CN102107179B - Method for controlling single-layer leather gluing based on binocular vision - Google Patents

Abstract

The invention provides a method for controlling single-layer leather gluing based on binocular vision. As a binocular camera and a three-shaft gluing manipulator are adopted, the method comprises the following steps of: 1) performing three-dimensional calibration on the fixed binocular camera; 2) collecting a binocular image of single-layer leather and extracting an edge outline of the binocular image; 3) extracting characteristics according to the edge outline image in the step 2); 4) matching according to the characteristic points in the step 3); 5) reducing real coordinate values of a gluing platform system corresponding to the characteristic points in the step 4) according to a weight projection matrix value; and 6) inputting the real coordinate values of the characteristic points into a gluing coordinate queue of the three-shaft gluing manipulator to control the manipulator to finish the gluing. By the method, the working efficiency is improved, the single-layer leather is conveniently integrated to the processing of a production line, the precision is high, and harms to the health of workers can be avoided.

Description

A kind of individual layer leather gluing control method based on binocular vision

Technical field

The present invention relates to leather glue spreading method field, relate to a kind of individual layer leather glue spreading method, especially a kind of comprising detection, discern, measure the individual layer leather gluing control method of gluing entire flow based on binocular vision based on binocular vision.

Background technology

The leather gluing is mainly finished by manual in the present actual production of China, and the toxic gas inhomogeneous and that coating distributes of manual gluing all is the subject matter of leather gluing.The inhomogeneous difficulty that causes the stitching of leather later stage of manual gluing; can have influence on the efficient that leather is produced; and because the toxic gas that the gluing process is distributed can greatly influence workers'health again; therefore adopt the gluing means of mechanization can not only improve efficient and the technological level that leather is produced, also can effectively protect workers'health.

In existing image processing techniques and the binocular vision technology, most of function is comprehensive inadequately, also there be not a cover aspect the individual layer leather gluing from detecting, discern, measure the complete method of gluing, it as application number 200810232122.8 invention, the pixel equivalent of utilizing the demarcation thing of known geometric area to try to achieve camera can be influenced by computational accuracy, the leather area of trying to achieve may not be suitable for further commercial Application, obviously obtains the edge of leather and the characteristic point of processing and is more suitable for commercial Application.Application number is 200710190470.9 invention, utilize vision technique to discern the classification and the existing parameter gluing of basis of windshield, just simply utilized the visual identity technology, need sample and the step of input feature vector database for unsampled windshield, comparatively loaded down with trivial details, the shape of windshield is comparatively simple simultaneously.Application number is 200710123727.9 invention, under the driving of three servo motion controllers, carry out gluing according to three-dimensional or two-dimensional points sequence, but numerous three-dimensional point or two-dimensional points can influence the efficient of gluing, thereby should get the quantity that its Origin And Destination reduces point for collinear two-dimensional points.

Summary of the invention

For the inefficiency that overcomes existing pure manual leather glue spreading method, be unfavorable for being integrated into that streamline processing, precision are low, the deficiency of harm workers ' health, the invention provides and a kind ofly promote operating efficiency, be convenient to the individual layer leather gluing control method that is integrated into streamline processing, precision height, avoids endangering workers ' health based on binocular vision.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of individual layer leather gluing control method based on binocular vision, this method adopts binocular camera and three gluing mechanical arms, and described individual layer leather gluing control method may further comprise the steps:

1), fixing binocular camera being carried out solid demarcates;

2), gather individual layer leather binocular image and extract its edge contour;

3), according to step 2) the edge contour image carry out feature extraction:

Individual layer leather edge contour by limited to two-dimensional coordinate point (x _i, y _i) constitute, i=1,2, L, n represent that this profile is made up of coordinate points n; For on the profile arbitrarily coordinate points to (x _iy _i), its curvature computing formula is calculated the right curvature K of all coordinate points as (1) _i, i=1,2, L, n represent that this profile has n to coordinate;

$K_i=\frac{\left(\frac{x_{i+1}-x_{i-1}}{2}\right){\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}^2-{\left(\frac{x_{i+1}-x_{i-1}}{2}\right)}^2\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}{\sqrt[3]{{[{\left(\frac{x_{i+1}-x_{i-1}}{2}\right)}^2+{\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}^2]}^2}},$ i＝1，2，L，n(1)

Wherein, (x _I+1, y _I+1) be that this coordinate points is to (x _i, y _i) next coordinate points right, (x _I-1, y _I-1) be that this coordinate points is to (x _i, y _i) a last coordinate points right.

Feature extraction is according to following steps:

(3.1), with

Calculate profile local curvature threshold value, B is a proportionality coefficient, is selected between [1,2];

(3.2), satisfy K _iThe coordinate points of 〉=T (i) can be selected into candidate feature point P _i, i=1,2, L, s＜n represent total s candidate feature point;

(3.3), calculate two formed angle theta of tangent line of each candidate feature point _I+1: with the some P of candidate feature point _iBe starting point, P _I+1Be terminal point, the mid point P of Origin And Destination place line segment _{(i+ (i+1))/2}, according to formula (2) theory of computation center of circle

In like manner, with P _I+1Be starting point, P _I+2Be terminal point, the mid point P of Origin And Destination place line segment _{((i+1)+(i+2))/2}, get its theoretical center of circle by formula (2)

$x_{i+1}^C=\frac{(x_{i+1}^2+y_{i+1}^2)(y_{\frac{i+(i+1)}{2}}-y_i)+(x_{\frac{i+(i+1)}{2}}^2+y_{\frac{i+(i+1)}{2}}^2)(y_i-y_{i+1})+(x_{\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="HDA0000038198400000031.png,HDA0000038198400000051.png"\; state="\{\{state\}\}">i</figure-callout>}}^2+y_i^2)(y_{i+1}-y_{\frac{i+(i+1)}{2}})}{2[x_{i+1}(y_{\frac{i+(i+1)}{2}}-y_i)+x_{\frac{i+(i+1)}{2}}(y_i-y_{i+1})+x_i(y_{i+1}-y_{\frac{i+(i+1)}{2}})]},---\left(2\right)$

$y_{i+1}^C=\frac{(x_{i+1}^2+y_{i+1}^2)(x_{\frac{i+(i+1)}{2}}-x_i)+(x_{\frac{i+(i+1)}{2}}^2+y_{\frac{i+(i+1)}{2}}^2)(x_i-x_{i+1})+({\mathrm{<figure-callout\; id="2"\; label="x\; i"\; filenames="HDA0000038198400000031.png,HDA0000038198400000051.png"\; state="\{\{state\}\}">x</figure-callout>}}_i^{}+y_i^2)(x_{i+1}-x_{\frac{i+(i+1)}{2}})}{2[y_{i+1}(x_{\frac{i+(i+1)}{2}}-x_i)+y_{\frac{i+(i+1)}{2}}(x_i-x_{i+1})+y_i(x_{i+1}-x_{\frac{i+(i+1)}{2}})]}$

Wherein, x _I+1, y _I+1) be P _I+1Coordinate, For

Coordinate,

For

Coordinate;

Calculate P with formula (3) then _I+1With

Line and P _I+1The angle of tangent line

And P _I+1With Line and P _I+1The angle of tangent line

${\mathrm{\&tau;}}_{i+1}^L=a\tan \frac{(y_{i+1}^C-y_{i+1})}{(x_{i+1}^C-x_{i+1})}$

${\mathrm{\&tau;}}_{i+1}^R=a\tan \frac{(y_{i+2}^C-y_{i+1})}{(x_{i+2}^C-x_{i+1})}---\left(3\right)$

Wherein, θ _I+1Be two angles With

Sum;

(3.4), according to θ _I+1Value judges whether it is line segment shape between required characteristic point and the marker characteristic point: if θ _I+1It is not obvious to think then that greater than 170 degree the residing local configuration curvature of this characteristic point changes, and is not required characteristic point, and the line segment between the characteristic point is shaped as Straight; Otherwise this characteristic point is required characteristic point, and the line segment between the characteristic point is shaped as Arc;

4), mate according to the characteristic point of step 3), characteristic matching is according to following steps:

(4.1), with the characteristic point of the left figure in the binocular image Be reference, i=0,1,2, L, m, wherein, m＜s, m represent that left figure has m characteristic point;

(4.2), the right figure in calculating and the binocular image is to all characteristic points in 2 pixels around there won't be any problem

Euclidean distance, i=0,1,2, L, k, wherein, k＜m, k represent k characteristic point arranged in 2 pixel coverages in the right figure m characteristic point;

(4.3), get the Euclidean distance minimum As the characteristic point that matches, thus the characteristic point one-to-one relationship of figure about definite individual layer leather edge contour;

5), according to formula (7) reduction step 4) characteristic point the true coordinate value of corresponding gluing platform system;

6), the true coordinate value of characteristic point is input in the gluing coordinate formation of three gluing mechanical arms, thus the control mechanical arm is finished the gluing process.

Further, in the described step 6), line segment shape and the formula (4) according to step 3) between the characteristic point carried out gluing control;

$L(P_i,P_{i+1})=\left\{\begin{array}{c}\mathrm{Straigt}\\ \mathrm{Arc}\end{array}\right.,i=\mathrm{0,1,2},\mathrm{Lm}-1---\left(4\right)$

Wherein, L (P _i, P _I+1) expression P _iWith P _I+1Individual layer leather edge line section shape between the individual characteristic point, Straight represents line segment, carries out gluing according to the linear interpolation mode during gluing, Arc represents arc, carries out gluing, i=0 according to the curve interpolating mode during gluing, 1,2, L m-1 represents to have m edge contour characteristic point.

Further again, in the described step 1), the demarcation thing of taking different attitudes obtains to be used to demarcate the image of analysis, obtains intrinsic parameter (the k l u of each camera according to linear camera model (5) ₀v ₀F θ) and outer parameter, 3 * 4 matrix M that obtain by camera inside and outside parameter dot product _{3 * 4}Be perspective projection matrix;

$\left(\begin{array}{c}\mathrm{<figure-callout\; id="1"\; label="u\; v"\; filenames="HDA0000038198400000031.png,HDA0000038198400000051.png"\; state="\{\{state\}\}">u</figure-callout>}& \\ v\\ 1\end{array}\right)=\frac{1}{\mathrm{\&lambda;}}\left(\begin{array}{cccc}\frac{f}{k}& -\frac{f}{k}\cot \mathrm{\&theta;}& u_0& 0\\ 0& \frac{f}{l\sin \mathrm{\&theta;}}& v_0& 0\\ 0& 0& 1& 0\end{array}\right)\left(\begin{array}{cc}{\mathrm{<figure-callout\; id="3"\; label="R"\; filenames="HDA0000038198400000031.png,HDA0000038198400000041.png"\; state="\{\{state\}\}">R</figure-callout>}}_{3\&times;3}& {\mathrm{<figure-callout\; id="3"\; label="t"\; filenames="HDA0000038198400000031.png,HDA0000038198400000041.png"\; state="\{\{state\}\}">t</figure-callout>}}_{3\&times;1}\\ 0^T& 1\end{array}\right)\left(\begin{array}{c}{X}_W\\ Y_{\mathrm{<figure-callout\; id="1"\; label="W\; Z\; W"\; filenames="HDA0000038198400000031.png,HDA0000038198400000051.png"\; state="\{\{state\}\}">W</figure-callout>}}& \\ Z_W{}_{}\\ 1\end{array}\right)={\mathrm{<figure-callout\; id="3"\; label="M"\; filenames="HDA0000038198400000031.png,HDA0000038198400000041.png"\; state="\{\{state\}\}">M</figure-callout>}}_{3\&times;4}\left(\begin{array}{c}{X}_W\\ Y_W\\ Z_W\\ 1\end{array}\right)---\left(5\right)$

Wherein, k, l are the pixel sizes of binocular camera, u ₀, v ₀Be the coordinate figure of the optical axis center of camera, f is a camera focus, and θ is the degree of skewness (being generally 90 °) of camera coordinates system, R _{3 * 3}Be the spin matrix of camera, t _{3 * 1}Then be the translation matrix of camera, (u v 1) ^TBe any point in the image, (X _WY _WZ _W1) ^TIt then is the pairing gluing coordinate system of any point coordinate in the image;

The process that image is corrected is as follows:

(1.1), the inside and outside parameter by formula (6) and left and right sides camera obtains rotation and the translation matrix of right camera with respect to left camera, and correct the left and right sides image trip aimed at;

R＝R _r(R _l)T

(6)

T＝T _r-RT _l

Wherein, R _l, R _rBe the spin matrix of left and right sides camera, T _l, T _rBe the translation matrix of left and right sides camera, R, T are spin matrix and the translation matrix of right camera with respect to left camera.

(1.2), the camera inside and outside parameter by formula (7) and front obtains re-projection matrix Q:

$Q=\left(\begin{array}{cccc}1& 0& 0& -c_x\\ 0& 1& 0& -c_y\\ 0& 0& 0& \mathrm{<figure-callout\; id="0"\; label="f"\; filenames="HDA0000038198400000081.png"\; state="\{\{state\}\}">f</figure-callout>}\\ 0& 0& 1/T_x& 0\end{array}\right)---\left(7\right)$

Wherein, T _xBe the x axle component that binocular camera is joined translation vector T outward, c _xAnd c _yBe the world coordinate system coordinate figure at left camera optics center, f is the focal length of left camera.

Described step 2) in, convert individual layer leather image to gray level image according to the different threshold filters of finishing of the rgb value of leather and noise, detailed process is as follows:

(2.1), use Gaussian function that image is carried out smoothly;

(2.2), the closed operation in the employing mathematical morphology;

(2.3), adopt the leather edge contour of the unidirectional element of Canny operator extraction.

Described step 2) in, around the gluing platform, light source is installed, directly over the center of described gluing platform, is arranged binocular camera.

Technical conceive of the present invention is: utilize vision and image technique, by binocular camera acquisition of image data and analyzing and processing, can obtain comparatively accurate leather characteristic point and change into the machining coordinate of glue spreading apparatus, the advantage of this method is contactless gluing, the enforcement difficulty is low, cost is low, and the accuracy height does not have harm to human body.

The unknown of individual layer leather edge shape, if treated leather then edge is comparatively regular, undressed leather then edge is comparatively complicated, when finishing rim detection, also need to discern the edge contour characteristic point based on the curvature method with what improve, when gluing, be the gluing control point with the characteristic point point sequence, with the edge line segment between the characteristic point serves as with reference to straight line or the curve gluing mode of determining three shaft mechanical arms, and the present invention is exactly according to above-mentioned purpose and requires effectively to have finished from detecting, discern, measure the complete individual layer leather glue spreading method of gluing.

This cognition technology based on vision is dissolved into conventional leather gluing process, the intelligent level and the automaticity of its process will effectively be promoted, can avoid because of the potential hazard of gluing process generation workers ' health, realize the closed-loop control flow process, not only ensured workers'health but also improved production efficiency.

Obtain the individual layer leather image for the treatment of gluing by the binocular camera that is fixed on directly over the gluing platform center, because the effect of light source provides higher picture quality, passing threshold filtering, Gaussian function and closed operation are carried out level and smooth noise reduction to image, obtain comparatively smooth leather gray level image, extract the leather edge contour that obtains single pixel by Canny operator, extract edge contour characteristic point and coupling by improved feature point extraction algorithm again based on curvature, at last, obtain the coordinate figure of its corresponding gluing coordinate system by the characteristic point reduction of coupling, finish the gluing process in conjunction with the line segment shape of edge contour between the characteristic point.

Beneficial effect of the present invention mainly shows: 1, entire method has comprised detection to leather, has discerned, measured the entire flow of gluing, helps the industrial automation of gluing, and does not need the workman to participate; 2, the glue spreading method based on binocular vision has noncontact, high-precision advantage; 3, the improved effective extract minutiae of feature point extraction algorithm energy based on curvature, and reduced the number of coordinates of importing gluing mechanical arm job sequence, improve gluing speed.

Description of drawings

Fig. 1 is automatic double surface gluer indication device figure of the present invention.

Fig. 2 is a glue spreading method flow chart of the present invention.

Fig. 3 is left and right sides camera edge extracting result's a schematic diagram, and wherein, left hand view is the edge extracting result of left camera, and right part of flg is the edge extracting result of right camera, first kind of leather example print of (1) expression; (2) second kind of leather example print of expression; (3) represent the third leather example print.

Fig. 4 is left and right sides camera Edge Gradient Feature result's a schematic diagram, wherein, the black surround frame table shows the feature extraction result, and left hand view is the Edge Gradient Feature result of left camera, right part of flg is the Edge Gradient Feature result of right camera, first kind of leather example print of (1) expression; (2) second kind of leather example print of expression; (3) represent the third leather example print.

Fig. 5 is two angle theta that tangent line forms of candidate feature point _I+1Schematic diagram.

Fig. 6 is left and right sides Feature Points Matching result's a schematic diagram, and wherein, black numbers represents to match characteristic point, and left hand view is left figure Feature Points Matching result, and right part of flg is right figure Feature Points Matching result, first kind of leather example print of (1) expression; (2) second kind of leather example print of expression; (3) represent the third leather example print.

Fig. 7 is the schematic diagram that reconstructed results symbol quadrant distributes.

Fig. 8 is the schematic diagram of the error analysis of reconstructed results and actual value.

The specific embodiment

Below in conjunction with accompanying drawing the present invention is further described.

With reference to Fig. 1～Fig. 8, a kind of individual layer leather gluing control method based on binocular vision, this method adopts binocular camera and three gluing mechanical arms, and described individual layer leather gluing control method may further comprise the steps:

1), fixing binocular camera being carried out solid demarcates;

2), gather individual layer leather binocular image and extract its edge contour;

3), according to step 2) the edge contour image carry out feature extraction:

Individual layer leather edge contour by limited to two-dimensional coordinate point (x _i, y _i) constitute, i=1,2, L, n represent that this profile is made up of coordinate points n; For on the profile arbitrarily coordinate points to (x _i, y _i), its curvature computing formula is calculated the right curvature K of all coordinate points as (1) _i, i=1,2, L, n represent that this profile has n to coordinate;

$K_i=\frac{\left(\frac{x_{i+1}-x_{i-1}}{2}\right){\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}^2-{\left(\frac{x_{i+1}-x_{i-1}}{2}\right)}^2\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}{\sqrt[3]{{[{\left(\frac{x_{i+1}-x_{i-1}}{2}\right)}^2+{\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}^2]}^2}},i=\mathrm{1,2},L,n---\left(1\right)$

Wherein, (x _I+1, y _I+1) be that this coordinate points is to (x _i, y _i) next coordinate points right, (x _I-1, y _I-1) be that this coordinate points is to (x _i, y _i) a last coordinate points right.

Feature extraction is according to following steps:

(3.1), with

Calculate profile local curvature threshold value, B is a proportionality coefficient, is selected between [1,2];

(3.2), satisfy K _iThe coordinate points of 〉=T (i) can be selected into candidate feature point P _i, i=1,2, L, s＜n represent total s candidate feature point;

(3.3), calculate two formed angle theta of tangent line of each candidate feature point _I+1: with the some P of candidate feature point _iBe starting point, P _I+1Be terminal point, the mid point P of Origin And Destination place line segment _{(i+ (i+1))/2}, according to formula (2) theory of computation center of circle In like manner, with P _I+1Be starting point, P _I+2Be terminal point, the mid point P of Origin And Destination place line segment _{((i+1)+(i+2))/2}, get its theoretical center of circle by formula (2)

$x_{i+1}^C=\frac{(x_{i+1}^2+y_{i+1}^2)(y_{\frac{i+(i+1)}{2}}-y_i)+(x_{\frac{i+(i+1)}{2}}^2+y_{\frac{i+(i+1)}{2}}^2)(y_i-y_{i+1})+({\mathrm{<figure-callout\; id="2"\; label="x\; i"\; filenames="HDA0000038198400000031.png,HDA0000038198400000051.png"\; state="\{\{state\}\}">x</figure-callout>}}_i^{}+y_i^2)(y_{i+1}-y_{\frac{i+(i+1)}{2}})}{2[x_{i+1}(y_{\frac{i+(i+1)}{2}}-y_i)+x_{\frac{i+(i+1)}{2}}(y_i-y_{i+1})+x_i(y_{i+1}-y_{\frac{i+(i+1)}{2}})]},---\left(2\right)$

$y_{i+1}^C=\frac{(x_{i+1}^2+y_{i+1}^2)(x_{\frac{i+(i+1)}{2}}-x_i)+(x_{\frac{i+(i+1)}{2}}^2+y_{\frac{i+(i+1)}{2}}^2)(x_i-x_{i+1})+({\mathrm{<figure-callout\; id="2"\; label="x\; i"\; filenames="HDA0000038198400000031.png,HDA0000038198400000051.png"\; state="\{\{state\}\}">x</figure-callout>}}_i^{}+y_i^2)(x_{i+1}-x_{\frac{i+(i+1)}{2}})}{2[y_{i+1}(x_{\frac{i+(i+1)}{2}}-x_i)+y_{\frac{i+(i+1)}{2}}(x_i-x_{i+1})+y_i(x_{i+1}-x_{\frac{i+(i+1)}{2}})]}$

Wherein, (x _I+1, y _I+1) be P _I+1Coordinate,

For

Coordinate,

For

Coordinate;

Calculate P with formula (3) then _I+1With

Line and P _I+1The angle of tangent line

And P _I+1With

Line and P _I+1The angle of tangent line

${\mathrm{\&tau;}}_{i+1}^L=a\tan \frac{(y_{i+1}^C-y_{i+1})}{(x_{i+1}^C-x_{i+1})}$

${\mathrm{\&tau;}}_{i+1}^R=a\tan \frac{(y_{i+2}^C-y_{i+1})}{(x_{i+2}^C-x_{i+1})}---\left(3\right)$

Wherein, θ _I+1Be two angles

With

Sum;

(3.4), according to θ _I+1Value judges whether it is line segment shape between required characteristic point and the marker characteristic point: if θ _I+1It is not obvious to think then that greater than 170 degree the residing local configuration curvature of this characteristic point changes, and is not required characteristic point, and the line segment between the characteristic point is shaped as Straight; Otherwise this characteristic point is required characteristic point, and the line segment between the characteristic point is shaped as Arc;

4), mate according to the characteristic point of step 3), characteristic matching is according to following steps:

(4.1), with the characteristic point of the left figure in the binocular image

Be reference, i=0,1,2, L, m, wherein, m＜s, m represent that left figure has m characteristic point;

(4.2), the right figure in calculating and the binocular image is to all characteristic points in 2 pixels around there won't be any problem

Euclidean distance, i=0,1,2, L, k, wherein, k＜m, k represent k characteristic point arranged in 2 pixel coverages in the right figure m characteristic point;

(4.3), get the Euclidean distance minimum

As the characteristic point that matches, thus the characteristic point one-to-one relationship of figure about definite individual layer leather edge contour;

5), according to formula (6) reduction step 4) characteristic point the true coordinate value of corresponding gluing platform system;

6), the true coordinate value of characteristic point is input in the gluing coordinate formation of three gluing mechanical arms, thus the control mechanical arm is finished the gluing process.

In the present embodiment, four light sources are housed around the gluing platform, can eliminate the shade for the treatment of gluing leather edge, distinct image more is provided, the binocular vision collecting device of directly over the platform center, forming by two cameras, this equipment is linked to each other with host computer by the USB line, and computer then can be finished IMAQ and analysis subsequently and finish individual layer leather gluing function.

1), fixing binocular camera being carried out solid demarcates and correcting image.Binocular camera is fixed in directly over the center of gluing platform, obtains to be used to demarcate the image of analysis by the demarcation thing (chequered with black and white gridiron pattern scaling board, each gridiron pattern size is all known on the plate) of taking different attitudes.According to (Heikkila, Silven, " A Four-step Camera Calibration Procedurewith Implicit Image Correction ", CVPR97, p:1106-1112) the linear camera model formation (4) in obtains intrinsic parameter (the k l u of each camera ₀v ₀F θ) and outer parameter (spin matrix R _{3 * 3}, translation matrix t _{3 * 1}).3 * 4 matrix M that obtain by camera inside and outside parameter dot product _{3 * 4}Be perspective projection matrix.

$\left(\begin{array}{c}u\\ v\\ 1\end{array}\right)=\frac{1}{\mathrm{\&lambda;}}\left(\begin{array}{cccc}\frac{f}{k}& -\frac{f}{k}\cot \mathrm{\&theta;}& {\mathrm{<figure-callout\; id="0"\; label="u"\; filenames="HDA0000038198400000081.png"\; state="\{\{state\}\}">u</figure-callout>}}_0& 0\\ 0& \frac{f}{l\sin \mathrm{\&theta;}}& {\mathrm{<figure-callout\; id="0"\; label="v"\; filenames="HDA0000038198400000081.png"\; state="\{\{state\}\}">v</figure-callout>}}_0& 0\\ 0& 0& 1& 0\end{array}\right)\left(\begin{array}{cc}{\mathrm{<figure-callout\; id="3"\; label="R"\; filenames="HDA0000038198400000031.png,HDA0000038198400000041.png"\; state="\{\{state\}\}">R</figure-callout>}}_{3\&times;3}& {\mathrm{<figure-callout\; id="3"\; label="t"\; filenames="HDA0000038198400000031.png,HDA0000038198400000041.png"\; state="\{\{state\}\}">t</figure-callout>}}_{3\&times;1}\\ 0^T& 1\end{array}\right)\left(\begin{array}{c}{X}_W\\ Y_W\\ Z_W\\ 1\end{array}\right)---\left(4\right)$

Wherein, k, l are the pixel sizes of binocular camera, u ₀, v ₀Be the coordinate figure of the optical axis center of camera, f is a camera focus, and θ is the degree of skewness (being generally 90 °) of camera coordinates system, R _{3 * 3}Be the spin matrix of camera, t _{3 * 1}Then be the translation matrix of camera, (u v 1) ^TBe any point in the image, (X _WY _WZ _W1) ^TIt then is the pairing gluing coordinate system of any point coordinate in the image.Calculate the M of left and right sides camera according to formula (4) _{3 * 4}Matrix is as follows:

${\mathrm{<figure-callout\; id="3"\; label="M"\; filenames="HDA0000038198400000031.png,HDA0000038198400000041.png"\; state="\{\{state\}\}">M</figure-callout>}}_{3\&times;4}=\left(\begin{array}{cccc}840.95380999999998& 0& 394.97588729858398& 0\\ 0& 840.95380999999998& 243.11664962768555& 0\\ 0& 0& 1& 0\end{array}\right)$

(1.1), the outer parameter by binocular camera, obtain rotation and the translation vector of right camera with formula (5), and correct the left and right sides image trip is aimed at respect to left camera.

R＝R _r(R _l)T

(5)

T＝T _r-RT _l

Wherein, R _l, R _rBe the spin matrix of left and right sides camera, T _l, T _rBe the translation matrix of left and right sides camera, R, T are spin matrix and the translation matrix of right camera with respect to left camera.As follows according to the result that formula (5) obtains:

$R=\left(\begin{array}{ccc}0.9995178471005625& 0.0215390900270048& -0.0223638308047886\\ -0.0220500783453662& 0.9994954940280076& -0.0228593845646005\\ 0.0218601777764921& 0.0233414870674015& 0.9994885230001706\end{array}\right)$

T＝(-110892622920883670.06314981629290070.0921180076963402)

(1.2), the camera inside and outside parameter by formula (6) and front obtains re-projection matrix Q:

$Q=\left(\begin{array}{cccc}1& 0& 0& -{\mathrm{<figure-callout\; id="0"\; label="c\; x"\; filenames="HDA0000038198400000081.png"\; state="\{\{state\}\}">c</figure-callout>}}_x\\ 0& 1& 0& -{\mathrm{<figure-callout\; id="0"\; label="c\; y"\; filenames="HDA0000038198400000081.png"\; state="\{\{state\}\}">c</figure-callout>}}_y\\ 0& 0& 0& \mathrm{<figure-callout\; id="0"\; label="f"\; filenames="HDA0000038198400000081.png"\; state="\{\{state\}\}">f</figure-callout>}\\ 0& 0& 1/T_x& 0\end{array}\right)---\left(6\right)$

Wherein,

Be the x axle component that binocular camera is joined translation vector T outward, c _xAnd c _yBe the world coordinate system coordinate figure at left camera optics center, f is the focal length of left camera.As follows according to the result that formula (6) obtains:

$Q=\left(\begin{array}{cccc}1& 0& 0& -394.97588729858398\\ 0& 1& 0& -243.11664962768555\\ 0& 0& 0& 840.95380999999998\\ 0& 0& 0.9022405163464954& 0\end{array}\right)$

2), gather individual layer leather binocular image and extract its edge contour.Profile extracts the result as shown in Figure 3

(2.1), because the light source effect of gluing platform, the present invention with individual layer leather image according to the different threshold filters of finishing of the rgb value of leather and noise.Threshold filter promptly is to choose suitable gray value in [0,1,2, L, 255] interval, can keep greater than the pixel of this gray value;

(2.2), the present invention uses Gaussian function that image is carried out smoothly, reduces the noise that may the exist influence to the leather edge with this.The present invention chooses 3 * 3 gaussian kernel function (G.Bradski, A.Kaebler, " Learning OpenCV ", 2008,110) to the gray-scale map smoothing processing through threshold filter;

(2.3), the present invention adopt closed operation in the mathematical morphology guarantee preserving edge when extracting complete edge as far as possible scrambling (list of references: G.Bradski, A.Kaebler, " Learning OpenCV ", 2008,120-121).Structural element in the closed operation is that 3 * 3 rectangular configuration is carried out interative computation according to the order of first expansion post-etching one time to level and smooth image, makes image more smooth;

(2.4), the present invention adopts is based on Tuscany (Canny) operator (Fleck, " Somedefects in finite-difference edge finders ", IEEE PAMI, 1992,14 (3): 337-345) individual layer leather image border profile is carried out single pixel extraction, wherein, the low threshold value and the high threshold of Canny operator are elected [0.3 0.35] as.

3), according to step 2) the edge contour image carry out feature extraction.In fact individual layer leather edge contour is exactly to two-dimensional coordinate point (x by limited _iy _i) constitute, the present invention adopt improvement based on the profile coordinate between the mode (X.C.He of curvature, " Curvature Scale SpaceCorner Detector with Adaptive Threshold and Dynamic Region ofSupport ", ICPR, 2004, vo1.2:791-794) extract the characteristic point of profile.For on the profile arbitrarily coordinate points to (x _i, y _i), i=1,2, L, n represent that this profile is made up of coordinate points n, its curvature computing formula is as (1).Feature point extraction result as shown in Figure 4, black surround frame point representation feature point.

$K_i=\frac{\left(\frac{x_{i+1}-x_{i-1}}{2}\right){\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}^2-{\left(\frac{x_{i+1}-x_{i-1}}{2}\right)}^2\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}{\sqrt[3]{{[{\left(\frac{x_{i+1}-x_{i-1}}{2}\right)}^2+{\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}^2]}^2}},i=\mathrm{1,2},L,n---\left(1\right)$

(3.1), with

Calculate profile local curvature threshold value, B is a proportionality coefficient, and it is comparatively reasonable to be selected between [1,2], elects 1.5 among the present invention as;

(3.2), satisfy K _iThe coordinate points of 〉=T (i) can be selected into candidate feature point P _i, i=1,2, L, s＜n represent total s candidate feature point;

(3.3), calculate two formed angle theta of tangent line of each candidate feature point _I+1: with the some P of candidate feature point _iBe starting point, P _I+1Be terminal point, the mid point P of Origin And Destination place line segment _{(i+ (i+1)/2}, according to formula (2) theory of computation center of circle In like manner, with P _I+1Be starting point, P _I+2Be terminal point, the mid point P of Origin And Destination place line segment _{((i+1)+(i+22))/2}, get its theoretical center of circle by formula (2)

$x_{i+1}^C=\frac{(x_{i+1}^2+y_{i+1}^2)(y_{\frac{i+(i+1)}{2}}-y_i)+(x_{\frac{i+(i+1)}{2}}^2+y_{\frac{i+(i+1)}{2}}^2)(y_i-y_{i+1})+({\mathrm{<figure-callout\; id="2"\; label="x\; i"\; filenames="HDA0000038198400000031.png,HDA0000038198400000051.png"\; state="\{\{state\}\}">x</figure-callout>}}_i^{}+y_i^2)(y_{i+1}-y_{\frac{i+(i+1)}{2}})}{2[x_{i+1}(y_{\frac{i+(i+1)}{2}}-y_i)+x_{\frac{i+(i+1)}{2}}(y_i-y_{i+1})+x_i(y_{i+1}-y_{\frac{i+(i+1)}{2}})]},---\left(2\right)$

$y_{i+1}^C=\frac{(x_{i+1}^2+y_{i+1}^2)(x_{\frac{i+(i+1)}{2}}-x_i)+(x_{\frac{i+(i+1)}{2}}^2+y_{\frac{i+(i+1)}{2}}^2)(x_i-x_{i+1})+({\mathrm{<figure-callout\; id="2"\; label="x\; i"\; filenames="HDA0000038198400000031.png,HDA0000038198400000051.png"\; state="\{\{state\}\}">x</figure-callout>}}_i^{}+y_i^2)(x_{i+1}-x_{\frac{i+(i+1)}{2}})}{2[y_{i+1}(x_{\frac{i+(i+1)}{2}}-x_i)+y_{\frac{i+(i+1)}{2}}(x_i-x_{i+1})+y_i(x_{i+1}-x_{\frac{i+(i+1)}{2}})]}$

Wherein, (x _I+1, y _I+1) be P _I+1Coordinate,

For Coordinate,

For

Coordinate;

Calculate P with formula (3) then _I+1With Line and P _I+1The angle of tangent line

And P _I+1With

Line and P _I+1The angle of tangent line

${\mathrm{\&tau;}}_{i+1}^L=a\tan \frac{(y_{i+1}^C-y_{i+1})}{(x_{i+1}^C-x_{i+1})}$

${\mathrm{\&tau;}}_{i+1}^R=a\tan \frac{(y_{i+2}^C-y_{i+1})}{(x_{i+2}^C-x_{i+1})}---\left(3\right)$

Wherein, θ _I+1Be two angles With

Sum;

(3.4), according to θ _I+1Value judges whether it is line segment shape between required characteristic point and the marker characteristic point: if θ _I+1It is not obvious to think then that greater than 170 degree the residing local configuration curvature of this characteristic point changes, and is not required characteristic point, and the line segment between the characteristic point is shaped as Straight; Otherwise this characteristic point is required characteristic point, and the line segment between the characteristic point is shaped as Arc.

4), mate according to the characteristic point of step 3).About figure characteristic matching result as shown in Figure 6, black surround frame point representation feature point.

(4.1), with the characteristic point of left figure

For with reference to (i=0,1,2, L, m＜s represent that left figure has m characteristic point);

(4.2), calculate figure right to all characteristic points in 2 pixels there won't be any problem on every side with it

Euclidean distance (i=0,1,2, L, k, k＜m represent k characteristic point arranged in 2 pixel coverages in the right figure m characteristic point);

(4.3), get the Euclidean distance minimum

As the characteristic point that matches, thus the characteristic point one-to-one relationship of figure about definite individual layer leather edge contour.The Euclidean distance computing formula as shown in Equation (8) (list of references: Yang Shuying, " pattern-recognition and intelligence computation---Matlab technology realize ", 2008:54).

${\mathrm{Euclidean}}_i=\sqrt{{(x_L^i-x_R^i)}^2+{(y_L^i-y_R^i)}^2}---\left(8\right)$

Wherein, Euclidean _iExpression

With

Minimum euclidean distance,

For

Coordinate figure,

For

Coordinate figure.

5), according to formula (6) reduction step 4) characteristic point the true coordinate value of corresponding gluing platform system.For each characteristic point, obtaining this characteristic point image among the figure about the individual layer leather by step 4) is coordinate figure and formula (6) re-projection matrix, reduction obtains the true coordinate value that its corresponding gluing platform is, the gluing platform coordinate system among the present invention promptly is the kinetic coordinate system of three gluing mechanical arms.Black numbers marker characteristic point shown in (1)-(3) in 6 with reference to the accompanying drawings, the true coordinate value that reduction obtains, wherein left figure, right side figure are the pixel coordinate value of characteristic point, reconstructed results is the true coordinate value:

Fig. 1	Left figure	Right figure	Reconstructed results (cm)
				1	(149，187)	(94，185)	(16.85，3.62，50.84)
2	(170，262)	(115，259)	(15.41，-1.22，50.84)
				3	(227，303)	(178，301)	(12.92，-4.33，57.07)
4	(225，316)	(178，312)	(13.63，-5.5，59.49)
				5	(502，364)	(448，361)	(-7.47，-7.94，51.78)
6	(511，357)	(454，355)	(-7.67，-7.09，49.05)
				7	(526，355)	(469，353)	(-8.66，-6.96，49.05)
8	(555，267)	(496，267)	(-10.22，-1.44，47.39)
				9	(283，170)	(227，167)	(7.54，4.63，49.93)
10	(559，186)	(501，185)	(-10.66，3.49，48.21)
				11	(540，93)	(484，92)	(-9.76，9.51，49.93)
12	(527，91)	(471，91)	(-8.88，9.63，49.93)

13	(519，79)	(464，79)	(-8.5，10.58，50.84)
				14	(251，87)	(197，87)	(9.86，10.07，50.84)
15	(243，101)	(185，100)	(9.87，8.69，48.21)
				16	(184，126)	(128，125)	(14.2，7.42，49.93)

Fig. 2	Left figure	Right figure	Reconstructed results (cm)
				1	(167，223)	(110，223)	(15.07，1.25，49.06)
2	(176，252)	(117，251)	(13.99，-0.53，47.39)
				3	(507，347)	(448，347)	(-7.16，-6.24，47.39)
4	(508，258)	(448，258)	(-7.1，-0.88，46.6)
				5	(287，250)	(229，242)	(7.02，-0.42，48.21)
6	(292，176)	(238，173)	(7.19，4.41，51.78)
				7	(326，162)	(259，162)	(3.88，4.29，41.73)
8	(514，188)	(454，187)	(-7.48，3.06，46.6)
				9	(521，99)	(459，99)	(-7.66，8.34，45.1)
10	(168，201)	(111，200)	(15.01，2.62，49.06)
				11	(197，213)	(141，212)	(13.32，1.91，49.93)

Fig. 3	Left figure	Right figure	Reconstructed results (cm)
				1	(86，252)	(30，249)	(20.79，-0.56，49.93)
2	(197，349)	(141，346)	(13.32，-6.71，49.93)
				3	(248，336)	(187，334)	(9.08，-5.4，45.84)
4	(536，423)	(479，420)	(-9.32，-11.19，49.06)
				5	(602，161)	(548，159)	(-14.45，5.39，51.78)
6	(249，129)	(194，126)	(10.00，7.36，50.84)
				7	(201，170)	(147，166)	(13.54，4.8，51.78)
8	(178，175)	(131，172)	(17.4，5.14，59.49)
				9	(117，167)	(62，164)	(19.05，4.91，50.84)

In the above-mentioned reconstructed results, according to the matrix Q value of formula (6), can find that the pairing pixel coordinate value of optical axis center of left camera is (394,243), symbol quadrant distribution is as shown in Figure 7 released in the definition of associate(d) matrix Q again.(+,+) mean x direction coordinate figure less than 394 pixel rebuild the back symbol for+, y direction coordinate figure rebuilds the back symbol less than 243 pixel and is+; (+,-) mean x direction coordinate figure less than 394 pixel rebuild the back symbol for+, y direction coordinate figure rebuilds the back symbol greater than 243 pixel and is-; (-,-) mean x direction coordinate figure greater than 394 pixel rebuild the back symbol for-, y direction coordinate figure rebuilds the back symbol greater than 243 pixel and is-; (-,+) mean x direction coordinate figure greater than 394 pixel rebuild the back symbol for-, y direction coordinate figure rebuilds the back symbol less than 243 pixel and is+.

Accompanying drawing 8 is the error analysis figure that obtain after 6 matching result and the reconstructed results with reference to the accompanying drawings, wherein, and in the reconstructed results of Fig. 1, vertical range between point 4 and the point 14 is 15.57, and actual range is 16, error 2.7%, vertical range between point 5 and the point 13 is 18.52, and actual range is 18.8, error 1.5%, vertical range between point 11 and the point 7 is 16.47, actual range is 17, error 3.2%, and the vertical range between point 10 and the point 8 is 4.93, actual range 5.1, error 3.4%; In the reconstructed results of Fig. 2, the vertical range between point 3 and the point 4 is 5.36, and actual range is 5.4, error 0.7%, vertical range between point 8 and the point 9 is 5.28, and actual range is 5.4, error 2.2%, distance between point 10 and the point 11 is 1.83, actual range is 1.88, error 2.7%, the distance 1.86 between point 11 and the point 1, actual range is 1.88, error 1.1%; In the reconstructed results of Fig. 3, the distance between point 1 and the point 2 is 9.69, and actual range is 10, error 3.2%, distance between point 4 and the point 5 is 17.36, and actual range is 17.9, error 3.1%, distance between point 9 and the point 1 is 5.74, and actual range is 5.9, error 2.8%; The 3rd coordinate in all coordinate figures is z axle value, i.e. the distance of this characteristic point and left camera optical axis center, and this is worth and is 48cm herein, and error about 3% is so this method is feasible.

6), the true coordinate value of characteristic point is input in the gluing coordinate formation of three gluing mechanical arms, thus the control mechanical arm is finished the gluing process.Because the gluing platform coordinate system is the coordinate system of gluing mechanical arm just again, therefore control card can convert each characteristic point coordinates value of this individual layer leather edge contour to job sequence, carries out gluing at characteristic point and line segment shape and the formula (7) put according to step 2.

$L(P_i,P_{i+1})=\left\{\begin{array}{c}\mathrm{Straigt}\\ \mathrm{Arc}\end{array}\right.,i=\mathrm{0,1,2},\mathrm{Lm}-1---\left(7\right)$

L (P _i, P _I+1) expression P _iWith P _I+1Individual layer leather edge line section shape between the individual characteristic point, Straight represents line segment, carries out gluing according to the linear interpolation mode during gluing, Arc represents arc, carries out gluing, i=0 according to the curve interpolating mode during gluing, 1,2, L m-1 represents to have m edge contour characteristic point.

Claims (5)

1. individual layer leather gluing control method based on binocular vision is characterized in that: this method adopts binocular camera and three gluing mechanical arms, and described individual layer leather gluing control method may further comprise the steps:

1), fixing binocular camera being carried out solid demarcates;

2), gather individual layer leather binocular image and extract its edge contour;

3), according to step 2) the edge contour image carry out feature extraction:

Individual layer leather edge contour by limited to two-dimensional coordinate point (x _i, y _i) constitute, i=1,2 ..., n represents that this profile is made up of coordinate points n; For on the profile arbitrarily coordinate points to (x _i, y _i), its curvature computing formula is calculated the right curvature K of all coordinate points as (1) _i, i=1,2 ..., n represents that this profile has n to coordinate;

$K_i=\frac{\left(\frac{x_{i+1}-x_{i-1}}{2}\right){\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}^2-{\left(\frac{x_{i+1}-x_{i-1}}{2}\right)}^2\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}{\sqrt[3]{{[{\left(\frac{x_{i+1}-x_{i-1}}{2}\right)}^2+{\left(\frac{y_{i+1}-y_{i-1}}{2}\right)}^2]}^2}},i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,n---\left(1\right)$

Wherein, (x _I+1, y _I+1) be that this coordinate points is to (x _i, y _i) next coordinate points right, (x _I-1, y _I-1) be that this coordinate points is to (x _i, y _i) a last coordinate points right;

Feature extraction is according to following steps:

(3.1), with

Calculate profile local curvature threshold value, B is a proportionality coefficient, is selected between [1,2];

(3.2), satisfy K _iThe coordinate points of 〉=T (i) can be selected into candidate feature point P _i, i=1,2 ... s, s＜n represents total s candidate feature point;

(3.3), calculate two formed angle theta of tangent line of each candidate feature point _I+1: with the some P of candidate feature point _iBe starting point, P _I+1Be terminal point, the mid point of Origin And Destination place line segment is P _{(i+ (i+1))/2}, according to following formula (2) theory of computation center of circle In like manner, with P _I+1Be starting point, P _I+2Be terminal point, the mid point P of Origin And Destination place line segment _{((i+1)+(i+2))/2}, get its theoretical center of circle by following formula (2)

$\begin{array}{c}{x}_{i+1}^C=\frac{(x_{i+1}^2+y_{i+1}^2)(y_{\frac{i+(i+1)}{2}}-y_i)+(x_{\frac{i+(i+1)}{2}}^2+y_{\frac{i+(i+1)}{2}}^2)(y_i-y_{i+1})+(x_i^2+y_i^2)(y_{i+1}-y_{\frac{i+(i+1)}{2}})}{2[x_{i+1}(y_{\frac{i+(i+1)}{2}}-y_i)+x_{\frac{i+(i+1)}{2}}(y_i-y_{i+1})+x_i(y_{i+1}-y_{\frac{i+(i+1)}{2}})]}\\ y_{i+1}^C=\frac{(x_{i+1}^2+y_{i+1}^2)(x_{\frac{i+(i+1)}{2}}-x_i)+(x_{\frac{i+(i+1)}{2}}^2+y_{\frac{i+(i+1)}{2}}^2)(x_i-x_{i+1})+(x_i^2+y_i^2)(x_{i+1}-x_{\frac{i+(i+1)}{2}})}{2[y_{i+1}(x_{\frac{i+(i+1)}{2}}-x_i)+y_{\frac{i+(i+1)}{2}}(x_i-x_{i+1})+y_i(x_{i+1}-x_{\frac{i+(i+1)}{2}})]}\end{array}---\left(2\right)$

Wherein, (x _I+1, y _I+1) be P _I+1Coordinate, (

) be

Coordinate, (

) be

Coordinate;

Calculate P with formula (3) then _I+1With

Line and P _I+1The angle of tangent line

And P _I+1With

Line and P _I+1The angle of tangent line

${\mathrm{\&tau;}}_{i+1}^L=a\tan \frac{(y_{i+1}^C-y_{i+1})}{(x_{i+1}^C-x_{i+1})}$

${\mathrm{\&tau;}}_{i+1}^R=a\tan \frac{(y_{i+2}^C-y_{i+1})}{(x_{i+2}^C-x_{i+1})}---\left(3\right)$

Wherein, θ _I+1Be two angles

With

Sum;

(3.4), according to θ _I+1Value judges whether it is line segment shape between required characteristic point and the marker characteristic point: if θ _I+1It is not obvious to think then that greater than 170 degree the residing local configuration curvature of this characteristic point changes, and is not required characteristic point, and the line segment between the characteristic point is shaped as line segment; Otherwise this characteristic point is required characteristic point, and the line segment between the characteristic point is shaped as arc;

4), mate according to the characteristic point of step 3), characteristic matching is according to following steps:

(4.1), with the characteristic point of the left figure in the binocular image

Be reference, i=0,1,2 ..., m, wherein, m＜s, m represent that left figure has m characteristic point;

(4.2), the right figure in calculating and the binocular image is to all characteristic points in 2 pixels around there won't be any problem

Euclidean distance, i=0,1,2 ..., k, wherein, k＜m, k represent k characteristic point arranged in 2 pixel coverages in the right figure m characteristic point;

(4.3), get the Euclidean distance minimum

As the characteristic point that matches, thus the characteristic point one-to-one relationship of figure about definite individual layer leather edge contour;

5), according to re-projection matrix reduction step 4) characteristic point the true coordinate value of corresponding gluing platform system;

6), the true coordinate value of characteristic point is input in the gluing coordinate formation of three gluing mechanical arms, thus the control mechanical arm is finished the gluing process.

2. a kind of individual layer leather gluing control method based on binocular vision as claimed in claim 1 is characterized in that: in the described step 6), line segment shape and the formula (4) according to step 3) between the characteristic point carried out gluing control;

$L(P_i,P_{i+1})=\left\{\begin{array}{c}\mathrm{Straigt}\\ \mathrm{Arc}\end{array}\right.,i=\mathrm{0,1,2},\&CenterDot;\&CenterDot;\&CenterDot;m-1---\left(4\right)$

Wherein, L (P _i, P _I+1) expression P _iWith P _I+1Individual layer leather edge line section shape between the individual characteristic point, Straight represents line segment, carries out gluing according to the linear interpolation mode during gluing, Arc represents arc, carry out gluing, i=0,1 according to the curve interpolating mode during gluing, 2 ... m-1, expression has m edge contour characteristic point.

3. a kind of individual layer leather gluing control method as claimed in claim 1 or 2 based on binocular vision, it is characterized in that: in the described step 1), the demarcation thing of taking different attitudes obtains to be used to demarcate the image of analysis, obtains intrinsic parameter k, 1, the u of each camera according to the formula (5) of linear camera model ₀, v ₀, f, θ and outer parameter, 3 * 4 matrix M that obtain by camera inside and outside parameter dot product _{3 * 4}Be perspective projection matrix;

$\left(\begin{array}{c}u\\ v\\ 1\end{array}\right)=\frac{1}{\mathrm{\&lambda;}}\left(\begin{array}{cccc}\frac{f}{k}& -\frac{f}{k}\cot \mathrm{\&theta;}& u_0& 0\\ 0& \frac{f}{l\sin \mathrm{\&theta;}}& v_0& 0\\ 0& 0& 1& 0\end{array}\right)\left(\begin{array}{cc}{R}_{3\&times;3}& t_{3\&times;1}\\ 0^T& 1\end{array}\right)\left(\begin{array}{c}{X}_W\\ Y_W\\ Z_W\\ 1\end{array}\right)=M_{3\&times;4}\left(\begin{array}{c}{X}_W\\ Y_W\\ Z_W\\ 1\end{array}\right)---\left(5\right)$

Wherein, k, l are the pixel sizes of binocular camera, u ₀, v ₀Be the coordinate figure of the optical axis center of camera, f is a camera focus, and θ is the degree of skewness of camera coordinates system, R _{3 * 3}Be the spin matrix of camera, t _{3 * 1}Then be the translation matrix of camera, (u v 1) ^ΤBe any point in the image, (X _WY _WZ _W1) ^ΤIt then is the pairing gluing coordinate system of any point coordinate in the image;

The process that image is corrected is as follows:

(1.1), the inside and outside parameter by formula (6) and left and right sides camera obtains rotation and the translation matrix of right camera with respect to left camera, and correct the left and right sides image trip aimed at;

R=R _r(R _l)T （6）

T=T _r-RT _l

Wherein, R _l, R _rBe the spin matrix of left and right sides camera, T _l, T _rBe the translation matrix of left and right sides camera, R, T are spin matrix and the translation matrix of right camera with respect to left camera.

(1.2), the camera inside and outside parameter by formula (7) and front obtains re-projection matrix Q:

$Q=\left(\begin{array}{cccc}1& 0& 0& {-c}_x\\ 0& 1& 0& -c_y\\ 0& 0& 0& f\\ 0& 0& 1/T_x& 0\end{array}\right)---\left(7\right)$

Wherein, T _xBe the x axle component that binocular camera is joined translation vector T outward, c _xAnd c _yBe the world coordinate system coordinate figure at left camera optics center, f is the focal length of left camera.

4. a kind of individual layer leather gluing control method as claimed in claim 1 or 2 based on binocular vision, it is characterized in that: described step 2), convert individual layer leather binocular image to gray level image according to the different threshold filters of finishing of the rgb value of leather and noise, detailed process is as follows:

(2.1), use Gaussian function that image is carried out smoothing processing;

(2.2), the closed operation in the employing mathematical morphology;

(2.3), adopt the leather edge contour of the unidirectional element of Canny operator extraction.

5. a kind of individual layer leather gluing control method based on binocular vision as claimed in claim 1 or 2 is characterized in that: described step 2), around the gluing platform light source is installed, is arranged binocular camera directly over the center of described gluing platform.

Images