CN103914675A - Garment QD code recognition method - Google Patents

Abstract

The invention discloses a garment QD code recognition method. The method includes the following steps that firstly, a two-dimension code image attached to a garment raw material is collected; secondly, the image is subjected to graying, fast median filtering and binarization processing; thirdly, the image having been subjected to binarization processing is subjected to edge extraction; fourthly, two side edge regions obtained after edge extraction of a QR code are separated out; fifthly, Hough transform is adopted to detect a side-edge imaginary line so as to obtain the deviation angle of the QR code; sixthly, the image having been subjected to binarization processing is rotated according to the bilinear interpolation method, and initial correction is carried out; seventhly, the QR code is positioned, the sequence of three view finding images is determined, and the QR code is adjusted to have a correct orientation according to the image matrix transposition turning method; eighthly, decoding is carried out according to the national two-dimension code standard. According to the method, due to the combination of projection cutting and the Hough transform method, the deviation angle of the QR code can be quickly and accurately acquired, the collected image is immune from noise pollution and uneven illumination, and the two-dimension code image easy to identify can be obtained through transform.

Description

A kind of 2 D code of quick response matrix recognition methods of clothes

Technical field

The invention discloses a kind of 2 D code of quick response matrix recognition methods of clothes, particularly relate to a kind of Quick Response Code recognition methods of 2 D code of quick response matrix offset correction of clothes.

Background technology

Modern wears the clothes and stresses fashion, individual character, clothes industry particularly fashionable dress producer also to walk on a large scale, the production route of short run, discrepancy cargo type class and the shipment frequency in warehouse also can continue to increase, if clothes enter, put in storage and the work of making an inventory still following traditional manually make out an invoice, the mode of manually checking over goods, its efficiency and accuracy all can not be guaranteed, wrong document may be received by treasurer's department, and administrative authority can not obtain required inventory data in time and exactly.

Barcode technology is one of data input means of current most convenient, practicality, has wide development and application prospect in China.But traditional bar code can only identify " article ", and can not describe " article ", and the use of bar code in every profession and trade is greatly limited.And the appearance of two-dimensional bar code has well solved the problem that one dimension code cann't be solved.Abroad, it just has the title of " move database " (Portable data file).It has advantages of that information density is high, information capacity is large, antijamming capability is strong, error correcting capability is strong.And these advantages can solve well due to one dimension code finite capacity, and the problem of product can not be described.

QR Code(2 D code of quick response matrix, is called for short QR code) be a kind of common Quick Response Code, it can only be distinguished by the mode of image recognition, and image recognition mode is subject to outside factor impact, as light inequality, excessively dark, the image that can cause CCD camera to take has noise.In addition, due in the time obtaining QR Code view data, may be different because of placement location, obtaining QR Code image has skew, cannot locate smoothly QR Code view finding graph position, directly sets up sampling grid.So solution noise pollution, uneven illumination, QR Code skew are unavoidable problems in many application.

Summary of the invention

In view of the above problems, technical matters to be solved by this invention is to provide a kind of Quick Response Code recognition methods of 2 D code of quick response matrix offset correction of clothes.

For solving the problems of the technologies described above, the present invention is achieved by the following technical solutions:

The 2 D code of quick response matrix recognition methods of a kind of clothes of the present invention, comprises the following steps:

A) image in 2 D code of quick response matrix pasting on CCD camera collection clothes raw material;

B) coloured image of gathered 2 D code of quick response matrix to be identified is converted to 256 color shade images, and with the ground unrest point of the above-mentioned gray level image of Fast median filtering algorithm elimination;

C) adopt maximum variance between clusters to find best segmentation threshold, this calculation ratio juris is, when the weighted sum of the prospect of being separated by threshold value and two pixel group variances of background reaches hour, this threshold value is best segmentation threshold;

D) according to best segmentation threshold, by the gray-scale value of each pixel of above-mentioned gray level image and threshold, if be greater than threshold value, assignment 1, if be less than threshold value, assignment 0, obtains binary image;

E) adopt Sobel edge detection operator to do edge extracting to binary image, between the adjacent area that two have different gray-scale values, always there is gray scale edge, gray scale edge is the discontinuous result of gray-scale value, this discontinuous conventionally can utilize to differentiate detect easily;

E1) Sobel edge detection operator is a kind of gradient operator, it uses two templates, one is level, another is vertical, two is all 3*3 masterplate, and each template is approached a partial derivative, with horizontal shuttering and vertical formwork, image is carried out to convolution respectively, obtain two matrixes, their difference presentation videos are at two partial derivatives at same position place;

E2), using the maximal value of two numbers of these two matrix correspondence positions as the output of giving point, obtain the edge of image in 2 D code;

F) adopt adaptive two side areas of isolating 2 D code of quick response matrix based on histogram projection cutting-out method;

F1) image after edge extraction carries out the projection that horizontal direction is X-direction, and the number of the pixel that on statistics X=i straight line, gray-scale value is 255, deposits array xProjection[i in] in.Start to scan from left to right from i=0, if condition (xProjection[i]==0) & & (xProjection[i+1] >0) & & (xProjection[i+2] >1) be true, think the horizontal ordinate on a summit finding 2 D code of quick response matrix, defining this point is A point, obtains x _a=i+1.Then scan from top to bottom straight line X=x _a, finding first gray-scale value is 255 pixel, the ordinate of this point for the ordinate of A, obtaining the coordinate that A orders is if condition (xProjection[i] >1) & & (xProjection[i+1] >0) & & (xProjection[i+2]==0) be true, think the horizontal ordinate on another summit of finding 2 D code of quick response matrix, defining this point is C point, obtains x _c=i+1.Then scan from the bottom up straight line X=x _c, finding first gray-scale value is 255 pixel, the ordinate of this point for the ordinate of C, obtaining C point coordinate is

F2) image after edge extraction carries out vertical square to the projection that is Y direction, the number of the pixel that on statistics Y=j straight line, gray-scale value is 255, deposit array yProjection[j in] in, then start to scan from left to right from j=0, if condition (yProjection[j]==0) & & (yProjection[j+1] >0) & & (yProjection[j+2] >1) be true, think the ordinate on the 3rd summit finding 2 D code of quick response matrix, defining this point is D point, obtain y _d=j+1, scanning straight line Y=y then turns left from the right side _d, finding first gray-scale value is 255 pixel, the horizontal ordinate of this point for the horizontal ordinate of D, obtaining the coordinate that D orders is if condition (yProjection[j] >1) & & (yProjection[j+1] >0) & & (yProjection[j+2]==0) be true, think the ordinate on the 4th summit finding 2 D code of quick response matrix, defining this point is B point, obtains y _b=j+1, then scans straight line Y=y from left to right _b, finding first gray-scale value is 255 pixel, the horizontal ordinate of this point for the horizontal ordinate of B, obtaining B point coordinate is

F3) analyze D point and B point coordinate position, set an offset direction decision criteria: work as x _d>=x _btime judge that image in 2 D code of quick response matrix is to left avertence, works as x _d<x _btime judge that QR code image is to right avertence;

F4) according to the 2 D code of quick response matrix offset direction of above-mentioned decision criteria gained, automatically select cutting method: in the time judging 2 D code of quick response matrix to left avertence, be embodied in the horizontal ordinate x that D is ordered _dbe more than or equal to the horizontal ordinate x that B is ordered _b, straight line X=x _bright side area and straight line Y=y _athe intersection area in region of below crop;

In the time judging 2 D code of quick response matrix to right avertence, be embodied in the horizontal ordinate x that D is ordered _athe horizontal ordinate x of ordering than B _blittle, straight line X=x _bleft field and Y=y _athe intersection area of upper area crop, isolate two side areas of 2 D code of quick response matrix;

The regulation of the cutting method of above-mentioned offset direction decision criteria and corresponding offset direction is in order farthest to crop 2 D code of quick response matrix interior zone, obtain two long and narrow side areas, calculated amount while reducing hough transform detection side sideline, has also avoided the interference of the inner line segment of Quick Response Code to detection side sideline simultaneously;

G) adopt hough transform algorithm to detect side dotted line and obtain 2 D code of quick response matrix deviation angle;

G1) polar coordinates ρ of initialization, the array in θ space, the quantification number in ρ direction is the pixel count of image diagonal direction wherein N ₁and N ₂be respectively the wide and high of image, the quantification number in θ direction is 1800, and angle is from 0 °～180 °, 0.1 ° of every lattice;

G2) set up a totalizer A (ρ, θ), juxtaposition finite element is 0;

G3) pixel (x that each gray-scale value on sequential search image is 255 _i, y _i) set up equation ρ=x _icos (θ)+y _isin (θ), calculates ρ value to each discrete value of θ parameter, obtains this pixel (x _i, y _i) sinusoidal curve on polar coordinates, if curve in too small lattice (ρ, θ), this totalizer A (ρ, θ) adds 1;

G4) find out totalizer A (ρ, θ) local maximum above, this value is the conllinear parameter of image collinear point, and its polar theta is the deviation angle of side dotted line;

H) adopt bilinear interpolation image rotating, tentatively proofread and correct image in 2 D code of quick response matrix;

H1) find the central point (a, b) of image after above-mentioned binaryzation, determine that postrotational centre coordinate is for (c, d), according to reverse rotation transformation matrix expression formula: $\left[\begin{array}{c}{x}_1\\ y_1\\ 1\end{array}\right]=\left[\begin{array}{ccc}1& 0& c\\ 0& -1& d\\ 0& 0& 1\end{array}\right]\left[\begin{array}{cc}\cos \left(\mathrm{\&theta;}\right)& \sin \left(\mathrm{\&theta;}\right)\\ -\sin \left(\mathrm{\&theta;}\right)& \cos \left(\mathrm{\&theta;}\right)\\ 0& 1\end{array}\right]\left[\begin{array}{ccc}1& 0& -a\\ 0& -1& b\\ 0& 0& 1\end{array}\right]\left[\begin{array}{c}{x}_0\\ y_0\\ 1\end{array}\right],$ Image is rotated, point (x ₀, y ₀) rotation θ angle becomes point (x ₁, y ₁), wherein (x ₀, y ₀) be any point of not rotating in front image, (x ₁, y ₁) be by point (x ₀, y ₀) rotation θ angle obtains take (a, b) as rotation center point;

H2) image is carried out to bilinear interpolation, image by rotational transform after (x ₁, y ₁) corresponding (x ₀, y ₀) not necessarily integer of coordinate, find (x ₀, y ₀) near four pixels in (x, y), (x+1, y), (x, y+1), (x+1, y+1), according to bilinear interpolation formula, in horizontal and vertical directions, to its interpolation, formula is: $\begin{array}{c}f(x_0,y_0)=f(x,y)+\mathrm{\&alpha;}[f(x+1,y)-f(x,y)]+\mathrm{\&beta;}[f(x,y+1)-f(x,y)]\\ +\mathrm{\&alpha;\&beta;}\left[\right]f(x+1,y+1)+f(x,y)-f(x,y+1)-f(x+1,y)]\end{array}$ Wherein α=x ₀-x, β=y ₀-y, obtains point (x ₀, y ₀) gray-scale value f (x ₀, y ₀), this value is exactly point (x ₁, y ₁) gray-scale value;

I) to 2 D code of quick response matrix location, determine the order of three view finding figures, propose a kind of method with the upset of image array transposition 2 D code of quick response matrix is adjusted to correct orientation, realize the correction of 2 D code of quick response matrix;

I1) first horizontal scanning image in 2 D code of quick response matrix from left to right, record the line segment that meets (1:1:3:1:1) view finding graphics proportion feature, the width of each pixel is permitted and had 0.5 deviation, the starting point Pr and the terminal Pl that the most important thing is to record this line segment, the horizontal ordinate of the centre coordinate of view finding figure is then along straight line from vertical scan direction, record the such line segment that meets 1:1:3:1:1, obtain the ordinate of the centre coordinate of view finding figure; Record according to this mode that image is carried out to horizontal and vertical directions scanning the such line segment that meets 1:1:3:1:1, just can obtain successively the center point coordinate of three view finding figures;

I2) central point of three view finding figures is coupled together, form an isosceles right triangle, the view finding figure in regulation left side is that view finding figure being positioned on isosceles right triangle right angle; Determine the position of view finding figure in left side in image by calculating the length of side and triangular angles;

I3) by the method for image array transposition upset, 2 D code of quick response matrix is adjusted to correct orientation, be as the criterion in a clockwise direction, if view finding figure in left side in the image upper left corner, judges that 2 D code of quick response matrix is correct orientation, without correction; If view finding figure in left side is in the image upper right corner, judge that 2 D code of quick response matrix has been offset 90 degree, by image array transposition (i, j) → (j, i) exchange (i with up-downgoing, j) → (n+1-i, j), proofreaies and correct 2 D code of quick response matrix to correct orientation; If view finding figure in left side in the image lower right corner, is judged 2 D code of quick response matrix and has been offset 180 degree, by the conversion of (i, j) → (m+1-j, n+1-i) matrix element transposition, proofread and correct 2 D code of quick response matrix to correct orientation; If left side view finding is in the image lower left corner, judge that 2 D code of quick response matrix has been offset 270 degree, by image array transposition (i, j) → (j, i) exchange (i with left and right row, j) → (i, m+1-j), proofreaies and correct 2 D code of quick response matrix to correct orientation;

J) to proofreading and correct to the 2 D code of quick response matrix in correct orientation, decode according to standard GB/T/T18284-2000, realize the 2 D code of quick response matrix identification of clothes.

The 2 D code of quick response matrix recognition methods of a kind of clothes as above, comprises title, the place of production, checker, specification, model and the unit price information of clothes raw material in described image in 2 D code.

The 2 D code of quick response matrix recognition methods of a kind of clothes as above, in described CCD camera, sensor pixel is 300,000, the coloured image highest resolution of its collection is 480*640;

The 2 D code of quick response matrix recognition methods of a kind of clothes as above, the formula that described coloured image is converted to 256 color shade images is:

Gray scale=0.3* red component+0.59* green component+0.11* blue component.

The 2 D code of quick response matrix recognition methods of a kind of clothes as above, the noise spot of described Fast Median Filtering elimination image background refers to, select a current pixel as central point, choose the matrix sampling window of a 3*3 size of putting centered by it, first every row in window are carried out to ascending sort, again every row is carried out to ascending sort, get afterwards cornerwise three numbers, sort, get intermediate value, this is worth as current output valve, and window moves up and down in image, to entire image filtering.

The present invention also provides a kind of 2 D code of quick response matrix recognition device of clothes, described recognition device comprises PC and CCD camera that video frequency collection card is housed, CCD camera connects PC by usb data line, CCD camera passes to PC by usb data line after obtaining the image in 2 D code of quick response matrix of clothes, in PC, the 2 D code of quick response matrix identification software of clothes is decoded to 2 D code information, and the information that decoding is obtained deposits in database; The 2 D code of quick response matrix identification software of above-mentioned clothes comprises that image in 2 D code acquisition module, image correction module, Quick Response Code decoder module and raw material information import database module;

Image in 2 D code acquisition module imports the image in 2 D code of quick response matrix of frame clothes of PC video flowing into by usb data line for obtaining CCD camera;

Image correction module is converted to 256 color shade images for the Quick Response Code coloured image that above-mentioned image in 2 D code acquisition module is obtained, and with Fast median filtering algorithm picture noise point, adopt maximum variance between clusters to find optimal segmenting threshold, image binaryzation, image after adopting Sobel edge detection operator to binaryzation does edge extracting, utilize adaptive two side areas of isolating 2 D code of quick response matrix based on histogram projection cutting-out method, adopt hough transform to detect side dotted line and obtain 2 D code of quick response matrix deviation angle, adopt bilinear interpolation image rotating, the preliminary image in 2 D code of quick response matrix of proofreading and correct, afterwards 2 D code of quick response matrix is located, determine the order of three view finding figures, 2 D code of quick response matrix is adjusted to correct orientation by the method for image array transposition upset, realize the correction of 2 D code of quick response matrix,

Quick Response Code decoder module for according to standard GB/T/T18284-2000 the image in 2 D code of quick response matrix after to above-mentioned correction decode, the information such as title, the place of production, checker, specification, model and unit price in the Quick Response Code that obtains pasting on clothes raw material;

Raw material information imports database module for by the 2 D code information automated import of data storehouse on above-mentioned clothes raw material, realizes informationization and paperless management.

Beneficial effect

The present invention is adaptive based on histogram projection cutting-out method by adopting, isolate exactly two side areas of QR Code, then obtain deviation angle with hough transform, the preliminary QR Code image of proofreading and correct, to QR Code location, determine the order of three view finding figures afterwards, QR Code is adjusted to correct orientation by the method that image array transposition overturns, realize the correction of QR Code, the object being correctly decoded.Adopt Quick Response Code automatic identification technology simultaneously, realize clothes raw material information automatic guide to enter in stock's Management System Data storehouse.Most important or treasurer's department, administrative authority, can obtain required inventory data quickly and accurately.Treasurer's department's error correction in time, corresponding adjustment can be made according to data by administrative authority, realizes automatic management.

Accompanying drawing explanation

Fig. 1 is system construction drawing of the present invention;

Fig. 2 is that QR Code of the present invention is offset the binaryzation schematic diagram while being less than 90 ° left;

Fig. 3 is that QR Code of the present invention is offset the binaryzation schematic diagram while being greater than 90 ° to the right;

Fig. 4 is the binaryzation schematic diagram of QR Code of the present invention while being offset 180 ° to the right;

Fig. 5 is vertical-horizontal perspective view after QR Code edge extracting in Fig. 2;

Fig. 6 is vertical-horizontal perspective view after QR Code edge extracting in Fig. 3;

Fig. 7 is vertical-horizontal perspective view after QR Code edge extracting in Fig. 4;

Fig. 8 is that two side areas that QR Code of the present invention is offset left while being less than 90 ° obtain schematic diagram;

Fig. 9 is that two side areas that QR Code of the present invention is offset to the right while being greater than 90 ° obtain schematic diagram;

Figure 10 is that the two piece side areas of QR Code of the present invention while being offset 180 ° to the right obtain schematic diagram;

Figure 11 is the schematic diagram having in Fig. 2 after preliminary correction of QR Code of skew;

Figure 12 is the schematic diagram having in Fig. 3 after preliminary correction of QR Code of skew;

Figure 13 is the schematic diagram having in Fig. 4 after preliminary correction of QR Code of skew;

Figure 14 is the QR Code image schematic diagram after cumulative correction after matrix transpose upset that has skew in Fig. 2;

Figure 15 is the QR Code image schematic diagram after cumulative correction after matrix transpose upset that has skew in Fig. 3;

Figure 16 is the QR Code image schematic diagram after cumulative correction after matrix transpose upset that has skew in Fig. 4.

Embodiment

Below in conjunction with embodiment, further set forth the present invention.Should be understood that these embodiment are only not used in and limit the scope of the invention for the present invention is described.In addition should be understood that those skilled in the art can make various changes or modifications the present invention after having read the content of the present invention's instruction, these equivalent form of values fall within the application's appended claims limited range equally.

Refer to system shown in Figure 1 structural drawing, comprising: image in 2 D code acquisition module S101, image correction module S102, Quick Response Code decoder module S103, decoded information import database module S104, image in 2 D code acquisition module S101 takes the image in 2 D code pasting on clothes raw materials by CCD camera, be transferred to PC with usb data line, with the image in 2 D code of quick response matrix of frame clothes of the C# program acquisition video flowing of writing, the image in 2 D code that image correction module S102 collects image in 2 D code acquisition module S101 carries out gray processing, Fast Median Filtering is removed picture noise point, find optimal segmentation threshold values with maximum between-cluster variance, realize image binaryzation, image after adopting Sobel edge detection operator to binaryzation does edge extracting, and with adaptive two side areas of isolating QR Code based on histogram projection cutting-out method, adopt hough transform to detect side dotted line and obtain QR Code deviation angle, adopt bilinear interpolation image rotating, the preliminary QR Code image of proofreading and correct, afterwards QR Code is located, determine the order of three view finding figures, QR Code is adjusted to correct orientation by the method for image array transposition upset, realize the correction of QR Code, Quick Response Code decoder module S103 is for QR Code image that image correction module S102 is obtained, and according to standard GB/T/T18284-2000, decoding obtains the information such as the title, the place of production, checker, specification, model, unit price of clothes raw material wherein, raw material information imports database module S104 for the information of S103 is imported in clothes inventory management system database automatically, realizes automatic management.

Embodiment 1

Introduce one embodiment of the present of invention below in conjunction with accompanying drawing 2,5,8,11,14, the image in 2 D code of quick response matrix that the skew left collecting on clothes raw material is less than to 90 ° is proofreaied and correct to correct identification direction.

Gathered coloured image is converted to 256 color shade images according to gray scale=0.3* red component+0.59* green component+0.11* blue component, and find best partition threshold T=82 with maximum between-cluster variance, by the gray-scale value of each pixel of above-mentioned gray level image and threshold, if be greater than threshold value, assignment 255, if be less than threshold value, assignment 0, change into binary image, as shown in Figure 2.

Again binary image is done to edge extracting with Sobel edge detection operator, with 3*3 horizontal shuttering and 3*3 vertical formwork, image is carried out to convolution respectively, obtain two matrixes, their difference presentation videos are at two partial derivatives at same position place, using the maximal value of two numbers of these two matrix correspondence positions as the output of giving point, realize the edge extracting of binary image, as shown in Figure 5.

Image after edge extraction carries out the projection that horizontal direction is X-direction again, and the number of the pixel that on statistics X=i straight line, gray-scale value is 255, deposits array xProjection[i in] in.Start to scan from left to right from i=0, if condition (xProjection[i]==0) & & (xProjection[i+1] >0) & & (xProjection[i+2] >1) be true, think the horizontal ordinate on a summit finding 2 D code of quick response matrix, defining this point is A point, obtains x _a=i+1.Then scan from top to bottom straight line X=x _a, finding first gray-scale value is 255 pixel, the ordinate of this point for the ordinate of A, obtaining the coordinate that A orders is .If condition (xProjection[i] >1) & & (xProjection[i+1] >0) & & (xProjection[i+2]==0) be true, think the horizontal ordinate on another summit of finding 2 D code of quick response matrix, defining this point is C point, obtains x _c=i+1.Then scan from the bottom up straight line X=x _c, finding first gray-scale value is 255 pixel, the ordinate of this point for the ordinate of C, obtaining C point coordinate is then, image after edge extracts carries out vertical square to the projection that is Y direction, the number of the pixel that on statistics Y=j straight line, gray-scale value is 255, deposit array yProjection[j in] in, then start to scan from left to right from j=0, if condition (yProjection[j]==0) & & (yProjection[j+1] >0) & & (yProjection[j+2] >1) be true, think the ordinate on the 3rd summit finding 2 D code of quick response matrix, defining this point is D point, obtain y _d=j+1, scanning straight line Y=y then turns left from the right side _d, finding first gray-scale value is 255 pixel, the horizontal ordinate of this point for the horizontal ordinate of D, obtaining the coordinate that D orders is if condition (yProjection[j] >1) & & (yProjection[j+1] >0) & & (yProjection[j+2]==0) be true, think the ordinate on the 4th summit finding 2 D code of quick response matrix, defining this point is B point, obtains y _b=j+1, then scans straight line Y=y from left to right _b, finding first gray-scale value is 255 pixel, the horizontal ordinate of this point for the horizontal ordinate of B, obtaining B point coordinate is .2 D code of quick response matrix four summit A, B, C, the D coordinate position (x in image are so just determined _a, y _a), (x _b, y _b), (x _c, y _c), (x _d, y _d), and can obtain x _d>x _b, as shown in Figure 5.

Again according to the offset direction decision criteria stipulating in technical scheme: x _d>=x _btime image in 2 D code of quick response matrix be left avertence, x _d<x _bspeed per hour response matrix image in 2 D code is right avertence.Judge that this image in 2 D code is as left avertence, and select left avertence cutting method:

Straight line X=x _bright side area and straight line Y=y _athe intersection area in region of below crop, as shown in Figure 8.The regulation of the cutting method of above-mentioned offset direction decision criteria and corresponding offset direction is in order farthest to crop 2 D code of quick response matrix interior zone, obtain two long and narrow side areas, the calculated amount while reducing hough transform detection side sideline has also been avoided the interference of the inner line segment of Quick Response Code to detection side sideline simultaneously.

Adopt again hough transform algorithm to detect the side dotted line in two long and narrow side areas, first polar coordinates ρ of initialization, the array in θ space, the quantification number in ρ direction is the pixel count of image diagonal direction wherein N ₁and N ₂be respectively the wide and high of image, the quantification number in θ direction is 1800, and angle is from 0 °～180 °, 0.1 ° of every lattice.Then, set up a totalizer A (ρ, θ), juxtaposition finite element is 0.Pixel (the x that each gray-scale value on sequential search image is 255 _i, y _i) set up equation ρ=x _icos (θ)+y _isin (θ), calculates ρ value to each discrete value of θ parameter, obtains this pixel (x _i, y _i) sinusoidal curve on polar coordinates, if curve in too small lattice (ρ, θ), this totalizer A (ρ, θ) adds 1.Finally, find out the maximal value on totalizer A (ρ, θ), this value is the conllinear parameter of image side dotted line, obtains θ=7.5 ^°the deviation angle that is Quick Response Code is 7.5 °.

Then, find the center point P of above-mentioned binary image, the coordinate of P is as shown in Figure 2.Take P point as rotation center, according to the reverse rotation transformation matrix expression formula in technical scheme and bilinear interpolation formula, with the bilinear interpolation image that turns clockwise, tentatively proofread and correct image in 2 D code of quick response matrix, as shown in figure 11.

Then, image in 2 D code of quick response matrix is tentatively proofreaied and correct in horizontal scanning from left to right, record the line segment that meets (1:1:3:1:1) view finding graphics proportion feature, the width of each pixel is permitted and had 0.5 deviation, the most important thing is to record the starting point P of this line segment _rwith terminal P _l, the horizontal ordinate of the centre coordinate of view finding figure is , then along straight line from vertical scan direction, record the such line segment that meets 1:1:3:1:1, obtain the ordinate of the centre coordinate of view finding figure; Record according to this mode that image is carried out to horizontal and vertical directions scanning the such line segment that meets 1:1:3:1:1, obtain successively the central point of three view finding figures, as the P1 in Figure 11, P2, P3.The center point P 1 of three view finding figures, P2, P3 couples together, and forms an isosceles right triangle, and the view finding figure in regulation left side is that view finding figure being positioned on isosceles right triangle right angle.By calculating length of side L12, L23, ratio between L13 two is determined the position of view finding figure in left side in image, ratio is approximately that folded point of two limits of 1 and is the central point of left side view finding figure, P1 is the position of view finding centre of figure in left side in image, as shown in figure 11.

Finally, 2 D code of quick response matrix is adjusted to correct orientation by the method for image array transposition upset, be as the criterion in a clockwise direction, left side view finding centre of figure point P1 is in the image upper left corner, judge that the 2 D code of quick response matrix after preliminary correction has been correct orientation, without again proofreading and correct, as shown in figure 14.Thereby subsequently can decode and realize the 2 D code of quick response matrix identification of pasting on clothes raw material 2 D code of quick response matrix by standard GB/T/T18284-2000.

Embodiment 2

Introduce one embodiment of the present of invention below in conjunction with accompanying drawing 3,6,9,12,15, the image in 2 D code of quick response matrix that the skew to the right collecting on clothes raw material is greater than to 90 ° is proofreaied and correct to correct identification direction, and method and embodiment 1 that this embodiment uses are identical.

According to the method identical with embodiment 1, gathered coloured image is changed into binary image, as shown in Figure 3.Adopt Sobel edge detection operator to do edge extracting to binary image, image after edge extracts carry out horizontal direction be the projection of X-direction with vertical square to the projection that is Y direction, then respectively from four direction scanning up and down, determine 2 D code of quick response matrix four summit A, B, C, the D coordinate position (x in image _a, y _a), (x _b, y _b), (x _c, y _c), (x _d, y _d), and can obtain x _d<x _b, as shown in Figure 6.Offset direction decision criteria according to stipulating in technical scheme: x _d>=x _btime image in 2 D code of quick response matrix be left avertence, x _d<x _bspeed per hour response matrix image in 2 D code is right avertence.Judge that this image in 2 D code is as right avertence, and select right avertence cutting method: straight line X=x _bleft field and straight line Y=y _athe intersection area in region of top crop, as shown in Figure 9.Adopting hough transform algorithm detection side dotted line to obtain 2 D code of quick response matrix deviation angle is 79.7 °.Find the center point P of above-mentioned binary image, the coordinate of P is as shown in Figure 3.Take P point as rotation center, with the bilinear interpolation image that turns clockwise, tentatively proofread and correct image in 2 D code of quick response matrix, as shown in figure 12.According to Quick Response Code GB decoding algorithm, utilize horizontal and vertical scanning, to 2 D code of quick response matrix location, determine the central point of three view finding figures, as the P1 in Figure 12, P2, P3.The center point P 1 of three view finding figures, P2, P3 couples together, and forms an isosceles right triangle, and the view finding figure in regulation left side is that view finding figure being positioned on isosceles right triangle right angle.By calculating length of side L12, L23, ratio between L13 two is determined the position of view finding figure in left side in image, ratio is approximately that folded point of two limits of 1 and is the central point of left side view finding figure, P1 is the position of view finding centre of figure in left side in image, as shown in figure 12.Finally, 2 D code of quick response matrix is adjusted to correct orientation by the method for image array transposition upset, be as the criterion in a clockwise direction, left side view finding centre of figure point P1, in the image lower right corner, judges that the 2 D code of quick response matrix after preliminary correction has been offset 180 degree, by (i, j) → (m+1-j, n+1-i) matrix element transposition conversion, proofreaies and correct 2 D code of quick response matrix to correct orientation, as shown in figure 15.Thereby subsequently can decode and realize the 2 D code of quick response matrix identification of pasting on clothes raw material 2 D code of quick response matrix by standard GB/T/T18284-2000.

Embodiment 3

Introduce one embodiment of the present of invention below in conjunction with accompanying drawing 4,7,10,13,16, the extremely correct identification direction of image in 2 D code of quick response matrix correction that is offset 180 ° to the right collecting on clothes raw material, method and embodiment 1 that this embodiment uses are identical.

According to the method identical with embodiment 1, gathered coloured image is changed into binary image, as shown in Figure 4, adopt Sobel edge detection operator to do edge extracting to binary image.Image after edge extracts carry out horizontal direction be the projection of X-direction with vertical square to the projection that is Y direction, then respectively from four direction scanning up and down, determine 2 D code of quick response matrix four summit A, B, C, the D coordinate position (x in image _a, y _a), (x _b, y _b), (x _c, y _c), (x _d, y _d), and can obtain x _d>x _b, as shown in Figure 7.Offset direction decision criteria according to stipulating in technical scheme: x _d>=x _btime image in 2 D code of quick response matrix be left avertence, x _d<x _bspeed per hour response matrix image in 2 D code is right avertence.Judge that this image in 2 D code is as left avertence, and select left avertence cutting method: straight line X=x _bright side area and straight line Y=y _athe intersection area in region of below crop, as shown in figure 10.Then, adopting hough transform algorithm detection side dotted line to obtain 2 D code of quick response matrix deviation angle is 90.1 °.Find the center point P of above-mentioned binary image, the coordinate of P is as shown in Figure 4.Take P point as rotation center, with the bilinear interpolation image that turns clockwise, tentatively proofread and correct image in 2 D code of quick response matrix, as shown in figure 13.According to Quick Response Code GB decoding algorithm, utilize horizontal and vertical scanning, to 2 D code of quick response matrix location, determine the central point of three view finding figures, as the P1 in Figure 13, P2, P3.The center point P 1 of three view finding figures, P2, P3 couples together, and forms an isosceles right triangle, and the view finding figure in regulation left side is that view finding figure being positioned on isosceles right triangle right angle.By calculating length of side L12, L23, ratio between L13 two is determined the position of view finding figure in left side in image, ratio is approximately that folded point of two limits of 1 and is the central point of left side view finding figure, P1 is the position of view finding centre of figure in left side in image, as shown in figure 13.Finally, 2 D code of quick response matrix is adjusted to correct orientation by the method for image array transposition upset, be as the criterion in a clockwise direction, left side view finding centre of figure point P1 is in the image lower left corner, judge that the 2 D code of quick response matrix after preliminary correction has been offset 270 degree, by image array transposition (i, j) → (j, i) exchange (i with left and right row, j) → (i, m+1-j), proofread and correct 2 D code of quick response matrix to correct orientation, as shown in figure 16.Thereby subsequently can decode and realize the 2 D code of quick response matrix identification of pasting on clothes raw material 2 D code of quick response matrix by standard GB/T/T18284-2000.

Claims (6)

1. a 2 D code of quick response matrix recognition methods for clothes, is characterized in that comprising the following steps:

A) image in 2 D code of quick response matrix pasting on CCD camera collection clothes raw material;

B) coloured image of gathered 2 D code of quick response matrix to be identified is converted to 256 color shade images, and with the ground unrest point of the above-mentioned gray level image of Fast median filtering algorithm elimination;

C) adopt maximum variance between clusters to find best segmentation threshold, this calculation ratio juris is, when the weighted sum of the prospect of being separated by threshold value and two pixel group variances of background reaches hour, this threshold value is best segmentation threshold;

D) according to best segmentation threshold, by the gray-scale value of each pixel of above-mentioned gray level image and threshold, if be greater than threshold value, assignment 1, if be less than threshold value, assignment 0, obtains binary image;

E) adopt Sobel edge detection operator to do edge extracting to binary image, between the adjacent area that two have different gray-scale values, always there is gray scale edge, gray scale edge is the discontinuous result of gray-scale value, and this discontinuous common utilization is differentiated and detected easily;

E1) Sobel edge detection operator is a kind of gradient operator, it uses two templates, one is level, another is vertical, two is all 3*3 masterplate, and each template is approached a partial derivative, with horizontal shuttering and vertical formwork, image is carried out to convolution respectively, obtain two matrixes, their difference presentation videos are at two partial derivatives at same position place;

E2), using the maximal value of two numbers of these two matrix correspondence positions as the output of giving point, obtain the edge of image in 2 D code;

F) adopt adaptive two side areas of isolating 2 D code of quick response matrix based on histogram projection cutting-out method;

F1) image after edge extraction carries out the projection that horizontal direction is X-direction, and the number of the pixel that on statistics X=i straight line, gray-scale value is 255, deposits array xProjection[i in] in; Start to scan from left to right from i=0, if condition (xProjection[i]==0) & & (xProjection[i+1] >0) & & (xProjection[i+2] >1) be true, think the horizontal ordinate on a summit finding 2 D code of quick response matrix, defining this point is A point, obtains x _a=i+1; Then scan from top to bottom straight line X=x _a, finding first gray-scale value is 255 pixel, the ordinate of this point for the ordinate of A, obtaining the coordinate that A orders is if condition (xProjection[i] >1) & & (xProjection[i+1] >0) & & (xProjection[i+2]==0) be true, think the horizontal ordinate on another summit of finding 2 D code of quick response matrix, defining this point is C point, obtains x _c=i+1; Then scan from the bottom up straight line X=x _c, finding first gray-scale value is 255 pixel, the ordinate of this point for the ordinate of C, obtaining C point coordinate is

F2) image after edge extraction carries out vertical square to the projection that is Y direction, the number of the pixel that on statistics Y=j straight line, gray-scale value is 255, deposit array yProjection[j in] in, then start to scan from left to right from j=0, if condition (yProjection[j]==0) & & (yProjection[j+1] >0) & & (yProjection[j+2] >1) be true, think the ordinate on the 3rd summit finding 2 D code of quick response matrix, defining this point is D point, obtain y _d=j+1, scanning straight line Y=y then turns left from the right side _d, finding first gray-scale value is 255 pixel, the horizontal ordinate of this point for the horizontal ordinate of D, obtaining the coordinate that D orders is if condition (yProjection[j] >1) & & (yProjection[j+1] >0) & & (yProjection[j+2]==0) be true, think the ordinate on the 4th summit finding 2 D code of quick response matrix, defining this point is B point, obtains y _b=j+1, then scans straight line Y=y from left to right _b, finding first gray-scale value is 255 pixel, the horizontal ordinate of this point for the horizontal ordinate of B, obtaining B point coordinate is

F3) analyze D point and B point coordinate position, set an offset direction decision criteria: work as x _d>=x _btime judge that image in 2 D code of quick response matrix is to left avertence, works as x _d<x _btime judge that QR code image is to right avertence;

F4) according to the 2 D code of quick response matrix offset direction of above-mentioned decision criteria gained, automatically select cutting method: in the time judging 2 D code of quick response matrix to left avertence, be embodied in the horizontal ordinate x that D is ordered _dbe more than or equal to the horizontal ordinate x that B is ordered _b, straight line X=x _bright side area and straight line Y=y _athe intersection area in region of below crop;

In the time judging 2 D code of quick response matrix to right avertence, be embodied in the horizontal ordinate x that D is ordered _athe horizontal ordinate x of ordering than B _blittle, straight line X=x _bleft field and Y=y _athe intersection area of upper area crop, isolate two side areas of 2 D code of quick response matrix;

The regulation of the cutting method of above-mentioned offset direction decision criteria and corresponding offset direction is in order farthest to crop 2 D code of quick response matrix interior zone, obtain two long and narrow side areas, calculated amount while reducing hough transform detection side sideline, has also avoided the interference of the inner line segment of Quick Response Code to detection side sideline simultaneously;

G) adopt hough transform algorithm to detect side dotted line and obtain 2 D code of quick response matrix deviation angle;

G1) polar coordinates ρ of initialization, the array in θ space, the quantification number in ρ direction is the pixel count of image diagonal direction wherein N ₁and N ₂be respectively the wide and high of image, the quantification number in θ direction is 1800, and angle is from 0 °～180 °, 0.1 ° of every lattice;

G2) set up a totalizer A (ρ, θ), juxtaposition finite element is 0;

G3) pixel (x that each gray-scale value on sequential search image is 255 _i, y _i) set up equation ρ=x _icos (θ)+y _isin (θ), calculates ρ value to each discrete value of θ parameter, obtains this pixel (x _i, y _i) sinusoidal curve on polar coordinates, if curve in too small lattice (ρ, θ), this totalizer A (ρ, θ) adds 1;

G4) find out totalizer A (ρ, θ) local maximum above, this value is the conllinear parameter of image collinear point, and its polar theta is the deviation angle of side dotted line;

H) adopt bilinear interpolation image rotating, tentatively proofread and correct image in 2 D code of quick response matrix;

H1) find the central point (a, b) of image after above-mentioned binaryzation, determine that postrotational centre coordinate is for (c, d), according to reverse rotation transformation matrix expression formula: $\left[\begin{array}{c}{x}_1\\ y_1\\ 1\end{array}\right]=\left[\begin{array}{ccc}1& 0& c\\ 0& -1& d\\ 0& 0& 1\end{array}\right]\left[\begin{array}{cc}\cos \left(\mathrm{\&theta;}\right)& \sin \left(\mathrm{\&theta;}\right)\\ -\sin \left(\mathrm{\&theta;}\right)& \cos \left(\mathrm{\&theta;}\right)\\ 0& 1\end{array}\right]\left[\begin{array}{ccc}1& 0& -a\\ 0& -1& b\\ 0& 0& 1\end{array}\right]\left[\begin{array}{c}{x}_0\\ y_0\\ 1\end{array}\right],$ Image is rotated, point (x ₀, y ₀) rotation θ angle becomes point (x ₁, y ₁), wherein (x ₀, y ₀) be any point of not rotating in front image, (x ₁, y ₁) be by point (x ₀, y ₀) rotation θ angle obtains take (a, b) as rotation center point;

H2) image is carried out to bilinear interpolation, image by rotational transform after (x ₁, y ₁) corresponding (x ₀, y ₀) not necessarily integer of coordinate, find (x ₀, y ₀) near four pixels in (x, y), (x+1, y), (x, y+1), (x+1, y+1), according to bilinear interpolation formula, in horizontal and vertical directions, to its interpolation, formula is:

$\begin{array}{c}f(x_0,y_0)=f(x,y)+\mathrm{\&alpha;}[f(x+1,y)-f(x,y)]+\mathrm{\&beta;}[f(x,y+1)-f(x,y)]\\ +\mathrm{\&alpha;\&beta;}[f(x+1,y+1)+f(x,y)-f(x,y+1)-f(x+1,y)]\end{array}$ Wherein α=x ₀-x, β=y ₀-y, obtains point (x ₀, y ₀) gray-scale value f (x ₀, y ₀), this value is exactly point (x ₁, y ₁) gray-scale value;

I) to 2 D code of quick response matrix location, determine the order of three view finding figures, propose a kind of method with the upset of image array transposition 2 D code of quick response matrix is adjusted to correct orientation, realize the correction of 2 D code of quick response matrix;

I1) first horizontal scanning image in 2 D code of quick response matrix from left to right, records the line segment that meets (1:1:3:1:1) view finding graphics proportion feature, and the width of each pixel is permitted and had 0.5 deviation, the most important thing is to record the starting point P of this line segment _rwith terminal P _l, the horizontal ordinate of the centre coordinate of view finding figure is then along straight line from vertical scan direction, record the such line segment that meets 1:1:3:1:1, obtain the ordinate of the centre coordinate of view finding figure; Record according to this mode that image is carried out to horizontal and vertical directions scanning the such line segment that meets 1:1:3:1:1, just can obtain successively the center point coordinate of three view finding figures;

I2) central point of three view finding figures is coupled together, form an isosceles right triangle, the view finding figure in regulation left side is that view finding figure being positioned on isosceles right triangle right angle; Determine the position of view finding figure in left side in image by calculating the length of side and triangular angles;

I3) by the method for image array transposition upset, 2 D code of quick response matrix is adjusted to correct orientation, be as the criterion in a clockwise direction, if view finding figure in left side in the image upper left corner, judges that 2 D code of quick response matrix is correct orientation, without correction; If view finding figure in left side is in the image upper right corner, judge that 2 D code of quick response matrix has been offset 90 degree, by image array transposition (i, j) → (j, i) exchange (i with up-downgoing, j) → (n+1-i, j), proofreaies and correct 2 D code of quick response matrix to correct orientation; If view finding figure in left side in the image lower right corner, is judged 2 D code of quick response matrix and has been offset 180 degree, by the conversion of (i, j) → (m+1-j, n+1-i) matrix element transposition, proofread and correct 2 D code of quick response matrix to correct orientation; If left side view finding is in the image lower left corner, judge that 2 D code of quick response matrix has been offset 270 degree, by image array transposition (i, j) → (j, i) exchange (i with left and right row, j) → (i, m+1-j), proofreaies and correct 2 D code of quick response matrix to correct orientation;

J) to proofreading and correct to the 2 D code of quick response matrix in correct orientation, decode according to standard GB/T/T18284-2000, realize the 2 D code of quick response matrix identification of clothes.

2. the 2 D code of quick response matrix recognition methods of a kind of clothes according to claim 1, is characterized in that, comprises title, the place of production, checker, specification, model and the unit price information of clothes raw material in described image in 2 D code.

3. the 2 D code of quick response matrix recognition methods of a kind of clothes according to claim 1, is characterized in that, in described CCD camera, sensor pixel is 300,000, and the coloured image highest resolution of its collection is 480*640.

4. the 2 D code of quick response matrix recognition methods of a kind of clothes according to claim 1, is characterized in that, the formula that described coloured image is converted to 256 color shade images is:

Gray scale=0.3* red component+0.59* green component+0.11* blue component.

5. the 2 D code of quick response matrix recognition methods of a kind of clothes according to claim 1, it is characterized in that, the noise spot of described Fast Median Filtering elimination image background refers to, select a current pixel as central point, choose the matrix sampling window of a 3*3 size of putting centered by it, first every row in window are carried out to ascending sort, again every row is carried out to ascending sort, get afterwards cornerwise three numbers, sort, get intermediate value, this is worth as current output valve, window moves up and down in image, to entire image filtering.

6. the 2 D code of quick response matrix recognition device of a kind of clothes of the recognition methods as described in any one in claim 1～5, it is characterized in that: described recognition device comprises PC and CCD camera that video frequency collection card is housed, CCD camera connects PC by usb data line, CCD camera passes to PC by usb data line after obtaining the image in 2 D code of quick response matrix of clothes, in PC, the 2 D code of quick response matrix identification software of clothes is decoded to 2 D code information, and the information that decoding is obtained deposits in database; The 2 D code of quick response matrix identification software of above-mentioned clothes comprises that image in 2 D code acquisition module, image correction module, Quick Response Code decoder module and raw material information import database module;

Image in 2 D code acquisition module imports the image in 2 D code of quick response matrix of frame clothes of PC video flowing into by usb data line for obtaining CCD camera;

Image correction module is converted to 256 color shade images for the Quick Response Code coloured image that above-mentioned image in 2 D code acquisition module is obtained, and with Fast median filtering algorithm picture noise point, adopt maximum variance between clusters to find optimal segmenting threshold, image binaryzation, image after adopting Sobel edge detection operator to binaryzation does edge extracting, utilize adaptive two side areas of isolating 2 D code of quick response matrix based on histogram projection cutting-out method, adopt hough transform to detect side dotted line and obtain 2 D code of quick response matrix deviation angle, adopt bilinear interpolation image rotating, the preliminary image in 2 D code of quick response matrix of proofreading and correct, afterwards 2 D code of quick response matrix is located, determine the order of three view finding figures, 2 D code of quick response matrix is adjusted to correct orientation by the method for image array transposition upset, realize the correction of 2 D code of quick response matrix,

Quick Response Code decoder module for according to standard GB/T/T18284-2000 the image in 2 D code of quick response matrix after to above-mentioned correction decode, the information such as title, the place of production, checker, specification, model and unit price in the Quick Response Code that obtains pasting on clothes raw material;

Raw material information imports database module for by the 2 D code information automated import of data storehouse on above-mentioned clothes raw material, realizes informationization and paperless management.