CN103593838B - A kind of cross-correlation gray level image matching method and device fast - Google Patents

Abstract

The invention discloses a kind of cross-correlation gray level image matching method and device fast, for searching position and the direction of target pattern in gray scale target image, described method comprises: carry out polar coordinates conversion to template image; Obtain the external connection box of target pattern in gray scale target image, be defined as the target pattern field of search; In the target pattern field of search, choose the matching area onesize with template image, polar coordinates conversion is carried out to described matching area, calculate the correlativity of described matching area one-dimensional data and described template image one-dimension array; Until all complete correlation calculations to all matching areas, choose the maximum matching area of correlativity as target pattern, obtain position and the direction of described target pattern.By the inventive method, thus reduce the hunting zone of coupling, improve matching speed, and calculated by the normalized crosscorrelation of polar form, realize target image ± 180 degree full angle coupling.

Description

A kind of cross-correlation gray level image matching method and device fast

Technical field

The invention belongs to technical field of image processing, more specifically, relate to a kind of cross-correlation gray level image matching method and device fast.

Background technology

In integrated circuit (IC) manufacturing industry, high-speed, high-precisionly carry out picking up, chip placement is the key affecting production efficiency.And pick up, chip placement depends on high speed, high-precision chip positioning technology.Machine vision technique utilizes video camera to take pictures to detected object, then undertaken calculating by corresponding image processing algorithm and analyze, thus complete target detection and location, as the non-contacting measurement means of one, machine vision technique is widely applied in IC manufacturing industry.In machine vision, image matching technology is the key realizing high speed, hi-Fix.

Traditional normalized crosscorrelation Gray-scale Matching method carries out search coupling to entire image, wastes a large amount of match times.Simultaneously due to the machine error of IC manufacturing equipment own, usually occur that the pattern in die plate pattern and target image exists situation about rotating, due to the existence rotated, make image matching algorithm more complicated.

Summary of the invention

For above defect or the Improvement requirement of prior art, the invention provides the quick full angle of a kind of energy and be normalized cross-correlation Gray-scale Matching method, its object is to improve Gray-scale Matching speed, solve IC manufacturing equipment because images match overlong time thus, affect the technical matters of production efficiency.

For achieving the above object, according to one aspect of the present invention, providing a kind of cross-correlation gray level image matching method fast, for searching position and the direction of target pattern in gray scale target image, comprising:

(1) polar coordinates conversion is carried out to template image, and the template image data after conversion is converted to one-dimension array, be defined as template image one-dimension array;

(2) obtain the external connection box of target pattern in gray scale target image, be defined as the target pattern field of search, wherein said external connection box refers to the minimum rectangle comprising target pattern; Be specially:

Corresponding gray threshold is set by gray scale target image binaryzation, gray scale target image after binaryzation becomes black-and-white two color, element marking or distance of swimming connectivity analysis methods is adopted to carry out connected domain analysis to the gray scale target image after binaryzation, calculate the external connection box of target pattern, wherein external connection box refers to the minimum rectangle comprising target pattern;

The pixel coordinate designing the upper left corner relative target image coordinate initial point of the external connection box calculated is (X _b, Y _b), the width of external connection box is W _b, be highly H _b, then the upper left corner, the target pattern field of search in former target image is X relative to the coordinate of target image true origin _s0=X _b-Offset, Y _s0=Y _b-Offset, the width W of the target pattern field of search _s=W _b+ 2 × Offset is highly H _s=H _b+ 2 × Offset, wherein Offset is integer, and span is 5 ~ 10 pixels;

(3) in the target pattern field of search, the matching area onesize with template image is chosen, polar coordinates conversion is carried out to described matching area, and the matching area view data after conversion is also converted to one-dimension array, be defined as matching area one-dimension array, calculated the correlativity of described matching area one-dimensional data and described template image one-dimension array;

(4) repeat step (3), until all complete correlation calculations to all matching areas, choose the maximum matching area of correlativity as target pattern, obtain position and the direction of described target pattern.

By method proposed by the invention, pre-service is carried out to the image collected, determine the external connection box position of pattern in target image, thus reduce the hunting zone of coupling, improve matching speed; Convert template image and target search area image to one-dimension array in addition, the normalized crosscorrelation then carrying out polar form calculates, realize target image ± 180 degree full angle coupling.

Preferably, described step (1) specifically comprises:

If the width of template image is W _m, the height of template image is H _m, the grey scale pixel value of template image point is Mode [Y] [X], wherein 0≤Y<H _m, 0≤X<W _m, then the center point coordinate of template image is Y _cM=(H _m-1)/2, X _cM=(W _m-1)/2, with template image center for the center of circle, with template W _m/ 2 and H _mthe minimum value of/2 is radius R _m, carry out polar coordinates conversion, the image after conversion is expressed as the one-dimension array of grey scale pixel value: ModePolar [P]=Mode [Y] [X], wherein Y=FLOOR (Y _cM-ρ sin θ _m), X=FLOOR (X _cM+ ρ cos θ _m), FLOOR symbology gets the maximum integer less than the number in bracket, ρ=1,2 ... R _m, θ _m=0, Δ θ, 2 Δ θ ... 2 π, Δ θ are the precision of angle automatching, N=CEIL ((2 π/Δ θ+1) R _m-1), the smallest positive integral larger than the number in bracket is got in CEIL representative, and ModePolar [P] array is copied 1 part, 2 arrays join end to end and form new one-dimension array ModePolarD [Q], wherein Q=0,1,2 ... 2N, and calculate normalized crosscorrelation parameter:

$SM={\mathrm{\&Sigma;}}_{P=0}^{P=N}ModePolar\&lsqb;P\&rsqb;$

$S\left(M^2\right)={\mathrm{\&Sigma;}}_{P=0}^{P=N}{(ModePolar\&lsqb;P\&rsqb;)}^2$

${\mathrm{SM}}^2={({\mathrm{\&Sigma;}}_{P=0}^{P=N}ModePolar\&lsqb;P\&rsqb;)}^2$

By converting two-dimensional image data to one dimensional image data through polar coordinate transform, thus by the matching problem having the anglec of rotation of complexity, be converted to the problem that simple one-dimension array calculates cross correlation value, some parameters simultaneously in calculated off-line part normalized crosscorrelation value, improve the speed of On-line matching.

Preferably, the assignment procedure of described ModePolar [P] is:

(A11) initiation parameter: ρ=1, θ _m=0, P=0, Y _c=(H _m-1)/2, X _c=(W _m-1)/2;

(A12) calculating parameter: Y=FLOOR (Y _cM-ρ sin θ _m), X=FLOOR (X _cM+ ρ co θ _m);

(A13) data of template image are carried out polar coordinates conversion: ModePolar [P]=Mode [Y] [X];

(A14)ρ＝ρ+1，P＝P+1；

(A15) judge whether ρ is greater than R _mif be greater than, perform (A16) step, if be not more than, return execution (A12) step;

(A16)ρ＝0，θ _M＝θ _M+Δθ；

(A17) θ is judged _mwhether be greater than 2 π, if be greater than, terminate, if be not more than, return execution (A12) step.

Preferably, described step (2) is specially:

Corresponding gray threshold is set by gray scale target image binaryzation, gray scale target image after binaryzation becomes black-and-white two color, element marking or distance of swimming connectivity analysis methods is adopted to carry out connected domain analysis to the gray scale target image after binaryzation, calculate the external connection box of target pattern, wherein external connection box refers to the minimum rectangle comprising target pattern;

The pixel coordinate designing the upper left corner relative target image coordinate initial point of the external connection box calculated is (X _b, Y _b), the width of external connection box is W _b, be highly H _b, then the upper left corner, the target pattern field of search in former target image is X relative to the coordinate of target image true origin _s0=X _b-Offset, Y _s0=Y _b-Offset, the width W of the target pattern field of search _s=W _b+ 2 × Offset is highly H _s=H _b+ 2 × Offset, wherein Offset is integer, and span is 5 ~ 10 pixels.

By calculating external connection box, thus fast target pattern is carried out coarse localization, thus reduced the hunting zone that next step carries out Cross Correlation Matching, improve images match speed.

Preferably, described step (3) specifically comprises:

From former target image, intercept the image in the target pattern field of search, be called searching image, then the width of searching image is W _s, be highly H _s, searching image X-coordinate is expressed as X _s, Y-coordinate is expressed as Y _s, MaxCcorr represents the maximum cross-correlation value of searching image and template image, is normalized cross-correlation Gray-scale Matching in searching image, and coupling flow process is as follows:

B1) initiation parameter: X _s=0, Y _s=0, MaxCcorr=0;

B2) Y is judged _swhether be less than H _sif be not less than, terminate, if be less than, perform B3);

B3) X is judged _swhether be less than W _sif be not less than, X _s=0, Y _s=Y _s+ 1, return and perform B2) step, if be less than, perform B4) step;

B4) carry out polar coordinates conversion to image-region identical with template size in searching image, the image in this region is called matching area image; The width of matching area image is identical with template image with height is respectively W _m, be highly H _m, in matching area, the grey scale pixel value of certain point is Image [Y] [X], wherein Y _s≤ Y<Y _s+ H _m, X _s≤ X<X _s+ W _m, then the center point coordinate of matching area is Y _cI=Y _s+ (H _m-1)/2, X _cI=X _s+ (W _m-1)/2, with matching area center for the center of circle, with W _m/ 2 and H _mthe minimum value of/2 is radius R _m, carry out polar coordinates conversion, the image after conversion is expressed as one-dimension array ImagePolar [P]=Image [Y] [X] of grey scale pixel value, Y=FLOOR (Y _cI-ρ sin θ _i), X=FLOOR (X _cI+ ρ cos θ _i), FLOOR representative gets the maximum integer less than the number in bracket, wherein ρ=1,2 ... R _m, θ _i=0, Δ θ, 2 Δ θ ... 2 π, Δ θ are the precision of angle automatching, P=0,1,2 ... N, N=CEIL ((2 π/Δ θ+1) R _m-1), the smallest positive integral larger than the number in bracket is got in CEIL representative;

B5) normalized crosscorrelation parameter is calculated:

$SI={\mathrm{\&Sigma;}}_{P=0}^{P=N}\mathrm{Im}agePolar\&lsqb;P\&rsqb;$

$S\left(I^2\right)={\mathrm{\&Sigma;}}_{P=0}^{P=N}{(\mathrm{Im}agePolar\&lsqb;P\&rsqb;)}^2$

${\mathrm{SI}}^2={({\mathrm{\&Sigma;}}_{P=0}^{P=N}\mathrm{Im}agePolar\&lsqb;P\&rsqb;)}^2$

B6) initial integer i=0;

B7) judge whether i is greater than N, if be greater than, X _s=X _s+ 1, return and perform B3), otherwise perform B8);

B8) the normalized crosscorrelation value Ccorr of calculation template image and matching area _i:

B9) Ccorr is judged _iwhether be greater than MaxCcorr, if be not more than, perform B10), if be greater than, by Ccorr _iassignment to MaxCcorr, and records the coordinate (X of now matching area central point _t, Y _t), wherein X _t=X _s+ (W _m-1)/2, Y _t=Y _s+ (H _m-1)/2, recording angular θ _t=Δ θ FLOOR ((P+R _m)/R _m);

B10) i=i+1 returns and performs B7) step.

By adopting, template image and target image are carried out the mode that polar coordinates convert one-dimension array to, thus by the rotation matching of template image and target image, be converted between two one-dimension array and carry out the problem that relative translation calculates cross correlation value, avoid and piece image is rotated once every certain angle, then problem is once mated, thus shorten match time, improve matching speed.

Preferably, the assignment procedure of described ImagePolar [P] is:

(B41) initiation parameter: ρ=1, θ _i=0, P=0, Y _cI=Y _s+ (H _m-1)/2, X _cI=X _s+ (W _m-1)/2;

(B42) calculating parameter: Y=FLOOR (Y _cI-ρ sin θ _i), X=FLOOR (X _cI+ ρ cos θ _i);

(B43) data of matching area are carried out polar coordinates conversion: ImagePolar [P]=Image [Y] [X];

(B44)ρ＝ρ+1，P＝P+1；

(B45) judge whether ρ is greater than R _mif be greater than, perform (B46) step, if be not more than, return execution (B42) step;

(B46)ρ＝0，θ _I＝θ _I+Δθ；

(B47) θ is judged _iwhether be greater than 2 π, if be greater than, terminate, if be not more than, return execution (B42) step.

Preferably, described step (4) is specially:

If at B2) in step, YS is no longer less than HS, then calculate end, after correlativity meter is all completed to all matching areas, obtain template image position (XF in the target image, and angle θ F YF), wherein XF=XT+Xs0, YF=YT+Ys0, θ F=θ T, wherein XT, YT and θ T is B1) ~ B10) calculate the value of finally preserving before terminating.

According to another aspect of the present invention, provide a kind of cross-correlation Image Matching device fast, it is characterized in that, for searching position and the direction of target pattern in gray scale target image, described device comprises:

First module, for carrying out polar coordinates conversion to template image, and being converted to one-dimension array by the template image data after conversion, being defined as template image one-dimension array;

Second module, for obtaining the external connection box of target pattern in gray scale target image, is defined as the target pattern field of search; Be specially:

Corresponding gray threshold is set by gray scale target image binaryzation, gray scale target image after binaryzation becomes black-and-white two color, element marking or distance of swimming connectivity analysis methods is adopted to carry out connected domain analysis to the gray scale target image after binaryzation, calculate the external connection box of target pattern, wherein external connection box refers to the minimum rectangle comprising target pattern;

The pixel coordinate designing the upper left corner relative target image coordinate initial point of the external connection box calculated is (X _b, Y _b), the width of external connection box is W _b, be highly H _b, then the upper left corner, the target pattern field of search in former target image is X relative to the coordinate of target image true origin _s0=X _b-Offset, Y _s0=Y _b-Offset, the width W of the target pattern field of search _s=W _b+ 2 × Offset is highly H _s=H _b+ 2 × Offset, wherein Offset is integer, and span is 5 ~ 10 pixels;

3rd module, for choosing the matching area onesize with template image in the target pattern field of search, polar coordinates conversion is carried out to described matching area, and the matching area view data after conversion is also converted to one-dimension array, be defined as matching area one-dimension array, calculated the correlativity of described matching area one-dimensional data and described template image one-dimension array;

Four module, when described 3rd module all completes correlation calculations to all matching areas, chooses the maximum matching area of correlativity as target pattern, obtains position and the direction of described target pattern.

By method provided by the present invention, the invention has the beneficial effects as follows:

1. the image chip pattern collected in manufacturing for IC differs larger feature with chip background gray levels, by adding the preprocess method image collected being carried out to binaryzation and connected domain analysis, determine the external connection box position of pattern in target image, thus reduce the hunting zone of coupling, improve matching speed.

2. adopt and convert template image and target search area image to one-dimension array, the normalized crosscorrelation then carrying out polar form calculates, thus realize target image ± 180 degree full angle coupling.

Accompanying drawing explanation

Fig. 1 is the overall flow figure of quick cross-correlation gray level image matching method of the present invention;

Fig. 2 is the polar coordinates flow path switch figure of template image in an embodiment constructed by the present invention;

Fig. 3 is target image search pattern image process flow diagram in an embodiment constructed by the present invention;

Fig. 4 is matching area image polar coordinates flow path switch figure in an embodiment constructed by the present invention.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.In addition, if below in described each embodiment of the present invention involved technical characteristic do not form conflict each other and just can mutually combine.

As shown in Figure 1, the invention provides a kind of cross-correlation gray level image matching method fast, for searching position and the direction of target pattern in gray scale target image, described method comprises:

(1) polar coordinates conversion is carried out to template image, and the template image data after conversion is converted to one-dimension array, be defined as template image one-dimension array;

(2) obtain the external connection box of target pattern in gray scale target image, be defined as the target pattern field of search;

(3) in the target pattern field of search, the matching area onesize with template image is chosen, polar coordinates conversion is carried out to described matching area, and the matching area view data after conversion is also converted to one-dimension array, be defined as matching area one-dimension array, calculated the correlativity of described matching area one-dimensional data and described template image one-dimension array;

(4) repeat step (3), until all complete correlation calculations to all matching areas, choose the maximum matching area of correlativity as target pattern, obtain position and the direction of described target pattern.

By method proposed by the invention, pre-service is carried out to the image collected, determine the external connection box position of pattern in target image, thus reduce the hunting zone of coupling, improve matching speed; Convert template image and target search area image to one-dimension array in addition, the normalized crosscorrelation then carrying out polar form calculates, realize target image ± 180 degree full angle coupling.

Concrete, followingly in the present invention, the matching process of target pattern in gray level image to be elaborated, the two-dimensional array in ranks direction can be expressed as gray level image, the value of each array element represents the grey scale pixel value of this point, with the initial point that the image upper left corner is image coordinate, vertical downward direction is Y positive dirction, simultaneously also for line number label increases progressively direction, represent the short transverse of image, horizontal right direction is X positive dirction, simultaneously also for columns label increases progressively direction, represent the Width of image.

This method is specifically divided into off-line phase and on-line stage.

Off-line phase, first carry out pre-service to template image, concrete steps are as follows:

A1) carry out polar coordinates conversion to template image, the data after conversion save as one-dimension array, and concrete steps are as follows:

Suppose that the width of template image is W _m, the height of template image is H _m, the grey scale pixel value of template image point is Mode [Y] [X], wherein 0≤Y<H _m, 0≤X<W _m, then the center point coordinate of template image is Y _cM=(H _m-1)/2, X _cM=(W _m-1)/2, with template image center for the center of circle, with template W _m/ 2 and H _mthe minimum value of/2 is radius R _m, carry out polar coordinates conversion, the image after conversion is expressed as the one-dimension array of grey scale pixel value:

ModePolar[P]＝Mode[Y][X]

Y＝FLOOR(Y _CM-ρsinθ _M)

X＝FLOOR(X _CM+ρcosθ _M)

Wherein FLOOR symbology gets the maximum integer less than the number in bracket, ρ=1,2 ... R _m, θ _m=0, Δ θ, 2 Δ θ ... 2 π, Δ θ are the precision of angle automatching, such as, require that the angle precision mated is 1 degree, then Δ θ=π/180, P=0,1,2 ... N, N=CEIL ((2 π/Δ θ+1) R _m-1), the maximum integer larger than the number in bracket is got in CEIL representative, and as shown in Figure 2, the assignment procedure of described ModePolar [P] is:

(A11) initiation parameter: ρ=1, θ _m=0, P=0, Y _c=(H _m-1)/2, X _c=(W _m-1)/2;

(A12) calculating parameter: Y=FLOOR (Y _cM-ρ sin θ _m), X=FLOOR (X _cM+ ρ co θ _m);

(A13) data of template image are carried out polar coordinates conversion: ModePolar [P]=Mode [Y] [X];

(A14)ρ＝ρ+1，P＝P+1；

(A15) judge whether ρ is greater than R _mif be greater than, performed the 6th step, if be not more than, return execution (2) step;

(A16)ρ＝0，θ _M＝θ _M+Δθ；

(A17) θ is judged _mwhether be greater than 2 π, if be greater than, terminate A1) perform A2), if be not more than, return execution (2) step.

A2) partial parameters in normalized crosscorrelation formula is calculated:

$SM={\mathrm{\&Sigma;}}_{P=0}^{P=N}ModePolar\&lsqb;P\&rsqb;$

$S\left(M^2\right)={\mathrm{\&Sigma;}}_{P=0}^{P=N}{(ModePolar\&lsqb;P\&rsqb;)}^2$

${\mathrm{SM}}^2={({\mathrm{\&Sigma;}}_{P=0}^{P=N}ModePolar\&lsqb;P\&rsqb;)}^2$

A3) ModePolar [P] array is copied 1 part, 2 arrays join end to end and form new one-dimension array ModePolarD [Q], wherein Q=0,1,2 ... 2N.

On-line stage, for search pattern picture pattern in the target image, concrete steps are as follows:

B1) former target image is copied, by arranging corresponding gray threshold, the image after copy is carried out binaryzation, image after binaryzation becomes black-and-white two color, because the image chip pattern collected in IC manufacture differs larger with chip background gray levels, after binaryzation, the major part of chip becomes a kind of look, background becomes look in contrast, and such as chip major part is black, and background is white.Adopt as element marking, distance of swimming connectivity analysis methods carry out connected domain analysis to the target image after binaryzation, calculate the external connection box in chip image region, wherein external connection box refers to the minimum rectangle comprising target area (referring to chip design here), and general rectangle is towards coordinate axis.The pixel coordinate supposing the upper left corner relative target image coordinate initial point of the external connection box calculated is (X _b, Y _b), the width of external connection box is W _b, be highly H _b, then the upper left corner, region of search in former target image is X relative to the coordinate of target image true origin _s0=X _b-Offset, Y _s0=Y _b-Offset, the width W of region of search _s=W _b+ 2 × Offset is highly H _s=H _b+ 2 × Offset, wherein Offset is integer, and span is 5 ~ 10 pixels.

From former target image, intercept the image in region of search, be called searching image, then the width of searching image is W _s, be highly H _s, searching image X-coordinate is expressed as X _s, Y-coordinate is expressed as Y _s, MaxCcorr represents the maximum cross-correlation value of searching image and template image, is normalized cross-correlation Gray-scale Matching in searching image, and as shown in Figure 3, described coupling flow process is as follows:

B2) initiation parameter: X _s=0, Y _s=0, MaxCcorr=0;

B3) Y is judged _swhether be less than H _sif be not less than, terminate, if be less than, perform B4).

B4) X is judged _swhether be less than W _sif be not less than, X _s=0, Y _s=Y _s+ 1, return and perform B3) step, if be less than, perform B5) step.

B5) carry out polar coordinates conversion to image-region identical with template size in searching image, the image in this region is called matching area image.The width of matching area image is identical with template image with height is respectively W _m, be highly H _m, in matching area, the grey scale pixel value of certain point is Image [Y] [X], wherein Y _s≤ Y<Y _s+ H _m, X _s≤ X<X _s+ W _m, then the center point coordinate of matching area is Y _cI=Y _s+ (H _m-1)/2, X _cI=X _s+ (W _m-1)/2, with matching area center for the center of circle, with W _m/ 2 and H _mthe minimum value of/2 is radius R _m, carry out polar coordinates conversion, the image after conversion is expressed as one-dimension array ImagePolar [P]=Image [Y] [X] of grey scale pixel value, Y=FLOOR (Y _cI-ρ sin θ _i), X=FLOOR (X _cI+ ρ cos θ _i), FLOOR representative gets the maximum integer less than the number in bracket, wherein ρ=1,2 ... R _m, θ _i=0, Δ θ, 2 Δ θ ... 2 π, Δ θ are the precision of angle automatching, identical with the precision of template image, such as, require that the angle precision mated is 1 degree, then Δ θ=π/180, P=0,1,2 ... N, N=CEIL ((2 π/Δ θ+1) R _m-1), the maximum integer larger than the number in bracket is got in CEIL representative, and as shown in Figure 4, the assignment procedure of described ImagePolar [P] is:

(B51) initiation parameter: ρ=1, θ _i=0, P=0, Y _cI=Y _s+ (H _m-1)/2, X _cI=X _s+ (W _m-1)/2;

(B52) calculating parameter: Y=FLOOR (Y _cI-ρ sin θ _i), X=FLOOR (X _cI+ ρ cos θ _i);

(B53) data of matching area are carried out polar coordinates conversion: ImagePolar [P]=Image [Y] [X];

(B54)ρ＝ρ+1，P＝P+1；

(B55) judge whether ρ is greater than R _mif be greater than, perform (6) step, if be not more than, return execution (2) step;

(B56)ρ＝0，θ _I＝θ _I+Δθ；

(B57) θ is judged _iwhether be greater than 2 π, if be greater than, terminate B5), perform B6), if be not more than, return execution (2) step.

B6) normalized crosscorrelation parameter is calculated:

$SI={\mathrm{\&Sigma;}}_{P=0}^{P=N}\mathrm{Im}agePolar\&lsqb;P\&rsqb;,$

$S\left(I^2\right)={\mathrm{\&Sigma;}}_{P=0}^{P=N}{(\mathrm{Im}agePolar\&lsqb;P\&rsqb;)}^2,$

${\mathrm{SI}}^2={({\mathrm{\&Sigma;}}_{P=0}^{P=N}\mathrm{Im}agePolar\&lsqb;P\&rsqb;)}^2.$

B7) initial integer i=0;

B8) judge whether i is greater than N, if be greater than, X _s=X _s+ 1, return and perform B4), otherwise perform B9);

B9) the normalized crosscorrelation value Ccorr of calculation template image and matching area _i:

B10) Ccorr is judged _iwhether be greater than MaxCcorr, if be not more than, perform B11), if be greater than, by Ccorr _iassignment to MaxCcorr, and records the coordinate (X of now matching area central point _t, Y _t), wherein X _t=X _s+ (W _m-1)/2, Y _t=Y _s+ (H _m-1)/2, recording angular θ _t=Δ θ FLOOR ((P+R _m)/R _m);

B11) i=i+1 returns and performs B8) step.

Through B1) ~ B11) after step, if at B3) and in step, Y _sno longer be less than H _s, then end is calculated.Template image position (X in the target image can be obtained _f, Y _f) and angle θ _f, wherein X _f=X _t+ X _s0, Y _f=Y _t+ Y _s0, θ _f=θ _t, wherein X _t, Y _tand θ _tfor B1) ~ B11) calculate the value of finally preserving before terminating.

Those skilled in the art will readily understand; the foregoing is only preferred embodiment of the present invention; not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (7)

1. a quick cross-correlation gray level image matching method, is characterized in that, for searching position and the direction of target pattern in gray scale target image, comprising:

(1) polar coordinates conversion is carried out to template image, and the template image data after conversion is converted to one-dimension array, be defined as template image one-dimension array;

(2) obtain the external connection box of target pattern in gray scale target image, be defined as the target pattern field of search, wherein said external connection box refers to the minimum rectangle comprising target pattern; Be specially:

Corresponding gray threshold is set by gray scale target image binaryzation, gray scale target image after binaryzation becomes black-and-white two color, element marking or distance of swimming connectivity analysis methods is adopted to carry out connected domain analysis to the gray scale target image after binaryzation, calculate the external connection box of target pattern, wherein external connection box refers to the minimum rectangle comprising target pattern;

The pixel coordinate designing the upper left corner relative target image coordinate initial point of the external connection box calculated is (X _b, Y _b), the width of external connection box is W _b, be highly H _b, then the upper left corner, the target pattern field of search in former target image is X relative to the coordinate of target image true origin _s0=X _b-Offset, Y _s0=Y _b-Offset, the width W of the target pattern field of search _s=W _b+ 2 × Offset is highly H _s=H _b+ 2 × Offset, wherein Offset is integer, and span is 5 ~ 10 pixels;

(3) in the target pattern field of search, the matching area onesize with template image is chosen, polar coordinates conversion is carried out to described matching area, and the matching area view data after conversion is also converted to one-dimension array, be defined as matching area one-dimension array, calculated the correlativity of described matching area one-dimensional data and described template image one-dimension array;

(4) repeat step (3), until all complete correlation calculations to all matching areas, choose the maximum matching area of correlativity as target pattern, obtain position and the direction of described target pattern.

2. the method for claim 1, is characterized in that, described step (1) specifically comprises: set the width of template image as W _m, the height of template image is H _m, the grey scale pixel value of template image point is Mode [Y] [X], wherein 0≤Y<H _m, 0≤X<W _m, then the center point coordinate of template image is Y _cM=(H _m-1)/2, X _cM=(W _m-1)/2, with template image center for the center of circle, with template W _m/ 2 and H _mthe minimum value of/2 is radius R _m, carry out polar coordinates conversion, the image after conversion is expressed as the one-dimension array of grey scale pixel value: ModePolar [P]=Mode [Y] [X], wherein Y=FLOOR (Y _cM-ρ sin θ _m), X=FLOOR (X _cM+ ρ cos θ _m), FLOOR symbology gets the maximum integer less than the number in bracket, ρ=1,2 ... R _m, θ _m=0, Δ θ, 2 Δ θ ... 2 π, Δ θ are the precision of angle automatching, N=CEIL ((2 π/Δ θ+1) R _m-1), the smallest positive integral larger than the number in bracket is got in CEIL representative, and ModePolar [P] array is copied 1 part, 2 arrays join end to end and form new one-dimension array ModePolarD [Q], wherein Q=0,1,2 ... 2N, and calculate normalized crosscorrelation parameter:

$SM={\mathrm{\&Sigma;}}_{P=0}^{P=N}ModePolar\&lsqb;P\&rsqb;$

$S\left(M^2\right)={\mathrm{\&Sigma;}}_{P=0}^{P=N}{(ModePolar\&lsqb;P\&rsqb;)}^2$

${\mathrm{SM}}^2={({\mathrm{\&Sigma;}}_{P=0}^{P=N}ModePolar\&lsqb;P\&rsqb;)}^2$

$corrDen=\sqrt{N\&CenterDot;S\left(M^2\right)-{\mathrm{SM}}^2}.$

3. method as claimed in claim 2, it is characterized in that, the assignment procedure of described ModePolar [P] is:

(A11) initiation parameter: ρ=1, θ _m=0, P=0, Y _c=(H _m-1)/2, X _c=(W _m-1)/2;

(A12) calculating parameter: Y=FLOOR (Y _cM-ρ sin θ _m), X=FLOOR (X _cM+ ρ co θ _m);

(A13) data of template image are carried out polar coordinates conversion: ModePolar [P]=Mode [Y] [X];

(A14)ρ＝ρ+1，P＝P+1；

(A15) judge whether ρ is greater than R _mif be greater than, perform (A16) step, if be not more than, return execution (A12) step;

(A16)ρ＝0，θ _M＝θ _M+Δθ；

(A17) θ is judged _mwhether be greater than 2 π, if be greater than, terminate, if be not more than, return execution (A12) step.

4. the method for claim 1, is characterized in that, described step (3) specifically comprises:

From former target image, intercept the image in the target pattern field of search, be called searching image, then the width of searching image is W _s, be highly H _s, searching image X-coordinate is expressed as X _s, Y-coordinate is expressed as Y _s, MaxCcorr represents the maximum cross-correlation value of searching image and template image, is normalized cross-correlation Gray-scale Matching in searching image, and coupling flow process is as follows:

B1) initiation parameter: X _s=0, Y _s=0, MaxCcorr=0;

B2) Y is judged _swhether be less than H _sif be not less than, terminate, if be less than, perform B3);

B3) X is judged _swhether be less than W _sif be not less than, X _s=0, Y _s=Y _s+ 1, return and perform B2) step, if be less than, perform B4) step;

B4) carry out polar coordinates conversion to image-region identical with template size in searching image, the image in this region is called matching area image; The width of matching area image is identical with template image with height is respectively W _m, be highly H _m, in matching area, the grey scale pixel value of certain point is Image [Y] [X], wherein Y _s≤ Y<Y _s+ H _m, X _s≤ X<X _s+ W _m, then the center point coordinate of matching area is Y _cI=Y _s+ (H _m-1)/2, X _cI=X _s+ (W _m-1)/2, with matching area center for the center of circle, with W _m/ 2 and H _mthe minimum value of/2 is radius R _m, carry out polar coordinates conversion, the image after conversion is expressed as one-dimension array ImagePolar [P]=Image [Y] [X] of grey scale pixel value, Y=FLOOR (Y _cI-ρ sin θ _i), X=FLOOR (X _cI+ ρ cos θ _i), FLOOR representative gets the maximum integer less than the number in bracket, wherein ρ=1,2 ... R _m, θ _i=0, Δ θ, 2 Δ θ ... 2 π, Δ θ are the precision of angle automatching, P=0,1,2 ... N, N=CEIL ((2 π/Δ θ+1) R _m-1), the smallest positive integral larger than the number in bracket is got in CEIL representative;

B5) normalized crosscorrelation parameter is calculated:

$SI={\mathrm{\&Sigma;}}_{P=0}^{P=N}\mathrm{Im}agePolar\&lsqb;P\&rsqb;$

$S\left(I^2\right)={\mathrm{\&Sigma;}}_{P=0}^{P=N}{(\mathrm{Im}agePolar\&lsqb;P\&rsqb;)}^2$

${\mathrm{SI}}^2={({\mathrm{\&Sigma;}}_{P=0}^{P=N}\mathrm{Im}agePolar\&lsqb;P\&rsqb;)}^2$

B6) initial integer i=0;

B7) judge whether i is greater than N, if be greater than, X _s=X _s+ 1, return and perform B3), otherwise perform B8);

B8) the normalized crosscorrelation value Ccorr of calculation template image and matching area _i:

$S_i\left(IM\right)={\mathrm{\&Sigma;}}_{P=i}^{P=N+i}(\mathrm{Im}agePolar\&lsqb;P-i\&rsqb;\&CenterDot;ModePolarD\&lsqb;P\&rsqb;)$

${\mathrm{Ccorr}}_i=\frac{N\&CenterDot;S_i\left(IM\right)-SI\&CenterDot;SM}{\sqrt{|N\&CenterDot;S\left(I^2\right)-{\mathrm{SI}}^2||N\&CenterDot;S\left(M^2\right)-{\mathrm{SM}}^2|}}$

B9) Ccorr is judged _iwhether be greater than MaxCcorr, if be not more than, perform B10), if be greater than, by Ccorr _iassignment to MaxCcorr, and records the coordinate (X of now matching area central point _t, Y _t), wherein X _t=X _s+ (W _m-1)/2, Y _t=Y _s+ (H _m-1)/2, recording angular θ _t=Δ θ FLOOR ((P+R _m)/R _m);

B10) i=i+1 returns and performs B7) step.

5. method as claimed in claim 4, it is characterized in that, the assignment procedure of described ImagePolar [P] is:

(B41) initiation parameter: ρ=1, θ _i=0, P=0, Y _cI=Y _s+ (H _m-1)/2, X _cI=X _s+ (W _m-1)/2;

(B42) calculating parameter: Y=FLOOR (Y _cI-ρ sin θ _i), X=FLOOR (X _cI+ ρ cos θ _i);

(B43) data of matching area are carried out polar coordinates conversion: ImagePolar [P]=Image [Y] [X];

(B44)ρ＝ρ+1，P＝P+1；

(B45) judge whether ρ is greater than R _mif be greater than, perform (B46) step, if be not more than, return execution (B42) step;

(B46)ρ＝0，θ _I＝θ _I+Δθ；

(B47) θ is judged _iwhether be greater than 2 π, if be greater than, terminate, if be not more than, return execution (B42) step.

6. method as claimed in claim 4, it is characterized in that, described step (4) is specially:

If at B2) in step, YS is no longer less than HS, then calculate end, after all completing correlativity meter to all matching areas, obtains template image position (X in the target image _f, Y _f) and angle θ _f, wherein X _f=X _t+ X _s0, Y _f=Y _t+ Y _s0, θ _f=θ _t,wherein X _t, Y _tand θ _tfor B1) ~ B10) calculate the value of finally preserving before terminating.

7. a quick cross-correlation Image Matching device, is characterized in that, for searching position and the direction of target pattern in gray scale target image, described device comprises:

First module, for carrying out polar coordinates conversion to template image, and being converted to one-dimension array by the template image data after conversion, being defined as template image one-dimension array;

Second module, for obtaining the external connection box of target pattern in gray scale target image, is defined as the target pattern field of search; Be specially:

Corresponding gray threshold is set by gray scale target image binaryzation, gray scale target image after binaryzation becomes black-and-white two color, element marking or distance of swimming connectivity analysis methods is adopted to carry out connected domain analysis to the gray scale target image after binaryzation, calculate the external connection box of target pattern, wherein external connection box refers to the minimum rectangle comprising target pattern;

The pixel coordinate designing the upper left corner relative target image coordinate initial point of the external connection box calculated is (X _b, Y _b), the width of external connection box is W _b, be highly H _b, then the upper left corner, the target pattern field of search in former target image is X relative to the coordinate of target image true origin _s0=X _b-Offset, Y _s0=Y _b-Offset, the width W of the target pattern field of search _s=W _b+ 2 × Offset is highly H _s=H _b+ 2 × Offset, wherein Offset is integer, and span is 5 ~ 10 pixels;

3rd module, for choosing the matching area onesize with template image in the target pattern field of search, polar coordinates conversion is carried out to described matching area, and the matching area view data after conversion is also converted to one-dimension array, be defined as matching area one-dimension array, calculated the correlativity of described matching area one-dimensional data and described template image one-dimension array;

Four module, when described 3rd module all completes correlation calculations to all matching areas, chooses the maximum matching area of correlativity as target pattern, obtains position and the direction of described target pattern.