CN102163336A - Method for coding and decoding image identification codes - Google Patents

Abstract

The invention discloses a method for coding and decoding an image identification code. The method for coding the image identification code comprises the following steps of: setting the identification code on a background image; and identifying the background image, wherein identification code color are determined by the following steps of: solving an average value of all pixel point colors in the background area image, finding out current neutral color difference from the background, and using the neutral color difference or the average value of all the colors as the color of the identification code. The method for decoding the image identification code comprises the following steps of: analyzing image color data; extracting edges of an image; seeking for a data positioning unit; adding an external edge frame to data of a source code map according to the position of the data positioning unit; and reading the identification code and judging a code value of the identification code. According to the method, a large number of character code sets are abandoned, so less coding quantity and small changes and damage to an original map are realized, and effective identification can be realized by using a 0.3 mega-pixel common camera. The method is applicable for various complicated background conditions, has abundant overall pattern expressions, and can be used for conveniently endowing pictures with digital information content under the condition of not destroying information expression of the background image.

Description

A kind of Code And Decode method of video identification code

Technical field

The present invention relates to a kind of image coding and decoding method.

Background technology

Need the image of commodity is scanned to obtain the information of these commodity in life, normally used is the two-dimension code of commodity surface printing.Two-dimension code is to encode by black, the different distributions of white pixel in matrix at a coffin.On matrix respective element position, represent binary one with the appearance of point (square point, round dot or other shapes), point binary " 0 " do not appear representing, the permutation and combination of point has been determined the meaning of matrix two-dimensional barcode representative.Two-dimension code is that a kind of graphical symbol automatically identifying and reading that is based upon on the bases such as computer image processing technology, assembly coding principle is handled code system.Representative two-dimension code has: Code One, Maxi Code, QR Code, Data Matrix etc.

Along with the quantity of information that two-dimension code comprised is increasing, can become becomes increasingly complex; And 2 D code information reads by mobile phone cam, have only the professional terminal of supporting the close-shot technology could satisfy user's the demand of initiatively discerning, but present professional terminal costs an arm and a leg, and most of CCD terminal (Charge-coupled Device) is as mobile phone cam, computer cameras etc. do not possess such function, therefore limited by bottlenecks such as information capacity, terminal, two-dimension code is not broken through the bottleneck of market development all the time, is rarely having conduct aspect the active identification especially.

The problem and shortage of two-dimension code:

First: the low user's obstacle that causes of discrimination

Because two-dimensional bar code is that information is carried sign indicating number, the information content is stored in the sign indicating number, coding is compact, the information encoding amount is big, cause the interior data point of unit area too much, need the high-precision CCD camera lens of band microspur to do code value collection and identification, and common mobile phone cam and computer camera are difficult to normal identification, therefore can't propagate universal at cellular network (mobile Internet) and internet arena.

The second, specialized equipment causes the cost obstacle

The special-purpose recognizing apparatus that two-dimension code adopts, buy and hiring cost all higher, 1000 yuan of rent per months/more than the platform, 8000 yuan of purchase costs/more than the platform.With high costsly cause existing in the marketing " cost obstacle ", directly have influence on popularization and development space that two-dimensional bar code is used.

Three, the single application limitation that causes of the form of expression

Two-dimension code is encoded by black, the different distributions of white pixel in matrix, can only accomplish coding and identification under the single background, can't be applicable to the application under the vast complex background condition.

Summary of the invention

The purpose of this invention is to provide the image recognition Code And Decode method that a kind of convenience is used under complex background condition.

To achieve these goals, by the following technical solutions: a kind of Code And Decode method of video identification code is characterized in that: described coding method is:

Identification code is set on background image, background image is identified, definite step of wherein said identification code color is:

One, asks for the mean value of each pixel color of this background area image

$P[R,G,B]=\frac{\underset{i=0}{\overset{i=n-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{j=m-1}{\mathrm{\Σ}}}{p}_{i,j}(r_{i,j},g_{i,j},b_{i,j})}{i\×j};---(1-1)$

Wherein: P[R, G, B]: color average;

R: red color component value;

G: green component values;

B: blue component value;

N: pixel is in the last maximum number of horizontal ordinate;

M: pixel is in the last maximum number of ordinate;

p _{I, j}(r _{I, j}, g _{I, j}, b _{I, j}): at horizontal ordinate is i, and ordinate is the value of each pixel component of this point of j;

Two, judge whether this background is pure color, if then find out current middle dyeing by hue circle;

If not, then carry out the following step:

1., ask for this regional color average earlier by the 1-1 formula;

Obtain the difference square of the color of each pixel and mean value 2., respectively,

p _Δij＝p _ij[(r _ij-r _p) ²+(g _ij-g _p) ²+(b _ij-b _p) ²]

r _Ij: the value of red component;

g _Ij: the value of green component;

b _Ij: the value of blue component; A

r _p: the mean value of red component;

g _p: the mean value of green component;

b _p: the mean value of blue component;

3., obtain poor square mean value:

$q_{\mathrm{\Δ}}=\underset{i=0}{\overset{i=n-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{j=n-1}{\mathrm{\Σ}}}{p}_{\mathrm{\Δij}}/i*j;$

4., with mean value q _ΔBe standard,, greater than q _ΔPixel be divided into a class, less than q _ΔPixel be divided into a class, according to step 1. to the q that 3. calculates the mean value of each class pixel respectively _{Δ 1}And q _{Δ 2}, then with they addition q _{Δ 12}=q _{Δ 1}+ q _{Δ 2}The q of match stop preceding pixel _ΔWith classification back q _{Δ 12}Size, if | q _Δ-q _{Δ 12}|≤δ, δ are the nicety of grading parameter, then need not classification, otherwise should carry out this classification;

5., 4. 1. sorted each group handled to step according to step, till need not to classify again;

6., add up all classification number and such color value, obtain the middle dyeing of every class color respectively;

Select in each color public dyeing as the color of identification code, as there not being common middle dyeing, with the mean value of the middle dyeing of each color color as identification code;

Wherein decoding algorithm is:

One, obtains identification code sign indicating number figure image;

Two, carry out the image color data analysis, image is carried out edge extracting;

Three, analyze marginal date, carry out curve description and signature analysis and ask for;

A, choose all boundary curves, carry out the boundary curve refinement;

B, select the curve of sealing, and remove the curve in the closed interval, only keep outline;

C, the curved profile of each sealing is judged;

1), with the size of slope variation, find out the straight line portion in the curve.2), judge whether straight line portion is parallel in twos; 3), by this parallel in twos straight line portion, prolong the sideline and form a parallelogram; 4), two figures relatively, calculate the ratio of intersection, ratio is greater than certain value, illustrate revise effective;

Four, be starting point with the closed curve lower left corner, along the length of curve with widely extend to original 3 times respectively, form a little square region, if with this starting point is the center, 2 times with the length and width of square region is that new length and width constitutes a new large square zone, if the curve of all sealings in the large square zone all in little square region, illustrates that then this closed curve is the data locking unit in the identification code; Otherwise, other closed curves are handled according to above-mentioned steps;

Five,, add outer rim for the true form diagram data according to the position of data locking unit;

Six, adjust the angle of image, make one side one-tenth level of image;

Seven, adjust image direction, identification code is read again, according to from left to right, order from top to bottom reads;

Eight, identification code is read, and judge code value.

Image is re-recognized and is meant the image spread effect in the human body sensory organ, and it is the process of a certain figure in the brain memory that human body is recognized it.In this process, human body should have the information that enters sense organ, also canned data in the memory will be arranged, and compares processing by canned data and the current information of obtaining, thereby realizes re-recognizing image.In order to re-recognize activity by the human image of computer simulation, the mankind have created various image recognition models, and " prototype Matching Model " is one of wherein the most basic model.

The prototype Matching Model thinks that what store is not the numerous image that human body lives through in memory, but composition diagram picture some " similarity ".By abstract " similarity " of coming out from image as prototype, take it to check the image that will discern, if can find a similar prototype, this image also just has been identified.This model meets the process that neural transmission and memory are sought, and can realize some irregularly, and position, shape, size change, and re-recognize at some local image similar to prototype.But because artificial intelligence field fails to obtain to break through, this model is difficult to realize by computer program is perfect at present.For the image intelligent identification of auxiliary prototype Matching Model, give image accurate data association simultaneously, the invention provides the image recognition Code And Decode method that a kind of convenience is used under complex background condition, be used for re-recognizing of assistant images.

The present invention follows a cover coding rule, can reduce to coded message by decoding, and it and prototype Matching Model complement one another, and can finish image accurately and fast and re-recognize.Its process is, by on the image that will discern, giving discernible coded message, and store image prototype feature value and coded message into prototype library, when computing machine is discerned when re-recognizing this image, obtain the coded message and the prototype feature value of image earlier by coding/decoding method,, promptly directly in storage, navigate to the memory location of this image at prototype library by coded message if the coded message of obtaining is complete, and do difference with the prototype feature value that obtains and compare, finish image and re-recognize.If the coded message of obtaining is imperfect, can in prototype library, find out the prototype feature value that all comprise this encode fragment by coded message, and do difference one by one relatively with the prototype feature value that obtains, finish image by screening and re-recognize.

This method greatly reduces the identification difficulty of prototype Matching Model by giving coding for the image that will discern, thereby has broken through the limitation because of the low a lot of image recognition application facet that cause of discrimination, and its outstanding advantage is as follows:

1, coded message has been abandoned a large amount of character set of conventional two-dimensional sign indicating number, and encoding amount is few, and is very little to the change and the destruction of the image that is encoded;

2, do not have the equipment bottleneck, adopt the CCD camera of common 300,000 pixels effectively to discern, nearly all mobile phone can both be used this service and do that image is re-recognized and data acquisition;

3, easy to identify, zmodem is torn into two by us such as a picture that is encoded, and has only half to keep, and we still can compare mutually by top residual coding and characteristics of image, accurately locate at prototype library, finish image and re-recognize;

4, when decoding is as the criterion with prototype library, the antifalsification height, if coding by authorizing, the prototype feature value that prototype library can the memory encoding image then can not be by resolving;

5, of many uses, be applicable to various complicated image situations, be applied in various packings easily, advertisement, the propaganda material is on the display screen.Can be related various IP links of image and data, services by prototype library, the mode of " sweeping the figure online " by the CCD camera realizes article Internet of Things inlet function.

Figure of description

Fig. 1 is the background image of embodiment among the present invention;

Fig. 2 is through the background image after dividing;

Fig. 3 is through the image behind the coding;

Fig. 4 is the unit rectangle among the present invention;

The direction code of Fig. 5 for defining among the present invention;

Fig. 6 is the synoptic diagram during the judgment data positioning unit among the present invention.

Embodiment

The present invention will be further described below in conjunction with the drawings and specific embodiments.

Embodiment:

For background image shown in Figure 1, at first it is divided, be divided into six zones as shown in Figure 2, forming final Fig. 3 as a result through the coding back.In the zone after each is divided in Fig. 3, added 6 different codings, each is encoded to a unit rectangle, and as shown in Figure 4, a unit rectangle is made of 3 * 3 cells.Between the adjacent unit rectangle distance length greater than 5 cell limits more than.In the unit rectangle, have or not color can be used for representing different information in each cell.Removing a lower left corner has color forever to be used for the carrying out location, data field, is called the data locking unit.Here the unit rectangle just is called identification code, and that is self-defining direction code for the lower left corner among Fig. 3 in 6 unit rectangles, as shown in Figure 5.Just have 6 identification code information like this in this background image, comprise 1 direction code, 5 information points sign indicating numbers.

The form that points to sign indicating number is: X1-X2-X3-X4-X5, wherein, X1, X2, X3, X4, X5 ∈ (0-255); These 5 this backgrounds of having pointed to the sign indicating number unique identifications are being stored these 5 and are being pointed to sign indicating number information pointed in the high in the clouds of cloud computing.

These point to sign indicating number and direction code need be printed on the Background, and simultaneously in order to reduce the influence to Background, the step of specific coding is:

Identification code is set on background image, background image is identified, definite step of wherein said identification code color is:

One, asks for the mean value of each pixel color of this background area image

$P[R,G,B]=\frac{\underset{i=0}{\overset{i=n-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{j=m-1}{\mathrm{\Σ}}}{p}_{i,j}(r_{i,j},g_{i,j},b_{i,j})}{i\×j};---(1-1)$

Wherein: P[R, G, B]: color average;

R: red component mean value;

G: green component mean value;

B: blue component mean value;

N: pixel is in the last maximum number of horizontal ordinate;

M: pixel is in the last maximum number of ordinate;

p _{I, j}(r _{I, j}, g _{I, j}, b _{I, j}): at horizontal ordinate is i, and ordinate is the value of each pixel component of this point of j;

Example: the image M of 3 * 3 pixel sizes is arranged,

$M=\left[\begin{array}{ccc}{p}_{00}\left(\mathrm{125,125,125}\right)& p_{01}\left(\mathrm{100,125,125}\right)& p_{02}\left(\mathrm{70,50,100}\right)\\ p_{10}\left(\mathrm{120,125,120}\right)& p_{11}\left(\mathrm{120,125,125}\right)& p_{12}\left(\mathrm{70,50,100}\right)\\ p_{20}\left(\mathrm{120,125,125}\right)& p_{21}\left(\mathrm{120,125,125}\right)& p_{22}\left(\mathrm{70,50,100}\right)\end{array}\right]$

Average color is:

: the average color of image is P (76,100,110).

Two, judge whether this background is pure color, if then find out the middle dyeing of this look;

Judge the pure color step:

1, first computed image each point and mean value is poor:

Δp _i，j＝p _i，j(r _i，j，g _i，j，b _i，j)-P[R，G，B]

2, in absolute value addition with individual difference;

$\mathrm{\ΔP}=\underset{j=0}{\overset{j=m-1}{\mathrm{\Σ}}}\underset{i=0}{\overset{i=n-1}{\mathrm{\Σ}}}\left|\mathrm{\Δ}{p}_{\mathrm{ij}}\right|$

3, according to purity requirement, specification error maximal value ε, if Δ P＜ε should the zone be a pure color then,

Otherwise just be not;

In dyeing: on hue circle with certain form and aspect at interval 4-7 look be dyeing in this look.Hue circle is referring to " color formation ", and Cui Shengguo is outstanding, the 2nd edition the 5th printing November in 2006, art photography press, Hubei, P14 page or leaf.

If not, then carry out the following step:

1., ask for this regional color average earlier by the 1-1 formula;

Obtain the difference square of the color of each pixel and mean value 2., respectively,

p _Δij＝p _ij[(r _ij-r _p) ²+(g _ij-g _p) ²+(b _ij-b _p) ²]

r _Ij: the value of red component;

g _Ij: the value of green component;

b _Ij: the value of blue component;

r _p: the mean value of red component;

g _p: the mean value of green component;

b _p: the mean value of blue component;

3., obtain poor square mean value

$q_{\mathrm{\Δ}}=\underset{i=0}{\overset{i=n-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{j=n-1}{\mathrm{\Σ}}}{p}_{\mathrm{\Δij}}/i*j;$

4., with mean value q _ΔBe standard, poor square greater than q _ΔPixel be divided into a class, less than q _ΔPixel be divided into a class, according to step 1. to the q that 3. calculates each class pixel respectively _{Δ 1}And q _{Δ 2}, then with they addition q _{Δ 12}=q _{Δ 1}+ q _{Δ 2}The q of match stop preceding pixel _ΔWith classification back q _{Δ 12}Size, if | q _Δ-q _{Δ 12}|≤δ, δ are the nicety of grading parameter, then need not classification, otherwise should carry out this classification;

5., 4. 1. sorted each group handled to step according to step, till need not to classify again;

6., add up all classification number and such color value, obtain the middle dyeing of every class color respectively;

Example: the image N of 4 * 4 pixel sizes is arranged,

$N=\left[\begin{array}{cccc}{p}_{00}\left(\mathrm{0,0,0}\right)& p_{01}\left(\mathrm{0,0,0}\right)& p_{02}\left(\mathrm{25,75,125}\right)& p_{03}\left(\mathrm{25,75,125}\right)\\ p_{10}\left(\mathrm{25,76,126}\right)& p_{11}\left(\mathrm{25,76,126}\right)& p_{12}\left(\mathrm{26,76,126}\right)& p_{12}\left(\mathrm{26,76,126}\right)\\ p_{20}\left(\mathrm{100,120,75}\right)& p_{21}\left(\mathrm{100,120,75}\right)& p_{22}\left(\mathrm{100,120,75}\right)& p_{23}\left(\mathrm{100,120,75}\right)\\ p_{30}\left(\mathrm{100,121,74}\right)& p_{31}\left(\mathrm{100,120,74}\right)& p_{32}\left(\mathrm{100,119,73}\right)& p_{33}\left(\mathrm{100,120,75}\right)\end{array}\right]$

Each pixel to image N is carried out color classification, nicety of grading: δ=100

Average color: P (75,105,92)

$q_{\mathrm{\Δ}}=\frac{{(0-75)}^2+{(0-105)}^2+{(0-92)}^2+...+{(100-75)}^2+{(120-125)}^2+{(75-92)}^2}{4\×4}$

$=5265$

With q _ΔBe standard, the pixel of image N is classified; Because P ₀₀And P ₀₁Square less than 5265, therefore remaining point is divided into two class N1, N2 all greater than 5265 to these two points to the difference of average color;

N1＝{p ₀₀，p ₀₁}，

N2＝{p ₀₂，p ₀₃，p ₁₀，p ₁₁，p ₁₂，p ₁₃，p ₂₀，p ₂₁，p ₂₂，p ₂₃，p ₃₀，p ₃₁，p ₃₂，p ₃₃}

Calculate: q _{Δ 1}=0, q ₂=2505, q _{Δ 12}=q _{Δ 1}+ q _{Δ 2}=2505

: | q _Δ-q _{Δ 12}|=2560＞δ needs to continue classification;

To subgraph N1, N2 is with the method classification in 1 to 4 step, up to satisfying nicety of grading respectively.

Last classification results is: N1, N2, N3;

N1＝{p ₀₀，p ₀₁}

N2＝{p ₀₂，p ₀₃，p ₁₀，p ₁₁，p ₁₂，p ₁₃}

N3＝{p ₂₀，p ₂₁，p ₂₂，p ₂₃，p ₃₀，p ₃₁，p ₃₂，p ₃₃}

Select in each color public dyeing as the color of identification code, as there not being common middle dyeing, with the mean value of the middle dyeing of each color color as identification code.

Example: average color point p ₁(0,60,255), corresponding middle dyeing is (255,0,0) and (0,255,200).

Just can encode by above coding method, and the color of the sensing sign indicating number that comes out of editor is little to the background color influence, does not hinder the overall appearance of background, helps a difference sign indicating number district and a background again to a width of cloth background image.

Two, analyze characteristics of image in the identification code district, coded image information is corresponding one by one image information and coding;

1, with image filtering, edge extracting, binaryzation then; (this algorithm can be rolled up for the 10th phase in firelight or sunlight Tsing-Hua University journal natural science edition referring to " a kind of new color images edge detection algorithm " in 2005 the 45th)

2, according to the image of binaryzation, according to from left to right, from top to bottom order, form single-row scale-of-two ordered series of numbers string, a pixel accounts for a byte, information coding front end adds the wide value of yard head of district;

3, ordered series of numbers is put in order, compression forms encoding image information.

To obtain information coding and identification code and be uploaded to the high in the clouds server, on the server of high in the clouds corresponding information be arranged like this.

When the image after the user uses camera to coding is taken pictures, will decode to image, the step of concrete decoding is:

1, uses CCD recognizing apparatus readout code figure, obtain identification code sign indicating number figure image;

2, sign indicating number figure image is carried out image restoration and enhancement process;

3, carry out the image color data analysis, image is carried out edge extracting;

4, analyze marginal date, carry out curve description and signature analysis and ask for;

A, choose all boundary curves, carry out the boundary curve refinement;

B, select the curve of sealing, and remove the curve in the closed interval, only keep outline;

C, the curved profile of each sealing is judged;

1), with the size of slope variation, find out the straight line portion in the curve.2), judge whether straight line portion is parallel in twos; 3), by this parallel in twos straight line portion, prolong the sideline and form a parallelogram; 4), two figures relatively, calculate the ratio of intersection, ratio is greater than certain value, illustrate revise effective; 5) if above-mentioned condition all satisfies, then this zone may be the data locking unit of code area;

5, be starting point with the closed curve lower left corner, along the length of curve with widely extend to original 3 times respectively, form a little square region, if with this starting point is the center, 2 times with the length and width of square region is that new length and width constitutes a new large square zone, if the curve of all sealings in the large square zone all in little square region, illustrates that then this closed curve is the data locking unit of identification code.Otherwise, other closed curves are handled according to above-mentioned steps.Here owing to more than the length of distance between the adjacent unit rectangle, therefore can not judge the situation that the adjacent cells rectangle disturbs greater than 5 cell limits.

As shown in Figure 6, because all closed curves all in little square region, therefore can judge that this closed curve is the data locking unit.

6,, add outer rim for the true form diagram data according to the position of data locking unit;

7, according to the limit of code pattern, adjust the angle of image, make one side one-tenth level of image;

8, search direction identity code looks at whether the direction of image is correct, if incorrect then image is rotated accordingly; After adjusting direction, identification code is read, according to from left to right, order from top to bottom reads again;

9, with template matching method with read compiling method and decode respectively, if two kinds of method unanimities illustrate that decoding is correct;

Read compiling method:

1., the code area on average is divided into nine zones;

2., respectively calculate each regional average gray, if mean value, illustrates then that this zone for black, is encoded to 1 greater than the gray-scale value of setting, otherwise coding just is 0, so obtains 9 binary codings.

3., table look-up, obtain the implication of encoding.

Template matching method:

1., prepare the two-value standard picture of each coding, remain on server;

2., the coding that cuts out and each template that will be from image be poor;

3., difference relatively, the template of difference minimum is exactly the template that we will look for;

4., table look-up, obtain the implication of encoding.

The present invention has abandoned a large amount of character sets, and encoding amount is few, former figure is changed and destroys little, and the camera of common 300,000 pixels can effectively be discerned.Be applicable to various complex background situations, the configuration performance is very abundant, it can not destroy under the information statement prerequisite of Background easily, give the picture digital contents, can be applied in various packings easily, advertisement, propaganda material, on the image display panel, realize the coexistence of data and picture and text.Adopt information points type coding, data content is kept at the data center of cloud service end, and data capacity and form are all unrestricted, safety, antifalsification height.

Claims (1)

1. the Code And Decode method of a video identification code, it is characterized in that: described coding method is:

Identification code is set on background image, background image is identified, definite step of wherein said identification code color is:

One, asks for the mean value of each pixel color of this background area image

$P[R,G,B]=\frac{\underset{i=0}{\overset{i=n-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{j=m-1}{\mathrm{\Σ}}}{p}_{i,j}(r_{i,j},g_{i,j},b_{i,j})}{i\×j};---(1-1)$

Wherein: P[R, G, B]: color average;

R: red color component value;

G: green component values;

B: blue component value;

N: pixel is in the last maximum number of horizontal ordinate;

M: pixel is in the last maximum number of ordinate;

p _{I, j}(p _{I, j}, g _{I, j}, b _{I, j}): at horizontal ordinate is i, and ordinate is the value of each pixel component of this point of j;

Two, judge whether this background is pure color, if then find out current middle dyeing by hue circle;

If not, then carry out the following step:

1., ask for this regional color average earlier by the 1-1 formula;

Obtain the difference square of the color of each pixel and mean value 2., respectively,

p _Δij＝p _ij[(r _ij-r _p) ²+(g _ij-g _p) ²+(b _ij-b _p) ²]

r _Ij: the value of red component;

g _Ij: the value of green component;

b _Ij: the value of blue component;

r _p: the mean value of red component;

g _p: the mean value of green component;

b _p: the mean value of blue component;

3., obtain poor square mean value:

$q_{\mathrm{\Δ}}=\underset{i=0}{\overset{i=n-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{j=n-1}{\mathrm{\Σ}}}{p}_{\mathrm{\Δij}}/i*j;$

4., with mean value q _ΔBe standard,, greater than q _ΔPixel be divided into a class, less than q _ΔPixel be divided into a class, according to step 1. to the q that 3. calculates the mean value of each class pixel respectively _{Δ 1}And q _{Δ 2}, then with they addition q _{Δ 12}=q _{Δ 1}+ q _{Δ 2}The q of match stop preceding pixel _ΔWith classification back q _{Δ 12}Size, if | q _Δ-q _{Δ 12}|≤δ, δ are the nicety of grading parameter, then need not classification, otherwise should carry out this classification;

5., 4. 1. sorted each group handled to step according to step, till need not to classify again;

6., add up all classification number and such color value, obtain the middle dyeing of every class color respectively;

Select in each color public dyeing as the color of identification code, as there not being common middle dyeing, with the mean value of the middle dyeing of each color color as identification code;

Wherein coding/decoding method is:

One, obtains identification code sign indicating number figure image;

Two, carry out the image color data analysis, image is carried out edge extracting;

Three, analyze marginal date, carry out curve description and signature analysis and ask for;

A, choose all boundary curves, carry out the boundary curve refinement;

B, select the curve of sealing, and remove the curve in the closed interval, only keep outline;

C, the curved profile of each sealing is judged;

1), with the size of slope variation, find out the straight line portion in the curve.2), judge whether straight line portion is parallel in twos; 3), by this parallel in twos straight line portion, prolong the sideline and form a parallelogram; 4), two figures relatively, calculate the ratio of intersection, ratio is greater than certain value, illustrate revise effective;

Four, be starting point with the closed curve lower left corner, along the length of curve with widely extend to original 3 times respectively, form a little square region, if with this starting point is the center, 2 times with the length and width of square region is that new length and width constitutes a new large square zone, if the curve of all sealings in the large square zone all in little square region, illustrates that then this closed curve is the data locking unit in the identification code; Otherwise, other closed curves are handled according to above-mentioned steps;

Five,, add outer rim for the true form diagram data according to the position of data locking unit;

Six, adjust the angle of image, make one side one-tenth level of image;

Seven, adjust image direction, identification code is read again, according to from left to right, order from top to bottom reads;

Eight, identification code is read, and judge code value.

Images