CN102184555A - Color clustering method for central color fringes of De Bruijn color structure light coding image - Google Patents

Abstract

The invention provides a color clustering method for the central color fringes of a De Bruijn color structure light coding image, which belongs to the field of a computer image processing technique. In the color clustering method provided by the invention, the color of each pixel point on each central color fringe of the color structure light coding image is clustered by adopting a clustering analysis method. The color clustering method provided by the invention is suitable for the color structure light coding image which has the pixel points of each color in linear distribution, and can effectively improve the accuracy for determining the color of the central color fringes.

Description

A kind of color cluster method of center color fringe of De Bruijn color structured light coded image

Technical field

The present invention relates to a kind of color cluster method of center color fringe of De Bruijn color structured light coded image, belong to the computer image processing technology field.

Background technology

The optical three-dimensional measurement technology refers to the pattern of the means Measuring Object of using optics, thereby obtains the three-dimensional coordinate data of object shape face, and with these data the shape face of object is made judgement and assessment, or the three-dimensional appearance of object is rebuild.And be a kind of improvement of conventional stereo vision system and extend based on the three-dimensional reconfiguration system of structured light.

The structured light method is the light that the light beam that laser instrument or projector etc. sends is formed certain form through optical system, comprise trend of purchasing scenery such as point, single line, multi-thread, Dan Yuan, with one heart more circles, grid, right-angled intersection, gray-coded pattern, color coding pattern and random grain projection, on scenery, form pattern and absorb by video camera, then by image according to trigonometry and and the structural parameters of relevant device calculate, obtain the depth image on scenery surface, further calculate the D coordinates value of object plane.Wherein the research to the coding method of structured light has had a lot, and time encoding is to use the most extensive first kind of coded system.Time encoding is a plurality of different coding patterns successively to be projected body surface chronologically obtain the respective coding image sequence, coded video sequences combined decode, thereby solve the correspondence problem of projective patterns and images acquired, it is the coding method that present resolution is the highest, measuring error is minimum, but its speed is lower, can not be reconstructed object in real time fast.Because coloured image provides than the gray level image information of horn of plenty more, and fast based on color structured light encoding measurement speed, can be reconstructed object in real time.Therefore, just being subjected to people based on the pattern method of coloud coding more and more pays close attention to.

People such as Lee of Harbin University of Science and Technology powder orchid have proposed a kind of continuous changes colour encoding scheme, the color of color coding pattern from red, to green, be linear change to indigo plant, utilize the relation between wavelength and the phase-shifted, set up the geometric position of its optical plane and the mathematical model between the projected angle, thereby realize using the coding of the wavelength (color) of light wave the projected angle of its place optical plane.Because Di Buluying (De Bruijn) sequence, have in advance confirmability and repeatability, so the color structured light coding method that is generated by the Di Buluying pseudo-random sequence has become one of method of the widespread use of color structured light three-dimensionalreconstruction.In research in the past, the color structured light coding method is used for space encoding and direct coding more, and great majority can both be used for the measurement of dynamic object, but owing to be subjected to the influence of the color fidelity and the resolution of ccd video camera, be easy to be subjected to the influence of ccd video camera and can't obtain desirable effect and precision, be not easy to find clearly, the correspondence of unambiguity, not high to the measuring accuracy of three-dimensional body, be difficult to the measurement of accurate object.Simultaneously, because the color of utilization occurs color easily and obscure problem not easy to identify generally more than three kinds, and usually sampling density and resolution are not high.The coding method of color structured light and decoding technique play crucial effects and affect measuring accuracy, speed and the reliability of system in the data of structured light are obtained.

Summary of the invention

Classify for the color to the center color fringe of De Bruijn color structured light coded image exactly, the present invention proposes a kind of color cluster method of center color fringe of De Bruijn color structured light coded image.

A kind of color cluster method of center color fringe of De Bruijn color structured light coded image is characterized in that, this method equipment needed thereby and performing step are as follows:

Equipment needed thereby: as shown in Figure 1, equipment needed thereby comprises projector 1, projection screen 2, video camera 3, computing machine 4, object under test 5; Wherein projector 1 all is connected with computing machine 4 by cable with video camera 3.

Performing step:

Step 1, in computing machine 4, utilization has random character and deterministic De Bruijn pseudo-random sequence, produce one 7 yuan 3 grades De Bruijn sequential color structure light coding images, this image background color is a black, and its rgb value is (0,0,0), the width of every color fringe is a pixel, all inequality and 3 pixels in interval of any two adjacent color fringe colors, and 7 kinds of colors are used a respectively ₁, a ₂, a ₃, a ₄, a ₅, a ₆, a ₇Expression, wherein color a _xRgb value use r respectively _x, g _x, b _xExpression, x=1,2 ..., 7; Color a ₁Rgb value be (0,0,1), color a ₂Rgb value be (0,1,0), color a ₃Rgb value be (0,1,1), color a ₄Rgb value be (1,0,0), color a ₅Rgb value be (1,0,1), color a ₆Rgb value be (1,1,0), color a ₇Rgb value be (1,1,1).

Step 2, computing machine 4 project on the projection screen 2 after by projector 1 the De Bruijn sequential color structure light coding image that produces in the step 1 being modulated through object under test 5, generate the projected image after modulating.

Step 3, computing machine 4 pass through the projected image after video camera 3 is taken the modulation of steps 2 generation, obtain the color structured light encoded picture.

The coboundary and the lower boundary of every color fringe in the color structured light encoded picture that step 4, the traditional canny algorithm extraction step 3 of computing machine 4 employings obtain, calculate the coboundary of every color fringe and the centre position of lower boundary again, obtain the center color fringe of every color fringe; If obtain m bar center color fringe altogether, every center color fringe has n pixel; On m is capable, be arranged in order the rgb value of n pixel of every center color fringe, obtain m * n rank matrix Z that the rgb value by these pixels constitutes; The capable j column element of i of matrix Z be Z (i, j)=(R (and i, j), G (i, j), B (i, j)), wherein R (i, j), G (i, j), B (i j) is the rgb value of j pixel of i bar center color fringe, i=1,2 ..., m, j=1,2 ..., n; Note Z (i, j, k) expression Z (i, k element j), k=1,2,3.

Step 5, according to formula Z (i, j)=Z (i j)/255, carries out normalization to each element among the matrix Z, i=1,2 ..., m, j=1,2 ..., n; If M _qRepresent to belong in all center color fringes the number of q kind color pixel point, establish M _q=0, q=1,2 ..., 7.

Step 6, the coordinate of O in the RGB color space of setting up an office are (0,0,0), with k the element of O (k) expression O, k=1,2,3; Structure 7 * 3 matrix class, its k row element is followed successively by r _k, g _k, b _k, k=1 wherein, 2 ..., 7; And with class (i, j) the capable j column element of i of representing matrix class, i=1,2 ..., 7; J=1,2,3; According to formula

Class (i, j)=class (i, j)/(class (i, 1) ²+ class (i, 2) ²+ clas (i, 3) ²) ^1/2Element on each row of matrix class carry out unitization.

Step 7, establish N _q=0, q=1,2 ..., 7; Structural matrix D, wherein D (i, j, s) expression Z (described distance is defined as follows for i, the j) distance of the straight line that is constituted to being linked to each other with the represented point of the capable vector of matrix class s by an O:

$D(i,j,s)=\sqrt{\underset{k=1}{\overset{3}{\mathrm{\Σ}}}{(Z(i,j,k)-O\left(k\right))}^2-{\left[\underset{k=1}{\overset{3}{\mathrm{\Σ}}}(\mathrm{class}(s,k)\×(Z(i,j,k)-O\left(k\right)))\right]}^2}$

I=1 wherein, 2 ..., m; J=1,2 ..., n; S=1,2 ..., 7.

Step 8, to fixing i, (i, j), (i, j q) are that minimum value and the q value in { (D (i, j, 1), D (i, j, 2), D (i, j, 3), D (i, j, 4), D (i, j, 5), D (i, j, 6), D (i, j, 7) } is minimum to the Z of j, N to establish D _q=N _q+ 1; If D (i, j, q)=0, so D (i, j, q)=0.001; Structural matrix minI _q, minI wherein _q(h y) is matrix minI _qThe capable y column element of h, y=1,2,3, minI _q(N _q, 1)=Z (i, j, 1), minI _q(N _q, 2)=Z (i, j, 2), minI _q(N _q, 3)=Z (i, j, 3); Construct vectorial minD _q, minD wherein _q(i) represent vectorial minD _qI element, minD _q(N _q)=D (i, j, q);

To i=1,2 ..., m, j=1,2 ..., n, the said process in the repeating step 8; If constructed t vectorial minD altogether _q

If for each q=1,2 ..., 7 all have N _q=M _q, so a kind of color cluster method of center color fringe of De Bruijn color structured light coded image finishes, otherwise continues execution in step 9.

Step 9, to q=1,2 ..., 7, if obtained minD _q, then according to following formula construction matrix E _q, E wherein _q(i j) is matrix E _qThe capable j column element of i,

$E_q\left(\mathrm{1,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,1)-O\left(1\right))}^2/{\min D}_q\left(i\right))$

$E_q\left(\mathrm{1,2}\right)=E_q\left(\mathrm{2,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,1)-O\left(1\right))\×(({\min I}_q(i,2)-O\left(2\right))/{\min D}_q\left(i\right))$

$E_q(1,3)=E_q\left(\mathrm{3,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,1)-O\left(1\right))\×(({\min I}_q(i,3)-O\left(3\right))/{\min D}_q\left(i\right))$

$E_q(2,2)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,2)-O\left(2\right))}^2/{\min D}_q\left(i\right))$

$E_q(2,3)=E_q\left(\mathrm{3,2}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,2)-O\left(2\right))\×(({\min I}_q(i,3)-O\left(3\right))/{\min D}_q\left(i\right))$

$E_q\left(\mathrm{3,3}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,3)-O\left(3\right))}^2/{\min D}_q\left(i\right))$

Compute matrix E _qEigenwert, establish λ q ₁, λ q ₂, λ q ₃Be matrix E _qThree eigenwerts, get the eigenwert of a maximum, and calculate the pairing unit character vector of this eigenwert, use P _qExpression; The capable vector of q of matrix class is revised as P _q

Step 10, according to following formula construction symmetric matrix A, wherein A (i is that matrix A i is capable j), the j column element,

$A\left(\mathrm{1,1}\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,2)}^2+\mathrm{class}{(q,3)}^2))$

$A\left(\mathrm{1,2}\right)=A\left(\mathrm{2,1}\right)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,1)\×\mathrm{class}(q,2))$

$A\left(\mathrm{1,3}\right)=A\left(\mathrm{3,1}\right)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,1)\×\mathrm{class}(q,3))$

$A(2,2)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,3)}^2+\mathrm{class}{(q,1)}^2))$

$A\left(\mathrm{2,3}\right)=A(3,2)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,2)\×\mathrm{class}(q,3))$

$A(3,3)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,2)}^2))$

Construct vectorial B, wherein B (i) is i element of vectorial B, according to following formula construction matrix B,

$B\left(1\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,1)\×(\mathrm{class}{(q,2)}^2+\mathrm{class}{(q,3)}^2)-$

${\min I}_q(i,2)\×\mathrm{class}(q,1)\×\mathrm{class}(q,2)-{\min I}_q(i,3)\×\mathrm{class}(q,1)\×\mathrm{class}(q,3)$

$B\left(2\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,2)\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,3)}^2)-$

${\min I}_q(i,1)\×\mathrm{class}(q,2)\×\mathrm{class}(q,1)-{\min I}_q(i,3)\×\mathrm{class}(q,2)\×\mathrm{class}(q,3)$

$B\left(3\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,3)\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,2)}^2)-$

${\min I}_q(i,1)\×\mathrm{class}(q,1)\×\mathrm{class}(q,3)-{\min I}_q(i,2)\×\mathrm{class}(q,2)\×\mathrm{class}(q,3)$

According to formula O=A ^-1B revises the coordinate of some O at the RGB color space.M _q=N _q, q=1,2 ..., 7; Then execution in step 7.

The color that has realized each pixel on each bar center color fringe of color structured light encoded picture by above step carry out clusterization.

Beneficial effect

The present invention proposes a kind of color cluster method of center color fringe of De Bruijn color structured light coded image, this method is applicable to that every class color pixel puts the color structured light encoded picture of linear distribution, can improve the accuracy of determining center color fringe color effectively.

Description of drawings

Fig. 1 is a kind of color cluster method equipment needed thereby pie graph of center color fringe of De Bruijn color structured light coded image.This figure also is the specification digest accompanying drawing.Wherein: 1 is projector, and 2 is projection screen, and 3 is video camera, and 4 is computing machine, and 5 is object under test.

Fig. 2 is the De Bruijn sequential color structure light coding image that generates.

Embodiment

A kind of color cluster method of center color fringe of De Bruijn color structured light coded image, this method equipment needed thereby and performing step are as follows:

Equipment needed thereby: as shown in Figure 1, equipment needed thereby comprises projector 1, projection screen 2, video camera 3, computing machine 4, object under test 5; Wherein projector 1 all is connected with computing machine 4 by cable with video camera 3.

Performing step:

Step 1, in computing machine 4, utilization has random character and deterministic De Bruijn pseudo-random sequence, produce one 7 yuan 3 grades De Bruijn sequential color structure light coding images, this image background color is a black, and its rgb value is (0,0,0), the width of every color fringe is a pixel, all inequality and 3 pixels in interval of any two adjacent color fringe colors, and 7 kinds of colors are used a respectively ₁, a ₂, a ₃, a ₄, a ₅, a ₆, a ₇Expression, wherein color a _xRgb value use r respectively _x, g _x, b _xExpression, x=1,2 ..., 7; Color a ₁Rgb value be (0,0,1), color a ₂Rgb value be (0,1,0), color a ₃Rgb value be (0,1,1), color a ₄Rgb value be (1,0,0), color a ₅Rgb value be (1,0,1), color a ₆Rgb value be (1,1,0), color a ₇Rgb value be (1,1,1).

Step 2, computing machine 4 project on the projection screen 2 after by projector 1 the De Bruijn sequential color structure light coding image that produces in the step 1 being modulated through object under test 5, generate the projected image after modulating.

Step 3, computing machine 4 pass through the projected image after video camera 3 is taken the modulation of steps 2 generation, obtain the color structured light encoded picture.

The coboundary and the lower boundary of every color fringe in the color structured light encoded picture that step 4, the traditional canny algorithm extraction step 3 of computing machine 4 employings obtain, calculate the coboundary of every color fringe and the centre position of lower boundary again, obtain the center color fringe of every color fringe; If obtain m bar center color fringe altogether, every center color fringe has n pixel; On m is capable, be arranged in order the rgb value of n pixel of every center color fringe, obtain m * n rank matrix Z that the rgb value by these pixels constitutes; The capable j column element of i of matrix Z be Z (i, j)=(R (and i, j), G (i, j), B (i, j)), wherein R (i, j), G (i, j), B (i j) is the rgb value of j pixel of i bar center color fringe, i=1,2 ..., m, j=1,2 ..., n; Note Z (i, j, k) expression Z (i, k element j), k=1,2,3.

Step 5, according to formula Z (i, j)=Z (i j)/255, carries out normalization to each element among the matrix Z, i=1,2 ..., m, j=1,2 ..., n; If M _qRepresent to belong in all center color fringes the number of q kind color pixel point, establish M _q=0, q=1,2 ..., 7.

Step 6, the coordinate of O in the RGB color space of setting up an office are (0,0,0), with k the element of O (k) expression O, k=1,2,3; Structure 7 * 3 matrix class, its k row element is followed successively by r _k, g _k, b _k, k=1 wherein, 2 ..., 7; And with class (i, j) the capable j column element of i of representing matrix class, i=1,2 ..., 7; J=1,2,3; According to formula

Class (i, j)=class (i, j)/(class (i, 1) ²+ class (i, 2) ²+ class (i, 3) ²) ^1/2Element on each row of matrix class carry out unitization.

Step 7, establish N _q=0, q=1,2 ..., 7; Structural matrix D, wherein D (i, j, s) expression Z (described distance is defined as follows for i, the j) distance of the straight line that is constituted to being linked to each other with the represented point of the capable vector of matrix class s by an O:

$D(i,j,s)=\sqrt{\underset{k=1}{\overset{3}{\mathrm{\Σ}}}{(Z(i,j,k)-O\left(k\right))}^2-{\left[\underset{k=1}{\overset{3}{\mathrm{\Σ}}}(\mathrm{class}(s,k)\×(Z(i,j,k)-O\left(k\right)))\right]}^2}$

I=1 wherein, 2 ..., m; J=1,2 ..., n; S=1,2 ..., 7.

Step 8, to fixing i, (i, j), (i, j q) are that minimum value and the q value in { (D (i, j, 1), D (i, j, 2), D (i, j, 3), D (i, j, 4), D (i, j, 5), D (i, j, 6), D (i, j, 7) } is minimum to the Z of j, N to establish D _q=N _q+ 1; If D (i, j, q)=0, so D (i, j, q)=0.001; Structural matrix minI _q, minI wherein _q(h y) is matrix minI _qThe capable y column element of h, y=1,2,3, minI _q(N _q, 1)=Z (i, j, l), minI _q(N _q, 2)=Z (i, j, 2), minI _q(N _q, 3)=Z (i, j, 3); Construct vectorial minD _q, minD wherein _q(i) represent vectorial minD _qI element, minD _q(N _q)=D (i, j, q);

To i=1,2 ..., m, j=1,2 ..., n, the said process in the repeating step 8; If constructed t vectorial minD altogether _q

If for each q=1,2 ..., 7 all have N _q=M _q, so a kind of color cluster method of center color fringe of De Bruijn color structured light coded image finishes, otherwise continues execution in step 9.

Step 9, to q=1,2 ..., 7, if obtained minD _q, then according to following formula construction matrix E _q, E wherein _q(i j) is matrix E _qThe capable j column element of i,

$E_q\left(\mathrm{1,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,1)-O\left(1\right))}^2/{\min D}_q\left(i\right))$

$E_q\left(\mathrm{1,2}\right)=E_q\left(\mathrm{2,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,1)-O\left(1\right))\×(({\min I}_q(i,2)-O\left(2\right))/{\min D}_q\left(i\right))$

$E_q\left(\mathrm{1,3}\right)=E_q\left(\mathrm{3,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,1)-O\left(1\right))\×(({\min I}_q(i,3)-O\left(3\right))/{\min D}_q\left(i\right))$

$E_q(2,2)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,2)-O\left(2\right))}^2/{\min D}_q\left(i\right))$

$E_q(2,3)=E_q\left(\mathrm{3,2}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,2)-O\left(2\right))\×(({\min I}_q(i,3)-O\left(3\right))/{\min D}_q\left(i\right))$

$E_q\left(\mathrm{3,3}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,3)-O\left(3\right))}^2/{\min D}_q\left(i\right))$

Compute matrix E _qEigenwert, establish λ q ₁, λ q ₂, λ q ₃Be matrix E _qThree eigenwerts, get the eigenwert of a maximum, and calculate the pairing unit character vector of this eigenwert, use P _qExpression; The capable vector of q of matrix class is revised as P _q

Step 10, according to following formula construction symmetric matrix A, wherein A (i is that matrix A i is capable j), the j column element,

$A\left(\mathrm{1,1}\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,2)}^2+\mathrm{class}{(q,3)}^2))$

$A\left(\mathrm{1,2}\right)=A\left(\mathrm{2,1}\right)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,1)\×\mathrm{class}(q,2))$

$A\left(\mathrm{1,3}\right)=A\left(\mathrm{3,1}\right)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,1)\×\mathrm{class}(q,3))$

$A(2,2)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,3)}^2+\mathrm{class}{(q,1)}^2))$

$A\left(\mathrm{2,3}\right)=A(3,2)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,2)\×\mathrm{class}(q,3))$

$A(3,3)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,2)}^2))$

Construct vectorial B, wherein B (i) is i element of vectorial B, according to following formula construction matrix B,

$B\left(1\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,1)\×(\mathrm{class}{(q,2)}^2+\mathrm{class}{(q,3)}^2)-$

${\min I}_q(i,2)\×\mathrm{class}(q,1)\×\mathrm{class}(q,2)-{\min I}_q(i,3)\×\mathrm{class}(q,1)\×\mathrm{class}(q,3)$

$B\left(2\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,2)\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,3)}^2)-$

${\min I}_q(i,1)\×\mathrm{class}(q,2)\×\mathrm{class}(q,1)-{\min I}_q(i,3)\×\mathrm{class}(q,2)\×\mathrm{class}(q,3)$

$B\left(3\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,3)\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,2)}^2)-$

${\min I}_q(i,1)\×\mathrm{class}(q,1)\×\mathrm{class}(q,3)-{\min I}_q(i,2)\×\mathrm{class}(q,2)\×\mathrm{class}(q,3)$

According to formula O=A ^-1B revises the coordinate of some O at the RGB color space.M _q=N _q, q=1,2 ..., 7; Then execution in step 7.

The color that has realized each pixel on each bar center color fringe of color structured light encoded picture by above step carry out clusterization.

Claims (1)

1. the color cluster method of the center color fringe of a De Bruijn color structured light coded image is characterized in that, this method equipment needed thereby and performing step are as follows:

Equipment needed thereby: equipment needed thereby comprises projector (1), projection screen (2), video camera (3), computing machine (4), object under test (5); Wherein projector (1) all is connected with computing machine (4) by cable with video camera (3);

Performing step:

Step 1, in computing machine (4), utilization has random character and deterministic De Bruijn pseudo-random sequence, produce one 7 yuan 3 grades De Bruijn sequential color structure light coding images, this image background color is a black, and its rgb value is (0,0,0), the width of every color fringe is a pixel, all inequality and 3 pixels in interval of any two adjacent color fringe colors, and 7 kinds of colors are used a respectively ₁, a ₂, a ₃, a ₄, a ₅, a ₆, a ₇Expression, wherein color a _xRgb value use r respectively _x, g _x, b _xExpression, x=1,2 ..., 7; Color a ₁Rgb value be (0,0,1), color a ₂Rgb value be (0,1,0), color a ₃Rgb value be (0,1,1), color a ₄Rgb value be (1,0,0), color a ₅Rgb value be (1,0,1), color a ₆Rgb value be (1,1,0), color a ₇Rgb value be (1,1,1);

Step 2, computing machine (4) project on the projection screen (2) after by projector (1) the De Bruijn sequential color structure light coding image that produces in the step 1 being modulated through object under test (5), generate the projected image after modulating;

Step 3, computing machine (4) pass through the projected image after video camera (3) is taken the modulation of step 2 generation, obtain the color structured light encoded picture;

The coboundary and the lower boundary of every color fringe in the color structured light encoded picture that step 4, the traditional canny algorithm extraction step 3 of computing machine (4) employing obtain, calculate the coboundary of every color fringe and the centre position of lower boundary again, obtain the center color fringe of every color fringe; If obtain m bar center color fringe altogether, every center color fringe has n pixel; On m is capable, be arranged in order the rgb value of n pixel of every center color fringe, obtain m * n rank matrix Z that the rgb value by these pixels constitutes; The capable j column element of i of matrix Z be Z (i, j)=(R (and i, j), G (i, j), B (i, j)), wherein R (i, j), G (i, j), B (i j) is the rgb value of j pixel of i bar center color fringe, i=1,2 ..., m, j=1,2 ..., n; Note Z (i, j, k) expression Z (i, k element j), k=1,2,3;

Step 5, according to formula Z (i, j)=Z (i j)/255, carries out normalization to each element among the matrix Z, i=1,2 ..., m, j=1,2 ..., n; If M _qRepresent to belong in all center color fringes the number of q kind color pixel point, establish M _q=0, q=1,2 ..., 7;

Step 6, the coordinate of O in the RGB color space of setting up an office are (0,0,0), with k the element of O (k) expression O, k=1,2,3; Structure 7 * 3 matrix class, its k row element is followed successively by r _k, g _k, b _k, k=1 wherein, 2 ..., 7; And with class (i, j) the capable j column element of i of representing matrix class, i=1,2 ..., 7; J=1,2,3; According to formula

Class (i, j)=class (i, j)/(class (i, 1) ²+ class (i, 2) ²+ class (i, 3) ²) ^1/2Element on each row of matrix class carry out unitization;

Step 7, establish N _q=0, q=1,2 ..., 7; Structural matrix D, wherein D (i, j, s) expression Z (i, j) to the distance of the straight line that is made of an O and the represented point of the capable vector of matrix class s, described distance is defined as follows:

$D(i,j,s)=\sqrt{\underset{k=1}{\overset{3}{\mathrm{\Σ}}}{(Z(i,j,k)-O\left(k\right))}^2-{\left[\underset{k=1}{\overset{3}{\mathrm{\Σ}}}(\mathrm{class}(s,k)\×(Z(i,j,k)-O\left(k\right)))\right]}^2}$

I=1 wherein, 2 ..., m; J=1,2 ..., n; S=1,2 ..., 7;

Step 8, to fixing i, (i, j), (i, j q) are that minimum value and the q value in { (D (i, j, 1), D (i, j, 2), D (i, j, 3), D (i, j, 4), D (i, j, 5), D (i, j, 6), D (i, j, 7) } is minimum to the Z of j, N to establish D _q=N _q+ 1; If D (i, j, q)=0, so D (i, j, q)=0.001; Structural matrix minI _q, minI wherein _q(h y) is matrix minI _qThe capable y column element of h, y=1,2,3, minI _q(N _q, 1)=Z (i, j, l), minI _q(N _q, 2)=Z (i, j, 2), minI _q(N _q, 3)=Z (i, j, 3); Construct vectorial minD _q, minD wherein _q(i) represent vectorial minD _qI element, minD _q(N _q)=D (i, j, q);

To i=1,2 ..., m, j=1,2 ..., n, the said process in the repeating step 8; If constructed t vectorial minD altogether _qIf for each q=1,2 ..., 7 all have N _q=M _q, so a kind of color cluster method of center color fringe of De Bruijn color structured light coded image finishes, otherwise continues execution in step 9;

Step 9, to q=1,2 ..., 7, if obtained minD _q, then according to following formula construction matrix E _q, E wherein _q(i j) is matrix E _qThe capable j column element of i,

$E_q\left(\mathrm{1,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,1)-O\left(1\right))}^2/{\min D}_q\left(i\right))$

$E_q\left(\mathrm{1,2}\right)=E_q\left(\mathrm{2,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,1)-O\left(1\right))\×(({\min I}_q(i,2)-O\left(2\right))/{\min D}_q\left(i\right))$

$E_q(1,3)=E_q\left(\mathrm{3,1}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,1)-O\left(1\right))\×(({\min I}_q(i,3)-O\left(3\right))/{\min D}_q\left(i\right))$

$E_q(2,2)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,2)-O\left(2\right))}^2/{\min D}_q\left(i\right))$

$E_q(2,3)=E_q\left(\mathrm{3,2}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}(({\min I}_q(i,2)-O\left(2\right))\×(({\min I}_q(i,3)-O\left(3\right))/{\min D}_q\left(i\right))$

$E_q\left(\mathrm{3,3}\right)=\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}({({\min I}_q(i,3)-O\left(3\right))}^2/{\min D}_q\left(i\right))$

Compute matrix E _qEigenwert, establish λ q ₁, λ q ₂, λ q ₃Be matrix E _qThree eigenwerts, get the eigenwert of a maximum, and calculate the pairing unit character vector of this eigenwert, use P _qExpression; The capable vector of q of matrix class is revised as P _q

Step 10, according to following formula construction matrix symmetry A, wherein A (i is that matrix A i is capable j), the j column element,

$A\left(\mathrm{1,1}\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,2)}^2+\mathrm{class}{(q,3)}^2))$

$A\left(\mathrm{1,2}\right)=A\left(\mathrm{2,1}\right)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,1)\×\mathrm{class}(q,2))$

$A\left(\mathrm{1,3}\right)=A\left(\mathrm{3,1}\right)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,1)\×\mathrm{class}(q,3))$

$A(2,2)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,3)}^2+\mathrm{class}{(q,1)}^2))$

$A\left(\mathrm{2,3}\right)=A(3,2)=-\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×\mathrm{class}(q,2)\×\mathrm{class}(q,3))$

$A(3,3)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}(N_q\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,2)}^2))$

Construct vectorial B, wherein B (i) is i element of vectorial B, according to following formula construction matrix B,

$B\left(1\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,1)\×(\mathrm{class}{(q,2)}^2+\mathrm{class}{(q,3)}^2)-$

${\min I}_q(i,2)\×\mathrm{class}(q,1)\×\mathrm{class}(q,2)-{\min I}_q(i,3)\×\mathrm{class}(q,1)\×\mathrm{class}(q,3)$

$B\left(2\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,2)\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,3)}^2)-$

${\min I}_q(i,1)\×\mathrm{class}(q,2)\×\mathrm{class}(q,1)-{\min I}_q(i,3)\×\mathrm{class}(q,2)\×\mathrm{class}(q,3)$

$B\left(3\right)=\underset{q=1}{\overset{t}{\mathrm{\Σ}}}\underset{i=1}{\overset{N_q}{\mathrm{\Σ}}}{\min I}_q(i,3)\×(\mathrm{class}{(q,1)}^2+\mathrm{class}{(q,2)}^2)-$

${\min I}_q(i,1)\×\mathrm{class}(q,1)\×\mathrm{class}(q,3)-{\min I}_q(i,2)\×\mathrm{class}(q,2)\×\mathrm{class}(q,3)$

According to formula O=A ^-1B revises the coordinate of some O at the RGB color space, M _q=N _q, q=1,2 ..., 7; Then execution in step 7;

The color that has realized each pixel on each bar center color fringe of color structured light encoded picture by above step carry out clusterization.

Images