two or three mutually orthogonal rigid planes usually, and it is provided with abundant calibration point.The stereo target difficulty of processing is big, and the cost height is difficult to reach degree of precision.In addition, the uneven problem of captured each face uncalibrated image quality by illumination is blocked, reason such as target surface local reflection causes will cause the calibration point image coordinate to extract precise decreasing.Plane target drone can provide the three dimensions point coordinate of one group of coplane, constitute by a rigid plane, as required in the not isolabeling of its surface working some so that calibration point to be provided.

At present, no matter be at stereo target or plane target drone, the calibration point that is provided with on the target surface is one of 6 kinds of forms as shown in Figure 1: intersection point, the summit of square and the lattice point of chessboard of the intersection point of the center of circle, square center, cross curve, rice word line.The perspective projection of circle shows as oval feature, the center of circle does not have the perspective projection unchangeability, be the center of space circle oval central point in the perspective projection point of the plane of delineation is inaccurate correspondence image, therefore can only be applied to the situation that the target plane is approximately perpendicular to camera optical axis.The calibration point limited amount that is obtained based on the target cellular construction of 6 kinds of forms shown in Figure 1, Dan Yuan, be that the single square of calibration point, single cross and single rice word line all can only obtain 1 calibration point with the center, the single square that with the summit is calibration point can produce 4 calibration points, and single checkerboard configuration unit also can only obtain 4 calibration points.Camera calibration needs 6 calibration points at least, therefore needs the combination of a plurality of cellular constructions just to be enough to realize Camera calibration.

Little at the visual field, the camera calibration of the small-scale vision measuring that operating distance is short need produce a large amount of high precision unique points among a small circle.The demarcation target of multi-unit structure need guarantee geometric accuracy of each cellular construction itself and positional precision each other in process, so difficulty of processing is bigger, and has improved processing cost.In addition, in calibration process, need the calibration point image of the target of assurance multi-unit structure to have high-quality, nothing is blocked, and could obtain high-precision calibration point image coordinate.The compound plane target drone that research has simple structure, the dominant character by limited quantity produce the video camera of recessive character for small-scale vision measuring, and flexibility and the precision that improves the small-scale vision measuring scaling method had great importance.

Summary of the invention

The objective of the invention is: the camera marking method that proposes a kind of small-scale vision measuring based on compound plane target drone, produce the recessive calibration points with perspective projection unchangeability by a spot of dominance calibration point, to realize the camera calibration of small-scale vision measuring more.

Technical scheme of the present invention is: a kind of camera marking method of the small-scale vision measuring based on compound plane target drone, it is characterized in that,

1, sets compound plane reference target, said target has the planar substrates of a carrying calibration point, planar substrates is made of glass or stupalith, a black square and a white circular spot are arranged in this substrate, and the center with the black square at the center of adularescent circle spot overlaps, and the black square length of side is 8mm～20mm, precision is 0.005mm, circle spot radius is 4mm～10mm, and precision is 0.005mm, and circle spot radius is less than the black square;

2, choosing 4 summits of black square on the target and 8 point of contacts of 4 summits and white circular spot is calibration point;

3, the world coordinates of obtaining step 2 said calibration points, concrete steps are as follows:

The first step is an initial point with any one summit in 4 summits of black square on the target, sets up world coordinate system, according to the concrete size of target, calculates the world coordinates on all the other 3 summits; Calculate the equation of dark circles spot corresponding circle under world coordinate system;

Second step, world coordinates and the equation of white circular spot corresponding circle under world coordinate system according to 4 summits of black square on the target, calculate 2 point of contacts of each summit and white circular spot respectively, obtain the world coordinates at 8 point of contacts altogether, add step 4 summits, obtain the world coordinates of 12 calibration points altogether;

4, in the field range of video camera, move freely at least 3 positions of target, whenever move a position, take piece image, be called uncalibrated image, require black square and white circular spot on the target to be included in the uncalibrated image;

5, extract the image coordinate of the calibration point of all uncalibrated images, concrete steps are as follows:

The first step, the perspective projection of the square in the space are quadrilateral, extract the image coordinate on tetragonal 4 summits in the uncalibrated image, and corresponding with the world coordinates on summit;

In second step, the perspective projection of space circle is oval, extracts the oval feature in the uncalibrated image, obtains oval equation under image coordinate system;

The 3rd step, under image coordinate system, calculate 4 summits and oval point of contact, obtain the image coordinate at 8 point of contacts, and corresponding with the world coordinates at point of contact;

6, utilize the image coordinate and the corresponding world coordinate calibrating camera parameters of all calibration points of step 5 extraction, camera parameters is kept in the Parameter File, call in order to measuring.

Advantage of the present invention is:

1, compound plane target drone of the present invention only needs to incorporate a white circular spot in a black square, and processing is simple, and cost is low, and is easier to guarantee machining precision.

2, by the point of contact of black square summit and summit and white circular spot, produce nearly 12 high-precision calibrating points with perspective projection unchangeability, make interior among a small circle calibration point quantity increase greatly, thereby be applicable to the camera calibration of small-scale vision measuring.

3, pass through the square and the circle spot size of the compound plane target drone of this invention of increase, can make this target be applicable to the demarcation of the vision measurement system of multiple different visual fields size and operating distance, thereby improve the target service efficiency, reduce cost.

Description of drawings

Fig. 1 is six kinds of target unique point synoptic diagram using always.The target unique point of (a) is the center of circle among the figure, the target unique point of (b) is the square center among the figure, the target unique point of (c) is the intersection point of cross curve among the figure, the target unique point of (d) is the intersection point of rice word line among the figure, the target unique point of (e) is the summit of square among the figure, and the target unique point of (f) is the lattice point of chessboard among the figure.

Fig. 2 is the mathematical model synoptic diagram of video camera.

Fig. 3 is the structural representation of compound type plane target drone of the present invention.

Fig. 4 is the synoptic diagram that compound plane target drone of the present invention produces calibration point.1～4 is the summit of square among the figure, and 5～12 are respectively the point of contact of 4 summits and white circular spot.

Embodiment

Below the present invention is made and further describing.Little at the visual field, the camera calibration of the small-scale vision measuring that operating distance is short need be at the high-precision calibrating point that produces among a small circle more than at least 6.

As shown in Figure 2, o _c-x _cy _cz _cBe camera coordinate system, o _w-x _wy _wz _wFor being based upon the world coordinate system on the target, o _u-x _uy _uBe image coordinate system, o _n-x _ny _nBe projected coordinate system.π _nBe projection plane, π _uBe the plane of delineation.Point o _pBe straight line o _cz _cWith π _uIntersection point, be called the principal point of video camera.Definition o _cx _c‖ o _ux _u‖ o _nx _n, o _cy _c‖ o _uy _u‖ o _ny _n, $o_c{z}_c\⊥{\mathrm{\π}}_u$ And π _n‖ π _uAny spatial point P _iWorld coordinates be X _Wi=(x _Wi, y _Wi, z _Wi) ^T, camera coordinates is X _Ci=(x _Ci, y _Ci, z _Ci) ^T, corresponding subpoint is p _iO then _w-x _wy _wz _wTo o _c-x _cy _cz _cBe transformed to:

$X_{\mathrm{ci}}=R_c^w{X}_{\mathrm{wi}}+T_c^w---\left[1\right]$

Wherein

Be 3 * 3 quadrature rotation matrix,

It is 3 * 1 translation vector.

P sets up an office _iProjection coordinate be X _Ni=(x _Ni, y _Ni) ^T, image coordinate is X _Ui=(x _Ui, y _Ui) ^T, the effective focal length of video camera on x, y direction is f _xAnd f _y, the principal point coordinate of video camera is (u ₀, v ₀), then have

X _ni＝[x _ci/Z _ci?y _ci/z _ci] ^T [2]

X _ui＝[f _xx _ni+u ₀?f _yy _ni+v ₀] ^T [3]

The projection model of spatial point to image coordinate represented in formula [1]～[3], and only image projection point can be determined in the arbitrfary point, space.Otherwise, if known intrinsic parameters of the camera, according to formula [1]～[3], can be by image coordinate in the hope of the projection coordinate of spatial point, thus obtain subpoint at o _c-x _cy _cz _cUnder coordinate.By camera model as can be known, the parameter of video camera mainly comprises the effective focal length f of video camera _xAnd f _y, the principal point (u of video camera ₀, v ₀) and the external parameter of video camera With

According to camera model, the concrete steps based on the camera marking method of compound plane target drone that the present invention proposes are as follows:

1, sets compound plane reference target, said target has the planar substrates of a carrying calibration point, planar substrates is made of glass or stupalith, a black square and a white circular spot are arranged in this substrate, and the center with the black square at the center of adularescent circle spot overlaps, and the black square length of side is 8mm～20mm, precision is 0.005mm, circle spot radius is 4mm～10mm, and precision is 0.005mm, and circle spot radius is less than the black square.

2, choosing 4 summits of black square on the target and 8 point of contacts of 4 summits and white circular spot is calibration point.

3, the world coordinates of obtaining step 2 said calibration points, concrete steps are as follows:

The first step is an initial point with any one summit in 4 summits of black square on the target, sets up world coordinate system, according to the concrete size of target, calculates the world coordinates on all the other 3 summits; Calculate the equation of dark circles spot corresponding circle under world coordinate system.

Second step, world coordinates and the equation of white circular spot corresponding circle under world coordinate system according to 4 summits of black square on the target, calculate 2 point of contacts of each summit and white circular spot respectively, obtain the world coordinates at 8 point of contacts altogether, add step 4 summits, obtain the world coordinates of 12 calibration points altogether.Put " how much instrument algorithms detailed annotations of computer graphics (GeometricTools for Computer Graphic) " of translating referring to Zhou Changfa, Electronic Industry Press, 2005 with the computing method at round point of contact.

4, in the field range of video camera, move freely at least 3 positions of target, whenever move a position, take piece image, be called uncalibrated image, require black square and white circular spot on the target to be included in the uncalibrated image.

5, extract the image coordinate of the calibration point of all uncalibrated images, concrete steps are as follows:

The first step, the perspective projection of the square in the space are quadrilateral, extract the image coordinate on tetragonal 4 summits in the uncalibrated image, and corresponding with the world coordinates on summit.The image coordinate extraction algorithm on tetragonal summit is referring to Zhou Fuqiang work " the gordian technique research that binocular stereo vision detects ", BJ University of Aeronautics ﹠ Astronautics's post-doctoral research work report, 2002.

In second step, the perspective projection of space circle is oval, extracts the oval feature in the uncalibrated image, obtains oval equation under image coordinate system.Oval extracting method is referring to the patent of invention " extraction method of oval characteristics of image in a kind of complex background image " of Zhou Fuqiang etc., application number: 200710303877.8, and Gao Kairi: on July 9th, 2008.

The 3rd step, under image coordinate system, calculate 4 summits and oval point of contact, obtain the image coordinate at 8 point of contacts, and corresponding with the world coordinates at point of contact.

If oval equation is that (x, y)=0, the coordinate of establishing oval outer arbitrfary point is (x to F ₁, y ₁), then the point of contact with ellipse is following solution of equations

$\left\{\begin{array}{c}F(x,y)=0\\ x.\frac{\∂F}{\∂x}(x_1,y_1)+y.\frac{\∂F}{\∂y}(x_1,y_1)=0\end{array}\right.---\left[4\right]$

6, utilize the image coordinate and the corresponding world coordinate calibrating camera parameters of all calibration points of step 5 extraction, camera parameters is kept in the Parameter File, call in order to measuring.

According to the video camera realistic model of formula [1]～[3] expression, utilize the world coordinates X of unique point _Wi, obtain the computed image coordinate X of calibration point _Mi=(x _Mi, y _Mi) ^TIf camera parameters is not accurate enough, then X _MiWith X _UiThere is error, can sets up the optimization aim function of camera calibration thus

$f(f_x,f_y,u_0,v_0,R_c^w,T_c^w)={\mathrm{\Σ}}_{i=1}^{12}{d}^2(X_{\mathrm{mi}},X_{\mathrm{ui}})---\left[5\right]$

D wherein ²(X _Mi, X _Ui) the computed image coordinate of expression calibration point is to the distance of the real image coordinate of calibration point.The initial value of camera parameters is estimated referring to Zhou Fuqiang work " the gordian technique research that binocular stereo vision detects ", BJ University of Aeronautics ﹠ Astronautics's post-doctoral research work report, 2002.

Can estimate whole intrinsic parameters of the camera according to formula [5], comprise f _x, f _y, u ₀, v ₀,

With

If intrinsic parameters of the camera is known, then according to formula [5], adopt the external parameter that the Levenberg-Marquardt nonlinear optimization method can the direct estimation video camera (

With

).The Levenberg-Marquardt algorithm is referring to " Optimum Theory and method ", (Yuan Yaxiang, Sun Wenyu work, Science Press, 1999).

Embodiment

Fig. 4 is the synoptic diagram that compound plane target drone of the present invention produces calibration point.Plated film, corroding method are made high-precision compound plane target drone on optical glass, and the length of side of the black square of target is 9mm, and the diameter of circle spot is 5mm, and precision is 0.005mm.For the compound target of making, be initial point with the upper left angle point of grid, horizontal direction is the x axle, vertical downward direction is set up world coordinate system for the y axle, calculates the world coordinates of all calibration points on the target, and is as shown in table 1.

(the unit: millimeter) of the world coordinates of calibration point on table 1 target

Image coordinate (the unit: pixel) of calibration point on the target that table 2 extracts

The image coordinate of the point corresponding with target plane characteristic point, as shown in table 2.

The world coordinates of the calibration point of several uncalibrated images that employing obtains and corresponding image coordinate thereof are demarcated one 1/6 inches video camera, and image acquisition resolution is a pal mode.The intrinsic parameters of the camera of demarcating is:

f _x＝1186.859pixels，f _y＝1171.695pixels

u ₀＝308.212pixels，v ₀＝286.413pixels

Calibrated error represents with the error of the coordinate of projected image again of unique point, is respectively in the RMS error of x and y direction: Δ _x=0.182pixels, Δ _y=0.131pixels.

Claims (1)

1, a kind of camera marking method of the small-scale vision measuring based on compound plane target drone is characterized in that,

1.1, set compound plane reference target, said target has the planar substrates of a carrying calibration point, planar substrates is made of glass or stupalith, a black square and a white circular spot are arranged in this substrate, and the center with the black square at the center of adularescent circle spot overlaps, and the black square length of side is 8mm～20mm, precision is 0.005mm, circle spot radius is 4mm～10mm, and precision is 0.005mm, and circle spot radius is less than the black square;

1.2, choose 4 summits of black square on the target and 8 point of contacts of 4 summits and white circular spot are calibration point;

1.3, the world coordinates of obtaining step 1.2 said calibration points, concrete steps are as follows:

The first step is an initial point with any one summit in 4 summits of black square on the target, sets up world coordinate system, according to the concrete size of target, calculates the world coordinates on all the other 3 summits; Calculate the equation of dark circles spot corresponding circle under world coordinate system;

Second step, world coordinates and the equation of white circular spot corresponding circle under world coordinate system according to 4 summits of black square on the target, calculate 2 point of contacts of each summit and white circular spot respectively, obtain the world coordinates at 8 point of contacts altogether, add step 4 summits, obtain the world coordinates of 12 calibration points altogether;

1.4, in the field range of video camera, move freely at least 3 positions of target, whenever move a position, take piece image, be called uncalibrated image, require black square and white circular spot on the target to be included in the uncalibrated image;

1.5, extract the image coordinate of the calibration point of all uncalibrated images, concrete steps are as follows:

The first step, the perspective projection of the square in the space are quadrilateral, extract the image coordinate on tetragonal 4 summits in the uncalibrated image, and corresponding with the world coordinates on summit;

In second step, the perspective projection of space circle is oval, extracts the oval feature in the uncalibrated image, obtains oval equation under image coordinate system;

The 3rd step, under image coordinate system, calculate 4 summits and oval point of contact, obtain the image coordinate at 8 point of contacts, and corresponding with the world coordinates at point of contact;

1.6, utilize the image coordinate and the corresponding world coordinate calibrating camera parameters of all calibration points that step 1.5 extracts, camera parameters is kept in the Parameter File, call in order to measuring.

Images