CN102201118B - Method for positioning boundaries of belt layers at tyre crown part of X-ray image of tyre - Google Patents

Abstract

The invention relates to a method for positioning the boundaries of belt layers at a tyre crown part of an X-ray image of a tyre. The method comprises the steps of: positioning belt layers of all the belt layers at the tyre crown part of the X-ray image of a tyre, carrying out frequency-spectrum analysis on the tyre crown part of the X-ray image of the tyre with overlapped multiple belt layer textures, respectively processing frequency-spectrum information by using a Gaussian wedge-shaped filter with the angle between 0 degree and 179 degrees, restoring the extracted frequency-spectrum information into images, carrying out grey statistics to obtain gray statistical information vectors, combining all the gray statistical information vectors and converting the combined gray statistical information vectors into a gray information image, observing and selecting the gray information vectors which can be used for measuring and calculating the boundary information of all the belt layers, and calculating the boundary position coordinates of all the belt layers according to the selected gray information vectors.

Description

Be applied to tire x-ray image crown part position belt boundary alignment method

[technical field]

The invention belongs to the digital image processing techniques field, the many texture images that are specifically related to be applied to crown position in the tire x-ray image forms because multi-band bundle layer texture superposes mutually carry out each belt boundary alignment.

[background technology]

Tire is the main action executing member of motor vehicles, and stablizing and whether meeting the safe design standard of tyre performance will directly determine to use the personal security of motor vehicles.Tire is the rotatable body of a kind of circular ring type of tubular section, is processed through applying, moulding, sulfuration typing by rubber preformed material, the compounded rubber preformed material of multilayer with all-steel cord.The tire of described structure, the constituent material at its crown position inevitably exist the problems such as all-steel cord density unevenness, wire of steel wire tire cord fracture, all-steel cord disappearance.Perspective imaging principle according to X-ray can obtain inside tires all-steel cord state and distribution situation, by to the state of all-steel cord and the analysis of distribution situation, obtains the information of defective, thereby realizes tire is declared level.

Existing tire X-ray machine test unit, in process of the test, tire is mounted, is clamped on four rotatable reference columns.In the time of the reference column rotation, drive tire rotation, at this moment, the X-ray generator tube emission X-ray line by being positioned at the tire inboard penetrates tire, is received by the U-shaped receiver that is positioned at the tire outside.

Existing tire X-ray machine test unit after obtaining the x-ray image of tire, take the x-ray image of tire as foundation, is declared level.Declare at present the main people of dependence of level by the observation of naked eyes, the tire x-ray image defective that is presented on the computer screen is carried out artificial cognition, by artificial intuitive judgment, confirm the grade of tire.At the crown position, owing to have a plurality of belts mutually to superpose, the skew of belt position or scarce layer, the defective that can have a strong impact on tire x-ray image crown position is declared level.Overlapping belt can be covered the distribution situation of each belt each other mutually, therefore at the crown position of tire x-ray image, with the naked eye observe intuitively the boundary position that is difficult to tell belt, the unascertainable words of the boundary position of belt, automatically declaring level for the defective of belt just is difficult to carry out, therefore need to each belt be positioned before the computing machine automatic defect declares level carrying out, follow-up like this declare level work and could shoot the arrow at the target.

[summary of the invention]

The present invention seeks to overcome the prior art above shortcomings, a kind of tire x-ray image crown part position belt boundary alignment method that is applied to is provided.

The method that the present invention adopts Digital Image Processing is carried out boundary alignment to the belt at crown position, mark off the effective coverage of each belt, at follow-up declaring in the level work, help in concrete zone, to declare targetedly level operation and defect recognition, defective is divided by diverse location, different ownership, simultaneously help also to judge whether belt is offset, whether have the layer of lacking phenomenon to occur.

The present invention is directed to the tire x-ray image crown position that multi-band bundle layer texture superposes mutually and carry out spectrum analysis, Gauss's wedge shape wave filter is accepted or rejected spectrum information when adopting respectively angle to be 0 ° to 179 °, the spectrum information that extracts is reduced to image and carries out gray-scale statistical, obtain the grey-level statistics vector, all grey-level statistics Vector Groups merging are converted into the half-tone information image, by observing the half-tone information vector of selecting to can be used for calculating each belt boundary information, according to the half-tone information vector that is selected, calculate the boundary position coordinate of each belt, thereby each belt Boundary Recognition that will mutually superpose out, realize the boundary alignment of each belt, implementation step is:

The first step is obtained a complete tire x-ray image I _X, at this x-ray image I _XInside also is simultaneously that its crown position intercepting size is 2 ^N* 2 ^NSquare-shaped image I _OThe value of N requires: N is the integer greater than 1, has simultaneously 2 ^NLess than tire x-ray image I _XWidth and 2 ^NLess than x-ray image I _XHeight, and so that I _OWidth just can cover whole crown position, image I simultaneously _OFour limits should be parallel to respectively tire x-ray image I _XFour limits;

Second step is to first step truncated picture I _OCarry out 2 dimension Fast Fourier Transform (FFT)s, obtain spectrum information matrix I _F, with spectrum information matrix I _FCarry out shifting function, and then obtain 2 dimension spectrum information matrix I of point centered by frequency (0,0) _FS, I _F, I _FSAll and image I _OWith size, I _F, I _FSWith image I _OMeet following relation:

I _F＝fft2(I _O)

I _FS＝fftshift(I _F)

Wherein fft2 represents to carry out quick 2 dimension Fourier transforms, and in the result images after the fft2 conversion, the corresponding value of information of frequency (0,0) is corresponding to spectrum information matrix I _FHeader element position, the upper left corner;

Fftshift represents spectrum information matrix I _FCarry out shifting function, namely take the matrix center as initial point, do the coordinate axis of horizontal and vertical, with 1,3 quadrant interchanges, the matrix that 2,4 quadrant interchanges are obtained; After the exchange, the corresponding value of information of frequency (0,0) is corresponding to the center of matrix;

In the 3rd step, utilize following Gauss's wedge shape wave filter formula, with spectrum information matrix I _FSCarry out the filtering of Gauss's wedge shape,

$G_1(x,y)=e^{-\frac{{(\mathrm{\θ}-{\mathrm{\θ}}_c)}^2}{2{\mathrm{\σ}}^2}}$

G ₂(x，y)＝G ₁(-x，-y)

G(x，y)＝G ₁(x，y)+G ₂(x，y)

Wherein, θ _CBe the angle center, its span is 0 °≤θ _c＜180 °

σ is standard deviation, has determined the width of Gaussian curve on the circumferential direction of certain radius

θ is spectrum information matrix I _FSIn certain first vegetarian refreshments be the summit with its frequency central point (0,0), take the angle value of horizontal dextrad direction as 0 °, as 0 °≤θ _cIn the time of＜180 °, the θ span is-90 °≤θ≤270 °, and the computing method of θ are:

Work as I _FSIn the coordinate (x, y) of certain element be positioned at 1, during 4 quadrant, have

$\mathrm{\θ}={\tan }^{-1}\left(\frac{y}{x}\right)$

Work as I _FSIn the coordinate (x, y) of certain pixel be positioned at 2, during 3 quadrant, have

$\mathrm{\θ}={\tan }^{-1}\left(\frac{y}{x}\right)+180$

G is 2 dimension Gauss wedge shape wave filters, and this is one is the Filtering Template of origin symmetry, i.e. G about frequency central point (0,0) ₁And G ₂About frequency central point (0,0) symmetry;

At first make θ _C=0 °, and according to the design parameter of each belt, the σ parameter of given Gauss's wedge shape wave filter is utilized Gauss's wedge shape Filtering Formula, generates the Gauss's wedge shape wave filter G about (0,0) origin symmetry;

The 4th step is with spectrum information matrix I _FSG carries out dot product with Gauss's wedge shape wave filter, obtains spectrum information matrix I _FL, this matrix has at utmost kept direction, the spectrum information of expectation texture, has abandoned direction, the spectrum information of other textures, and the dot product formula is as follows;

I _FL＝I _FS·G

The 5th step is with spectrum information matrix I _FLCarry out shifting function, and carry out inverse fast Fourier transform, thereby with I _FLBe reduced to image I _R, I _FLWith I _RSatisfy following relation:

I _R＝ifft2(fftshift(I _FL))

Wherein, fftshift represents spectrum information matrix I _FLCarry out shifting function, namely take the matrix center as initial point, do the coordinate axis of horizontal and vertical, with 1,3 quadrant interchanges, 2,4 quadrant interchanges are obtained new matrix; Before the exchange, the corresponding value of information of frequency (0,0) is corresponding to spectrum information matrix I _FLThe center; After the exchange, the corresponding value of information of frequency (0,0) is corresponding to the header element position, the upper left corner of matrix of consequence,

Ifft2 represents to carry out quick 2 dimension inverse Fourier transforms,

Through the image I after the filtering reduction _R, farthest kept the texture of certain direction, and the texture information of other directions is weakened;

In the 6th step, calculate I _RPixel grey scale average on the vertical direction of each horizontal ordinate position obtains one 1 vector of tieing up

Computing formula is as follows:

$V_{{\mathrm{\θ}}_C}\left(i\right)=\frac{1}{2^N}\underset{m=1}{\overset{2^N}{\mathrm{\Σ}}}{I}_R(i,m),i=\mathrm{1,2},...,2^N$

Wherein, I _R(i, m) presentation video I _RIn the grey scale pixel value of certain pixel (i, m), i represents abscissa value, m represents ordinate value;

In the 7th step, make again θ _CBe respectively 1 °, 2 °, 3 ° ... 179 °, repeated for the 3rd step to the 6th step, obtain

V _θ(i) θ＝0，1，…，179；i＝1，2，…2 ^N

And with following formula vector set V _θ(i) θ=0,1 ..., 179; I=1,2 ... 2 ^NBe rewritten as 2 dimension matrixes

V(θ，i) θ＝0，1，…，179；i＝1，2，…2 ^N

Being that matrix V is one 2 dimension matrix, highly is 180, and width is 2 ^N, this 2 dimension matrix is shown with image mode, can observe each belt boundary position distribution situation;

The 8th step by the observation to 2 dimension image V, can pick out and each belt horizontal strip region of some high gray-scale values one to one, and hypothesis has M belt here, and then the horizontal strip region of high gray-scale value is made as Zone _iI=1,2, M, the number of the horizontal strip region of high gray-scale value is consistent with the belt number, simultaneously the horizontal strip region of each high gray-scale value is corresponding with some belts, and the width of the horizontal strip region of high gray-scale value equals the width of the belt corresponding with it, and the left and right sides boundary coordinate of the horizontal strip region of high gray-scale value equals the belt left and right sides border corresponding with it at I _OIn coordinate, will with the horizontal strip region Zone of high gray-scale value _iI=1,2 ..., the abscissa value β of M _iI=1,2 ..., M extracts, can extract with each belt one to one I=1,2 ..., the M vector is for M

I=1,2 ..., the M vector can calculate M belt boundary information;

In the 9th step, calculate

I=1,2 ..., the average of M

I=1,2 ..., M is with average

I=1,2 ..., M multiplication by constants coefficient lambda _iI=1,2 ..., M obtains threshold value T _iI=1,2 ..., M, namely satisfy following relational expression:

$T_i={\mathrm{\λ}}_i{\stackrel{\‾}{V}}_{{\mathrm{\β}}_i},i=\mathrm{1,2},...,M$

For each vector I=1,2 ..., M, (being x=0) begins search from the left side, when the first fit condition

$V_{{\mathrm{\β}}_i}\left(x\right)\≤T_i\≤V_{{\mathrm{\β}}_i}(x+1),x=\mathrm{1,2},...,2^N-1;i=\mathrm{1,2},...,M$

The time, then the x of this moment is designated as the left margin F on required border _LiI=1,2 ..., M;

In like manner, from vector

I=1,2 ..., the right side of M (is x=2 ^N) beginning search left, when the first fit condition

$V_{{\mathrm{\β}}_i}\left(x\right)\≤T_i\≤V_{{\mathrm{\β}}_i}(x-1),x=2^N,2^N-1,...,2;i=\mathrm{1,2},...,M$

The time, then the x of this moment is designated as the right margin E on required border _RiI=1,2 ..., M;

At gray level image I _OIn, respectively with I _O(E _Li, y), I _O(E _Ri, y) i=1,2 ..., M; Y=1,2 ..., 2 ^NThe pixel value gray-scale value is set to (255,255,255), namely draws the boundary line of each belt with white lines;

So far, finish the belt boundary alignment.

Advantage of the present invention and good effect

The present invention can pick out the border of each belt accurately, mark off the effective coverage of each belt, at follow-up declaring in the level work, help in concrete zone, to declare targetedly level operation and defect recognition, defective is divided by diverse location, different ownership, simultaneously help also to judge whether belt is offset, whether have the layer of lacking phenomenon to occur; What the method was conducive to the tire x-ray image accurately declares level, improves accuracy rate and efficient that tire is declared level.

[description of drawings]

Fig. 1 is tire x-ray image crown position intercepting schematic diagram;

Fig. 2 is Gauss's wedge shape Filtering Template schematic diagram;

Fig. 3 is the process flow diagram of described belt Texture Boundaries localization method;

Fig. 4 is the average statistical graph that obtains after filtering under certain angle direction;

Fig. 5 is after adopting respectively 0 °～179 ° Gauss's wedge shape wave filters that the spectrum information of tire x-ray image crown part bit image is carried out filtering, the grey-level statistics matrix that shows with image format that obtains;

Fig. 6 is boundary alignment final effect figure.

[embodiment]

As shown in Figure 1, the tyre crown position x-ray image multi-band bundle layer texture separation method that is applied to of the present invention, its effective object is the subimage of the suitable size in tire x-ray image crown position.The size of subimage is 2 ⁹* 2 ⁹, subimage has covered left hand edge and the right hand edge at crown position; Four limits of subimage are parallel to four limits of tire x-ray image simultaneously;

As shown in Figure 2, this is to adopt Gauss's wedge shape Filtering Formula, Gauss's wedge shape Filtering Template schematic diagram of generation.This Gauss's wedge shape Filtering Template is about central point as can be known.

As shown in Figure 3, the concrete steps of tire x-ray image crown part position belt boundary alignment method are as follows:

The first step is obtained a complete tire x-ray image I _X, at this x-ray image I _XInside also is simultaneously that its crown position intercepting size is 2 ^N* 2 ^NSquare-shaped image I _OThe value of N requires: N is the integer greater than 1, has simultaneously 2 ^NLess than tire x-ray image I _XWidth and 2 ^NLess than x-ray image I _XHeight, and so that I _OWidth just can cover whole crown position, image I simultaneously _OFour limits should be parallel to respectively tire x-ray image I _XFour limits;

Second step is to first step truncated picture I _OCarry out 2 dimension Fast Fourier Transform (FFT)s, obtain spectrum information matrix I _F, with spectrum information matrix I _FCarry out shifting function, and then obtain 2 dimension spectrum information matrix I of point centered by frequency (0,0) _FS, I _F, I _FSAll and image I _OWith size, I _F, I _FSWith image I _OMeet following relation:

I _F＝fft2(I _O)

I _FS＝fftshift(I _F)

Wherein fft2 represents to carry out quick 2 dimension Fourier transforms, and in the result images after the fft2 conversion, the corresponding value of information of frequency (0,0) is corresponding to spectrum information matrix I _FHeader element position, the upper left corner.

Fftshift represents I _FCarry out shifting function, namely take the matrix center as initial point, do the coordinate axis of horizontal and vertical, with 1,3 quadrant interchanges, the matrix that 2,4 quadrant interchanges are obtained.After the exchange, the corresponding value of information of frequency (0,0) is corresponding to the center of matrix.

In the 3rd step, utilize following Gauss's wedge shape wave filter formula, with spectrum information matrix I _FSCarry out the filtering of Gauss's wedge shape,

$G_1(x,y)=e^{-\frac{{(\mathrm{\θ}-{\mathrm{\θ}}_c)}^2}{2{\mathrm{\σ}}^2}}$

G ₂(x，y)＝G ₁(-x，-y)

G(x，y)＝G ₁(x，y)+G ₂(x，y)

Wherein, θ _CBe the angle center, its span is 0 °≤θ _c＜180 °

σ is standard deviation, has determined the width of Gaussian curve on the circumferential direction of certain radius

θ is I _FSIn certain first vegetarian refreshments be the summit with its frequency central point (0,0), take the angle value of horizontal dextrad direction as 0 °, as 0 °≤θ _cIn the time of＜180 °, the θ span is-90 °≤θ≤270 °, and the computing method of θ are:

Work as I _FSIn the coordinate (x, y) of certain element be positioned at 1, during 4 quadrant, have

$\mathrm{\θ}={\tan }^{-1}\left(\frac{y}{x}\right)$

Work as I _FSIn the coordinate (x, y) of certain pixel be positioned at 2, during 3 quadrant, have

$\mathrm{\θ}={\tan }^{-1}\left(\frac{y}{x}\right)+180$

G is 2 dimension Gauss wedge shape wave filters, and this is one is the Filtering Template of origin symmetry, i.e. G about frequency central point (0,0) ₁And G ₂About frequency central point (0,0) symmetry;

At first make θ _C=0 °, and according to the design parameter of each belt, the σ parameter of given Gauss's wedge shape wave filter is utilized Gauss's wedge shape Filtering Formula, generates the Gauss's wedge shape wave filter G about (0,0) origin symmetry;

The 4th step is with spectral matrix I _FSG carries out dot product with Gauss's wedge shape wave filter, obtains spectrum information matrix I after the filtering _FL, this matrix has at utmost kept direction, the spectrum information of expectation texture, has abandoned direction, the spectrum information of other textures, and the dot product formula is as follows;

I _FL＝I _FS·G

The 5th step is with filtered spectrum information matrix I _FLCarry out shifting function, and carry out inverse fast Fourier transform, thereby with I _FLBe reduced to image I _R, I _FLWith I _RSatisfy following relation:

I _R＝ifft2(fftshift(I _FL))

Wherein, fftshift represents I _FLCarry out shifting function, namely take the matrix center as initial point, do the coordinate axis of horizontal and vertical, with 1,3 quadrant interchanges, 2,4 quadrant interchanges are obtained new matrix.Before the exchange, the corresponding value of information of frequency (0,0) is corresponding to I _FLThe center; After the exchange, the corresponding value of information of frequency (0,0) is corresponding to the header element position, the upper left corner of matrix of consequence,

Ifft2 represents to carry out quick 2 dimension inverse Fourier transforms,

Through the image I after the filtering reduction _R, farthest kept the texture of certain direction, and the texture information of other directions is weakened;

In the 6th step, calculate I _RPixel grey scale average on the vertical direction of each horizontal ordinate position obtains one 1 vector of tieing up

Computing formula is as follows:

$V_{{\mathrm{\θ}}_C}\left(i\right)=\frac{1}{2^N}\underset{m=1}{\overset{2^N}{\mathrm{\Σ}}}{I}_R(i,m),i=\mathrm{1,2},...,2^N$

Wherein, I _R(i, m) presentation video I _RIn the grey scale pixel value of certain pixel (i, m), i represents abscissa value, m represents ordinate value;

In the 7th step, make again θ _CBe respectively 1 °, 2 °, 3 ° ... 179 °, repeated for the 3rd step to the 6th step, obtain

V _θ(i) θ＝0，1，…，179；i＝1，2，…2 ^N

And with following formula vector set V _θ(i) θ=0,1 ..., 179; I=1,2 ... 2 ^NBe rewritten as 2 dimension matrixes

V(θ，i) θ＝0，1，…，179；i＝1，2，…2 ^N

Being that matrix V is one 2 dimension matrix, highly is 180, and width is 2 ^N, this 2 dimension matrix is shown with image mode, can observe each belt boundary position distribution situation;

The 8th step by the observation to 2 dimension image V, can pick out and each belt horizontal strip region of some high gray-scale values one to one, and hypothesis has M belt here, and then the horizontal strip region of high gray-scale value is made as Zone _i

I=1,2, M, the number of the horizontal strip region of high gray-scale value is consistent with the belt number, simultaneously the horizontal strip region of each high gray-scale value is corresponding with some belts, and the width of the horizontal strip region of high gray-scale value equals the width of the belt corresponding with it, and the left and right sides boundary coordinate of the horizontal strip region of high gray-scale value equals the belt left and right sides border corresponding with it at I _OIn coordinate, will with the horizontal strip region Zone of high gray-scale value _iI=1,2 ..., the abscissa value β of M _i

I=1,2 ..., M extracts, can extract with each belt one to one

I=1,2 ..., the M vector is for M

I=1,2 ..., the M vector can calculate M belt boundary information;

In the 9th step, calculate

I=1,2 ..., the average of M I=1,2 ..., M is with average

I=1,2 ..., M multiplication by constants coefficient lambda _iI=1,2 ..., M obtains threshold value T _iI=1,2 ..., M, namely satisfy following relational expression:

$T_i={\mathrm{\λ}}_i{\stackrel{\‾}{V}}_{{\mathrm{\β}}_i},i=\mathrm{1,2},...,M$

For each vector

I=1,2 ..., M, (being x=0) begins search from the left side, when the first fit condition

$V_{{\mathrm{\β}}_i}\left(x\right)\≤T_i\≤V_{{\mathrm{\β}}_i}(x+1),x=\mathrm{1,2},...,2^N-1;i=\mathrm{1,2},...,M$

The time, then the x of this moment is designated as the left margin E on required border _LiI=1,2 ..., M;

In like manner, from vector

I=1,2 ..., the right side of M (is x=2 ^N) beginning search left, when the first fit condition

$V_{{\mathrm{\β}}_i}\left(x\right)\≤T_i\≤V_{{\mathrm{\β}}_i}(x-1),x=2^N,2^N-1,...,2;i=\mathrm{1,2},...,M$

The time, then the x of this moment is designated as the right margin E on required border _RiI=1,2 ..., M;

At gray level image I _OIn, respectively with I _O(E _Li, y), I _O(E _Ri, y) i=1,2 ..., M; Y=1,2 ..., 2 ^NThe pixel value gray-scale value is set to (255,255,255), namely draws the boundary line of each belt with white lines;

So far, finish the belt boundary alignment.

As shown in Figure 4, this is after adopting certain Gauss's wedge shape wave filter that filtering is carried out at tire x-ray image crown position, the gray-scale statistical curve of the filtering result images that obtains; Horizontal ordinate has represented the width of image, and ordinate represents the pixel grey scale average at certain horizontal ordinate place of filtering result images.

As shown in Figure 5, this is the gray-scale statistical matrix that is comprised of 180 gray-scale statisticals vector that shows with image format, 1,2, No. 3 high gray-scale value zone correspondence wherein 3 different belts.Can find out, the boundary position in 3 high grey value profile zones is that significantly the border, the left and right sides of extracting these 3 high gray-scale value zones can obtain the border, the left and right sides of 3 belts.

As shown in Figure 6, draw 3 pairs of vertical straight lines of white at the crown part bit image of tire x-ray image, the corresponding border, the left and right sides of 3 belts.

Claims (2)

1. one kind is applied to tire x-ray image crown part position belt boundary alignment method, it is characterized in that: carry out spectrum analysis for the tire x-ray image crown position that multi-band bundle layer texture superposes mutually, Gauss's wedge shape wave filter is accepted or rejected spectrum information when adopting respectively angle to be 0 ° to 179 °, the spectrum information that extracts is reduced to image and carries out gray-scale statistical, obtain the grey-level statistics vector, all grey-level statistics Vector Groups merging are converted into the half-tone information image, by observing the half-tone information vector of selecting to can be used for calculating each belt boundary information, according to the half-tone information vector that is selected, calculate the boundary position coordinate of each belt, thereby each belt Boundary Recognition that will mutually superpose out, realize the boundary alignment of each belt, implementation step is:

The first step is obtained a complete tire x-ray image I _X, at this x-ray image I _XInside also is simultaneously that its crown position intercepting size is 2 ^N* 2 ^NSquare-shaped image I _OThe value of N requires: N is the integer greater than 1, has simultaneously 2 ^NLess than tire x-ray image I _XWidth and 2 ^NLess than x-ray image I _XHeight, and so that I _OWidth just can cover whole crown position, image I simultaneously _OFour limits should be parallel to respectively tire x-ray image I _XFour limits;

Second step is to first step truncated picture I _OCarry out 2 dimension Fast Fourier Transform (FFT)s, obtain spectrum information matrix I _F, with spectrum information matrix I _FCarry out shifting function, and then obtain 2 dimension spectrum information matrix I of point centered by frequency (0,0) _FS, I _F, I _FSAll and image I _OWith size, I _F, I _FSWith image I _OMeet following relation:

I _F=fft2(I _O)

I _FS=fftshift(I _F)

Wherein fft2 represents to carry out quick 2 dimension Fourier transforms, and in the result images after the fft2 conversion, the corresponding value of information of frequency (0,0) is corresponding to spectrum information matrix I _FHeader element position, the upper left corner;

Fftshift represents spectrum information matrix I _FCarry out shifting function, namely take the matrix center as initial point, do the coordinate axis of horizontal and vertical, with 1,3 quadrant interchanges, the matrix that 2,4 quadrant interchanges are obtained; After the exchange, the corresponding value of information of frequency (0,0) is corresponding to the center of matrix;

In the 3rd step, utilize following Gauss's wedge shape wave filter formula, with spectrum information matrix I _FSCarry out the filtering of Gauss's wedge shape,

$G_1(x,y)=e^{-\frac{{(\mathrm{\θ}-{\mathrm{\θ}}_c)}^2}{2{\mathrm{\σ}}^2}}$

G ₂(x,y)=G ₁(-x,-y)

G(x,y)=G ₁(x,y)+G ₂(x,y)

Wherein, θ _CBe the angle center, its span is 0 °≤θ _c＜180 °

σ is standard deviation, has determined the width of Gaussian curve on the circumferential direction of certain radius

θ is spectrum information matrix I _FSIn certain first vegetarian refreshments be the summit with its frequency central point (0,0), take the angle value of horizontal dextrad direction as 0 °, as 0 °≤θ _cIn the time of＜180 °, the θ span is-90 °≤θ≤270 °, and the computing method of θ are:

Work as I _FSIn the coordinate (x, y) of certain element be positioned at 1, during 4 quadrant, have

$\mathrm{\θ}={\tan }^{-1}\left(\frac{y}{x}\right)$

Work as I _FSIn the coordinate (x, y) of certain pixel be positioned at 2, during 3 quadrant, have

$\mathrm{\θ}={\tan }^{-1}\left(\frac{y}{x}\right)+180$

G is 2 dimension Gauss wedge shape wave filters, and this is one is the Filtering Template of origin symmetry, i.e. G about frequency central point (0,0) ₁And G ₂About frequency central point (0,0) symmetry;

At first make θ _C=0 °, and according to the design parameter of each belt, the σ parameter of given Gauss's wedge shape wave filter is utilized Gauss's wedge shape Filtering Formula, generates about the symmetrical Gauss's wedge shape wave filter G of frequency central point (0,0);

The 4th step is with spectrum information matrix I _FSG carries out dot product with Gauss's wedge shape wave filter, obtains spectrum information matrix I _FL, this matrix has at utmost kept direction, the spectrum information of expectation texture, has abandoned direction, the spectrum information of other textures, and the dot product formula is as follows;

I _FL=I _FS·G

The 5th step is with spectrum information matrix I _FLCarry out shifting function, and carry out inverse fast Fourier transform, thereby with I _FLBe reduced to image I _R, I _FLWith I _RSatisfy following relation:

I _R=ifft2(fftshift(I _FL))

Wherein, fftshift represents spectrum information matrix I _FLCarry out shifting function, namely take the matrix center as initial point, do the coordinate axis of horizontal and vertical, with 1,3 quadrant interchanges, 2,4 quadrant interchanges are obtained new matrix; Before the exchange, the corresponding value of information of frequency (0,0) is corresponding to spectrum information matrix I _FLThe center; After the exchange, the corresponding value of information of frequency (0,0) is corresponding to the header element position, the upper left corner of matrix of consequence,

Ifft2 represents to carry out quick 2 dimension inverse Fourier transforms,

Through the image I after the filtering reduction _R, farthest kept the texture of certain direction, and the texture information of other directions is weakened;

In the 6th step, calculate I _RPixel grey scale average on the vertical direction of each horizontal ordinate position obtains one 1 vector of tieing up

Computing formula is as follows:

$V_{{\mathrm{\θ}}_C}\left(i\right)=\frac{1}{2^N}\underset{m=1}{\overset{2^N}{\mathrm{\Σ}}}{I}_R(i,m),i=\mathrm{1,2},\·\·\·,2^N$

Wherein, I _R(i, m) presentation video I _RIn the grey scale pixel value of certain pixel (i, m), i represents abscissa value, m represents ordinate value;

In the 7th step, make again θ _CBe respectively 1 °, 2 °, 3 ° ... 179 °, repeated for the 3rd step to the 6th step, obtain

V _θ(i)θ=0,1,…,179;i＝1,2,…2 ^N

And with following formula vector set V _θ(i) θ=0,1 ..., 179; I=1,2 ... 2 ^NBe rewritten as 2 dimension matrixes

V(θ，i)θ=0,1,…,179;i＝1,2,…2 ^N

Being that matrix V is one 2 dimension matrix, highly is 180, and width is 2 ^N, this 2 dimension matrix is shown with image mode, can observe each belt boundary position distribution situation;

The 8th step by the observation to 2 dimension image V, can pick out and each belt horizontal strip region of some high gray-scale values one to one, and hypothesis has M belt here, and then the horizontal strip region of high gray-scale value is made as Zone _iI=1,2, M, the number of the horizontal strip region of high gray-scale value is consistent with the belt number, simultaneously the horizontal strip region of each high gray-scale value is corresponding with some belts, and the width of the horizontal strip region of high gray-scale value equals the width of the belt corresponding with it, and the left and right sides boundary coordinate of the horizontal strip region of high gray-scale value equals the belt left and right sides border corresponding with it at I _OIn coordinate, with the horizontal strip region Zone of high gray-scale value _iI=1,2 ..., the abscissa value β of M _i

I=1,2 ..., M extracts, can extract with each belt one to one

I=1,2 ..., the M vector is for M

I=1,2 ..., the M vector can calculate M belt boundary information;

In the 9th step, calculate I=1,2 ..., the average of M

I=1,2 ..., M is with average

I=1,2 ..., M multiplication by constants coefficient lambda _iI=1,2 ..., M obtains threshold value T _iI=1,2 ..., M, namely satisfy following relational expression:

$T_i={\mathrm{\λ}}_i{\stackrel{\‾}{V}}_{{\mathrm{\β}}_i},i=\mathrm{1,2},\·\·\·,M$

For each vector

I=1,2 ..., M, (being x=0) begins search from the left side, when the first fit condition

$V_{{\mathrm{\β}}_i}\left(x\right)\≤T_i\≤V_{{\mathrm{\β}}_i}(x+1),x=\mathrm{1,2},\·\·\·,2^N-1;i=\mathrm{1,2},\·\·\·,M$

The time, then the x of this moment is designated as the left margin E on required border _LiI=1,2 ..., M;

In like manner, from vector I=1,2 ..., the right side of M (is x=2 ^N) beginning search left, when the first fit condition

$V_{{\mathrm{\β}}_i}\left(x\right)\≤T_i\≤V_{{\mathrm{\β}}_i}(x-1),x=2^N,2^N-1,\·\·\·,2;i=\mathrm{1,2},\·\·\·,M$

The time, then the x of this moment is designated as the right margin E on required border _RiI=1,2 ..., M;

At gray level image I _OIn, respectively with I _O(E _Li, y), I _O(E _Ri, y) i=1,2 ..., M; Y=1,2 ..., 2 ^NThe pixel value gray-scale value is set to (255,255,255), namely draws the boundary line of each belt with white lines;

So far, finish the belt boundary alignment.

2. method according to claim 1 is characterized in that: be applied to separability that tire x-ray image crown part position belt boundary alignment method utilized spectrum information to realize the belt boundary alignment, separability refers to:

Crown position at the tire x-ray image, a plurality of belt textures superpose mutually, therefore the crown position of the x-ray image of tire demonstrates is the texture of mutually overlapping each belt all-steel cord, each belt texture that the border of each belt is superposeed is mutually covered, tire x-ray image crown part bit image I _OIn each belt include different texture informations, contain frequency spectrum, direction etc., to tire x-ray image crown part bit image I _OAfter carrying out Fast Fourier Transform (FFT) and displacement, the spectrum information matrix I that obtains _FSIn, the texture information correspondence that different belts comprise spectrum information matrix I _FSMiddle different local location, the spectrum information that is in different local locations is separable, extracts spectrum information matrix I by Gauss's wedge shape wave filter _FSIn the information of certain regional area, just can obtain the texture information of expectation.

Images