CN105021127B - A kind of benchmark camera calibration method of chip mounter - Google Patents

Abstract

A kind of benchmark camera calibration method of chip mounter, the present invention relates to the main trimming process of benchmark camera in the accuracy correction field of chip mounter, specifically chip mounter.The present invention is to solve existing chip mounter, camera can not realize that attachment plane is parallel with camera plane during mechanical erection, the problem of application condition is big.(1) determine and set up device coordinate system and benchmark camera coordinates system position each other and rotation relationship, derive the coordinate conversion relation between two coordinate systems；(2) exact position using circle detection algorithm detection reference in benchmark camera coordinates system；(3) on the basis of datum mark, using circle detection algorithm, the correction to the benchmark camera of chip mounter is realized, that is, completes a kind of benchmark camera calibration method of chip mounter.The present invention is applied to accuracy correction field.

Description

A kind of benchmark camera calibration method of chip mounter

Technical field

The present invention relates to the accuracy correction field of chip mounter, the benchmark camera of specifically chip mounter was mainly corrected Journey.

Background technology

In chip mounter, benchmark camera and placement head are fixedly mounted on movable framework, benchmark camera coordinates system and Position relationship between device coordinate system is detailed to be provided.During chip mounter works, benchmark camera calibration is utilized The position of MARK points and the anglec of rotation, obtained result represent that in benchmark camera coordinates system the unit of wherein position coordinates is Pixel.Expect position coordinates and the anglec of rotation of the MARK points relative to device coordinate system, first have to calculate benchmark camera One pixel of exact scale, i.e. camera represents how many centimetres；Its is secondary to calculate rotation of the benchmark camera relative to device coordinate system Gyration, to correct the actual anglec of rotation of MARK points.

Existing chip mounter, camera can not realize that attachment plane is parallel with camera plane during mechanical erection, application condition Greatly.

The content of the invention

Attachment plane and camera plane can not be realized the present invention is to solve camera during existing chip mounter mechanical erection It is parallel, the problem of application condition is big, and there is provided a kind of benchmark camera calibration method of chip mounter.

A kind of benchmark camera calibration method of chip mounter, it is realized according to the following steps：

(1) determine and set up device coordinate system and benchmark camera coordinates system position each other and rotation relationship, derive The coordinate conversion relation gone out between two coordinate systems；

(2) exact position using circle detection algorithm detection reference in benchmark camera coordinates system；

(3) on the basis of datum mark, using circle detection algorithm, realize the correction to the benchmark camera of chip mounter, i.e., it is complete Into a kind of benchmark camera calibration method of chip mounter.

Invention effect：

The origin of device coordinate system and the origin of benchmark camera coordinates system are simultaneously misaligned, and exist one between the two admittedly Fixed anglec of rotation θ and scale X_Scale、Y_Scale.Camera is relative to the anglec of rotation of device coordinate system, X on the basis of wherein θ_Scale And Y_ScaleOn the basis of the scale of camera horizontally and vertically.Here it is the parameter of the invention to be corrected.

Present invention is primarily intended to the scale of correction reference camera and the anglec of rotation, more accurately to detect MARK points Actual position coordinate and the anglec of rotation, can finally effectively improve the placement accuracy of chip mounter.

In the present invention, datum mark is circular, circle detection algorithm is designed, for benchmark camera calibration datum mark in base Position under quasi- camera coordinates system.

By designing the scale of camera and the trimming process of the anglec of rotation, benchmark camera X-direction and Y are accurately calculated The anglec of rotation of scale size and benchmark camera coordinates system relative to device coordinate system on direction.

In the case of benchmark camera, benchmark camera shoots and obtains datum mark picture, is obtained by circle detection algorithm process To the position of datum mark, so that the scale and the anglec of rotation of benchmark camera after being corrected.

Brief description of the drawings

Fig. 1 is the relative position relation schematic diagram of benchmark camera coordinates system and device coordinate system；

Fig. 2 is the flow chart of the present invention；

Fig. 3 be placement head in initial position datum mark in the magazine picture of benchmark；

Fig. 4 be when placement head moves along the x-axis a/2 datum mark in the magazine picture of benchmark；

Fig. 5 be when placement head moves along the x-axis-a datum mark in the magazine picture of benchmark；

Fig. 6 is that placement head moves along the x-axis a/2, and along y-axis move a/2 when datum mark in the magazine picture of benchmark；

Fig. 7 be placement head along y-axis move-a when datum mark in the magazine picture of benchmark；

Fig. 8 is chip mounter benchmark camera calibration data.

Embodiment

With reference to Fig. 1~Fig. 8 explanations：

Embodiment one：A kind of benchmark camera calibration method of chip mounter of present embodiment, it is according to the following steps Realize：

(1) determine and set up device coordinate system and benchmark camera coordinates system position each other and rotation relationship, derive The coordinate conversion relation gone out between two coordinate systems；

(2) exact position using circle detection algorithm detection reference in benchmark camera coordinates system；

(3) on the basis of datum mark, using circle detection algorithm, realize the correction to the benchmark camera of chip mounter, i.e., it is complete Into a kind of benchmark camera calibration method of chip mounter.

Embodiment two：Present embodiment from unlike embodiment one：Step (1) is specially：

Provide the coordinate transformation formula between device coordinate system and benchmark camera coordinates system here as shown in Figure 1：

Assuming that the coordinate of benchmark camera calibration to certain point is (x, y), the anglec of rotation of the benchmark camera relative to device coordinate system Spend for θ, and the scale of benchmark camera horizontally and vertically is respectively X_ScaleAnd Y_Scale, x_{Horizontal departure}And y_{Horizontal departure}Represent that benchmark camera is sat The point, then be converted to the coordinate of device coordinate system by the horizontal departure in horizontally and vertically direction between mark system and device coordinate system (x ', y ') expression formula is as follows：

X '=X_Scale(xcosθ-ysinθ)+x_{Horizontal departure}

Y '=Y_Scale(xcosθ+ysinθ)+y_{Horizontal departure}；

Wherein, the X_Scale、Y_ScaleRespectively need the scale value of the benchmark camera of correction horizontally and vertically, x_{Horizontal departure}With y_{Horizontal departure}For known value (x_{Horizontal departure}With y_{Horizontal departure}It is parallel with x and y respectively but not related between size).

Other steps and parameter are identical with embodiment one.

Embodiment three：Present embodiment from unlike embodiment one or two：Step (2) is specially：

(1) original digital image data is first obtained, using maximum between-cluster variance algorithm to original image processing, binary image is obtained；

(2) binary image profile is screened：

The first step：Binary image profile is obtained first, and generates the circular shuttering profile of known radius；

Second step：The binary image profile obtained to the first step, using pixel sum on profile as profile length, if The threshold value of minimum and maximum length is put, is screened by length, the profile between minimum and maximum length is obtained；

3rd step：Profile after being screened to second step length, calculates Hu squares, and compared with the Hu squares of circular shuttering profile Compared with the Hu squares difference retained with circular shuttering profile is less than 0.05 profile, thus obtains the profile after the screening of Hu squares；

4th step：Select contour area and the immediate profile of circular shuttering contour area after step 3 screening；

(3) step (2) is screened to the centre coordinate and length and width data of the minimum enclosed rectangle of an obtained profile, outwards Expand the 20% of rectangular aspect pixel, the centre coordinate and length and width of the rectangle after record expansion, so as to obtain binary image ROI region；

(4) morphological dilation is carried out to the profile in the ROI region of the binary image of acquisition, sets up filter graph Picture；

(5) ROI region to original image uses canny rim detections, and record falls in the side of step (4) median filter image Edge point set, when marginal point number is more than 50, carries out step (6), otherwise jumps out；

(6) minimum enclosed rectangle of (5) step edge point set is obtained, if the length and width difference of minimum enclosed rectangle is more than 20, Return to mistake；If meet, marginal point is subjected to Distance Filter, by each marginal point with minimum enclosed rectangle centre distance with most The marginal point that the difference of small boundary rectangle length and width sum a quarter is more than 4 is filtered；

(7) marginal point that the satisfaction to step (6) is required carries out least square fitting, obtains the centre bit of circular horizon point Put and radius；

Least square fitting is carried out to the marginal point that step (6) is obtained, elliptic parameter is obtained, according to oval general side Journey：

Ax²+2Bxy+Cy²+ Dx+Ey+1=0

Edge point set (the x obtained after filtering_i,y_i), its residual sum of squares (RSS) is：

According to the principle of least square, when above formula reaches minimum value, parameters are the parameter of institute's fitted ellipse；Respectively To A, B, C, D, E seeks local derviation, and makes its partial derivative be 0, then can obtain 5 by (x_i,y_i) composition equation：

Equation group is solved, that is, obtains 5 parameters A, B, C, D, E of ellipse, for arbitrary ellipse in plane, it is converted For：

A(x-u)²+2B(x-u)(y-v)+C(y-v)²+ f=0

Above formula is deployed, u can be tried to achieve according to the method for undetermined coefficients, v value, f is constant, that is, obtain the circle of circular horizon point The heart, wherein (u, v) is the oval center i.e. centre coordinate of datum mark；

Benchmark camera utilizes above-mentioned circle detection algorithm detection reference, obtains the center of circle of datum mark in benchmark camera coordinates Coordinate (x, y) in system, i.e., equal to (u, v).

Other steps and parameter are identical with embodiment one or two.

Embodiment four：Unlike one of present embodiment and embodiment one to three：Step (3) has Body is：

(1) by benchmark image center alignment fiducials point, the datum mark center of circle that step 2 is calculated is in the magazine position of benchmark The data that (x, y) has been deposited with reference to chip mounter are put, are converted to using step one in device coordinate system, then datum mark is in benchmark camera The actual deviation of the heart is (x_Base,y_Base):

x_Base=X '_Scale*(x*cosθ_{Benchmark camera}-y*sinθ_{Benchmark camera})+x_{Horizontal departure}

y_Base=Y '_Scale*(x*sinθ_{Benchmark camera}+y*cosθ_{Benchmark camera})+y_{Horizontal departure}

Wherein, the data that the chip mounter has been deposited are：X′_Scale、Y′_Scale、θ_{Benchmark camera}、x_{Horizontal departure}With y_{Horizontal departure}；

(2) benchmark camera is moved into (- x_Base,-y_Base) so that benchmark image center is moved to set datum mark Center；

(3) benchmark camera is moved along the x-axis into a/2, (a>0)；

(4) image detection algorithm of invocation step two, obtains the center of circle of datum mark in the magazine position of benchmark, is designated as (x₄,y₄)；

(5) benchmark camera moves along the x-axis-a, repeats (four), obtains benchmark dot center in the magazine position of benchmark, note For (x₃,y₃)；

(6) benchmark camera moves along the x-axis a/2, and moves a/2 along y-axis, repeats (four), obtains benchmark dot center in base Accurate magazine position, is designated as (x₂,y₂)；

(7) benchmark camera moves-a along y-axis, repeats (four), obtains benchmark dot center in the magazine position of benchmark, note For (x₁,y₁)；

(8) anglec of rotation of the calculating benchmark camera in device coordinate system

Calculate 1,2 points and 3 first, the 4 points of angles of composition straight lines in device coordinate system：

Then the anglec of rotation of the calculating benchmark camera relative to device coordinate system again：

The scale of calculating benchmark camera：

The scale of pixel is on benchmark camera x directions：

The scale of pixel is on benchmark camera y directions：

Other steps and parameter are identical with one of embodiment one to three.

Claims (3)

1. a kind of benchmark camera calibration method of chip mounter, it is characterised in that it is realized according to the following steps：

(1) determine and set up device coordinate system and benchmark camera coordinates system position each other and rotation relationship, derive two Coordinate conversion relation between individual coordinate system；

(2) exact position using circle detection algorithm detection reference in benchmark camera coordinates system；

(3) on the basis of datum mark, using circle detection algorithm, the correction to the benchmark camera of chip mounter is realized, that is, is completed A kind of benchmark camera calibration method of chip mounter；

Step (2) is specially：

(1) original digital image data is first obtained, using maximum between-cluster variance algorithm to original image processing, binary image is obtained；

(2) binary image profile is screened：

The first step：Binary image profile is obtained first, and generates the circular shuttering profile of known radius；

Second step：The binary image profile obtained to the first step, pixel sum on profile, as profile length, is set most The threshold value of small and maximum length, is screened by length, obtains the profile between minimum and maximum length；

3rd step：Profile after being screened to second step length, calculates Hu squares, and is compared with the Hu squares of circular shuttering profile, The Hu squares difference retained with circular shuttering profile is less than 0.05 profile, thus obtains the profile after the screening of Hu squares；

4th step：Select contour area and the immediate profile of circular shuttering contour area after step 3 screening；

(3) step (2) is screened to the centre coordinate and length and width data of the minimum enclosed rectangle of an obtained profile, expanded outwardly The 20% of rectangular aspect pixel, the centre coordinate and length and width of the rectangle after record expansion, so as to obtain the ROI areas of binary image Domain；

(4) morphological dilation is carried out to the profile in the ROI region of the binary image of acquisition, sets up filter graph picture；

(5) ROI region to original image uses canny rim detections, and record falls the marginal point in step (4) median filter image Collection, when marginal point number is more than 50, carries out step (6), otherwise jumps out；

(6) minimum enclosed rectangle of (5) step edge point set is obtained, if the length and width difference of minimum enclosed rectangle is more than 20, is returned Mistake；If meet, marginal point is subjected to Distance Filter, by each marginal point with the centre distance and minimum of minimum enclosed rectangle outside Marginal point of the difference more than 4 for connecing rectangular aspect sum a quarter is filtered；

(7) to step (6) satisfaction require marginal point carry out least square fitting, obtain circular horizon point center and Radius；

Least square fitting is carried out to the marginal point that step (6) is obtained, elliptic parameter is obtained, according to oval general equation：

Ax²+2Bxy+Cy²+ Dx+Ey+1=0

Edge point set (the x obtained after filtering_i,y_i), its residual sum of squares (RSS) is：

${\mathrm{\ϵ}}^2=\underset{i}{\Σ}{({{\mathrm{Ax}}_i}^2+2{\mathrm{Bx}}_i{y}_i+{{\mathrm{Cy}}_i}^2+{\mathrm{Dx}}_i+{\mathrm{Ey}}_i+1)}^2$

According to the principle of least square, when above formula reaches minimum value, parameters are the parameter of institute's fitted ellipse；Respectively to A, B, C, D, E seek local derviation, and make its partial derivative be 0, then can obtain 5 by (x_i,y_i) composition equation：

$\frac{\∂{\mathrm{\ϵ}}^2}{\∂A}=\underset{i}{\Σ}2*({{\mathrm{Ax}}_i}^2+2{\mathrm{Bx}}_i{y}_i+{{\mathrm{Cy}}_i}^2+{\mathrm{Dx}}_i+{\mathrm{Ey}}_i+1)*{x_i}^2=0$

$\frac{\∂{\mathrm{\ϵ}}^2}{\∂B}=\underset{i}{\Σ}2*({{\mathrm{Ax}}_i}^2+2{\mathrm{Bx}}_i{y}_i+{{\mathrm{Cy}}_i}^2+{\mathrm{Dx}}_i+{\mathrm{Ey}}_i+1)*2x_i{y}_i=0$

$\frac{\∂{\mathrm{\ϵ}}^2}{\∂C}=\underset{i}{\Σ}2*({{\mathrm{Ax}}_i}^2+2{\mathrm{Bx}}_i{y}_i+{{\mathrm{Cy}}_i}^2+{\mathrm{Dx}}_i+{\mathrm{Ey}}_i+1)*{y_i}^2=0$

$\frac{\∂{\mathrm{\ϵ}}^2}{\∂D}=\underset{i}{\Σ}2*({{\mathrm{Ax}}_i}^2+2{\mathrm{Bx}}_i{y}_i+{{\mathrm{Cy}}_i}^2+{\mathrm{Dx}}_i+{\mathrm{Ey}}_i+1)*x_i=0$

$\frac{\∂{\mathrm{\ϵ}}^2}{\∂E}=\underset{i}{\Σ}2*({{\mathrm{Ax}}_i}^2+2{\mathrm{Bx}}_i{y}_i+{{\mathrm{Cy}}_i}^2+{\mathrm{Dx}}_i+{\mathrm{Ey}}_i+1)*y_i=0$

Equation group is solved, that is, obtains 5 parameters A, B, C, D, E of ellipse, for arbitrary ellipse in plane, it is converted into：

A(x-u)²+2B(x-u)(y-v)+C(y-v)²+ f=0

Above formula is deployed, u can be tried to achieve according to the method for undetermined coefficients, v value, f is constant, that is, obtain the center of circle of circular horizon point, Wherein (u, v) is the oval center i.e. centre coordinate of datum mark；

Benchmark camera utilizes above-mentioned circle detection algorithm detection reference, obtains the center of circle of datum mark in benchmark camera coordinates system Coordinate (x, y), i.e., equal to (u, v).

2. the benchmark camera calibration method of a kind of chip mounter according to claim 1, it is characterised in that step (1) is specific For：

Assuming that the coordinate of benchmark camera calibration to certain point is (x, y), benchmark camera is relative to the anglec of rotation of device coordinate system θ, and the scale of benchmark camera horizontally and vertically is respectively X_ScaleAnd Y_Scale, x_{Horizontal departure}And y_{Horizontal departure}Represent benchmark camera coordinates system The horizontally and vertically horizontal departure in direction between device coordinate system, then by the point be converted to device coordinate system coordinate (x ', Y ') expression formula is as follows：

X '=X_Scale(xcosθ-_ysinθ)+x_{Horizontal departure}

Y '=Y_Scale(xcosθ+ysinθ)+y_{Horizontal departure}；

Wherein, the X_Scale、Y_ScaleRespectively need the scale value of the benchmark camera of correction horizontally and vertically, x_{Horizontal departure}With y_{Horizontal departure} For known value.

3. the benchmark camera calibration method of a kind of chip mounter according to claim 2, it is characterised in that step (3) is specific For：

(1) by benchmark image center alignment fiducials point, the datum mark center of circle that step 2 is calculated is in the magazine position of benchmark The data that (x, y) has been deposited with reference to chip mounter, are converted in device coordinate system, then datum mark is in benchmark image center using step one Actual deviation be (x_Base,y_Base):

x_Base=X '_Scale*(x*cosθ_{Benchmark camera}-y*sinθ_{Benchmark camera})+x_{Horizontal departure}

y_Base=Y '_Scale*(x*sinθ_{Benchmark camera}+y*cosθ_{Benchmark camera})+y_{Horizontal departure}

Wherein, the data that the chip mounter has been deposited are：X′_Scale、Y′_Scale、θ_{Benchmark camera}、x_{Horizontal departure}With y_{Horizontal departure}；

(2) benchmark camera is moved into (- x_Base,-y_Base) so that benchmark image center is moved to the center of set datum mark；

(3) benchmark camera is moved along the x-axis into a/2, (a>0)；

(4) image detection algorithm of invocation step two, obtains the center of circle of datum mark in the magazine position of benchmark, is designated as (x₄, y₄)；

(5) benchmark camera moves along the x-axis-a, repeats (four), obtains benchmark dot center in the magazine position of benchmark, be designated as (x₃, y₃)；

(6) benchmark camera moves along the x-axis a/2, and moves a/2 along y-axis, repeats (four), obtains benchmark dot center in benchmark phase Position in machine, is designated as (x₂,y₂)；

(7) benchmark camera moves-a along y-axis, repeats (four), obtains benchmark dot center in the magazine position of benchmark, be designated as (x₁, y₁)；

(8) anglec of rotation of the calculating benchmark camera in device coordinate system

Calculate 1,2 points and 3 first, the 4 points of angles of composition straight lines in device coordinate system：

${\mathrm{\θ}}_1={\tan }^{-1}\frac{x_1-x_2}{y_2-y_1}$

${\mathrm{\θ}}_2={\tan }^{-1}\frac{y_4-y_3}{x_3-x_4}$

Then the anglec of rotation of the calculating benchmark camera relative to device coordinate system again：

$\mathrm{\θ}=\frac{{\mathrm{\θ}}_1+{\mathrm{\θ}}_2}{2}$

The scale of calculating benchmark camera：

The scale of pixel is on benchmark camera x directions：

$X_{Scale}=\frac{a}{\sqrt{{(x_2-x_1)}^2+{(y_2-y_1)}^2}}$

The scale of pixel is on benchmark camera y directions：

$Y_{Scale}=\frac{a}{\sqrt{{(x_4-x_3)}^2+{(y_4-y_3)}^2}}.$